The Handbook of Rational Choice Social Research 9780804785501

Offers the first comprehensive overview of how the rational choice paradigm can inform empirical research within the soc

224 111 4MB

English Pages 624 [614] Year 2013

Report DMCA / Copyright

DOWNLOAD FILE

Polecaj historie

The Handbook of Rational Choice Social Research
 9780804785501

Citation preview

The Handbook of Rational Choice Social Research

The HANDBOOK of RATIONAL CHOICE SOCIAL RESEARCH edited by Rafael Wittek Tom A. B. Snijders Victor Nee

Stanford Social Sciences An Imprint of Stanford University Press

Stanford University Press Stanford, California ©2013 by the Board of Trustees of the Leland Stanford Junior University. All rights reserved. No part of this book may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying and recording, or in any information storage or retrieval system, without the prior written permission of Stanford University Press. Library of Congress Cataloging-in-Publication Data The handbook of rational choice social research / edited by Rafael Wittek, Tom A.B. Snijders, and Victor Nee p. cm. Includes bibliographical references and index. ISBN 978-0-8047-8418-4 (cloth : alk. paper) 1. Rational choice theory. I. Wittek, Rafael, editor of compilation. II. Snijders, T. A. B., editor of compilation. III. Nee,Victor, editor of compilation. HM495.H36 2013 330.01—dc23 Typeset at Stanford University Press in 10.5/12 Bembo

2012033972

Acknowledgments

The editors would like to thank the Russell Sage Foundation for generous support of an author conference at New York City in November 2007, Suzanne Nichols for guidance during the early phase of this project, and four anonymous reviewers for their constructive suggestions. We also would like to thank Kate Wahl from Stanford University Press for her encouragement, patience, and formidable support. Saskia Simon from the Department of Sociology at the University of Groningen deserves a very special thanks for her high-quality secretarial support in all phases of this project, including the final preparation of the manuscript. Rafael Wittek also gratefully acknowledges funding from the Netherlands’ Organization for Scientific Research (NWO) (NWO 016005-052 and 400-05-704).

Contents

Contributors

ix

Introduction: Rational Choice Social Research Rafael Wittek,Tom A. B. Snijders, and Victor Nee

1

Part I: Rationality and Decision-making 1 Rationality, Social Preferences, and Strategic Decision-making from a Behavioral Economics Perspective 33 Simon Gächter 2 Social Rationality, Self-Regulation, and Well-Being: The Regulatory Significance of Needs, Goals, and the Self 72 Siegwart Lindenberg 3 Rational Choice Research on Social Dilemmas: Embeddedness Effects on Trust 113 Vincent Buskens and Werner Raub 4 Modeling Collective Decision-making 151 Frans N. Stokman, Jelle Van der Knoop, and Reinier C. H.Van Oosten Part II: Networks and Inequality 5 Social Exchange, Power, and Inequality in Networks Karen S. Cook and Coye Cheshire 6 Social Capital 220 Henk Flap and Beate Völker 7 Network Dynamics Tom A. B. Snijders

252

185

viii

Contents

Part III: Communities and Cohesion 8 Rational Choice Research in Criminology: A Multi-Level Framework 283 Ross L. Matsueda 9 Secularization: Theoretical Controversies Generating Empirical Research 322 Nan Dirk De Graaf 10 Assimilation as Rational Action in Contexts Defined by Institutions and Boundaries 355 Victor Nee and Richard Alba Part IV: States and Conflicts 11 Terrorism and the State Ignacio Sánchez-Cuenca

381

12 Choosing War: State Decisions to Initiate and End Wars and Observe the Peace Afterward 411 James D. Morrow 13 Rational Choice Approaches to State-Making Edgar Kiser and Erin Powers

443

Part V: Markets and Organizations 14 Market Design and Market Failure 473 Carlos Cañón, Guido Friebel, and Paul Seabright 15 Organizational Governance 513 Nicolai J. Foss and Peter G. Klein 16 Rational Choice and Organizational Change Rafael Wittek and Arjen Van Witteloostuijn Index

589

556

Contributors

richard alba: Distinguished Professor of Sociology at the Center for Urban Research, City University of New York (CUNY). His fields of interest are race, ethnicity, and immigration. His recent books include Blurring the Color Line: The New Chance for a More Integrated America (Harvard University Press, 2009), The Next Generation: Immigrant Youth in a Comparative Perspective (coedited with Mary Waters, NYU Press, 2011), and the award-winning Remaking the American Mainstream: Assimilation and Contemporary Immigration (with V. Nee, Harvard University Press, 2003). vincent buskens: Professor of Theoretical Sociology, Utrecht University and Interuniversity Center for Social Science Theory and Methodology (ICS) and Professor of Empirical Legal Studies, Erasmus School of Law, Erasmus University Rotterdam. His main research areas are social dilemmas, cooperation, trust, social networks, mathematical sociology, complexity, experimental research, empirical studies on regulation and institutions, sociological applications of neuroscience. He is co-editor of Micro-Macro Links and Microfoundations (with W. Raub and M. van Assen, Routledge, 2012) and eTrust: Forming Relations in the Online World (with K. Cook, C. Snijders and C. Cheshire, Russell Sage, 2009). carlos cañón: Economist at the Financial Stability Division of Banco de México (Mexican Central Bank). He received his PhD from the Toulouse School of Economics with a dissertation on the Theory and Econometric Models of Platforms. His main area of interest is industrial organization, market design, and econometrics. coye cheshire: Associate professor at the UC Berkeley School of Information. He studies social exchange, interpersonal relationships, and paths of participation in computer-mediated environments. Recent articles have appeared in Personality and Social Psychology Bulletin, Daedalus, and Proceedings of the ACM Conference of Computer-Supported Cooperative Work (CSCW). karen s. cook: Ray Lyman Wilbur Professor of Sociology at Stanford University. Her fields of interest are social psychology, organizational

x

Contributors

behavior, group processes, social networks, and health care. For the Russell Sage Foundation, she has edited Trust in Society (2001), Trust and Distrust in Organizations: Emerging Perspectives (with R. Kramer, 2004), eTrust: Forming Relations in the Online World (with C. Snijders, V. Buskens and C. Cheshire, 2009), and Whom Can We Trust? (with M. Levi and R. Hardin, 2009).

nan dirk de graaf : Professor of Sociology and Official Fellow, Nuffield College, University of Oxford. His main research areas are social stratification, cultural sociology, sociology of religion, pro-social behavior, and political sociology. He edited with Geoff Evans Political Choice Matters: Explaining the Strength of Class and Religious Cleavages in Cross-National Perspective (Oxford University Press, 2013). Recent articles have appeared in American Sociological Review, Sociology of Health and Illness, Journal of the Scientific Study of Religion, Research in Social Stratification and Mobility, and Journal of Quantitative Criminology. henk flap: Professor of Sociology Utrecht University and Interuniversity Center for Social Science Theory and Methodology (ICS). His main research interest is social capital. He is editor (with B.Völker) of Creation and Returns of Social Capital: A New Research Program (2005). Recent articles have appeared in Social Networks, European Sociological Review, and Social Science & Medicine. nicolai j. foss: Professor of Strategy and Organization at the Copenhagen Business School and a part-time Professor of Knowledge-based Value Creation at the Norwegian School of Economics and Business Administration. His main areas of interest are the theory of the firm, economic organization, and strategy. Among his recent book publications are Management Innovation (with T. Pedersen, J. Pyndt and M. Schultz), Opportunity Discovery and Economic Organization: Entrepreneurship and the Theory of the Firm (with P. Klein, Cambridge University Press, 2011), and Knowledge Governance: Perspectives from Different Disciplines (with Snejina Michailova, Oxford University Press, 2009). guido friebel: Professor, Department of Economics and Business, Goethe University Frankfurt. His main area of interest is applied contract theory (organizations and personnel economics), industrial organization and regulation (in particular of railroads), and institutional and transition economics. Recent articles have appeared in American Economic Journal, Journal of Economic Behavior and Organization, and Economic Journal. simon gächter: Professor of the Psychology of Economic Decision Making at the University of Nottingham, UK. His main area of interest is the role of social norms in economic decision making. He has published in Science, Nature, and economic journals such as the American Economic Review and Econometrica. edgar kiser: Professor of Sociology and Political Science at the University of Washington. His main interests are the bureaucratization of states, the development of voting institutions, the bureaucratization and centralization of tax administration, and the methodology of historical sociology. His recent articles have appeared in the Annals of the Academy of Political and Social Science, American Sociological Review, and Political Power and Social Theory. peter g. klein: Associate Professor of Applied Social Sciences at the University of Missouri. His research focuses on the economics of organization,

Contributors

xi

entrepreneurship, and corporate strategy. His recent books include Organizing Entrepreneurial Judgment: A New Approach to the Firm (with N. Foss, Cambridge University Press, 2012), The Elgar Companion to Transaction Cost Economics (coedited with M. E. Sykuta, Edward Elgar, 2010), and The Capitalist and the Entrepreneur: Essays on Organizations and Markets (Ludwig von Mises Institute, 2010).

siegwart lindenberg: Professor of Cognitive Sociology at the Interuniversity Center for Social Science Theory and Methodology (ICS), University of Groningen. and at the Tilburg Institute for Behavioral Economics (TIBER), Tilburg University. His main interests lie in the areas of micro-foundations for theories on collective phenomena, self-regulation, pro- and antisocial behavior, groups and relationships, and governance structures in organizations. Recent articles have appeared in Science, Academy of Management Review, Criminology, PLoS ONE, and Aggressive Behavior. ross l. matsueda: Blumstein-Jordan Endowed Professor of Sociology at the University of Washington. His current research develops new methods of estimating trajectories of crime and drug use, examines social capital and informal control in neighborhoods, and estimates the effects of life course transitions on criminal trajectories. Recent articles have appeared in Journal of the American Statistical Association, American Sociological Review, Criminology, and Annals of the American Academy of Political and Social Science. james d. morrow: Professor of Political Science and Research Professor at the Center for Political Studies at the University of Michigan. His current research addresses the role of election institutions in domestic and foreign policy and the effects of norms on international politics. He is the author of Game Theory for Political Scientists (Princeton University Press, 1994) and the forthcoming Order within Anarchy: The Laws of War as an International Institution, and co-author of the award-winning The Logic of Political Survival (with B. Bueno de Mesquita, A. Smith and R. M. Siverson, MIT Press, 2003). victor nee: Frank and Rosa Rhodes Professor of Sociology, Cornell University. His areas of interest are in economic sociology, networks and institutions, and immigration. Recent articles have appeared in the Management Science, Daedalus, Social Forces, Kyklos, Journal of Institutional and Theoretical Economics, Management and Organization Review, and Research in the Sociology of Work. His books include Capitalism from Below: Markets and Institutional Change in China (with Sonja Opper, 2012), On Capitalism (co-edited with Richard Swedberg, 2007), Economic Sociology of Capitalism (coedited with Swedberg, 2005) and Remaking the American Mainstream: Assimilation and the New Immigration (with Richard Alba, 2003). erin powers: PhD candidate in Sociology at the University of Washington, is writing a dissertation on the development and effects of group solidarity and social capital in self-help groups. Her areas of interest include political sociology, collective action, and community. werner raub: Professor of Sociology, Utrecht University and Interuniversity Center for Social Science Theory and Methodology (ICS). His main research areas are theoretical sociology, organization theory and economic sociology,

xii

Contributors

mathematical sociology, experiments and the use of complementary research designs in the social sciences, and sociological applications of neuroscience. He is co-editor of Micro-Macro Links and Microfoundations (with V. Buskens, and M. van Assen, Routledge, 2012).

ignacio sánchez-cuenca: Research Director of the Center for Advanced Study in the Social Sciences, Juan March Institute (Madrid). His areas of study include political violence, comparative politics, and democratic theory. He is the author of Controlling Governments.Voters, Institutions, and Accountability (co-edited with J. M. Maravall, Cambridge University Press, 2008) and articles published in The Journal of Conflict Resolution, The Journal of Peace Research, Politics & Society, The Annual Review of Political Science, and several others. paul seabright: Professor of Economics at the Toulouse School of Economics and Director of the Institute for Advanced Study in Toulouse. His current research lies in three areas of microeconomics: industrial organization and competition policy; the economics of networks and the digital society; and behavioral economics (especially the integration of evolutionary biology and anthropology with an understanding of the development of economic institutions in the very long run). Among his publications are The Company of Strangers: A Natural History of Economic Life (Princeton University Press, 2nd edition, 2010) and The War of the Sexes: How Conflict and Cooperation Have Shaped Men and Women from Prehistory to the Present (Princeton, 2012). tom a. b. snijders: Professor of Statistics in the Social Sciences at the University of Groningen and the University of Oxford, and Interuniversity Center for Social Science Theory and Methodology (ICS). His main research interests are in statistical methods for social networks, multilevel research / random coefficient models, mathematical sociology, and item response theory. Among his recent book publications is Multilevel Analysis: An Introduction to Basic and Advanced Multilevel Modeling (with R. Bosker, Sage, 2011). frans n. stokman: Professor of Social Science Research Methodology at the University of Groningen, The Netherlands. His main fields of interest are models of and strategic intervention in collective decision-making, social network analysis, and policy networks. Among his recent book publications is The European Union Decides (co-edited with R. Thomson, C. Achen, and T. Koenig, Cambridge University Press, 2006). jelle van der knoop: Senior consultant at dutch. Holds a PhD in sociology. His main areas of interest are in the field of collective decision making processes and negotiation. reinier c. h. van oosten: Expert for computer simulation at dutch. He holds a degree in econometrics. His main areas of interest are game theory and collective decision making. arjen van witteloostuijn: Professor of Organization and Strategy, University of Tilburg, Research Professor of Economics and Management at the University of Antwerp in Belgium, and Professor of Institutional Economics at Utrecht University in the Netherlands. His main research interest lies in the field of organizational ecology. Among his recent book publications is Nations

Contributors

xiii

and Firms in the Global Economy: An Introduction into International Economics and Business (co-authored with S. Brakman, H. Garretsen and C. van Marrewijk, Cambridge University Press, 2006).

beate völker: Professor of the Sociology of Social Capital at Utrecht University and Interuniversity Center for Social Science Theory and Methodology (ICS). Her main field of interest is the influence of contexts on social network patterns. She is editor (with H. Flap) of Creation and Returns of Social Capital: A New Research Program (Routledge, 2005). rafael wittek: Professor of Theoretical Sociology, Department of Sociology, University of Groningen and Interuniversity Center for Social Science Theory and Methodology (ICS). His main research areas are social network analysis, organization studies, and social theory. Recent articles have appeared in the Journal of Applied Psychology, Social Networks, and the Journal of Public Administration Research and Theory.

Introduction: Rational Choice Social Research rafael wittek, tom a. b. snijders, and victor nee

During the past two decades or so, rational choice theory has significantly advanced in refining its theoretical core and its empirical applications, and has made a respectable contribution to a large variety of substantive research areas (Hechter and Kanazawa 1997; Hedström and Stern 2008; Kronenberg and Kalter 2012; Macy and Flache 1995;Voss and Abraham 2000). This volume presents an overview of some of the achievements of what we call rational choice social research—empirical investigations that were guided by rational choice reasoning. In this introductory essay, we first sketch what could be described as the “Rational Choice Paradox”—that it is actually the strengths of the approach that have inhibited its further advancement. We then sketch some of the major criticisms against the approach, and then provide a very brief summary of the theoretical core of the rational choice approach. Next we outline the analytical structure and chapters of this Handbook. In the concluding section, we discuss some future perspectives for the approach, in particular its potential to develop into a full-fledged interlevel, interfield research program (Kuipers 2001).

The Rational Choice Paradox Proponents of rational choice reasoning often argue that the rational choice approach, unlike any other paradigm in the social sciences, can be characterized by a well-developed, highly consistent, and widely shared set of formalized core assumptions (Coleman 1990). They praise its emphasis on parsimonious model building, conceptual rigor, and explicit attention to micro-macro problems for theory formation (Raub, Buskens, and Van Assen 2011)—qualities that, in the eyes of its proponents, warrant claims of a “privileged role” of rational choice modeling above other approaches attempting to explain social phenomena in terms of individual action (Goldthorpe 2007: 172; Abell 1992). The rational choice approach indeed continues to attract use by increasing numbers of scholars. In more and more subfields of the social sciences, scholars realize the usefulness of the rational choice approach as a tool for theory-driven social research and interventions. It is not uncommon that empirical research,

2 Wittek, Snijders, and Nee from studies of residential segregation to warfare, draws on the rational choice approach as its implicit theory. Paradoxically enough, however, there are at least two reasons why the strengths of the RC approach (RCA) seem also to inhibit its further advancement. First, with its traditional emphasis on theory building and formal modeling, the RC approach for a long time has been associated mainly with sophisticated but arcane and highly abstract modeling efforts. As a result, one of the major criticisms against RCA has been that rational choice scholars would excel in formal modeling but fail to provide empirical evidence to support their models. Therefore, the RC approach would neither have produced or be backed by relevant empirical insights, nor would it be useful in guiding empirical social research in the first place. While RC research indeed had a strong theoretical focus in the past, this statement is certainly increasingly less true. In the last two decades, RC research has been translated into Mertonian middle-range theory oriented toward empirical social research, often with impressive results. Second, the RC approach meanwhile is probably the only paradigm that has been applied to almost all subdisciplines and subfields of the social and behavioral sciences, ranging from the modeling of markets to the study of immigration, assimilation, ethnic enterprise, race relations, trust, networks, institutions, religious behavior, emotions, terrorism, and a huge variety of other phenomena. In this sense, Goldthorpe’s call that rational choice theory would benefit from “concentrating more on the application of RAT to specific explanatory tasks, rather than on theory development for its own sake” (Goldthorpe 2007: 134) was certainly heeded. Often the use of RC models has triggered fierce but fruitful controversies in these substantive and specialized fields of application.Through these discussions with subfield-specific audiences, rational choice researchers not only advanced our substantive knowledge on specific social phenomena but also significantly enhanced and refined the RC approach itself. Nevertheless, the insights generated within the subdisciplines only rarely diffused across subfield-specific boundaries. The result is that RC research, though being one of the few paradigms with a coherent set of formalized and widely shared core assumptions, remains fragmented, with RC scholars as well as their critics in one field often remaining unaware of the empirical and theoretical progress achieved in other subfields. Consequently, both RC scholars and their critics miss important refinements and corrections of the approach as they have taken place in the past two decades. In sum, though the past two decades have seen major theoretical and empirical advancements guided by rational choice reasoning in a large variety of subfields, most of these advancements are still fragmented. Combined with the ongoing criticism against the approach as such, this fragmentation prohibits a more objective assessment of its merits and limitations.

Criticism of Rational Choice Theory There is probably no aspect of rational choice theory that has not been criticized: its model of human nature, its reductionism, its inability to deal with culture and identity, its neglect of social embeddedness. Most of these issues have been discussed in Green and Shapiro’s (1994) book Pathologies of Rational Choice

Introduction: Rational Choice Social Research

3

Theory and the subsequent debate. One of their major complaints was that rational choice theory has not produced novel, empirically sustainable findings: “[S]uccessful empirical applications of rational choice models have been few and far between. . . . Part of the difficulty stems from the sheer paucity of empirical applications” (ix–x). What is more, if empirical research was inspired by rational choice theory, it is “marred by methodological defects.” When discussing criticism against rational choice reasoning, it is important to distinguish between criticism based on misconceptions and criticism directed toward real problems (Goldthorpe 2007). Much criticism indeed rests on often serious misconceptions. This holds for the assumptions that rational choice theory equals neoclassical economics, that the approach would be normative in nature, that rational choice approaches would acknowledge only instrumental rationality (ibid.: ch. 8), or that formalism would be an essential requirement of the approach (Cox 1999).The majority of these misconceptions could be invalidated in the debate following Green’s and Shapiro’s publication (Friedman 1996). Also the claim that the set of empirical tests of nonobvious rational choice hypotheses is almost empty—what most would consider the least controversial criticism—did not withstand closer scrutiny even at the time of Green’s and Shapiro’s publication (for a demonstration of this point for the political sciences, see Cox 1999). This Handbook collects additional evidence proving this point. Nevertheless, no one would deny that rational choice theory, like any other theoretical framework, has some real unresolved problems to address. Most of them are related to the highly stylized assumptions of neoclassical economics, in particular the assumption of atomized interaction between rational and selfish actors with full information, taking place in perfect markets. Rational choice scholars always acknowledged that deviations from this ideal typical construct of rationality were possible. Four different strategies to deal with such deviations can be discerned. They differ with regard to how they treat individual level deviations from rationality and its aggregate effects. cognitive anomalies The first solution, and the one usually invoked by proponents of neoclassical strong rationality assumptions, consists in classifying these deviations as “cognitive anomalies” at the level of individual actors. Such anomalies would be idiosyncratic and randomly distributed in the population, which is why they would not substantively affect the aggregated outcomes predicted by rational choice models. Adherents of this position therefore consider the neoclassical set of rationality assumptions as a valid foundation for model building, and see no need to increase cognitive complexity. Its fiercest proponents, such as Gary Becker, consider principles of strong rationality as applicable for decisionmaking in general, independent of the context in which it takes places. This position has become known as economics imperialism (Fine and Milonakis 2008): humans are general-purpose problem solvers acting according to rationality criteria, be it in choosing a mating partner or in buying a car. strong vs. weak rationality The second group of scholars suggested distinguishing between “strong” and “weak” rationality assumptions (“hyperrationality” vs. “bounded rationality”).

4 Wittek, Snijders, and Nee Strong rationality would assume, for example, perfect information of all actors, unlimited cognitive capacity to deal with information, and maximization as the decision-making criterion. Bounded rationality assumes unequal access to information, selective attention, and satisficing. Proponents of strong rationality acknowledge that the assumptions are abstractions that need not match with real-life individual decision-making processes, but emphasize that these assumptions nevertheless result in good models of aggregate outcomes (Coleman 1990; see also Buskens’s and Raub’s contribution to this volume). Bounded rationality proponents doubt this and urge scholars to make more realistic assumptions about human nature.They argue that individual deviations from strong rationality are not idiosyncratic but systematic. As a result, models that do not take such systematic deviations into account will also produce wrong aggregate level predictions.The bounded rationality approach refines the full rationality perspective by delineating a specific set of cognitive limitations. This does not mean that its proponents aim “at the construction of models of choice that are incompatible with rationality” (Rubinstein 1998: 25). market vs. nonmarket contexts The third group of scholars suggested that strong rationality assumptions would hold only in specific, usually market-related decision situations, whereas nonrational motives would dominate in noneconomic settings, such as transactions in families or within close-knit communities.This would justify the division of labor between economics studying markets and economic behavior and exchange, and the other social sciences studying social behavior. Here, increasing the “cognitive complexity” of the actor model would be considered an adequate strategy only if the phenomenon to be explained would be outside of the market or economic sphere. This approach acknowledges that there might be systematic individual-level deviations from strong rationality in specific noneconomic domains. Consequently, rational choice theory is not applicable to model behavior in these settings, since it would also lead to wrong aggregate-level predictions, but it definitely is adequate to model behavior in other settings. social rationality Finally, the fourth group of scholars opts for expanding assumptions about rationality. This builds on mounting evidence collected during the past two decades by cognitive neurosciences, behavioral economics, evolutionary psychology, and related fields. Rather than treating deviations from a strong rationality model as idiosyncratic cognitive anomalies of individuals, as applicable to only specific societal domains, or as simply limited by cognitive capacities, they should be conceived as systematic reflections and hence predictable characteristics of human nature (Ariely 2008; Thaler and Sunstein 2008; Camerer, Loewenstein, and Rabin 2004; see also Lindenberg’s contribution to this volume). This requires a refinement of microfoundations. At many points in this Handbook, two such strategies of refining the microfoundations are applied: the systematic incorporation of assumptions about human goals and preferences on the one hand, and about identities and beliefs on the other. The common denominator of these strategies consists in refining the cognitive,

Introduction: Rational Choice Social Research

5

motivational, and even neurophysiological ingredients of individual decisionmaking processes. By elaborating on the cognitive foundations of human decision-making (that is, rendering assumptions about the intrapersonal antecedents of behavior more complex), research following this strategy arrives at surprising hypotheses and insights, which sometimes are at odds with the predictions of the standard model, and sometimes can be incorporated into it. Extensions of the actor model of economics—for example, through “fast and frugal” heuristics, the incorporation of loss aversion and reciprocity effects, or the assumption that actors derive utility from punishment (Fehr and Gächter 2000)—were successfully applied to explain cooperative vs. selfish behavior, such as the decision to free-ride or to allocate sanctions for noncooperation. Crucial antecedents dealt with in this context are nonpecuniary incentives, reciprocity, and social incentives (for example, altruistic punishment) in general. In sum, the past decades have witnessed many attempts to refine, specify, or relax these assumptions. The next section provides a brief structured overview of these attempts.

Core Assumptions of Rational Choice Theory Following Goldthorpe (2007), we define the rational choice approach broadly as a family of theories explaining social phenomena as outcomes of individual action that can—in some way—be construed as rational. Simon refers to substantive rationality as behavior that “is appropriate to the achievement of given goals within the limits imposed by given conditions and constraints.” In this perspective, irrational behavior is an outcome of impulsive responses without adequate intervention of thought (Rubinstein 1998: 21; but see Lindenberg in this Handbook for a different approach). As in most other theory traditions, there are many variations in how rational choice theories are constructed. Rational choice scholars differ, often considerably so, with regard to the type of assumptions they make, their behavioral “microfoundations.” Yet they also share a common core. Though explicating the behavioral microfoundations underlying a proposed explanation is crucial for any theoretical endeavor, social scientists more often than not leave many of these assumptions implicit, and rational choice scholars are no exception. For mainstream economists, this is usually believed to be not too problematic, given a widely shared consensus on the assumptions of the canonical model to which the majority of economists adhere. In these cases, tacit assumptions can often be easily reconstructed by referring to textbook knowledge. Lacking explications of assumptions is problematic in those branches of the social sciences without a consensus about the theoretical core, and sociology is certainly one of them. Misconceptions are often the consequence, with resulting debates not addressing real problems (Goldthorpe 2000: ch. 8) and creating wrong divides between theoretical approaches. The importance of explicating the microfoundations underlying an explanation has been repeatedly demonstrated, and becomes most visible in those situations in which the same explanatory variable is hypothesized to have opposite effects on an outcome, depending on the type of microfoundation that is taken as

6 Wittek, Snijders, and Nee ta bl e 0 . 1 Varieties of Rational Choice Microfoundations Assumption

Thin or strong rationality

Rationality

Full rationality

Bounded rationality

Procedural rationality

Social rationality

Opportunism

Egoism

Linked-utility

Solidarity

Tangible resources Natural

Intangible resources Social

Physical wellbeing Institutional

Social wellbeing Structural

Preferences: Selfishness Preferences: Materialism Individualism

Thick or weak rationality

a starting point (Torsvik 2000). Take the example of explaining variations in employee commitment or performance as a consequence of the amount of pay. If the microfoundation assumes that individuals care only about the amount of their own salary, the resulting prediction is that pay raises should have a positive effect on individual performance, independently of the pay raises received by other employees in the firm. If the assumption is that individuals care about relative status, a pay raise may actually have detrimental effects on performance if it compares unfavorably with what one’s colleagues earn (Frank). In the case of this second microfoundation, the assumption of “atomistic” actors underlying the first hypothesis is relaxed by assuming that individuals know what their colleagues earn (a condition that is not necessary in the first model), that they make social comparisons, and that they are driven by social motives (that is, to increase their relative status). Note that the latter is still compatible with the selfishness assumption of the standard model, since individuals are assumed to maximize not only material gain but also relative status. This thought experiment can even be extended further. Assume that individuals again care about what others earn, but that “caring” means that they are guided mainly by fairness considerations, rather than by the drive to increase their own status. In that case, the positive effect of a pay raise on individual performance would be predicted only if either the pay raise is perceived to be legitimate, or if other employees in comparable positions also receive a pay raise. In this model, one element in the microfoundation deviates from the canonical rational choice model: it drops the assumption of selfish preferences. As the contributions to this Handbook amply illustrate, many contemporary rational choice models relax some of the assumptions of the canonical model while retaining others. In what follows, we sketch three dimensions on which rational choice scholars usually differ, and along which rational choice theories can be characterized (see also Goldthorpe 2007). We refer to these domains as rationality, preference, and individualism assumptions (see Table 0.1). rationality Rationality assumptions span the range from full (or hyper-) rationality, bounded rationality, procedural rationality to social or ecological rationality. In models of full rationality, the assumption is that individuals are fully informed about all their decision alternatives, the probabilities of their outcomes, and

Introduction: Rational Choice Social Research

7

their consequences. Individual decision-makers do not face any cognitive limitations or biases in perceiving or processing this information. Alternatives are evaluated against each other according to cost-benefit criteria, and actors choose the alternative with the highest (subjective) expected utility. Where outcomes depend on the decisions of other actors, full rationality is assumed to be strategic (rather than parametric) rationality, and modeled with game theoretical tools. Models of bounded rationality (Simon 1957; Rubinstein 1998) make two key assumptions in which they deviate from full rationality models. First, decision-makers are usually not fully informed about all available options: their perception of information is biased through selective attention (framing processes). Second, humans have limited cognitive capacities for processing the information that is available to them: rather than maximizing, boundedly rational actors satisfice—that is, once they detect a course of action that in their eyes is good enough to reach a goal, they won’t go on searching for a better one, even if they know that a better solution would be available. Models of procedural rationality share the assumption that many individual decisions and much behavior are guided by past experiences leading to imitation and “automatic” responses, rather than by conscious and deliberate evaluation of future costs and benefits. Such learning models consider trial and error mechanisms as crucial strategies, in particular under conditions of (radical) uncertainty where individuals do not know all possible outcomes (Knight 1921). Hedström (1998) refers to this strategy as “rational imitation.” Finally, Lindenberg’s social rationality and Todd’s and Gigerenzer’s ecological rationality approach suggest that rationality should be treated as an explanandum rather than an explanans. Building on insights from social and cognitive psychology and the evolutionary and neurocognitive sciences, which document the modularity of the brain, social rationality models try to specify under which conditions gain-maximization and other rationality traits contained in full or bounded rationality approaches will guide human decision-making, and under which conditions other processes such as learning or automatic responses will guide behavior (Lindenberg 2001; Todd and Gigerenzer 2007). The social rationality model goes furthest in relaxing the traditional rational choice core. It suggests that gain seeking is only one of three overarching goal frames, in between hedonic and normative frames. A key argument is that not gain seeking but hedonic goals—directed toward the immediate realization of pleasure, not necessarily of material gain—are the “natural default condition” (see Lindenberg’s chapter in this Handbook). preferences The second dimension on which rational choice models differ is preference assumptions. In the canonical, neoclassical rational choice model, preferences are assumed to be exogenously given and stable, and individuals are selfish egoists striving toward the maximization of material gain. Rational choice models can be distinguished based on the degree to which they relax these two preference assumptions. We refer to them as the selfishness and the materialism assumptions.

8 Wittek, Snijders, and Nee Selfishness With regard to selfishness assumptions, the following four positions can be discerned. First, on one extreme of the continuum,Williamson urged sharpening the selfishness assumption by incorporating opportunism (that is, self-seeking with guile) as an extreme form of egoism (Williamson 1975). Opportunism implies that exchange partners may deliberately break rules and cheat in order to increase their own benefits at the expense of the other party (see Foss’s and Klein’s chapter in this Handbook for a discussion of the implications of opportunism). Second, the opportunism assumption differs from a pure egoism assumption, in which contracting parties are assumed to respect the rules of the game. For example, complete contracting theories assume that rational and forwardlooking actors design and enforce these rules in such a way that they align the selfish interests between the exchange parties so that it does not pay to cheat (Milgrom and Roberts 1992). Third, some rational choice models explicitly incorporate the assumption that it might be in the best interest of an individual to take the well-being of other actors into account—that is, to link his or her utility to the utility of exchange partners. Approaches building on such linked utility assumptions argue that individuals might hold moral or partially altruistic preferences. This assumption does not require relaxing other rationality assumptions. In fact, selfishness is not a necessary component of rational choice models at all (see Gächter in this Handbook). Finally, at the other end of the selfish preference continuum, some rational choice scholars invoke goal-framing theories to argue that (social) preferences dominant in a given situation need to be endogenized. This approach suggests that under specific circumstances, humans may act in a strong solidarity frame in which no tangible direct personal benefit results from their actions. In other situations, weak solidarity (for example, in the form of direct reciprocity) may govern the behavior of individuals. For example, in many economic transactions the salient individual gain frame is tempered by fairness considerations, resulting in the more powerful exchange partner not squeezing the maximum possible out of the other party. We refer to these assumptions as solidarity assumptions (see Lindenberg in this Handbook). Materialism Although there is nothing inherent in rational choice theory that makes the assumption of material gain a necessary one, many rational choice models seem at least to build implicitly on the assumption that what drives human decisionmaking is the maximization of tangible, material resources. We refer to this as the tangible resources assumption. Such tangible resources are usually assumed to be money or other goods that may be accumulated. Somewhat less restrictive, but still in the same spirit is the intangible resources assumption: some resources are not tangible but can still constitute very valuable intangible assets, such as intellectual capital or capabilities and competencies (see, for example, Daum 2003). Social rationality approaches (Lindenberg 2001) have further relaxed the assumption that resources—be they tangible or intangible—are the major

Introduction: Rational Choice Social Research

9

objective individuals strive to maximize.They argue that humans put great value on their physical and social well-being, often at the expense of material gain. The physical well-being assumption proposes that individuals will seek stimulation and comfort, whereas the social well-being assumption states that individuals may also strive for the maximization of different types of social goals such as social approval, status and prestige, or affection. In this perspective, tangible and intangible resources are seen as instrumental lower-level means of production (“instrumental goals”) that individuals use to produce the higher-level goal of well-being. individualism Although rational choice scholars might differ in the type of microfoundation they use, what unites them all is the conviction that societal phenomena at the meso or macro level can be explained in a satisfactory way only by descending to the level of the individual and specifying the microlevel mechanisms— assumptions about individual decision-makers and the decision rules they use to make their choices—that generate the macrolevel outcome. This analytical strategy is referred to as individualism (for a discussion of the ambiguities of this term and varieties of individualism, see Hodgson 2007). For most social researchers in the rational choice tradition, the theoretical primacy—the phenomena that have to be explained—is situated on the meso or macro level, whereas the analytical primacy—the social mechanisms leading to behavior of individual actors—has to be connected to the micro level of individual choices (Raub, Buskens, and Van Assen 2011; Hedström and Bearman 2009). While some form of individualism underlies all rational choice approaches, individualism assumptions come in many varieties, and different labels have been coined to characterize these variations—such as methodological, institutional, holistic, or structural individualism. Drawing on his extensive analysis of methodological individualism, Udehn (2001: 354) considers the following approach as the core of explanatory methodological individualism: “Social phenomena must be explained in terms of individuals, their physical and psychic states, actions, interactions, social situation, and physical environment.” As in the case of rationality assumptions, strong and weak versions of methodological individualism can be distinguished (Udehn 2001). They differ to the degree that macro- or mesolevel conditions (such as institutions or social structures) are incorporated as part of the explanans. Strong versions require that exogenous variables and conditions (that is, the explananda) must refer only to individuals, but not to social institutions. In weak versions of methodological individualism, this rule is replaced by the requirement that social phenomena are allowed to enter the antecedents. Rational choice models subscribing to the latter view require specifying three steps in their social mechanism explanations: a macro-micro step or “situational mechanism,” a micro-micro step or “action generating mechanism,” and a micro-macro step or “transformation mechanism” (Hedström and Swedberg 1998).This analytical strategy is of course not restricted to rational choice theory (see, for example, Gross 2009), though rational choice scholars probably had a strong impact on the refinement of social mechanism approaches. The type of rationality and preference assumptions characterize the degree

10 Wittek, Snijders, and Nee to which the canonical model’s assumptions about cognitive abilities are relaxed—that is, they introduce different degrees of cognitive complexity in the microfoundation of a rational choice explanation.Varieties of individualism reflect the degree to which the canonical model’s idea of isolated, “atomized” actors is relaxed—that is to say, they add structural complexity (Lindenberg 1992) in the form of different types of social embeddedness. Again, the assumption space can be described as a continuum, ranging from strong to weak forms of individualism (Udehn 2001: ch. 12). At one end reigns the atomism assumption of neoclassical economics and theories of general equilibrium, which model exchanges as atomized interactions on perfect markets. The most prominent assumption of this natural methodological individualism is that all parties have equal access to information, and that patterns of (social) relationships are not relevant as opportunities or constraints for economic behavior. All information is contained in the prices for the exchanged goods. This approach has been called “natural” individualism, “since nothing socio-cultural enters the explanans, or exogenous variables, of its explanations” (ibid.: 347). This unrealistic assumption has frequently been challenged, and many efforts have been made to develop a more realistic set of assumptions that can be incorporated into rational choice models. Udehn (2001) suggests that “Austrian” or social methodological individualism based on the work of Menger, Weber, von Mises, Hayek, and Schumpeter, though still representing a strong version of individualism, departs from the atomistic model in that it emphasizes the importance of subjective meaning that individuals attach to their actions. In this version, society is seen as an intersubjective reality, and humans are considered as social and cultural beings. Representatives of this approach acknowledge that social institutions, which they situate in the minds of individuals, may influence individual preferences and actions: “[S]ocial institutions are the subjective meaning individuals attach to social actions or social things like money” (ibid.: 125). A third and weak version of methodological individualism has been labeled institutional individualism. It is considered the dominant version in new institutional economics and political sciences. Here, institutions are conceptualized as objective phenomena and accepted as exogenous variables, though there are also many attempts within this tradition to endogenize institutions. More specifically, institutional embeddedness refers to rules affecting opportunities, constraints, incentives, and information of actors and their exchange partners. Institutions can have a formal or an informal basis; they can contain ambiguities, may not be known by all actors, and may or may not be enforced. Fourth, sociologists Coleman, Lindenberg, Raub, and Wippler have elaborated what now is known as structural individualism, the version most frequently applied by sociologists (for an early statement, see, for example, Wippler and Lindenberg 1987). Udehn (2001) considers it as the weakest form of methodological individualism because it leaves room for a broad set of societal-level conditions to influence individual-level choice and behavior. Social structural and institutional embeddedness enters this model in several ways.

Introduction: Rational Choice Social Research

11

First, social structure—in the sense of positions in a system of relations—is assumed to influence individual preferences and beliefs. Scholars following this strategy incorporate fine-grained information on social-structural network characteristics of individual actors or their settings, and introduce learning (that is, differential access to information) and control effects (differential exposure to monitoring and sanctioning capacities) into their models. Structural embeddedness comes in two varieties (see Buskens and Raub in this Handbook): in dyadic and network forms. Dyadic embeddedness assumptions state that past, ongoing, or expected future interactions with specific other exchange parties will affect decisions, behavior, and exchanges. The focus is on the dyad—for example, the two parties in a contract governing a supplier relationship. The “tie” can take many forms, ranging from the completion of a business transaction in the past, to a friendship and family bond. Depending on the preference assumptions invoked (see above), the effects of dyadic embeddedness can be limited to, for example, information benefits (“learning”) and control (sanctioning opportunities), or can affect an individual’s inclination toward prosocial behavior as in linked utility or social rationality approaches. Ties to third parties do not affect these interactions. Network embeddedness assumptions further increase structural complexity by considering the potential influence of third parties on exchanges between two actors, because one or both of them are tied to the third party. Again, the nature of the “ties” as well as the mechanism can vary, depending on the rationality assumptions used. For example, third parties can act as intermediaries or guarantors where a potential trustor is uncertain about the trustworthiness of a trustee. Similarly, network closure may enhance sanctioning opportunities and therefore facilitate collective action. The second way to conceptualize structural individualism is in terms of social roles influencing preferences and actions. For example, individuals occupying a managerial position in a firm are supposed to maximize profits for their employer. Finally, structural individualism also acknowledges the potential impact of culture on individual preferences and beliefs.

The Handbook of Rational Choice Social Research The focus of this Handbook is on Rational Choice Social Research, with the link between theory and empirical research being of central concern. This emphasis has a number of implications for the structure and content of the Volume. In this section, we first outline the rationale and overall structure of the Handbook. We then briefly sketch the content of each chapter. purpose of the handbook This Handbook attempts to provide a state-of-the-art overview of current social research guided by rational choice reasoning.The contributions structure their problem area, assess what kind of empirical regularities have been confirmed, critically discuss the scope and explanatory power of rational choice explanations of these phenomena, and sketch fruitful areas for future research in their domain. In order to achieve this aim, the chapters were written with the following guidelines in the background. First, each chapter addresses both

12 Wittek, Snijders, and Nee theoretical and empirical issues. Put differently, we did not include chapters with a purely theoretical character (for example, discussions and comparisons of theoretical debates or refinements), or chapters with a purely empirical focus (such as literature reviews or summaries of findings). Each chapter provides a sound reflection and discussion of the major theoretical and empirical achievements in a subfield. It structures this subfield, thereby providing an analytical frame of reference that allows identifying underlying overarching themes in existing research. At the same time, each chapter functions as an organizing device and point of departure for promising new research efforts by going into depth with regard to the respective rational choice models relevant for the problem under investigation. Chapters carefully reconstruct the major assumptions, causal mechanisms, and theoretical propositions of the models, point out eventual problems and discuss possible solutions. The purpose of this reconstruction is to reach a maximum of transparency, thereby enabling future researchers to apply, refine, and test the models. Hence we have opted for an intensive rather than an extensive setup of each chapter. The result is an indepth treatment of the rational choice models related to a substantive problem, rather than a general overview of and comparison with alternative theories dealing with a specific problem. Second, the Handbook covers a broad range of topics in which the rational choice approach has proven to be a powerful tool of analysis. This implies that the primary intended audiences of each chapter are scholars dealing with a wellestablished and recognizable substantive problem. Such substantive core problems will be identified by the kind of phenomena to be explained.We opted for such a problem-based approach (for example, “explaining terrorism,” “explaining war”) rather than a discipline-based approach (“political sociology”), not only because specific problems are usually studied by various disciplines but also because one of the strengths of the rational choice approach is to provide a unifying analytical framework into which insights from different fields and disciplines can be integrated. Third, the purpose of the Handbook is to provide insights into concrete societal processes and issues, and to provide templates that stimulate future research. The chapters therefore emphasize what can be learned from earlier research, and what constitutes cutting-edge theoretical and methodological tools to advance research into societal developments. That is, the theory part of the chapters is analytical in nature, rather than historical (in the sense of reconstructing or summing up intellectual history of theories). In sum, each chapter has a theoretical core in which elements of rational choice reasoning are discussed, and a substantive domain of application, reviewing studies in which the resulting hypotheses were empirically put to a test. structure of the handbook The Handbook is divided into five parts. The four chapters of Part I (“Rationality and Decision-making”) address the microfoundations of rational choice theory. They present different versions and extended discussions of rationality assumptions, and also offer an analysis of how structural embeddedness affects cooperation among rational egoists. The main “outcome domain” addressed here is individual decision-making.

Introduction: Rational Choice Social Research

13

Part II (“Networks and Inequality”) provides examples for rational choice models adding structural complexity to their toolkit. The chapters share a focus on inequality as an antecedent and outcome of social processes. The main purpose of the three chapters is to explain how differences in exchange structures affect variations in power or access to valued resources. The remaining three parts focus on the role of institutional contexts governing behavior in communities, markets, organizations, and states. Individual behavior and choice differ significantly depending on the type of social context in which they take place. Rational choice scholars equate each context with a distinct form of governance—that is, a specific set of institutional rules and definitions of the situation regulating the exchanges between actors. An often-invoked distinction contrasts “spontaneous” and “constructed” social orders, considering primordial social orders such as groups and communities and markets as representing the former category, and states as well as formal organizations as representatives of the latter. While conceptually useful, this distinction also bears some risk of oversimplification: neither do markets as such emerge spontaneously, nor can states or organizations be adequately understood by looking only at their formal blueprint. Markets are designed and regulated, they are subject to institutional change, and can fail. The governance of organizations has to take into consideration that contracts tend to be incomplete. States emerge as a result of complex power struggles between and among internal and external parties. In market exchanges, gain seeking and competition are considered to be legitimate motives for exchange partners, whereas exchanges in constructed social orders are characterized by principles of authority ranking and hierarchical control. The concept of community represents the idea of primordial social orders in which the guiding principles of exchanges are social norms of communal sharing and equality. Rational choice scholars explicitly recognize that each of these types of governance can fail and that the different forms can interfere with or substitute for each other. For example, organizational governance may fail because principles of primordial social orders like friendship ties prevail or dominate the exchanges inside the organization. Likewise, where market failures occur because of externalities, organizational hierarchies or social control based on informal social relations can contribute to the solution of the resulting social dilemmas. Given their crucial role as constraints on behavior and choice, explaining the emergence, change, and eventual failure or success of each form of governance becomes an important task in itself. As a consequence, rational choice scholars have recognized the necessity of endogenizing the different forms of governance, resulting in numerous models explaining, for example, the emergence of norms or hierarchies in market settings. The three remaining sections in the Handbook are designed to account for these complications. Parts III and IV focus on the two extremes on the continuum between spontaneous and constructed social orders, addressing, respectively, communities and cohesion, and states and conflicts. Part V focuses on markets and organizations. In what follows, we will provide a structured overview of the content of each part and chapter. Drawing on the four core assumptions as they were

14 Wittek, Snijders, and Nee discussed in Part III, we will assess which type and combination of assumptions characterizes rational choice research in each of the fields. rationality and decision-making The four contributions in this part focus on the microfoundation of rational choice theory, and investigate recent advances of rational choice social research on cooperation, individual and collective decision-making, and well-being. Simon Gächter’s chapter (“Rationality, Social Preferences, and Strategic Decision-making from a Behavioral Economics Perspective”) starts with a concise introduction to the hard core of rational choice theory, the “canonical rational choice model.” It portrays strong rationality assumptions and how they can be fruitfully used for modeling purposes despite their unrealistic empirical content. In a second step, the chapter discusses the implications of relaxing one of rational choice theory’s key assumptions: the selfishness assumption. Gächter argues that rational choice theory does not require assuming self-regarding preferences, and points toward the low prediction accuracy of rational choice models based on this assumption. Using different types of social dilemma situations—including Trust, Dictator, Ultimatum, and Public Goods Games— the chapter then provides a systematic overview of the role of social preferences in game theoretical models of prosocial behavior. Most of these models have been tested in laboratory settings. The findings consistently show outcomes that would be considered at odds with selfishness assumptions in the canonical rational choice model, as in the case of altruistic punishment. Vincent Buskens’s and Werner Raub’s chapter, “Rational Choice Research on Social Dilemmas: Embeddedness Effects on Trust,” takes a different approach than Gächter’s chapter. Building on the Trust Game as their focus of analysis, the authors investigate the conditions under which selfish actors are inclined to trust others, and when this trust is likely to be abused or honored. Buskens and Raub explicitly stick to full rationality assumptions: actors in their model are assumed to be fully informed and selfish gain maximizers. Working in the tradition of structural individualism, Buskens and Raub add complexity by replacing the assumption of atomized interactions on perfect markets by assumptions specifying different types of dyadic and structural embeddedness. Embeddedness implies that actors had, have, or expect to have interactions with other actors.The chapter analyzes how embeddedness affects trust through two mechanisms—control (that is, sanctioning possibilities for trustor) and learning (availability of information about trustee). They then use game theoretical modeling to derive hypotheses about how control and learning affect trust behavior under different embeddedness conditions. Based on a systematic review of empirical findings from experimental research and survey studies— most of which related to the acquisition of tangible resources—they find strong evidence for learning effects on both the dyad and the network level. Findings for control effects were less clear-cut; in particular, research on network control produced ambiguous results. Siegwart Lindenberg’s chapter (“Social Rationality, Self-Regulation, and Well-Being”) pleads for a redefinition of the rationality concept in terms of self-regulation. Pointing toward many recent insights gathered among others within behavioral economics or evolutionary psychology, the chapter starts by

Introduction: Rational Choice Social Research

15

outlining the key components of such a “social rationality” framework and its inter-relationship with self-regulation. He argues that the human brain developed as a social brain to handle three types of self-regulatory behavior. Need-related, goal-related, and self-related self-regulation are necessary to deal with complex interdependencies related to reproduction and living in groups. While sticking to the microeconomic assumption of the importance of relative prices, the social rationality framework goes beyond microeconomics by incorporating goal framing. Three master goal frames are distinguished: hedonic, gain, and normative. It is assumed that these frames are selectively— and sometimes automatically—activated. The chapter outlines the mechanism leading to the activation of goal frames, the inter-relationships between them, and their subsequent influence on (prosocial) behavior. Another key component added to the microeconomic model are variations in self-regulation ability and the idea that individuals strive for tangible resources only to the degree that they are instrumental for the realization of the higher-order goals of physical and social well-being. The chapter illustrates the implications of these assumptions with findings from recent research on prosocial behavior and sanctioning. The chapter on “Modeling Collective Decision-making” by Stokman, Van der Knoop, and Van Oosten provides a detailed reconstruction of the theoretical assumptions underlying cooperative and noncooperative bargaining models of collective decision-making, and sketches the operational steps for empirical tests of an integrated model. According to this framework, relatively accurate predictions of the outcomes of collective decision-making processes can be made based on a limited amount of information: the relevant issues, stakeholders, their policy position, their power to affect collective outcomes, and their interest in each of the issues. The chapter then describes three different bargaining processes through which collective decisions are usually reached: persuasion, logrolling, and enforcement.The model specifies under which institutional and structural conditions each of these processes is likely to be dominant. Examples from decision-making in the European Union and in firms are used to illustrate the different aspects of the model. Though building on a game theoretical and exchange theoretical framework in its core, the proposed model makes a strong point for adding cognitive and structural complexity to this core, thereby replacing the full rationality and atomism assumptions of natural individualism by a structural individualist approach based on social rationality. For example, persuasion strategies are strongly tied to framing processes and trust; differences in policy positions between stakeholders can be due to conflicting cognitive maps, and the power position of a stakeholder is likely to affect the weight other stakeholders assign to his or her opinion. The four papers vary in the degree of realism in rationality assumptions. Gächter uses a full rationality model, sticks to the atomism assumption of natural individualism, assumes tangible resources as the major goal of individuals, but relaxes the selfishness postulate in favor of linked-utility assumptions: individuals show inequality aversion and have a tendency to reciprocate. Buskens and Raub stick to the full rationality and selfishness (egoism) assumption but relax the atomistic natural individualism assumption by explicitly incorporating dyadic and network embeddedness. They implicitly assume that actors strive mainly for tangible resources. Lindenberg relaxes both

16 Wittek, Snijders, and Nee the full rationality and the selfishness assumptions. His goal-framing model assumes social rationality, which systematically incorporates different types of goal frames into the theoretical core of the approach. This allows endogenizing (selfish or prosocial) preferences. This approach emphasizes physical and social well-being as higher-level goals, and considers tangible or intangible resources as lower-level instrumental goals or endowments that can be used to realize higher-level goals. Lindenberg’s social rationality approach does not provide a fine-grained elaboration of embeddedness assumptions, though social relations and institutions are acknowledged as key context conditions influencing goal frames and the relative prices of achieving different types of goals. Both Buskens and Raub as well as Lindenberg represent structural individualism. With the incorporation of structural and institutional embeddedness as well as cognitive processes, the structural individualist model of collective decisionmaking of Stokman et al. provides an example for a thick version of rationality. With regard to rationality assumptions, Stokman et al. incorporate elements of Lindenberg’s social rationality framework when building on the distinction between ultimate and instrumental goals. The model also endogenizes preference assumptions, suggesting that under conditions of joint production, cognitive interdependence will increase the likelihood for trust and persuasion. The model is not restricted to (in)tangible resources as goals, thereby relaxing the materialism assumption. Furthermore, though it does not elaborate on finegrained variations in network embeddedness, it explicitly provides a framework for considering opportunity structures allowing for logrolling. Finally, models of collective decision-making always incorporate the institutional context, since each context has different decision-making rules under which decisions have to be taken. networks and inequality In many settings, be they markets or social groups, some actors usually have a more advantageous position in the network of (social or economic) exchanges than others. For example, they have friends in high places, or can act as a broker between otherwise unconnected players, which allows them to control the resource flow between other actors. Such advantageous positions in exchange structures allow them to make better deals than their partners, making them materially better off in the long run. The chapters in Part II (“Networks and Inequality”) deal with the inter-relationship between individual positions in exchange structures, and the differential payoffs this generates. The focus is on structural opportunities and constraints as they result from social network embeddedness. Karen Cook’s and Coye Cheshire’s chapter (“Social Exchange, Power, and Inequality in Networks”) explicates the assumptions behind different theories of social exchange, in particular power-dependence and network-exchange approaches. Their contribution reflects the structuralist perspective according to which differences in power or inequalities in resource distribution derive from an actor’s structural position. Based on a long tradition of experimental empirical studies, this research line elaborated fine-grained distinctions between different types of exchange structures and their consequences in terms of resource distributions. Exchange structures are taken as exogenously given, and

Introduction: Rational Choice Social Research

17

rationality assumptions in the baseline models are straightforward: individuals are motivated by gain maximization or loss avoidance, and payoffs are subject to diminishing marginal utility. There is also no further differentiation in the type of resources: all experiments are tied to some material payoff for the subjects. The chapter also discusses the implications that different exchange structures have for cohesion and solidarity: both decrease to the degree that power is unequally distributed. The authors conclude that structuralism comes to its limits in explaining these findings, and suggest that future research might benefit from incorporating group identity and emotional reactions into their models. Henk Flap’s and Beate Völker’s chapter (“Social Capital”) introduces the social capital research program. Research in this tradition differs from exchange theory in several respects. First, it considers different types of material and immaterial resources and their inter-relationships. Issues related to the conceptualization and measurement of social capital occupy a central place in this research line. Social capital research has a strong record in empirical field research, and the chapter reviews findings related to many different types of outcomes related to inequality, ranging from occupational career to social support. Second, it also addresses the question of the creation of social relations through investment in others. The rational choice core of this program is the social resources hypothesis and the investment hypothesis: people with more social capital are better able to realize their objectives, and people will invest in others if this promises some return. Actor assumptions are again straightforward: (dis)investment in social capital depends on the expected future rewards of the tie, available alternatives, and earlier investments in the tie. With regard to context conditions, research in this program tends to focus on key characteristics of personal networks, in particular the size or density of ego-networks. Unlike power-dependence or network-exchange theory, social capital research is more sensitive to the question of why structural explanations meet so many exceptions. Pointing toward the strong influences of spatial and institutional contexts, it suggests that “pure” structural effects are likely to be the exception. Of the three chapters on social networks, Tom Snijders’s chapter on “Network Dynamics” addresses the widest range of assumptions on rationality and structural constraints. Statistical models endogenizing social network structures can be considered as one of the major advancements in social network research of the past two decades. The chapter discusses agent-based simulations, game-theoretical approaches, and stochastic actor-oriented models, all of which allow the simultaneous incorporation of change in actor characteristics and social networks. These models are very flexible with regard to the actor assumptions and structural forms that can be investigated. For example, structural positions that would be considered advantageous from a power-dependence perspective—such as being the only person linking two tightly knit cliques—can yield disutility based on the psychological mechanisms involved, and vice versa.The utility functions can be flexibly determined by the researcher, allowing to systematically test competing mechanisms. The chapter also reviews the recent and growing body of research on network formation games, where properties of whole networks—such as transitivity or centerperiphery structures—are the outcome variable.

18 Wittek, Snijders, and Nee Building on structural individualism, all three chapters go to great lengths in differentiating dyadic and network embeddedness, but they differ in their rationality and preference assumptions. Most of Cook’s and Cheshire’s models build on full rationality assumptions, assume egoistic actors, and focus on the acquisition of tangible resources as an actor’s major goal in their experimental studies. Flap’s and Völker’s work is based on straightforward rationality assumptions, including selfish preferences. Most of their assumptions in this regard remain implicit, and their position can best be characterized as somewhere between full and bounded rationality: (dis-)investments in social relations are made based on the direct costs and rewards related to the tie, and the past and expected costs and rewards. This leaves open whether the payoff is in terms of (in)tangible or physical and social well-being. Unlike the other two chapters, they also incorporate the potential influence of institutional embeddedness. Finally, the statistical models for network dynamics presented in Tom Snijders’s chapter are flexible with regard to all four dimensions. Actor-oriented models can be built by assuming utility functions with different degrees of rationality or selfishness, or social goals. Similarly, these models also allow incorporating dyadic or network embeddedness. communities and cohesion Why do people believe in a god and join a religion? How do social networks of immigrants explain their assimilation into a host culture? How useful are rational choice models of criminal behavior? The contributions in this part focus on the inter-relationship between aspects of communities (such as neighborhoods) on the one hand, and cohesion, integration, and (social control of) deviant behavior on the other. The focus of the chapters is, respectively, on joining or leaving religious communities, assimilation of migrants into their host cultures, or compliance to (legal) norms. All three phenomena can be seen as exemplary indicators of social “cohesion.” The integration of micro and macro levels of explanation is the core question underlying Ross Matsueda’s chapter on “Rational Choice Research in Criminology.” Matsueda starts with presenting the standard rational choice model of individual criminal behavior. Expected utility of the crime, the probability of getting arrested and punished, the return from crime, and the cost of punishment are the key parameters of this model. Empirical studies based on longitudinal surveys and vignettes provide consistent support for this model, in particular for the effects of the certainty of sanctions. The chapter then explores how this model of individual behavior can be integrated into an explanation of macrolevel phenomena (for example, crime rates). It pursues the idea that the degree to which a group or society is organized against rather than in favor of crime largely determines crime rates, and that a rational choice micromodel can explain variations in societal-level organization in favor of or against crime. Matsueda’s theoretical microfoundation departs from neoclassical full rationality assumptions. He assumes that individuals have limited access to information, have different initial beliefs, and are boundedly rational in the sense of responding to incentives rather than being able to engage in complex calculations of optimal solutions. This micromodel is then used to explain macrolevel variations in social capital and collective efficacy on the one hand,

Introduction: Rational Choice Social Research

19

and informal norms conducive to crime on the other. Collective efficacy is conceived as a characteristic of communities (such as neighborhoods) and results from the combination of social cohesion and informal social control. Arguments related to information asymmetry and signaling are used to explain the effectiveness of organization in favor of crime. These ideas are illustrated with empirical research on the “protection industry” of the Sicilian mafia. Nan Dirk De Graaf ’s chapter on “Secularization: Theoretical Controversies Generating Empirical Research” starts with a discussion of the microfoundations and macropropositions used by Stark and other rational choice scholars to explain religious participation. Religious goods are treated as nonverifiable compensators or “otherworldly rewards,” which are limited in supply. Humans are seen as agents formulating explanations about how to gain rewards and avoid costs. On the macrolevel, a key factor is the degree of competition between religions or churches. According to the influential “supply side approach,” demand for religions is stable, and religious participation is expected to be higher to the degree that competition between religions is stronger.The chapter provides a critical discussion of the supply side approach and its major rival, the secularization thesis.The latter predicts declining religious participation resulting from modernization. It also highlights the importance of (changes in) social embeddedness (for example, devout friends) and exposure to cultural beliefs (through, for example, proreligious governmental policies). De Graaf argues that the two approaches can be compatible. The chapter concludes with an overview of empirical studies. Available evidence so far does not corroborate the assumed link between religious pluralism and religious participation; religious demand seems to be stable, though Catholic societies form an exception. Many of the behavioral microassumptions of rational choice models of secularization discussed by De Graaf remain implicit, and largely build on the full rationality and selfish preference assumptions of the canonical model. Furthermore, though referring to social embeddedness as an explanatory factor, the models also do not make an elaborate effort to incorporate different types of embeddedness. Finally, the models clearly depart from the materialism assumption by acknowledging the role of intangible resources as potentially important goals. Taking rational choice institutionalism as a point of departure, Nee’s and Alba’s chapter on “Assimilation as Rational Action in Contexts Defined by Institutions and Boundaries” develops three overarching propositions, each of which explicates one specific mechanism underlying immigrants’ assimilation decisions and practices.The three propositions build on structural individualism as an overarching framework. The purposive action proposition assumes immigrants to assimilate if opportunities are more extensive in the mainstream economy than in ethnic enclaves. The network proposition adds structural complexity, assuming that where discriminatory barriers block individual social mobility, assimilation depends on social capital–based ethnic collective action. The institutional proposition adds cognitive complexity, suggesting that if political actors credibly signal their commitment to nondiscriminatory policies and equal opportunity, the resulting reinforcement of cultural beliefs will stimulate assimilation. Evidence from a recent large-scale study on second-generation immigrants in New York City is congruent with these propositions.

20 Wittek, Snijders, and Nee Taken together, all three chapters in this part retain the full rationality and selfish preference assumptions of the canonical rational choice model, but depart from this model by taking into consideration structural and institutional embeddedness as well as nonmaterial goals. In all three chapters, the micro-macro link is a central point of concern. Matsueda’s chapter contains a formal micromodel; De Graaf ’s chapter, and that by Nee and Alba, discuss a set of micro- and macrolevel propositions. states and conflicts Why do states go to war, despite the large costs this decision usually entails? Why do some wars take longer than others? Why do people join terrorist organizations, and even commit suicide attacks in their name? These and related questions are addressed by the three chapters in this part. All of them focus on conflictive relations within or between states. States are deliberately constructed social orders that govern the interdependencies between many different types of stakeholders. Since states are also a source of revenue and a primary source for the legitimate control of resources, their creation and functioning is accompanied by conflict.The three chapters in this part focus on the antecedents of conflicts within and between states. The chapter on “Terrorism and the State” by Ignacio Sánchez-Cuenca starts by providing a theory-driven taxonomy of different forms of political violence. The chapter then critically discusses current rational choice explanations of individual motivations to join and to contribute actively to (for example, in the form of suicide actions) terrorist organizations. These models are based on assumptions either of altruistic preferences or of selective incentives. While showing that suicide may not be irrational if preferences are assumed to have specific characteristics, such assumptions are of limited use for specifying under which conditions these motivations will arise. Based on the assumption that the ultimate goals terrorist organizations are striving for can be reduced to either regime change or territorial independence, Sánchez-Cuenca subsequently elaborates formal rational choice models explaining the different strategies that follow from each of these goals: if mobilization for regime change is the major goal, the use of violence is designed to influence followers; if attrition to achieve territorial independence is strived for, violence is used as a signaling device toward the state, demonstrating the terrorist organization’s power. The chapter proceeds with a game theoretical investigation of the effectiveness of different counterterrorist policies and concludes with an assessment of the contribution of rational choice theories to the study of terrorism. The theoretical core of Sánchez-Cuenca’s approach is rooted in full rationality assumptions. It criticizes previous rational choice models for “tinkering arbitrarily with preferences” and suggests that the selfishness assumption—though not required—is a useful one to start with in this case. The formation of terrorist organizations is assumed to be subject to the classical collective action dilemma. Cognitive complexity is limited to the assumption of two ultimate goals for terrorist organizations, and the acknowledgment of ideological benefits that terrorists may derive from their actions. The models refer to information asymmetries or the degree of support for terrorist organizations in the population, but do not systematically elaborate on the role of social network embeddedness. Hence,

Introduction: Rational Choice Social Research

21

structural complexity is kept low in this approach, which works in the tradition of institutional individualism. Jim Morrow’s chapter “Choosing War: State Decisions to Initiate and End Wars and Observe the Peace Afterward” starts with a sketch of three general empirical trends related to the occurrence of wars: wars are rare, escalation of disputes to war is uncommon, and frequency of wars over the long run is declining. The chapter then elaborates two rational choice theories, which are subsequently applied to model the outbreak of war, bargaining during wartime, and consequences in postwar situations. According to bargaining theory, war is the result of bargaining failure, and the chapter provides a detailed discussion of two different sources for bargaining failure: signaling problems and commitment problems. According to principal agent theory, a leader’s war-related decisions are subject to internal politics. The theoretical core behind both perspectives is noncooperative game theory. Consequently, Morrow’s structural individualistic actor model builds on full (strategic) rationality, selfishness, and materialism assumptions, and therefore keeps cognitive complexity at a minimum. Structural complexity is incorporated in form of the internal political structure shaping the principal agent relationship between the leaders and their supporters and citizens. In “Rational Choice Approaches to State-Making,” Edgar Kiser and Erin Powers use a rational choice perspective to analyze the conditions leading to the creation of states.Their chapter is organized chronologically, covering initial state formation, state formation in medieval Europe, and the formation of early modern, bureaucratic, and contemporary states. In many models of statemaking, conflict and war play a pivotal role. Explanations of initial state-making as well as of the formation of medieval states are often either based on a model of conflict over power in zero-sum games, or on models of mixed cooperation and conflict in positive sum games. The latter seem to match better with the empirical evidence than the former: rulers are assumed to be interested in both wealth and security, which they achieve by building political institutions that make possible the production of public goods in collaboration with their subjects. An agency theory is presented to model the structure of bureaucratic states. Here, rulers are seen as the principals who delegate authority to state officials for policy implementation, in particular the collection of taxes. Path dependence is considered as a major factor in the formation of contemporary states. Depending on the approach, actor assumptions in rational choice models of state-making vary from fully rational and selfish actors—as in agency theoretic models that assume self-interested, gain maximizing parties—to the more elaborate microfoundations of social production function theory, which suggests adding cognitive complexity by differentiating types of goals (for example, physical and social well-being) and fairness considerations (such as willingness to incur costs for punishing those who violate fairness rules). Though some of this work considers the role of social network embeddedness, in particular with regard to the solution of collective action problems for revolts or revolutions, fine-grained differentiation of social network structures is not at the heart of rational choice models of state-making and unmaking, the majority of which is informed by an institutional individualistic perspective. With regard to assumptions about microfoundations, both Morrow and

22 Wittek, Snijders, and Nee Sánchez-Cuenca retain the full rationality assumption, and stick to the selfish preference assumption of the canonical rational choice model. Their actors attempt mainly to maximize access to tangible resources, but their models also allow for the incorporation of intangible resources (for example, ideological benefits in the case of terrorists) as potentially important goals. Whereas Sánchez-Cuenca uses an institutional individualist framework that keeps structural embeddedness assumptions to a minimum and essentially assumes that social network structures can be neglected, Morrow’s approach, which acknowledges that principal agent relations and internal political structures matter, has more affinity with structural individualism. markets and organizations How can industrial pollution be reduced? Is there an effective way to increase compliance with hygiene requirements in restaurants? Why is there such a large variety in how business firms are structured? Why do not more firms implement High Performance Human Resource Management? The three chapters in this part address the question of how rational actors shape the institutional contexts governing behavior and interaction in markets, organizations, and states, and under which conditions these efforts succeed or fail. The difficulty of explaining institutional failure has long been considered one of the major challenges for rational choice theory: if actors are far-sighted and rational, they design institutions and governance structures that anticipate potential problems and guarantee the seamless functioning of transactions. Consequently, much previous and current rational choice theorizing attributes institutional failure to exogenous shocks. More recent approaches acknowledge cognitive limitations as potential antecedents. In “Market Design and Market Failure,” Carlos Cañón, Guido Friebel, and Paul Seabright start with an overview concerning how markets came to be established and evolved over time. Modern mass markets are a very recent innovation in this process, but one with the most far-reaching consequences, since well-being of individuals as well as the economic performance of states increasingly depends on the performance of other economies. The chapter then sketches the neoclassical assumptions about markets as they are represented by the two fundamental theorems of welfare economics. In this context they also briefly discuss the recent efforts to accommodate the limited cognitive capabilities of humans in this model. The chapter then proceeds with a description of well-known conditions for market failure: market power, contracting problems, information asymmetries, and externalities. Market design can be viewed as one of several ways to deal with market failures—the other solutions being, for example, bypassing markets through networks or organizations or regulation through government. The main part of the chapter is devoted to the analysis of different types of market failures and conscious efforts to solve these failures by comprehensive market design. The examples cover three types of market design: those related to the solution of asymmetric information problems (for example, those about product quality or personal characteristics), those for rights to inflict externalities (such as pollution), and those involving matching markets (for example, professional placements). The examples provide a vast range of different solutions, some working well, others

Introduction: Rational Choice Social Research

23

being less effective. In particular, two conclusions are worth emphasizing. First, the success of market design depends on accounting for each market’s most important details. Second, once a market is created, there will be ingenious entrepreneurs who learn how to take advantage of the new market conditions. At the same time there will be political entrepreneurs who will exploit the need to “fine tune” the original market design.With regard to actor assumptions, the contribution by Cañón, Friebel, and Seabright sticks to the full rationality and selfishness assumptions of neoclassical economics, though they acknowledge cognitive limitations of actors as a potential source of market failure. Working within the tradition of institutional individualism, they pay much attention to different forms of institutional embeddedness. They assume markets with a large number of exchange partners whose individual actions do not affect prices: dyadic or network embeddedness through which actors could replace or support the market is not considered as relevant. A key problem of research on organizational governance is to explain the emergence, boundaries, and internal organization of firms. Starting with the development of a general definition of organizational governance, the first part of Nicolai Foss’s and Peter Klein’s chapter on “Organizational Governance” discusses the emergence of organizational governance, as well as its various types of problems and failures. Whereas many of the conditions causing organization failure are not specific to organizations but apply also to markets, the “costs of authority” in the form of rent seeking and selective intervention can be regarded as two causes for governance failures that are specific to organizations. The authors then summarize what they consider to be the overall characteristics of rational choice approaches to organizational governance. Their description adheres to a strong rationality view: actors are assumed to be fully rational in the sense of being selfish, being extrinsically motivated, and being able to make complex cost-benefit calculations that are not hampered by selective attention. Social network embeddedness is considered as a phenomenon needing explanation rather than a condition that should be used to explain organizational outcomes. Working within the framework of institutional individualism, the chapter proceeds with a systematic review of current theories of the firm, ranging from the nexus of contract view to formal agency theory and different versions of incomplete contract theories, and also discusses extensions as well as recent attempts toward synthesis. In their overview of applications and evidence, the authors focus on empirical research on organizational boundaries, the internal structure of the firm, mergers and acquisitions, antitrust and regulation, and public bureaucracies. The chapter concludes with a discussion of current critiques of rational choice approaches to organizational governance that urges a relaxation of the strong assumptions with regard to human cognition and (absent) network embeddedness. Rafael Wittek’s and Arjen Van Witteloostuijn’s chapter, “Rational Choice and Organizational Change,” starts with a sketch of five stylized empirical trends characterizing organizational change in advanced capitalist societies. They argue that large organizations increase in size, engage in an increasing number of mergers and acquisitions as well as internal change projects, and pay increasingly high salaries to their top-level functionaries while at the same time being reluctant to implement high-performance human resource management

24 Wittek, Snijders, and Nee practices. The chapter then sketches the little explicit theoretical work that has been done to model organizational change from a rational choice perspective, mainly from an economic point of view. The bulk of this work focuses on micro- or mesolevel antecedents and outcomes, neglecting the societal level. Their subsequent structured review of available empirical research inspired by rational choice ideas covers the antecedents and consequences of strategic change (for example, diversification), corporate restructuring (changes in form, size, and structure), and workplace transformation (such as the introduction of high-performance human resource management). Each of these sections relates to the general trends by specifying major hypotheses and empirical findings. With a sociological rational choice model of organizational change being absent, the chapter concludes with a sketch of such a theory.The guiding idea behind this model is that power is part of a manager’s utility function. The model captures most of the empirical trends discussed in the chapter, and provides many testable hypotheses for future research. The theoretical core of their model builds on straightforward rational choice reasoning, in which individual managers are assumed to maximize income and power. Apart from the inclusion of power in the utility function, the model is congruent with full rationality and selfishness assumptions, and extends the materialism assumption by incorporating power as a social goal. In sum, all three chapters in this part are concerned with institutional emdeddedness and its effects. However, with regard to the other microfoundations, most rational choice models of institutional design and failure prefer to stick closely to the canonical model: they assume selfish preferences, build on strong or bounded rationality assumptions, keep structural complexity and embeddedness assumptions to a minimum, and seldom explicate the relative importance of (in)tangible resources or well-being as goals motivating behavior.

Perspectives for Rational Choice Social Research The chapters in this Handbook amply demonstrate that the rational choice approach has produced a sizable number of empirical studies across a large variety of substantive areas of application. Like any other theoretical paradigm in the social sciences, the approach produced counterintuitive as well as more straightforward hypotheses; empirical support for these hypotheses varies from full over partial corroboration to nonconfirmation or outright rejection; topics cover areas at the center and at the margins of the current social science mainstream: there is no such thing as “the rational choice approach.” As the sixteen contributions demonstrate, there is substantial variation in the assumptions applied by different rational choice scholars. Framed in terms of our coarse-grained four-dimensional typology of microfoundations presented in Table 0.1, we find almost any combination of assumptions, ranging from hardcore neoclassical assumptions about rationality, preferences, and atomism, to full-fledged social rationality models that depart from selfishness assumptions and introduce social goals, network and institutional embeddedness, and automatic frame activation as their behavioral foundations. Two emerging strategies to deal with this variety can be discerned: the first

Introduction: Rational Choice Social Research

25

one, rational choice institutionalism, opts for refining assumptions about the institutional context while keeping the rational choice microfoundations intact. The second one, the social rationality approach, consists in refining the very core of the behavioral microfoundations. rational choice institutionalism Rational Choice Institutionalism (Shepsle 2006) analyzes how institutions— the “rules of the game”—emerge and how they affect behavior and societal outcomes. It acknowledges that a large variety of formal and informal constraints shape individual decisions. “Institutional embeddedness” includes, for example, the obligations inherent in networks of social relationships or spontaneously emerging informal norms of conduct.The growing convergence between historical and rational choice institutionalism (Katznelson and Weingast 2005; Greif 2006), and between new institutional economics and sociological institutionalism (Nee and Ingram 1998), opens up fruitful areas for research. For example, by pointing to the importance of (mis-)alignment between formal rules and informal norms, rational choice institutionalists can account for outcomes—for example, organizational performance—that are otherwise difficult to explain in the context of the canonical rational choice model. Rational choice institutionalism has also successfully informed formal models of institutional change, as in Greif ’s analysis (2006) of medieval trade, which combines historical research and institutional analysis with game theoretical modeling. social rationality microfoundations The second strategy consists of relaxing the behavioral microfoundations. Many scholars would argue that such “thick” or extended rational choice models do not deserve the label “rational choice,” and question the fruitfulness of such an extension. After all, doesn’t increasing the complexity of the microfoundation of rational choice theory deprive it of its major advantage—that is, parsimony in the theoretical core? This question is at the center of a long-standing debate about the merits and limits of methodological individualism, and the key issue is of course what “parsimony” actually means in this context. The following issues seem particularly relevant here. First, a distinction should be made between ontological and methodological individualism (Udehn 2001). Ontological individualism makes statements about the nature of social reality, whereas methodological individualism is about how one should explain this reality—that is, it represents a set of rules or strategies about how to investigate social phenomena. The latter does not require explaining phenomena in terms of individuals alone. As the contributions to this Handbook amply illustrate, there are several varieties of methodological individualism, all of which share the minimum requirement that individuals should be part of the explanation. These methodological postulates are also likely to constitute the core of rational choice social research in the future. Much of the more recent research on human (ir)rationality in fact relates to ontological statements about how individual rationality and cognition work. These findings pose a challenge to the canonical model or “as if ” approaches to rational choice explanations, and may make a revision of the rational actor

26 Wittek, Snijders, and Nee model necessary. The reason is that in many situations, as some of the chapters in this Handbook have shown, rationality, preference, and individualism assumptions of the canonical model simply do not stand up any more against the state of the art in research dealing with human cognition, decision-making, emotions, or related factors. In addition, there is also increasing awareness within the field of economics about situations in which models based on the canonical microfoundations simply produce wrong predictions. This does not mean that humans are not goal oriented, or that core principles of rational choice reasoning should be discarded in their entirety. It means that rationality models need to be carefully redefined and adjusted, in order to accommodate these insights about the modular nature of the human brain. Second, the methodological individualists’ key strategy of dealing with inaccurate predictions and developing better models—in the sense of fit with the empirical data—has always been guided by what Lindenberg has called the “method of decreasing abstraction.” The guiding principle of this strategy is to start with very simple model assumptions (in the sense of an actor model), and add structural and cognitive complexity only in subsequent stages, when it becomes clear that the initial model fails to produce a satisfactory fit with the data. Many of the contributions in this Handbook implicitly or explicitly made use of this strategy. A condition that is often overlooked, however, is the principle of sufficient complexity, which needs to be met in all cases: simple model assumptions always need to be realistic enough to make possible a description of the phenomenon to be explained (Lindenberg). For example, if uncertainty is part of the explanandum, the theoretical core should not assume that actors are fully informed.

Toward an Interlevel, Interfield Research Program Rational choice social research can be seen as a developing explanatory research program in Lakatos’s sense (Kuipers 2001: ch. 1). Research programs have a hard core and are equipped with a “positive heuristic”: the hard core bears directly on the solution of the problem, whereas the positive heuristic deals with the question of how the hard core can be defended against attacks. The fruitful attempts to incorporate state of the art insights of cognitive sciences into economic reasoning can be seen as one example of such a positive heuristic (see, for example, Gächter’s and Lindenberg’s contributions to this Handbook). Furthermore, it should not be forgotten that rational choice social research—in particular its applications in social sciences other than economics—is a comparatively young program. Research programs develop in phases, and the rational choice approach is no exception. Usually, programs first go through an “internal” phase, consisting of a heuristic and an evaluation step, and then enter an “external,” or “application,” phase. The heuristic step is characterized by an elaboration and evaluation of the core idea and first attempts to develop positive heuristics to protect the core idea. In the evaluation step, the core idea is elaborated into specific theories for a limited number of subdomains or contexts. If the evaluation step yields positive results—in terms of explanatory and predictive success of the research program—“this usually leads to the more or less general acceptance of the core theory of the

Introduction: Rational Choice Social Research

27

program and it becomes clear for which domain and in what sense and to what extent the core theory can be assumed to be true. It should be stressed that many, if not most, programs in the empirical sciences, not to mention philosophy, do not reach this point” (Kuipers 2001). To what degree the core theory of the rational choice research program has reached this point is of course a matter of perspective, but it cannot be denied that within the field of economics, the core theory has become widely accepted and has also generated many empirical successes. This is considered one of the conditions favoring the transition of a research program into the external or application phase: the program is directed toward questions that are prima facie independent of the program itself. In the case of rational choice research, this means application of the core to the solution of social rather than just economic problems. With this step, the program crossed the boundary to other fields of research, fields that have traditionally been covered by disciplines such as sociology, political sciences, anthropology, or related subdisciplines—in which consensus about a theoretical core so far tends to be lower, and where many research programs have not yet reached the external application phase. Perceived as an illegitimate and ungrounded expansion into forbidden territories by many, and fueled by perceived disciplinary, ideological, and existential threats (Goldthorpe 2007: 164), this move gave rise to the well-known fears about “economics imperialism” (Fine and Milonakis 2008) and “colonization” (Archer and Tritter 2000). These kinds of descriptions frame the relationship between the rational choice research program and research programs in other social science disciplines as competitive. Depending on the phases in which the competing programs are found (Kuipers 2001), competition is about the adequacy of core ideas (if both programs are in the internal phase), the suitability of a program for the solution of problems external to science (if both programs are in the external phase), or the validity or degree of accuracy of a program’s external application (if one program is in the internal and the other the external phase). Such competition can in principle be fruitful if it focuses on the solution to real problems rather than on attacks based on misconceptions, to stay with Goldthorpe’s characterization. In the former case, interaction between research programs often leads to convergence, cooperation, and synthesis. Critics of the rational choice program are suspicious of this approach, because they have perceived—and often rightly so—this cooperation to be asymmetrical in nature, with economics claiming to be able to deliver the supply program for the solution of problems that the other social sciences are unable to solve. But the past two decades have substantially transformed the major points of reference for this debate. We believe that this transformation provides a much better ground for fruitful cooperation of the rational choice research program with other social science research programs than was the case during the time when Gary Becker launched his project of economics imperialism. There are at least three reasons why there now is a better ground for fruitful cooperation of the rational choice research program with other social science research. One reason lies within economics: research especially in behavioral economics, and analyses of the developments leading to the recent economic crisis, have empirically demonstrated the limits of the canonical RC model. Another reason is within rational choice social research: the combination

28 Wittek, Snijders, and Nee of relaxations of the RC model and empirical research, as presented in this Handbook, has shown that this adapted RC approach can be empirically fruitful and conceptually plausible. A third reason is developments in the cognitive (neuro-)sciences, which have shown that human goal-directed behavior is organized not only by rational thought but also, and often more, by semiautomated processing—which moderates the goal-directed nature but does not altogether do away with it. Research programs in these fields differ from the rational choice program and its (sociological) competitors in that they focus on a different level of analysis than economics or the other social sciences, such as brain activity, hormone activity, or neural linkages: the micro-micro level, as it has also been called (Ainslie 1992). The quick rise of fields such as behavioral economics or neuroeconomics, where findings from this type of research are actively and carefully scrutinized for their potential added value (Rubinstein 2008) and are incorporated into economic models (for example, Ross et al. 2008), shows that economics as a field takes these developments seriously. The interaction between economic research programs and these cognitive research programs is a good example of interlevel asymmetric cooperation between different research programs, with the (neuro)cognitive sciences acting as supplier. The importance of the insights generated by this cooperation for the rational choice research program cannot be overestimated, since they allow us more strongly to position its actor assumptions in empirical research. Though this interfield, interlevel cooperation is just in the beginning stages, its efforts have already resulted in remarkable insights, refinements, and corrections concerning our assumptions about rationality, individualism, and preference. It may eventually serve as an empirical foundation for a revised theoretical core of rational choice social research.

References Abell, P. 1992. “Is Rational Choice Theory a Rational Choice of Theory?” In Rational Choice Theory: Advocacy and Critique, edited by J. S. Coleman and T. Fararo, 183–206. Newbury Park, CA: Sage. Ainslie, G. 1992. Picoeconomics: The Strategic Interaction of Successive Motivational States within the Person. Cambridge: Cambridge University Press. Archer, M., and J. Tritter. 2000. Rational Choice Theory: Resisting Colonization. London and New York: Routledge. Ariely, Dan. 2008. Predictably Irrational: The Hidden Forces That Shape Our Decisions. London: Harper and Collins. Camerer, C., G. Loewenstein, and M. Rabin, eds. 2004. Advances in Behavioral Economics. New York: Russell Sage Foundation. Coleman, J. 1990. Foundations of Social Theory. Cambridge: Belknap. Cox, G. 1999. “The Empirical Content of Rational Choice Theory: A Reply to Green and Shapiro.” Journal of Theoretical Politics 11, no. 2: 147–69. Daum, J. H. 2003. Intangible Assets and Value Creation. Chichester: John Wiley and Sons. Fehr, E., and S. Gächter. 2000. “Cooperation and Punishment in Public Goods Experiments.” American Economic Review 90, no. 4: 980–94. Fine, B., and D. Milonakis. 2008. From Economics Imperialism to Freakonomics. London: Routledge. Friedman, J., ed. 1996. The Rational Choice Controversy. New Haven: Yale University Press.

Introduction: Rational Choice Social Research

29

Goldthorpe, J. 2007 [2000]. On Sociology. Vol. 1: Critique and Program. Stanford: Stanford University Press. Green, D., and I. Shapiro. 1994. Pathologies of Rational Choice Theory. New Haven: Yale University Press. Greif, A. 2006. Institutions and the Path to the Modern Economy. Cambridge: Cambridge University Press. Gross, N. 2009. “A Pragmatist Theory of Social Mechanisms.” American Sociological Review 74, no. 3: 358–79. Hechter, M., and S. Kanazawa. 1997. “Sociological Rational Choice Theory.” Annual Review of Sociology 23: 191–214. Hedström, P. 1998. “Rational Imitation.” In Social Mechanisms: An Analytical Approach to Social Theory, edited by P. Hedström and R. Swedberg. Cambridge: Cambridge University Press. Hedström, P., and P. Bearman. 2009. “What Is Analytical Sociology All About? An Introductory Essay.” In The Oxford Handbook of Analytical Sociology, edited by P. Hedström and P. Bearman, 3–24. Oxford: Oxford University Press. Hedström, P., and C. Stern. 2008. “Rational Choice and Sociology.” In The New Palgrave Dictionary of Economics, edited by S. N. Durlauf and L. E. Blume, 872–77. Basingstoke, UK: Palgrave Macmillan. Hedström, P., and R. Swedberg, eds. 1998. Social Mechanisms. An Analytical Approach to Social Theory. Cambridge: Cambridge University Press. Hodgson, G. 2007. “Meanings of Methodological Individualism.” Journal of Economic Methodology 14, no. 2: 211–26. Katznelson, I., and B. Weingast, eds. 2005. Preferences and Situations: Points of Intersection between Historical and Rational Choice Institutionalism. New York: Russell Sage. Knight, F. 1971 [1921]. Risk, Uncertainty, and Profit. Chicago: Chicago University Press. Kronenberg, C., and F. Kalter. 2012. “Rational Choice Theory and Empirical Research: Methodological and Theoretical Contributions in Europe.” Annual Review of Sociology 38: 73–92. Kuipers, T. 2001. Structures in Science. Dordrecht: Springer. Lindenberg, S. 1992. “The Method of Decreasing Abstraction.” In Rational Choice Theory: Advocacy and Critique, edited by J. S. Coleman and T. J. Fararo, 3–20. Newbury Park, NJ: Sage. ———. 2001. “Social Rationality versus Rational Egoism.” In Handbook of Sociological Theory, edited by J. Turner, 635–68. New York: Kluwer Academic/Plenum. Macy, M., and A. Flache. 1995. “Beyond Rationality in Models of Choice.” Annual Review of Sociology 21: 73–91. Milgrom, P., and P. Roberts. 1992. Economics, Organization, and Management. Englewood Cliffs, NJ: Prentice Hall. Nee, V., and P. Ingram. 1998. “Embeddedness and Beyond: Institutions, Exchange, and Social Structure.” In The New Institutionalism in Sociology, edited by M. Brinton and V. Nee, 19–45. New York: Russell Sage. Raub,W.,V. Buskens, and M. van Assen. 2011.“Micro-Macro Links and Microfoundations in Sociology.” Journal of Mathematical Sociology 35, no. 1: 1–25. Ross, D., C. Sharp, R. E. Vuchinich, and D. Spurret. 2008. Midbrain Mutiny: The Picoeconomics and Neuroeconomics of Disordered Gambling: Economic Theory and Cognitive Science. Boston: MIT Press. Rubinstein, A. 1998. Modeling Bounded Rationality. Cambridge, MA, and London: MIT Press. ———. 2008. “Comments on Neuroeconomics.” Economics and Philosophy 24: 485–94. Shepsle, K. A. 2006. “Rational Choice Institutionalism.” In Oxford Handbook of Political Institutions, edited by S. Binder, R. Rhodes, and B. Rockman, 23–38. Oxford: Oxford University Press.

30 Wittek, Snijders, and Nee Simon, Herbert. 1957. “A Behavioral Model of Rational Choice.” In Models of Man, Social and Rational: Mathematical Essays on Rational Human Behavior in a Social Setting. New York: Wiley. Thaler, R., and C. Sunstein. 2008. Nudge: Improving Decisions about Health, Wealth, and Happiness. London: Penguin. Todd, P., and G. Gigerenzer. 2000. “Précis of Simple Heuristics That Make Us Smart.” Behavioral and Brain Sciences 23, no. 5: 727–80. ———. 2007. “Environments That Make Us Smart: Ecological Rationality.” Current Directions in Psychological Science 16, no. 3: 167–71. Torsvik, G. 2000. “Social Capital and Economic Development—A Plea for the Mechanisms.” Rationality and Society 12, no. 4: 451–76. Udehn, L. 2001. Methodological Individualism: Background, History, and Meaning. London: Routledge. Voss, T., and M. Abraham. 2000. “Rational Choice Theory in Sociology: A Survey.” In The International Handbook of Sociology, edited by S. R. Quah and A. Sales, 50–83. London: Sage. Williamson, O. 1975. Markets and Hierarchies. New York: Free Press. Wippler, R., and S. Lindenberg, S. 1987. “Collective Phenomena and Rational Choice.” In The Micro Macro Link, edited by J. C. Alexander, B. Giesen, R. Münch, and N. J. Smelser, 135–52. Berkeley: University of California Press.

chapter

Rationality, Social Preferences, and Strategic Decision-making from a Behavioral Economics Perspective

1

simon gächter

Introduction The central assumption of the rational choice approach is that decisionmakers have logically consistent goals (whatever they are), and, given these goals, choose the best available option.This model, in particular its extension to interactive decision-making (game theory), has had a tremendous impact in the social sciences, in particular economics, and has allowed for great theoretical progress in the latter. For instance, the model and its formalization have led to important insights into the functioning of economic systems (see Mas-Colell, Whinston, and Green [1995] for a leading textbook account). They have also shed new light on numerous nonmarket processes, such as crime, addiction, family behavior, political decision-making, and organizational behavior (Coleman 1990; Becker 1993; Hechter and Kanazawa 1997; Gintis, Bowles, Boyd, and Fehr 2005). Rational choice theory in the form of game theory is now the core theoretical tool of economics. It is therefore certainly justified to speak of the rational choice model as a “success story.” Rational choice models often assume that agents are “unboundedly rational” and always know what is best for them. This assumption has long found many critics (Simon 1982). A further assumption that is not part of the canonical rational choice model but is frequently invoked in applications is that agents are primarily self-regarding. This assumption has been challenged in particular in the last decade through the accumulation of empirical, in particular experimental, evidence that has shown substantial and robust deviations from what selfishness predicts. In the past the discussions about the fruitfulness of the rational choice approach were based largely on philosophical convictions and less on facts. In this chapter I will argue on the basis of insights from behavioral economics that the usefulness of the rational choice approach is also an empirical question and not just a philosophical one. My approach is related to that of Hechter and Kanazawa (1997), who argue for the fruitfulness of rational choice explanations by discussing its empirical successes across a large variety of interesting

34 Simon Gächter sociological domains. My arguments are based on data from many laboratory experiments which all share the feature that the theoretical predictions are derived from rational choice models (typically game-theoretic models) and that decisions in the experiments have financial consequences for the participants, which give them an incentive to take decisions seriously. Specifically, I will use experimental results to argue that one can acknowledge that humans are boundedly rational and nevertheless appreciate the predictions made by rational choice models that at times rest on unrealistically strong assumptions. Moreover, I argue that a rational choice approach does not imply assuming that everyone is selfish. I start by giving a short characterization of the canonical rational choice model, including the selfishness assumption. With the help of one famous research program on the functioning of experimental markets I will illustrate one main point that will recur several times. I argue that one can fruitfully use rational choice theory to predict social outcomes even if the assumptions entering the model are empirically invalid. I will first discuss the success and limits of the rational choice approach. Then, I focus attention on the empirical investigation of the selfishness assumption. I will present evidence from several economic experiments that have been used as tools to uncover the structure of people’s “social preferences.” Numerous experiments have shown that the selfishness assumption is invalid as a description of typical behavior, although in all experiments we do find selfish people.The results on social preferences raise the challenge of how to model them, and I will sketch some attempts.

Rational Choice and Homo Economicus the rational choice model My main goal here is to set the stage for the subsequent analysis. For fuller accounts of various aspects and criticisms of the rational choice approach I refer to Coleman (1990); Hedström (2005); Elster (2007); Lindenberg (2008); and Gintis (2009). The rational choice model aims to explain the decisions of individuals and the individual and, in particular, the social consequences of those decisions. Figure 1.1 illustrates the core elements of the rational choice model. It is useful to distinguish two conceptual buildings blocks: decision theory and equilibrium theory. Decision theory describes how decision-makers should make decisions (normative approach), or actually do make decisions (positive approach). In case individuals interact, a so-called solution concept describes the predicted social outcome. The usual assumption is that the resulting outcome will be an equilibrium. Decision theory typically makes a conceptual distinction among preferences, beliefs, and constraints. Preferences describe how individuals rank the available alternatives according to their subjective tastes. Beliefs are the second conceptual building block behind the rational choice model. For instance, in which stocks you invest will probably depend on your expectations about the future earnings potential of that stock. However, choice depends not only on one’s subjective taste and beliefs but also on constraints. Constraints are the set of alternatives that are available to an individual. For instance, you might prefer a Ferrari to

Rationality, Social Preferences, and Strategic Decision-making

35

Social interactions

Preferences Optimal decision

Constraints Beliefs

Outcome (Prediction): Equilibrium

Decision theory

figure 1.1. The rational choice framework.

a Ford, but your available income might constrain you from buying a Ferrari. After specifying preferences, beliefs, and constraints, the analysis proceeds by assuming that individuals choose a feasible option that best satisfies the individual’s preferences (“the utility-maximizing choice”). The rationality assumption enters the picture on the conditions that are placed on preferences.The typical rationality assumptions are that (i) preferences are complete, which means an individual can compare all relevant alternatives and rank them, (ii) preferences are transitive—that is, if an individual prefers alternative A over B, and B over C, then that individual should also prefer A over C, and (iii) preferences are independent of irrelevant alternatives—that is to say, the relative attractiveness of two options does not depend on other options available to the decision-maker. This rationality assumption ensures that preferences are noncyclical and therefore contain a “best element,” which will be chosen if available. Notice that the rationality assumption is nothing more than a consistency requirement and is completely mute on the content of preferences. This has not gone without criticism. For instance, Amartya Sen, in an influential article, has mocked this conception of economic man as being a “rational fool” (Sen 1977). Rational choice theory becomes a theory of social interactions if the individual decision-makers interact with each other. The social sciences are typically interested in the social outcomes of the interaction of individuals. One particularly important approach, especially in economics, is to analyze social outcomes from an equilibrium perspective. An equilibrium is a situation in which, given the decisions of all other decision-makers, no agent has an incentive to change behavior. If there is still an incentive to change behavior for at least one decision-maker, then the resulting outcome of the social interaction cannot be an equilibrium. Two important classes of social interaction are markets and (small) group interactions. In markets decision-makers face prices and have to decide how much they want to produce or buy at given prices and income. In the prototypical case a single individual cannot affect prices and thus takes them as given. An equilibrium is a situation in which at given prices agents want to change neither their production plans nor their demands (for a comprehensive textbook account, see Mas-Colell et al. 1995). In strategic situations an equilibrium is reached if, given the strategic behavior of other players, no one wants to change strategy. Thus, in many social science applications the analysis

36 Simon Gächter does not stop after looking at individual decisions but proceeds to predicting social consequences under the assumption that the resulting outcome will be an equilibrium. Of course, it is an empirical question whether the equilibrium prediction (which is derived for specific assumptions about people’s preferences) is consistent with the data. the selfishness assumption The rational choice approach is a flexible framework that can account for any preferences as long as they obey the consistency axioms. This flexibility is also a weakness, because if almost any preferences are permissible almost anything and therefore nothing can be explained. For that reason economists very often made additional preference assumptions to give the rational choice analysis more structure or assume that people have the same preferences (see, for example, the influential discussion by Stigler and Becker 1977). The most common and long-standing assumption is that decision-makers are selfregarding. Self-regarding agents will take the welfare of others into account only for instrumental reasons to increase their own well-being. In essence there are two justifications for the selfishness assumption. A first justification is tractability in modeling, as selfishness can simplify the analysis considerably. The selfishness assumption often allows for exact predictions, which can be confronted with appropriate data that might refute the selfishness assumption. Moreover, it is often of independent interest to understand what would happen if everyone were selfish. Understanding the consequences of selfishness serves therefore as an important benchmark for understanding nonselfish behavior. I have already alluded to a second justification of the selfishness assumption, that assuming nonselfish preferences, although logically possible, is tantamount to opening “Pandora’s Box”—which in this case means that any bizarre behavior can be rationalized. This argument has considerable merit in the absence of empirical means to assess the structure of people’s social preferences. Yet, as I will demonstrate below, the development of experimental tools allows us to observe people’s social preferences under controlled conditions. Behavioral scientists have lately even added neuroscientific tools to understand the neural mechanisms behind people’s social preferences (Fehr 2009b). five methodological remarks on the rational choice model I conclude this discussion with five methodological remarks and refer the interested reader to Gintis (2009) for a more in-depth discussion of the issues raised here. The first remark concerns the selfishness assumption, to which we will return in more detail below. Nothing in the rational choice framework dictates that preferences have to be self-regarding; the only necessary assumption is that preferences obey some consistency axiom such that optimal choices are logically possible. Thus, a rational decision-maker can have other-regarding preferences and still obey all the rationality axioms. As we will show below, there is substantial experimental evidence for both that many people are not purely self-regarding and that other-regarding preferences often do obey consistency axioms (at least in simple situations).We will also show evidence that behavioral

Rationality, Social Preferences, and Strategic Decision-making

37

patterns in experiments in which social preferences are important are consistent with predictions derived from rational choice models. Second, one may criticize the rational choice approach as being unrealistic in the sense that human decision-makers do not reason and behave like those portrayed in the theory.This argument has a lot going for it, as has been pointed out emphatically from different angles by Simon (1982); Gigerenzer and Selten (2001); or Loewenstein (2007). The limits of the rational choice approach can be discussed with the help of Figure 1.1. Psychologists and other behavioral scientists have long argued that people are faulty statisticians, as they often do not update information rationally and therefore hold nonrational beliefs, they overweigh small probabilities and underweigh large probabilities, and they fall prey to various framing effects (see, for example, Kahneman and Tversky 2000; Lindenberg 2008). Furthermore, people are often swayed by their emotions and evaluate options differently depending on whether they are in a “hot” or a “cold” state (see, for example, Vohs, Baumeister, and Loewenstein 2007). People make mistakes in predicting their future utility (for example, Loewenstein, O’Donoghue, and Rabin 2003) and are often too impatient and show weaknesses of will (Loewenstein, Read, and Baumeister 2003). These are all important findings.They have been possible because exact model predictions as derived from formalized rational choice theories were confronted with appropriate data. These findings have stimulated extensive research in all social sciences. In economics they have helped to pave the way for “behavioral economics,” which is the integration of psychological and sociological insights into economics. Behavioral economics is now branching out into the various subdisciplines of economics and transforming them by integrating psychological and sociological insights into rational choice frameworks (see Camerer, Loewenstein, and Rabin 2004; Diamond and Vartiainen 2007).This development would not have been possible without the rational choice approach, which helps pinpoint the problems in standard theory. Yet, as I will demonstrate by way of an example below, a rational choice analysis often makes surprisingly accurate predictions despite considerable violations of some assumptions that underlie the theory. This point has already been made by Becker (1962). Such a viewpoint does not necessarily imply discarding the importance of the empirical findings mentioned above, or a methodology that cares only for prediction accuracy and not so much for the empirical validity of the underlying assumptions. Quite the contrary, I will discuss evidence that points to the violation of the selfishness assumption as being a cause of a failure of prediction accuracy in several games of interest to social scientists. It is an empirical question how sensitive theoretical predictions are to particular violations of underlying assumptions. The rational choice approach provides theoretical rigor in uncovering which violations matter for prediction accuracy and which not. Moreover, very often we are interested in the comparative statics prediction of a model and, as we will see throughout this chapter, these predictions are often empirically validated. Third, and somehow relatedly, equilibrium theory does not explain how a particular equilibrium is actually reached.Yet evolutionary game theorists have shown that equilibria may be reached through a process of trial and error or

38 Simon Gächter other adaptive processes (Gintis 2000a). Moreover, behavioral game theorists, who study actual human decision-making in strategic contexts, have devised theoretical and empirical learning models that help us understand and predict how and when equilibria are approached under a given learning dynamic (see Camerer 2003 for a comprehensive overview). Fourth, equilibria need not be socially optimal, despite the fact that they arise from all players choosing their individually optimal action. The “tragedy of the commons” or the famous prisoner’s dilemma are prime examples. Moreover, even in games with multiple equilibria (“coordination games”), inefficient equilibria can result from individually optimal interaction, and players can even be fully aware that they are playing an inefficient equilibrium. In general, a very important task of modern social science is to understand inefficient social outcomes (Bowles 2003). Fifth, the rational choice approach does not necessarily advocate methodological individualism, although most defenders of rational choice probably believe that social phenomena should be explained solely from the actions of individuals. Gintis (2009) argues convincingly against this doctrine. In his view individuals’ common understanding (coordinated beliefs) is crucial to understand many important social phenomena such as social norms, which coordinate the interaction of rational individuals.

The Success of the Rational Choice Model and an Illustrative Example Despite some important limits I have discussed briefly above, I believe the rational choice model is a considerable success story, both theoretically and empirically. First, the breakup of the model into three conceptual building blocks, coupled with an equilibrium analysis (and in some cases powerful mathematical tools), has allowed for the development of a consistent (unified) theoretical framework to address many important topics of social interactions and economic decision-making in formalized theories. This is certainly true in economics, where modern theory provides a framework for the analysis of such diverse topics as international trade, public finance, consumer behavior, production theory and investment behavior, financial decision-making, retirement decisions, labor market behavior, and so forth. This framework has led to a unified discourse in economics, and prevented a fragmentation into disconnected subdisciplines. On a related note, the rational choice approach has also led to the development of game theory (discussed in greater detail by Buskens and Raub, this volume; for textbook accounts see, for example, Colman 1999; Gintis 2000a). Game theory is particularly promising, as it can serve to unify the discourse about human behavior among all the behavioral sciences (Gintis 2009). The interdisciplinary discourse based on experimental games as tools for empirical investigations has already started. Experimental games are used by political scientists (Morton and Williams 2010), anthropologists (for example, Henrich et al. 2004), primatologists (such as Jensen, Call, and Tomasello 2007), cognitive neuroscientists (Camerer 2009), and evolutionary biologists (Hagen and Hammerstein 2006).

Rationality, Social Preferences, and Strategic Decision-making

39

Second and well known to (empirical) sociologists working in the rational choice framework, the scope for applying rational choice approaches extends way beyond economics (Hechter and Kanazawa 1997). Rational choice approaches have been applied to issues that were long seen as outside the realm of economics, such as criminal behavior, human capital formation, education, family decision-making, addiction, organizational behavior, and the like. Thus, the rational choice model has enhanced our understanding of social processes way beyond market transactions (Coleman 1990; Becker 1993; Bowles 2003; Gintis et al. 2005). an illustrative example In this section, I illustrate the empirical success of the rational choice approach by briefly discussing an influential research program initiated by experimental economist and Nobel laureate Vernon Smith half a century ago. This exemplary discussion serves two purposes. It illustrates how a theoretical paradigm can be tested in the laboratory and how theoretical predictions can be contrasted with the data. It also illustrates that the rational choice predictions are met surprisingly well by the data, although the assumptions that enter the model are invalid. Thus this example illustrates one argument of this chapter, that rational choice predictions can be useful even if one believes that agents are boundedly rational. The example concerns a research program on one of the most important models in economics: the model of supply and demand and the theory of perfect and imperfect competition that builds on it. The goal is to explain the coordination of supply and demand under complete contracts. Complete contracts are contracts in which the terms of the exchange are comprehensively specified and are enforceable by third parties, such as the courts. One of the most important results of modern economic theory is that under complete contracts and the assumptions of rationality and price-taking behavior (agents have no market power and therefore have to take prices as given), coordination of prices is possible such that on all markets supply equals demand. Such an allocation is efficient in the sense that no economic agent can be made better off without making someone else worse. This model is the building block of much of modern economic theory and policy advice. It is therefore of considerable interest to understand whether this theory predicts correctly, and whether unregulated competitive markets actually clear—that is, whether prices are such that supply equals demand. Testing this model is a big challenge, as in reality prices are formed in market institutions, like stock exchanges or posted-offer markets (where suppliers announce a price and customers decide whether they buy at that price or not). Moreover, to test the model the researcher would need to know supply and demand, which rest on unobservable preference and cost parameters. Finally, if one has just price and quantity data, an identification problem exists because, if the theory holds, all price and quantity combinations are on both the demand and supply curve. Smith (1962) found an ingenious solution to solve the problem of knowing supply and demand. His idea was to endow experimental participants with valuations for an artificial commodity for which people have no “homegrown” preference that is unobservable to the experimenter. These induced valuations

40 Simon Gächter given to the buyer can be interpreted as demand (“maximum willingness to buy”) and those given to sellers as supply (“minimum selling prices”). Since the experimenter knows these induced values, the experimenter can calculate the market clearing price and quantity and therefore compare the results in an experiment with these predictions. Furthermore, as mentioned, trade takes place within a market institution, and the experimenter can fix this institution, to test what impact institutional rules have on price formation and market clearing, independent of supply and demand parameters (see Smith 1986 for a comprehensive discussion of this methodology). Smith’s experimental methodology started a long stream of experimental papers that investigated how prices are formed in diverse market institutions and to what extent the theory of supply and demand actually can predict final allocations. Discussing this literature is beyond the scope of this chapter. The interested reader should consult Kagel and Roth (1995). The upshot of many of these studies is that the competitive market equilibrium predicts the results in experimental markets surprisingly well. Figure 1.2 illustrates the results of an experiment I ran with the purpose of replicating a particularly interesting experiment by Davis, Harrison, and Williams (1993). The top part of the figure shows the “induced” supply and demand schedules. Take the demand schedule labeled “lowest demand” and the supply schedule labeled “supply” first. For instance, there was a buyer whose maximal value for one unit was 250, so this buyer had an incentive to buy for a price of at most 250 because the buyer’s earnings were 250 minus the price paid; for another buyer the maximal willingness to buy was 245, and so on; the buyer with the lowest maximal willingness to pay had a valuation of only 175. Similarly, there was a seller for whom the induced reservation price was 100 and any price above that resulted in a gain of “price—100.” Other sellers had higher reservation prices, which together constitute the supply curve. The market equilibrium (where supply equals demand) predicts a price of 225 and that six units will be traded. The market was run as a so-called double auction market whereby both buyers and sellers can submit prices according to some prespecified rules; each can accept the other’s offer at any time. The experimental participants traded in this market for fifteen periods. The supply function stayed constant during the whole experiment—that is, participants in the role of a seller always kept their given valuations. By contrast, in the third trading period the participants in the role of buyers received new valuations, such that the whole demand function was shifted upward. This shifted the predicted prices and quantities to 250 and 7, respectively. In the subsequent periods, demand was shifted up every period, until in period 8 the highest demand illustrated in Figure 1.2 was implemented. From then on the demand function was shifted down every period until the lowest demand was implemented again in period 15. This up- and downward shifting of the demand function resulted in predicted prices and quantities illustrated in the lines without symbols in the bottom panel of Figure 1.2. It is important to notice that participants in this market did not know the whole supply and demand schedules and how demand was shifted; in every trading period they knew only their own valuation. The results, illustrated in the bottom panel, are striking and replicate

Price

Rationality, Social Preferences, and Strategic Decision-making 500 450 400 350 300 250 200 150 100 50 0

Induced supply and demand

41

Supply highest demand lowest demand

0

5

10

15

20

Quantity Market outcomes in the experiment

25

400 350 300 250

15

200 10 actual quantity predicted quantity actual price predicted price

5

Price

Quantity

20

150 100 50 0

0 1

2

3

4

5

6

7

8

9

10 11 12 13 14 15

Trading period

figure 1.2. The supply and demand model and its predictive success. The supply function remained fixed during the experiment. The demand function was shifted between the lowest and highest demand during the fifteen trading periods. Demand was increased in the first half and then decreased, resulting in the price and quantity predictions indicated in lines without markers in the bottom panel.

the findings of Davis et al. (1993). Transaction prices were very close to the equilibrium price in all periods; and the traded quantities were near to their equilibrium quantities as well, with the exception of the first two trading periods. This is just one result in an impressive stream of research initiated by Smith (1962). Smith and many collaborators as well as other researchers investigated the robustness of these results in various directions (see Kagel and Roth 1995 for an overview).The surprising finding is that with some small caveats the data are consistent with the equilibrium predictions. This is a remarkable result because the equilibrium prediction was derived under the assumption that supply and demand are known to all traders (and the “auctioneer”), and everyone is a price-taker. By contrast, in the experiments (and in many real world markets), traders were as much price-makers as price-takers and knew only their own valuations, and not the whole supply and demand schedule. Moreover, learning

42 Simon Gächter opportunities were rather limited. Thus the conditions assumed in theory are not necessary for achieving equilibration. In summary, rational choice predictions can organize the data and predict changes even if not all assumptions in the underlying theory are met. The relevance of this argument extends beyond markets. For instance, mixed strategy equilibrium predictions are often surprisingly good predictors of (aggregate) behavior, even though the mixed strategy equilibrium rests on psychologically implausible assumptions (see, for example, Camerer 2003: ch. 3). Other examples of when the data confirm surprisingly closely to rational choice predictions are oligopoly games (Huck, Normann, and Oechssler 2004), or tax-subsidy mechanisms for the provision of public goods (Falkinger et al. 2000).

Homo Economicus Put to Test—Social Preferences and Rationality At face value the selfishness assumption behind many applied rational choice models seems to be violated already by the fact that many people vote even in anonymous situations, take part in collective actions, often manage not to overuse common resources, care for the environment, mostly do not evade taxes on a large scale, donate to public radio as well as to charities, and so forth. Yet the selfishness assumption cannot be dismissed so easily. In reality many factors might give even selfish individuals an incentive to behave prosocially, although they are not so motivated. Thus the exact incentives people face must be fully controlled to gain conclusive evidence about the validity of the selfishness assumption. In principle, methods such as controlled attitude surveys using vignettes (see, for example, Kahneman, Knetsch, and Thaler 1986), or interviews (Bewley 1999) are possible sources of evidence for tests of the selfishness assumption. However, the drawback is that these instruments do not measure behavior but attitudes that might even be clouded by a social desirability bias. Laboratory experiments with decision-dependent financial incentives have the advantage that they measure people’s behavior in situations in which true opportunity costs for their decisions exist and are known by the researcher. laboratory experiments as test instruments and behavioral models The most important advantage of an experiment is that relevant parameters can be controlled by the experimenter. This allows for stringent tests of the selfishness assumption (and any other assumption of any model one wants to test) (see Friedman and Sunder 1994 for practical details; and Guala 2005; Bardsley et al. 2009; Falk and Heckman 2009; and Morton and Williams 2010 for the methodology of [laboratory] experiments; Croson and Gächter 2010 describes my own views). Experimental practices differ somewhat across disciplines (see, for example, Hertwig and Ortmann 2001). Here I briefly describe the standards that have emerged in experimental economics, as most of the experiments I will discuss below have been conducted according to those standards. First, the experiment runs according to a fixed protocol. Second, the participants receive written rules of the game. Third, decisions are usually anonymous, at least among participants. Most of the time experiments are computerized and

Rationality, Social Preferences, and Strategic Decision-making

43

take place in laboratories in which subjects are separated by partitions. Fourth, subjects get paid depending on their decisions (that is, they do not simply receive a flat payment for their participation). The level of payments is usually commensurable with what a participant could earn elsewhere in about the time it takes to participate in an experiment. Finally, there is no deception. The major reason for not using deception is to ensure the internal validity of the experiment, which might be compromised if the participants were to become suspicious about the experiment and a reputation that deception is being used were to develop (ibid.).The internal validity is compromised because subjects might effectively play another game than the experimenter thinks they are playing. The experiments I will discuss below are now classic instruments or behavioral models to measure various aspects of people’s social preferences.The Dictator Game measures the simplest form of other-regarding preferences. The Ultimatum Game has been used to study norms of fair sharing and negative reciprocity. The Gift Exchange Game measures positive reciprocity. The Trust Game helps us to study trust and trustworthiness (and positive reciprocity). The Public Goods Game is a tool for studying cooperation. The Public Goods Game with costly punishment allows for negative reciprocity. A skeptic might argue that the information we can gain from artificial laboratory experiments with undergraduates is limited. These skeptics should keep in mind that the most important results I will report below have been replicated across many participant pools, sometimes in representative surveys (Fehr, Fischbacher, Rosenbladt et al. 2002; Bellemare and Kröger 2007; Ermisch et al. 2009); and increasingly in field experiments (for example, Frey and Meier 2004; Falk 2007; and Shang and Croson 2009). Moreover, the findings from the lab are consistent with survey evidence (see, for example, Dohmen et al. 2009). Lab experiments can also be used to predict field experimental outcomes (for example, Karlan 2005; Benz and Meier 2008; Rustagi, Engel, and Kosfeld 2010; Carpenter and Seki 2011). My discussion here will be necessarily selective and my criteria are as follows. I will highlight evidence from one-shot experiments that were designed to test for the relevance of social preferences in an environment in which confounds with strategic incentives are excluded by design. These experiments help testing the selfishness assumption discussed above. I will also briefly discuss the role of sociodemographic background characteristics in experimental choices. A second emphasis is on experiments where the relevance of strategic effects can be gauged from comparisons with behavior in one-shot games. Thirdly, I will discuss findings from experiments that highlight the tradeoffs people make between behaving proselfishly and prosocially. These latter two selection criteria help me to shed light on the relevance of rational choice approaches in explaining the data. For more comprehensive treatments and discussions about the various purposes for which these experiments have been used, I refer the reader to Camerer (2003): ch. 2; and to Chaudhuri (2009). altruism the dictator game One of the first studies of the Dictator Game is by Forsythe et al. (1994). It is a two-player game in which players are randomly allocated to one of two

44 Simon Gächter roles: the “dictator” receives an amount of money and is asked how much he or she wants to give to a passive recipient, who has to accept what is offered. A rational and selfish dictator will keep all the money. Positive sharing is seen as evidence for altruism, or other-regarding preferences in general. The results in Forsythe et al. (ibid.) do not support the selfishness prediction. Twenty-two percent of the participants transferred a positive amount to the powerless recipient. Across a large set of dictator experiments, Engel (2011) finds that dictators share about 28 percent of their allocated sum with the recipient.The results vary greatly across different treatments, however. It matters strongly who the recipient is (see, for example, Carpenter, Connolly, and Knowles Myers 2008). Transfers are higher if the recipient is a charity rather than another participant. The sociodemographic background characteristics matter as well. In particular, as Carpenter et al. (ibid.) show from data of representatively selected participants, older “dictators” transfer more money than younger ones. Another variable that matters is the degree of social distance to the experimenter (Hoffman, McCabe, and Smith 1996). The availability of various outside options (Dana,Weber, and Kuang 2007) also influences transfers substantially. From the evidence that many contextual variables matter one might conclude that the Dictator Game does not provide a coherent measure of altruism. Such a conclusion would be premature, however. Behavior might be very sensitive to many contextual details and, nevertheless, for a given contextual frame behavior follows rationality principles. The data of Andreoni and Miller (2002) are clearly consistent with altruism being a taste that obeys important rationality axioms. fair sharing the ultimatum game The Ultimatum Game was first studied by Güth, Schmittberger, and Schwarze (1982). It is a simple bargaining game between two players, a proposer and a responder. As in the Dictator Game, the proposer receives a sum of money, let’s say $10, to split with the responder. But unlike the Dictator Game, here the responder may reject or accept the offer. In case the responder accepts, the offered division is implemented; if the offer is rejected, both get nothing. In the usual implementation of this game, the proposer does not know who the responder is, and vice versa, so all decisions are anonymous, to control for social approval effects. A rational choice analysis under the assumptions of rationality and selfishness makes an unambiguous prediction: if the responder is motivated solely by monetary payoffs, he or she will accept every offer. Therefore, the proposer will offer only the smallest money unit. The results across a wide range of Western subject pools reject this prediction unequivocally (see Camerer 2003; Oosterbeek, Sloof, and van de Kuilen 2004; and Chaudhuri 2009 for overviews). On average, proposers offer 30 to 40 percent of the available amount. The median and the mode are at 40 and 50 percent, respectively. Very few offers are at 10 percent, or above 50 percent. Offers below 20 percent or less will be rejected with a high probability, while equal splits are almost always accepted. Contextual variables are much less relevant than in the Dictator Game. These results also hold if substantial amounts are at stake (Cameron 1995), or if the experiment is played with

Rationality, Social Preferences, and Strategic Decision-making

45

nonstudents across different age groups (Güth, Schmidt, and Sutter 2003). Güth et al. (ibid.) found these results in a particularly innovative study with more than a thousand readers of a German newspaper.They also elicited expectations about the acceptance probability of a given offer. It turned out that people have quite good estimates about the actual rejection rates. The results of the Ultimatum Game appear inconsistent with rational choice theory, as recipients reject offers and proposers offer nonminimal amounts. However, these results are only partially inconsistent with rational choice theory. This is easiest to see for the proposer. If the proposer expects that a low offer might be rejected, it is rational to make an offer that is more likely to be accepted. Since people have quite realistic expectations about rejection rates, low offers should be less frequent in the Ultimatum Game than in the Dictator Game.This is the case, as shown by Forsythe et al. (1994), who found that offers in the Ultimatum Game are substantially higher than in the Dictator Game.The fact that people reject offers poses a bigger challenge. One explanation is that people make errors that are, however, much less likely the more costly it would be. To see this, notice that rejecting a low offer is a “cheap mistake,” whereas offering little is a big one. Thus, recipients might learn to reject and proposers might learn to make sufficiently high offers. In a modified Ultimatum Game, Eckel and Grossman (1996) also showed that rejections are less likely the more costly they are. People might reject for emotional reasons, because they feel unfairly treated and therefore want to punish the greedy intention (after all, the role allocation was random). This hypothesis is hard to test from behavioral data alone (although attempts exist; see, for instance, Abbink et al. 2001). Data from emotion self-reports support the emotion hypothesis (Pillutla and Murnighan 1996). Sanfey et al. (2003) applied neuroscientific methods to understand rejections in the Ultimatum Game. They found that low offers activated areas of the brain associated with anger and disgust, but also areas involved in information processing and decision-making. The strength of the activation of these areas also predicts the probability of rejections quite well. This evidence suggests that the interpretation of rejections in terms of learning and errors is much less appropriate than explanations in terms of social preferences: People reject low offers because they consciously want to reject them. positive reciprocity the gift exchange game The Gift Exchange Game was developed by Fehr, Kirchsteiger, and Riedl (1993). In the Gift Exchange Game there are two roles, employers and employees. Each employer can hire only one employee, and there are more employees than employers. The sequence of events is as follows. Participants in the role of employer make wage offers in a competitive market institution. Wages are between 20 and 120. Employees see these wage offers and can accept any wage offer that is still available. Acceptance of a wage offer concludes an employment contract. Contracts are incomplete because employers can offer only a fixed wage but cannot specify a particular effort; effort is therefore not contractible. In the next step employees choose their effort and the game ends. There are ten different effort levels. “Effort” in this experiment means choosing a number with the consequence that the higher the chosen number, the higher

10

9

8

Effort level 5 6 7

4

3

2

20

30

40

50 60 Wage

Fitted values

70

80

90

1

2

3

4

5 6 Period

7

Gift Exchange Market Complete Contracts Market

8

9

10

figure 1.3. Left panel: Positive reciprocity: wage and effort are positively correlated. Right panel: Market wages when reciprocity is possible (Gift Exchange Market) and when it is not possible (Complete Contracts Market). The predicted (market clearing wage) is 20. Source: Fehr, Kirchler, Weichbold and Gächter (1998).

1

80 70 Average wage 50 60 40 30 20

Rationality, Social Preferences, and Strategic Decision-making

47

is the employer’s profit and the higher are the employee’s effort costs. The payoff of employers increases with the effort of the employee and decreases with the wage paid to the employee. The employees’ payoffs increase in wage and decrease in effort. Parameters are such that maximal effort would maximize the total surplus available. Thus the Gift Exchange Game can be seen as a generalized version of a sequential Prisoner’s Dilemma. The experiment is conducted under anonymity and the market interaction described above is repeated for ten periods. This setup ensures that there are no strategic reasons for gift exchange. If we assume that all players are rational and self-interested payoff maximizers, then employees will, irrespective of the wage, choose the minimum effort because effort is costly. Employers therefore have an incentive to pay the lowest wage, because higher than minimal wages cannot trigger gift exchange from selfish employees. Since there are fewer employers than employees, employers are in a stronger position and should be able to push wages down to the lowest possible level of 20. However, the norm of reciprocity (Gouldner 1960) predicts that effort and wages should be positively correlated. Numerous experiments have been conducted in this framework. Figure 1.3 shows the results of the gift exchange experiments of Fehr et al. (1998). The left panel shows a bubble plot of the wage effort combinations and a regression line. Overall, the self-interest prediction is clearly refuted, because there is a highly significantly positive relation between wage and effort. Despite the monetary incentive to choose minimal effort, experimental subjects in the role of the employee tend to reward generous wage offers by high efforts. This is unambiguous evidence for positive reciprocity, found in numerous experiments (for an overview, see Fehr and Gächter 2000b). However, the figure also shows that there is substantial heterogeneity. Irrespective of the wage paid by the firm there is always a fraction of workers who choose minimal effort. I chose to discuss Fehr et al. (1998) because this experiment allows me to connect to the market experiments described below. Recall that in these experiments contracts were always complete and prices converged to the predicted levels. The right-hand panel of Figure 1.3 shows that contractual incompleteness makes a decisive difference: wages are far above the predicted level of 20 and even increase over time. Further analyses showed that the average wage observed in these markets was indeed the profit-maximizing wage, given the employees’ average wage-effort relationship. But are gift exchange and contractual incompleteness the cause behind this finding? To investigate this crucial question, Fehr et al. (ibid.) included a further treatment in their design, called the “Complete Contracts Market.” In this treatment effort is fully contractible (set exogenously at the highest effort level). Thus, the contract is complete because there is no effort discretion any longer. The results, also depicted in the right-hand panel of Figure 1.3, show that wages are dramatically different when contracts are complete: wages are substantially lower from the start and, as in numerous comparable experiments, converge toward the predicted level of 20 by the end of the experiment.

48 Simon Gächter trust and trustworthiness the trust game In economics, the now classic experiment to measure trust and trustworthiness is the work of Berg, Dickhaut, and McCabe (1995). The Trust Game (or the “investment game,” as Berg et al. call it) is a two-player game in which subjects are randomly and anonymously allocated to their roles as trustors and trustees (often called investor and recipient). The investor has an endowment of, say, $10. The investor’s task is to decide how much of this endowment to transfer to the recipient. Any amount x the investor transfers gets tripled by the experimenter—that is, the recipient receives 3x. The recipient has then to decide on the amount y (between 0 and 3x) to back transfer to the investor, who receives y. Selfish recipients would not return anything irrespective of the amount received; rational and selfish investors would foresee this and invest nothing. Why does this game measure trust? Given the fact that any amount invested is tripled, transferring the whole endowment would maximize the joint income of both players. Yet transferring x pays off to the investor only if he or she receives at least x back. Since communication is not possible, and, even more important, any promised back transfer would not be enforceable, sending a positive amount signals trust. Whether trust pays off depends on the back transfer. Therefore, the back transfer is a measure of trustworthiness. The Trust Game by Berg et al. (ibid.) offers therefore a measure of both trust and trustworthiness. This argument is not without problems, and I briefly mention some of them below. Berg et al. (ibid.) ran this experiment with sixty-four participants. The participants in the role of a trustor sent on average $5.16, although the whole range of possible transfers (between $0 and $10) was observed. Only two trustors chose to transfer $0.The trustees returned on average $4.66. Among the thirty trustees who received a positive transfer, twenty-four returned a positive amount and fourteen returned an amount that exceeded the sent amount. Because the Trust Game gives an intuitive and simple measure of trust and trustworthiness, it has been replicated numerous times in various treatment variations.The main result described above was always confirmed (see Camerer 2003; Chaudhuri 2009; and Johnson and Mislin 2011 for overviews). I highlight two studies for their methodological inventiveness. Sutter and Kocher (2007) used the Trust Game to study whether trust and trustworthiness change with age.They conducted experiments with more than six hundred participants from six different age cohorts. It turned out that both trust and trustworthiness exist in all age cohorts, even the youngest one (six- to eight-year-olds). However, both trust and trustworthiness increase with age, although the increase beyond the twenty-five- to thirty-five-year-olds is small. Sutter’s and Kocher’s study is not drawn from a representative sample, however. It is therefore an interesting question how trust and trustworthiness are distributed in representative subject pools.The pioneering studies of Fehr, Fischbacher, Rosenbladt et al. (2002) and Bellemare and Kröger (2007) integrated trust experiments into representative surveys in Germany and the Netherlands, respectively. Ermisch et al. (2009) investigated people’s trust in the British population. In all studies age emerges as an important demographic variable, in particular with regard to trusting behavior: older people trust more.

Rationality, Social Preferences, and Strategic Decision-making 49

I conclude with three remarks. First, the observation that people across all age cohorts and many social groups (see, for example, Buchan, Croson, and Dawes 2002; Fehr and List 2004; Carpenter, Daniere, and Takahashi 2004) show trust and trustworthiness even in one-shot situations is a stark result.Yet in daily life the largest part of trust is probably created through repeated interactions or within networks—that is, trust occurs in embedded relationships. If trust is combined with reputational incentives, trust and trustworthiness should increase.The reason is that now also selfish individuals have an incentive to trust and to behave trustworthily. The presence of reputational incentives should not undermine trust and trustworthiness of nonselfish individuals. This is indeed the case across different versions of Trust Games (Cochard, Van-Nguyen, and Willinger 2004; Bohnet and Huck 2004) and the related gift exchange games (Gächter and Falk 2002). Buskens and Raub (this volume) provide a comprehensive discussion of this line of research. Second, consistent with rational choice prediction, trust and trustworthiness respond to the parameters of the Trust Game. If trustworthiness becomes more costly, it occurs less and people trust less (Snijders and Keren 2001; Buskens and Raub, this volume). Third, researchers have challenged the view that this version of the Trust Game really measures trust and trustworthiness. In economics, for example, on the basis of Dictator Game results Cox (2004) argues that the fact that people send money to a recipient or return money to a trustor could also be the result of altruism. Another argument has to do with the riskiness of trusting decisions. Does a trusting decision reflect the trustor’s risk attitude, or do trusting people require a “trust risk premium” because they are “betrayal averse”? In a series of experiments Iris Bohnet and her coworkers demonstrate that trusting decisions are closely related to betrayal aversion (see, for example, Bohnet et al. 2008). Thus, trust and trustworthiness are more than just calculative behavior and reciprocity. Further illuminating discussions from various angles about the nature of trust can be found in, for example, Bacharach and Gambetta (2001); Ostrom and Walker (2003); Fehr (2009a); and Vieth (2009). cooperation and free riding the prisoners’ dilemma and public goods game The most important vehicles for studying cooperation problems in controlled laboratory experiments are the “prisoner’s dilemma” and the “public goods experiment.” The Prisoner’s Dilemma Game is probably one of the most extensively investigated games and a comprehensive discussion of the experimental results exceeds the scope of this chapter. The interested reader should consult Colman (1999), who discusses the results at great length. I concentrate on two aspects of main interest in this chapter: the extent of cooperation in one-shot games and the importance of strategic incentives. Figure 1.4 illustrates the results of two studies that highlight these issues (Cooper et al. 1996; Andreoni and Miller 1993). In both studies the participants played the game ten times under two different conditions. In one condition, called the “Stranger” condition, each player was matched with a new player in each of the ten periods. In the second condition, called “Partner,” the opponent stayed the same throughout all repetitions of the

0%

10%

20%

30%

40%

50%

"Partner"

Opponent is a ...

"Stranger"

Andreoni & Miller (1993)

0%

10%

20%

30%

40%

50%

60%

"Partner" Opponent is a ...

"Stranger"

Cooper et al. (1996)

figure 1.4. Cooperation rates in the prisoner’s dilemma among “Strangers” and “Partners.” Source: Andreoni and Miller (1993) and Cooper, DeJong, Forsythe, and Ross (1996).

Percent cooperation rate

60%

Percent cooperation rate

Rationality, Social Preferences, and Strategic Decision-making

51

game. The subjects were informed about this. Thus, under the assumption of selfishness and rationality, all players in both conditions are predicted to defect. In the “Stranger” condition this prediction holds because each play of the game is against a new opponent and hence “one-shot.” In the “Partner” condition the prediction holds with backward induction: in the last period both players (who are assumed to be rational and selfish) will defect.Therefore, in the penultimate period, there is no incentive to cooperate, since players will surely defect in the last period. Hence, there is also no incentive to cooperate in the period prior to the penultimate one. Continuing this logic implies that rational and selfish players will defect throughout. By contrast, if people are not completely sure that everyone is selfish, then it might pay to build up a reputation by cooperating if others cooperate until the final rounds, where a selfish player should defect for sure (see Kreps et al. 1982 for a game-theoretical explanation, and Selten and Stoecker 1986 for a bounded rationality approach). In both studies the results in the “Stranger” condition are at odds with this prediction. People cooperate on average in slightly more than 20 percent of the cases. A common future, if only for ten rounds, increases cooperation substantially. In the “Partner” condition, the average cooperation rate is at least 50 percent. Thus, (i) people are prepared to cooperate even in one-shot games, and (ii) the possibility to behave strategically strongly increases cooperation. This evidence is consistent with the findings from Trust Games discussed above. Clark and Sefton (2001) studied an interesting variation of the game of Figure 1.4. Instead of playing the game simultaneously, their subjects played the game sequentially—that is, player 1 first made a choice, which was then observed by player 2 before deciding whether to cooperate or to defect. The subjects also played the game for ten rounds in the “Stranger” setup. Clark and Sefton (ibid.) found that between 37 and 42 percent of the subjects cooperate conditionally on the cooperation of others. Such conditional cooperation is also observed in two further treatments—“double temptation,” in which the defection payoff was doubled, and “double stakes,” in which all payoffs were doubled. A statistical analysis shows that “double stakes” did not significantly affect the extent of conditional cooperation. By contrast, under “double temptation” the fraction of conditional cooperation is reduced relative to the baseline. This latter finding is consistent with rational choice theory: as cooperating becomes more expensive, it will occur less. The results from Prisoner Dilemma Games are interesting, because the prisoner’s dilemma is such a simple cooperation game. The fact that people cooperate even in one-shot games casts doubt on the selfishness assumption. The observation that there are strong effects of repeated interaction suggests that straightforward strategic incentives are very helpful for successful cooperation. There can thus be no doubt that the strategic gains from cooperation that come from repeated interactions are a powerful force in explaining real-world cooperation in small and stable groups. However, the success of repeated game incentives in sustaining cooperation may be limited if groups become larger.The intuition is as follows. In the bilateral prisoner’s dilemma a player can punish a defector by defecting as well. In larger groups such targeted punishment is not possible: defection punishes not only defectors but also other cooperators, who,

52 Simon Gächter as a consequence, might then defect as well. For this reason it is worthwhile to move beyond dyadic relationships. The most commonly used game for studying n-person cooperation problems is the Public Goods Game. In contrast to a private good, a public good is a good that can be consumed even if one has not contributed to its provision. Common examples of public goods are clean air, environmental quality, and national security, but also collective reputations or team output. The Public Goods Game is an economic model of public good provision. This game underlies many experiments that study cooperation for the provision of public goods. In a typical public goods experiment, four people form a group. All group members are endowed with 20 tokens. Each member i has to decide independently how many tokens gi (between 0 and 20) to contribute to a common project (the public good).The contributions of the whole group are summed up. The experimenter then multiplies the sum of contributions by 1.6 and distributes the resulting amount equally among the four group members. Thus each group member’s payoff is πi = 20 − gi +

1.6 4 ∑g 4 j =1 j

The first term (20 – gi) indicates the payoff from the tokens not contributed to the public good (the “private payoff ”). The second term is the payoff from the public good. Each token contributed to the public good becomes worth 1.6 tokens. The resulting amount is distributed equally among the four group members—irrespective of how much an individual has contributed. Thus, individuals benefit from the contributions of other group members, even if they have contributed nothing to the public good. Therefore, a rational and selfish individual has an incentive to keep all tokens, since the “return” (that is, the personal benefit) per token from the public good for him- or herself is only 0.4 (1.6/4), whereas it is 1 from keeping the token. By contrast, the group as a whole is best off if everybody contributes all 20 tokens. Since the Public Goods Game is an n-person cooperation problem that is easy to implement, and since it reflects as well the tension between individual incentives and collective benefits, it has been frequently used in experimental studies (see Gächter and Herrmann 2009; and Chaudhuri 2011 for overviews). Again, my review of results is selective and illustrates the main variables of interest. Figure 1.5 depicts a typical finding from a public good experiment, where the same game is repeated ten times and subjects know this. In each period subjects receive 20 tokens and decide how many of them to keep or to contribute to the public good. After each round subjects are informed about what the other group members have contributed. Figure 1.5 shows the resulting cooperation patterns in a “Stranger” condition, where group members change randomly from round to round, and a “Partner” condition, in which groups remain the same for all rounds. Figure 1.5 illustrates three stylized facts from dozens of public goods experiments. First, as in the prisoner’s dilemma experiments reported above, people make positive contributions to the public good even in one-shot

Rationality, Social Preferences, and Strategic Decision-making

53

20 18

Mean contribution

16 14 12

Partner

10 8 6 4

Stranger

2 0 1

2

3

4

5 6 Period

7

8

9

10

figure 1.5. Contributions to a public good in stable (“Partner”) and randomly changing groups (“Strangers”) over ten repetitions. Source: Fehr and Gächter (2000).

games. Sociodemographic variables influence this baseline cooperativeness: older people are somewhat more cooperative than younger ones, and rural residents contribute more to the public good than urban residents (see, for example, Gächter and Herrmann 2011). Second, strategic incentives matter also in the Public Goods Game: “Partners” (in most experiments) contribute more than “Strangers” (see Andreoni and Croson 2008 for an overview). A third stylized fact is that contributions decline over time to very low levels.This is true in almost every subject pool studied (see, for example, the cross-cultural experiments of Herrmann, Thöni, and Gächter 2008). The question why this happens is not yet settled in the literature. I will address some explanations below. altruistic punishment One important reason why contributions in the Public Goods Game deteriorate is that the only way a duped cooperator can avoid being “suckered” is by reducing cooperation, thereby punishing everyone, even other cooperators. This raises the question of whether targeted punishment (whereby group members can identify a defector and punish him or her) actually can solve the free rider problem and prevent the breakdown of cooperation. Yamagishi (1986) and Ostrom, Walker, and Gardner (1992) were among the first to allow for punishment in interesting games. Yamagishi (1986) looked at people’s willingness to provide a sanctioning system that itself is a public good. Ostrom et al. (1992) studied punishment in a common pool extraction system. Both studies found a substantial willingness to punish defectors. Fehr and Gächter (2000a) developed an experimental design that allowed studying punishment in a one-shot and repeated Public Goods Game. After subjects had made their contributions to the public good, they entered a second stage, in

54 Simon Gächter 20

Partner

18 16 Stranger

Mean contribution

14 12 10 8 6 4 2 0 1

2

3

4

5

6

7

8

9

10

Period

figure 1.6. Mean contributions to the public good in the presence of a punishment opportunity. The figure shows that contributions are substantially higher among “Partners” than among “Strangers.” Source: Fehr and Gächter (2000).

which they were informed about each individual group member’s contribution. They could then assign up to ten punishment points to each individual group member. Punishment was costly for the punishing subject, and each received punishment point reduced the punished subject’s income from the first stage by 10 percent. Fehr and Gächter (ibid.) played this experiment under two treatment conditions—a “Partner” treatment, in which group members knew that they were playing the game with the same four group members for ten periods, and a “Stranger” treatment, where group composition was changed from period to period. Fehr and Gächter (ibid.) also ran control experiments in which punishment was not possible (see Figure 1.5). Figure 1.6 shows the results in the treatments with punishment. Compared with the data in Figure 1.5, contributions to the public good are strongly increased in the presence of a costly punishment opportunity. This is true for both the “Partner” and the “Stranger” treatment. In the case of the “Partner” treatment, contributions approach almost 100 percent of the endowment; in the “Stranger” treatment contributions amount to 60 percent of the endowment. Thus, again we see that “Partners” contribute more than “Strangers.” From the very first period onward contributions are significantly higher in the “Partner” treatment than in the “Stranger” treatment. Strategic incentives also matter in the presence of punishment. A theoretically important question concerns the relevance of future interactions. In the “Partner” treatment, the likelihood of future interaction is 1; in the “Stranger” treatment, where groups are randomly rematched, it is much smaller (depending on the size of the pool from which groups are rematched), but still positive. An interesting benchmark case is the situation

Rationality, Social Preferences, and Strategic Decision-making 20

1. punishment 2. no punishment

1. no punishment 2. punishment

18

55

Mean contribution

16 14 12 10 8 6 4 2 0 1

2

3

4

5

6

1

2

3

4

5

6

Period

figure 1.7. Mean contributions to the public good among “Perfect Strangers” in the absence and presence of a punishment option. N = 236 participants. Open squares refer to the condition where participants first played six periods of the punishment condition and then changed to the no-punishment condition. The black triangles refer to the reverse sequence. Source: Fehr and Gächter (2002).

in which the likelihood of future interaction is zero—that is, groups play a one-shot game. This situation is interesting because evolutionary theories of cooperation (see Nowak 2006 for a succinct summary) predict no cooperation in this case. Therefore, Fehr and Gächter (2002) set up a socalled Perfect Stranger design where in each of the six repetitions all groups were composed of completely new members, and participants knew this. Participants played six one-shot games with no punishment and six one-shot games with punishment. Half of the participants started without a punishment opportunity and then were introduced to the punishment condition. For the other half this order was reversed. Figure 1.7 contains the results on the cooperation rates achieved. When punishment is not available, cooperation collapses, as in all previous experiments.The picture changes dramatically when punishment is possible. For instance, in the experiments that started with the punishment option (symbolized by open squares), contributions in the first period were significantly higher than in the experiment that started with no punishment option (symbolized by black triangles). In the experiments in which punishment was introduced in the second sequence, cooperation jumped up immediately. This is remarkable because in this sequence subjects experienced a strong decline in the games with no punishment. Still, after punishment had been introduced cooperation jumped up to a level that even exceeded cooperation in the first period. In both

56 Simon Gächter 10 Mean punishment expenditures

9 8 Partner Stranger Perfect Stranger

7 6 5 4 3 2 1 0 [-20,-14) [-14,-8) [-8,-2)

[-2,2]

(2,8]

(8,14]

(14,20]

Mean deviation from other group member's contribution

figure 1.8. (Altruistic) punishment among “Partners,” “Strangers,” and “Perfect Strangers.” Mean expenditures on punishment as a function of the deviation of the punished group member’s cooperation from the average cooperation of the other members. Source: Fehr and Gächter (2000) and Fehr and Gächter (2002).

sequences, cooperation in the presence of a punishment opportunity strongly increased over time. Thus, contrary to theoretical predictions, cooperation can flourish in the presence of punishment even in one-shot interactions. Is this result of sustained cooperation inconsistent with rational choice predictions? The answer is no, if enough people are prepared to punish. Figure 1.8 shows the punishment expenditures for a given deviation from the other group members’ average contributions. I distinguish between the “Partner,” the “Stranger,” and “Perfect Stranger” experiments. A couple of observations can be made from Figure 1.8. First, the more a subject’s contribution falls short of the average contribution of the other group members, the stronger is punishment for the deviating group member. This holds true in all treatments. Second, with the exception of very strong negative deviations (which constitute only a few cases, however), punishment is very similar across treatments. This is remarkable because cooperation levels differ strongly between the “Partner,” “Stranger,” and “Perfect Stranger” treatments. Thus, punishment is to a large degree nonstrategic: people punish deviations, and punishment is largely independent of the absolute level of cooperation. This view is also corroborated by the fact that the punishment pattern of Figure 1.8 is temporally stable—that is, even in the final periods some people are prepared to harm a free rider. By now these results have been replicated by many researchers (see Gächter and Herrmann 2009; Chaudhuri 2011; and Balliet, Mulder, and van Lange 2011 for overviews). But why do people punish free riders? Several motives may underlie punishment. First, punishment reduces the inequality between the “sucker” and the free riders (Fehr and Schmidt 1999). Second, free riding indicates a greedy intention, which is punished for reasons of negative reciprocity (Falk, Fehr, and Fischbacher 2005).Third, being duped might trigger negative emotions (Fehr and Gächter 2002). Recent neuroscientific evidence

Rationality, Social Preferences, and Strategic Decision-making

57

suggests furthermore that taking “revenge” by punishing a free rider might be experienced as rewarding (de Quervain et al. 2004). Why is punishment so successful in increasing cooperation? The most important reason is probably that it gives the selfish people—who care most about their individual payoff—a material incentive to cooperate. Since altruistic punishment is frequent it apparently is a credible threat and induces selfish individuals to cooperate. This feature makes punishment altruistic—a punished free rider might in the next encounter abstain from defecting, which benefits future interaction partners. A rational choice analysis of the consequences of punishment predicts that the severity of punishment should matter for the cooperation level that can be achieved. The reason is that (rational) free riders will weigh costs and benefits of contributing to the public good. If punishment is low, free riding pays off; otherwise it is better to cooperate. Nikiforakis and Normann (2008) tested this argument by varying the “punishment effectiveness”—that is, the income reduction per punishment unit inflicted on a punished subject. Nikiforakis and Normann (ibid.) had four levels of punishment effectiveness—ranging from 1 to 4. The unit cost of punishment was always 1. The results are in line with the predictions: the higher the punishment effectiveness is, the higher are cooperation rates. Egas and Riedl (2008) have a related design (they also vary the cost of punishment). Their results are consistent with the findings of Nikiforakis and Normann (2008). The final issue I discuss here is whether punishment also follows some rationality principles or whether it is an impulse that is immune to cold calculations of costs and benefits. To appreciate this question, notice that one of the most fundamental concepts in economics is the “Law of Demand,” according to which people will demand less of a certain commodity or activity the higher its price. Does the Law of Demand also hold for punishment? Figure 1.8 and all papers that studied punishment in the context of a cooperation game confirm that many people actually do have a “demand for punishment,” in the sense that they are willing to pay a certain amount of money to inflict punishment on others (that is, they “buy” punishment). The more a subject free rides, the higher is the demand for punishment. However, studying the Law of Demand requires a systematic variation of the cost of punishment. This is what Anderson and Putterman (2006) and Carpenter (2007) did. Their subjects played the cooperation and punishment game in the “Stranger” setup to minimize strategic effects. In each of the games, subjects faced different costs for inflicting a punishment unit on the punished subject. The results confirm that for a given amount of free riding people demand less punishment the higher the costs of punishing are. Thus punishment, although most likely emotion-driven, follows the Law of Demand, a hallmark principle of rational choice economics. micromotives and macrobehavior The juxtaposition of results in the Public Goods Game without punishment (where cooperation invariably declines) and with punishment (where cooperation is stabilized or even increases over time) also allows us to shed some light on a long-standing question in the social sciences: what is

58 Simon Gächter the relationship between micromotives and macrobehavior (Schelling 1978; Coleman 1990)? The starting point for my illustrative discussion is the observation that cooperation in repeated public good experiments tends to collapse with repeated interactions (see Figure 1.5). Why is this so? One explanation is that people have to learn how to play this game. Since errors can go in only one direction, any erroneous decision looks like a contribution. Over time, people learn and commit fewer errors, which is why contributions decline. The problem with this explanation is that it is inconsistent with the fact that after a so-called restart (after the tenth-round participants are told that they will play another ten rounds) cooperation jumps up again and basically starts at the same level as in the first period (see Cookson 2000 for a particularly impressive illustration). If learning would explain the decay in cooperation, then after the restart, cooperation should have continued at the level at which cooperation was in the tenth round. Another explanation is that people are heterogeneous with respect to their cooperative inclinations. Some people are free riders who try to maximize their monetary income, irrespective of other group members’ contribution. Other people are “conditional cooperators,” who cooperate if others cooperate. These differences in microlevel motivations produce a macrolevel outcome in which everyone eventually free rides. Fischbacher and Gächter (2010) test this idea with the goal of tracking the influence of micromotivations on the macrolevel outcome. To measure individual cooperative motivations, Fischbacher and Gächter (ibid.) use a design developed by Fischbacher, Gächter, and Fehr (2001).Their design allows measuring the “type” of a player by observing each participant’s contribution to the public good as a function of other group members’ contribution. Specifically, subjects are asked to indicate for each possible average contribution of the other group members how much they would like to contribute to the public good. Thus, participants submit a whole contribution schedule, not just one contribution. The payoff function is the same as in the other public goods experiments—that is, incentives are such that, given others’ average contribution, the monetary income is always highest if one contributes nothing. Thus a free rider type will always contribute zero to the public good. A conditional cooperator type will increase the contribution according to the average contribution of others. Fischbacher et al. (2001) found that 30 percent of the subjects are “free rider types” who contribute nothing for all contributions of the other group members. Fifty percent show contributions that increase with others’ contributions—that is, they are “conditional cooperators.” The rest show other, more complicated patterns. Fischbacher and Gächter (2010) replicate these findings. How can the heterogeneity of individual motivations explain the fragility of cooperation (see Figure 1.5) that is so typical of repeatedly played cooperation experiments? The idea is simple. Conditional cooperators are prepared to cooperate if others cooperate. If they realize that others take a free ride, they reduce their contribution because they do not want to be “suckered.” Moreover, most people are also “imperfect conditional cooperators” who contribute more the more others contribute, but with a self-serving bias. Therefore, cooperation

Rationality, Social Preferences, and Strategic Decision-making

59

is bound to be fragile, even if most people are conditional cooperators. Free riders will speed up the decline of cooperation. Fischbacher and Gächter (ibid.) test this argument in two steps. They first show that players in a subsequent Public Goods Game make contributions that are largely consistent with their elicited contribution schedule. In the second step Fischbacher and Gächter (ibid.) use simulation methods. In the simulations, players are assumed to behave exactly according to their elicited contribution schedule. The interaction structure in the simulation is the same as in the experiment. The results show that the simulated contributions track the actual contributions surprisingly well. The relevance of this result is twofold. First, it shows that social preferences, rather than bounded rationality and learning, can explain the decay of cooperation. Second, the conceptual separation of motivations and outcomes, given the interaction structure in this public good experiment, has revealed that even though not everyone is motivated selfishly, the aggregate outcome is nevertheless one in which eventually everyone behaves selfishly. In the public goods experiment with punishment things are turned around. Here a few dedicated punishers suffice to give the free riders a selfish incentive to cooperate. Thus, in this case, it is the altruistic punishers who shape the aggregate outcome. As a last example we turn to the gift exchange experiments discussed above. Given the sequential structure of interactions in the Gift Exchange Game (contracts are concluded before effort decisions are made), profit-maximizing employers are well advised not to pay too low wages, for in that case effort levels will be low. Notice that employers do not have to be motivated by fairness, or any other-regarding motive to come to this conclusion; profit maximization suffices. As a consequence of these microlevel motivations the macrolevel outcome will be wage rigidity at nonmarket clearing levels (Figure 1.3). If contracts are complete, wages do converge toward market clearing levels. This observation, in combination with the findings under incomplete contracts, suggests two things. The microstructure of interaction (complete vs. incomplete contracts in this case) is decisive for the aggregate outcome. A second insight of these and numerous other experiments is that under complete contracts, social preferences do not matter for shaping the aggregate outcomes in competitive markets (recall Figures 1.2 and 1.3). Put differently, despite the fact that on average people are not selfish but motivated by gift exchange, markets with complete contracts are likely to converge to predictions derived under selfishness. See Gächter and Thöni (2011) for further examples and a more extensive discussion of these issues. discussion I have reviewed evidence from the most important games used to understand people’s social preferences and the interplay with material (selfish) incentives. The broad picture that emerges is as follows: 1) Many people behave in a nonselfish way even in one-shot games, although selfishness clearly exists. People transfer money in the Dictator Game, reject low offers in the Ultimatum Game, are trustworthy in the Trust Game, cooperate

60 Simon Gächter in the prisoner’s dilemma, contribute to public good, and punish free riders. These findings are evidence for the existence of “strong reciprocity”—that is, reciprocal behavior (rewarding nice acts and punishing mean ones) that is not motivated by strategic concerns (Gintis 2000b; Fehr, Gächter, and Fischbacher 2002). However, in all games, a non-negligible fraction of people also behave in a selfish way. Sociodemographic variables, in particular age, matter for the extent of prosocial behavior. Older people are on average more prosocial than younger ones. Put differently, since most experiments are done with undergraduates, results with students most likely measure a lower bound of the prevalence of prosocial motivations in the broader population (Falk, Meier, and Zehnder (forthcoming). 2) Strategic incentives matter. In sequential games, first movers have an incentive to take the motivations of their followers into account, irrespective of their own motivation. In games people play repeatedly with the same player or the same set of players, more prosocial behavior is typically observed, because selfish people also have an incentive to cooperate. Such effects are easily predicted within a rational choice framework. 3) If prosocial behavior becomes more expensive, it is less likely to occur. This observation is entirely consistent with a rational choice approach. The reason is that almost all people also value the payoff they can achieve in an experiment— that is, their utility is a function of their own payoff, as well as the payoffs of others (and possibly some other motivation, such as “doing the right thing,” being honest, and so forth). Thus there is often a tradeoff involved between one’s own welfare and the welfare of others; the more costly this tradeoff in favor of others becomes, the less likely it is going to be made. Taken together, observations 1 to 3 suggest first, substantial evidence for the relevance of social preferences and, second, that the existence of social preferences is not incompatible with a rational choice approach. Given the regularities, it is a fruitful avenue to develop theories of social preferences within a rational choice approach, because this approach allows investigating how social preferences and incentives to behave selfishly interact. In the next section, I will therefore discuss some influential attempts at modeling social preferences.

Modeling Social Preferences in a Rational Choice Framework Before embarking on specific theories of social preferences it is worth considering what these theories should be able to explain. The biggest challenge is to show why in some situations the outcomes of interactions appear consistent with predictions derived under the selfishness assumption, whereas in other situations behavior is clearly inconsistent with selfishness. For instance, in competitive market experiments, outcomes correspond largely to rational choice predictions under selfishness. Similarly, in the Public Goods Game without punishment, cooperation collapses and everyone appears as a free rider. By contrast, people are often trustworthy in the Trust Game, reject unfair offers, and punish free riders. These facts appear inconsistent, and any theory of social preferences should be able to address them in one framework. Theories of social preferences are typically rational choice theories that

Rationality, Social Preferences, and Strategic Decision-making

61

assume that people behave according to the rational choice framework. The only assumption that is changed is the selfishness assumption. The evidence reported above that behavior in simple experimental games is largely consistent with rational choice predictions, but inconsistent with selfishness in many games, justifies the methodological choice to retain the rationality framework and to amend the preference assumption. Changing preference assumptions is potentially problematic from a methodological point of view (Stigler and Becker 1977). However, unlike in the past, where no experimental data were available on which to base modeling assumptions, modelers can now use experimental data to justify their assumptions. Amending preferences requires not only basing them on psychologically plausible mechanisms but also on keeping the parameters fixed across games, at least for theoretical purposes (it is an empirical question whether this is a tenable assumption). In my discussion I will concentrate on models of inequity aversion, for two reasons: first, they are the most popular models to date, and second, they are easier to understand than other models developed later. An important element of scientific progress is testing theories and improving them in light of the empirical findings. After discussing inequity aversion I will therefore sketch some attempts to test elements of this theory and how these tests have influenced newer models. I refer the reader to Camerer (2003): ch. 2; Sobel (2005); and Fehr and Schmidt (2006) for comprehensive discussions. theories of inequity aversion The two major models of inequity aversion (which have been developed simultaneously and independently of each other) are due to Fehr and Schmidt (1999), and Bolton and Ockenfels (2000). I discuss the issues with the help of the widely used Fehr-Schmidt model. For simplicity, I confine my attention to two-player games, but note that the theory has been formulated for n-person games. The central assumption of models of inequity aversion is that players draw utility from their own payoff and disutility from unequal payoffs between comparison partners (“inequity aversion”). Inequity aversion exists in two forms, aversion to disadvantageous inequity and aversion to advantageous inequity. Aversion to disadvantageous inequality means that a player suffers from getting a lower payoff than the player’s opponent receives. Aversion to advantageous inequity means that a player also suffers from getting a higher payoff. Based on research by psychologists interested in social comparisons (Loewenstein, Thompson, and Bazerman 1989), the Fehr-Schmidt model assumes that the aversion to disadvantageous inequity is stronger than the aversion to advantageous inequity. Formally, the Fehr-Schmidt utility function for two players looks as follows: U i (πi, πj) = πi – αi max[πj – πi, 0] – βi max[πi – πj, 0], where πi (πj) denotes player i’s (player j’s) payoff, αi is the parameter that captures the strength of player i’s aversion to disadvantageous inequality, and βi measures the aversion to advantageous inequality. According to the results

62 Simon Gächter of Loewenstein et al. (ibid.) it is plausible to assume that αi ≥ βi. The model contains selfish preferences as a special case (αi = βi = 0). Notice also that the preference parameters can be different across individuals. The ambitious goal of this model is to explain the stylized facts outlined above. Going through all of them is beyond the scope of this chapter. I refer the reader to Fehr and Schmidt (1999) for further details. The idea of the model can be most easily explained with the help of the Ultimatum Game. Since the model retains the rationality assumption, the game is solved by backward induction. That is, we first analyze the responder’s behavior. Suppose the proposer (player j) offers a share s < 0.5 (of a pie of size 1). Will the responder i accept? According to the general Fehr-Schmidt utility function, player i’s utility is s – s] if i accepts. {s −α [1– 0 if i rejects.

Ui =

i

i

If player i rejects, that player’s utility will be zero. Accepting yields player i utility from the share s but disutility from receiving a lower payoff than the proposer j. The total disutility is the payoff difference weighted by the strength of aversion to disadvantageous inequity. The payoff difference between the responder and the proposer is πj – πi = 1 – s – s = 1 – 2s; the total disutility is therefore αi [1 – 2s]. Player i will accept if s – αi [1 – 2s] ≥ 0 ⇔ s / (1 – 2s) ≥ αi. If player i is not inequality averse (αi = 0), he or she will accept any offer s. If player i is inequality averse, αi will put a lower bound on which offer to accept. A rational proposer will therefore offer a share s that makes player i just indifferent between accepting and rejecting the offer. One can also look at the inverse problem of what the minimal acceptable offer s* is as a function of αi. It is easy to see that s* = αi / (1 + 2αi); s* = 0 if αi = 0 and s* approaches 0.5 if αi becomes very large. Thus a very strongly inequality averse player will accept nothing less than the equal split, whereas a selfish player will accept every offer. The Fehr-Schmidt model can explain not only the Ultimatum Game data; it can also explain why people punish free riders (to reduce their unfair payoff advantage) and why cooperation is increased (because free riders are better off cooperating to avoid punishment). But can it explain why competition in experimental markets seems to make social preferences irrelevant (Figure 1.2)? More generally, in games of competition very unequal payoffs are frequently accepted that are clearly unacceptable in two-player interactions. As an example, take the Ultimatum Game with responder competition. In this game there is one proposer and there are several responders. The rules are as follows. The proposer makes an offer and the responders decide simultaneously whether to accept or reject the offer. If more than one responder accepts the offer, it is randomly allocated to one of the accepting responders. The subgame perfect Nash equilibrium outcome of this game is the same as in the twoplayer case—the proposer reaps almost the whole pie and the responders all accept the lowest share possible. Fischbacher, Fong, and Fehr (2009) compared an Ultimatum Game with five responders to one with only one responder.

Rationality, Social Preferences, and Strategic Decision-making

63

The results are striking: in the treatment with five responders, responders accept very low offers (20 percent of the pie), which are clearly unacceptable in bilateral Ultimatum Games. The intuition for how the Fehr-Schmidt model explains these results is simple (the math is more involved). In the bilateral Ultimatum Games an inequity averse responder can punish a greedy proposer by rejecting the offer, which in turn gives the proposer an incentive to make a sufficiently high offer. The threat of punishment is much less severe if there are more responders. The reason is that people differ in their degree of inequity aversion—some are very inequality averse, whereas others are selfish and accept any positive amount. Since rejecting the offer will not change the fact that there will be inequity (given that some less inequity-averse responder will accept), even a very inequality averse player has an incentive to accept an unfair offer, to secure at least the utility from the material payoff. testing and theory development The impressive feature of the theories of inequity aversion by Fehr and Schmidt (1999) and Bolton and Ockenfels (2000) is the large range of experimental results they can explain. The parsimonious assumption of these models is that in addition to their own payoffs, decision-makers care only about inequity. How restrictive is this assumption? Charness and Rabin (2002) and Engelmann and Strobel (2004) designed simple allocation tasks in which participants have to choose between allocations with different payoff consequences.Take the following simple example: you have to choose between allocation A = (400, 400) and allocation B = (750, 400)—that is, in allocation A both get 400, and in B you get 400 but your coplayer gets 750. Inequity aversion suggests that you will choose A, because in B you get the same material payoff of 400 but you have to suffer from disadvantageous inequity. However, by choosing B you can help your coplayer at no cost to yourself, and this maximizes total payoff. Thus, concerns for efficiency might induce people to disregard inequity aversion. And indeed, Charness and Rabin (ibid.), who ran this experiment, found that 69 percent of their subjects chose B. The upshot of these experiments is that concerns for efficiency and “minimax preferences” (helping the least well off) are sometimes at least as important as inequity aversion. Another fundamental assumption in the theory explained above is that inequity aversion is solely “outcome-oriented,” by which I mean that players care only about what they get and not what they could have gotten (relative to others). Take the following example. Suppose that you are the recipient in a “mini-Ultimatum Game.” The proposer can choose between two allocations (8 for the proposer, 2 for you) or (5 for both). Consider a second game, in which the proposer can choose between (8,2) and (10,0). Now imagine you are confronted with the (8,2) offer in both cases. How do you decide in each of these cases? The theory of inequity aversion outlined above suggests that whether you accept or reject depends on your degree of disadvantageous inequity aversion (“your α” in the Fehr-Schmidt model).You will reject if your α > 0.33 and accept if α ≤ 0.33, and this is true in both games, irrespective of whether the alternative offer was (5,5) or (10,0); after all, the relevant outcome is (8,2) in both games. The data (by Falk, Fehr, and Fischbacher 2003) show

64 Simon Gächter that this prediction is clearly refuted: when the alternative is (5,5), 44.4 percent of people reject the (8,2) offer, whereas only 8.9 percent of people reject the exact same (8,2) offer if the alternative is (10,0). Thus, contrary to the FehrSchmidt model, alternatives matter, not just outcomes. How can we interpret this finding? When the alternative is (5,5) an (8,2) offer appears quite greedy, but the very same offer appears generous if the alternative is (10,0): choices from available alternatives reveal intentions and intentions matter. Theories of reciprocity build on the intuition that intentions and the perceived kindness of actions matter. Reciprocity means that a kind action is matched by a kind action and an unkind action with an unkind one. Rabin (1993) was the first to formalize this idea. For example, suppose I believe that you will cooperate in a prisoner’s dilemma I play with you. I might perceive your cooperation as a nice act and might reciprocate by cooperating as well. If our reciprocal motivations are strong enough, cooperation can be an equilibrium outcome. Defection remains an equilibrium too: if I believe you will defect, I think that is unkind and I reciprocate by defecting as well. Rabin’s theory is confined to static games such as the prisoner’s dilemma, and Dufwenberg and Kirchsteiger (2004) generalize it to dynamic games, such as the Ultimatum Game or the sequential prisoner’s dilemma.These approaches are motivated by experimental findings, but unlike the theories of inequity aversion by Fehr and Schmidt (1999) and Bolton and Ockenfels (2000), the theories of reciprocity are not meant to explain various games.The main goal of these theoretical approaches is to provide a coherent theoretical formalization of kindness and “reciprocity.” By contrast, Falk and Fischbacher (2006) developed a theory of reciprocity that is intended to predict better across various games than theories of inequity aversion by incorporating reciprocal motivations in addition to inequity aversion. For example, Falk and Fischbacher (ibid.) can explain the difference in rejections in the mini-Ultimatum Games reported above, whereas theories of inequity aversion cannot. However, the additional explanatory power comes at a cost of increased complexity of the theoretical framework. Probably the most important achievement of all these formal theories of social preferences is that they provide a precise language within a rational choice framework with which to talk about social preferences that did not exist prior to these theories. More generally, theories are useful not only for their predictive power but also for the insights they produce. These theories also demonstrate that one can do rational choice analyses of various social situations of interest and that such an analysis does not imply that rational agents are selfish. Rationality and selfishness are orthogonal to one another.

Conclusions The gold standard of science is empirical evidence. In this chapter I have discussed the usefulness of a rational choice approach for collecting empirical knowledge in various social situations of interest to behavioral and social scientists. I have concentrated on the empirical relevance of the selfishness assumption behind many applied models. My approach was mainly experimental.

Rationality, Social Preferences, and Strategic Decision-making 65

I hope that I have convinced the reader of two things: first, the rational choice approach is foremost a useful framework with which to analyze social situations of interest. It is useful, despite sometimes being descriptively inaccurate because this framework forces the researcher to ask the right questions. Without an exact theory about what selfishness entails and what would refute selfishness, this research program could not have succeeded. The debates would still be mainly philosophical rather than empirical. Second, the experiments have uncovered numerous regularities. As is common in the natural sciences, empirical regularities that are in contradiction to a particular theory can be used to improve theory. I have presented one approach and mentioned a few others, theories of nonselfish “social preferences” that have been developed on the basis of experimental regularities. These theories can be tested as well and improved on the basis of new evidence. The endeavor continues.

Note I am grateful for comments I have received from Siegwart Lindenberg, Manuela Vieth, and workshop participants at the Russell Sage Foundation in New York.

References Abbink, K., G. Bolton, A. Sadrieh, and F. F. Tang. 2001. “Adaptive Learning versus Punishment in Ultimatum Bargaining.” Games and Economic Behavior 37, no. 1: 1– 25. Anderson, C. M., and L. Putterman. 2006. “Do Non-strategic Sanctions Obey the Law of Demand? The Demand for Punishment in the Voluntary Contribution Mechanism.” Games and Economic Behavior 54, no. 1: 1–24. Andreoni, James, and Rachel Croson. 2008. “Partners versus Strangers: The Effect of Random Rematching in Public Goods Experiments.” In The Handbook of Experimental Economic Results, edited by Charles Plott and Vernon Smith, 776–83. Amsterdam: Elsevier. Andreoni, James, and John H. Miller. 1993. “Rational Cooperation in the Finitely Repeated Prisoners Dilemma—Experimental Evidence.” Economic Journal 103, no. 418: 570–85. ———. 2002. “Giving According to GARP: An Experimental Test of the Consistency of Preferences for Altruism.” Econometrica 70, no. 2: 737–53. Bacharach, Michael, and Diego Gambetta. 2001. “Trust in Signs.” In Trust in Society, edited by Karen Cook, 148–84. New York: Russell Sage Foundation. Balliet, D., L. Mulder, and P. A. M. van Lange. 2011. “Reward, Punishment, and Cooperation: A Meta-Analysis.” Psychological Bulletin 137, no. 4: 594–615. Bardsley, Nicholas, Robin Cubitt, Graham Loomes, Peter Moffatt, Chris Starmer, and Robert Sugden. 2009. Experimental Economics: Rethinking the Rules. Princeton: Princeton University Press. Becker, G. S. 1962. “Irrational Behavior and Economic-Theory.” Journal of Political Economy 70, no. 1: 1–13. ———. 1993. “Nobel Lecture—The Economic Way of Looking at Behavior.” Journal of Political Economy 101, no. 3: 385–409. Bellemare, C., and S. Kröger. 2007. “On Representative Social Capital.” European Economic Review 51, no. 1: 183–202. Benz, Matthias, and Stephan Meier. 2008. “Do People Behave in Experiments as in the Field?—Evidence from Donations.” Experimental Economics 11, no. 3: 268–81.

66 Simon Gächter Berg, Joyce, John Dickhaut, and Kevin McCabe. 1995. “Trust, Reciprocity, and Social History.” Games and Economic Behavior 10, no. 1: 122–42. Bewley, Truman. 1999. Why Wages Don’t Fall in a Recession. Cambridge: Harvard University Press. Bohnet, Iris, Fiona Greig, Benedikt Herrmann, and Richard Zeckhauser. 2008. “Betrayal Aversion: Evidence from Brazil, China, Oman, Switzerland, Turkey, and the United States.” American Economic Review 98, no. 1: 294–310. Bohnet, Iris, and Steffen Huck. 2004. “Repetition and Reputation: Implications for Trust and Trustworthiness When Institutions Change.” American Economic Review 94, no. 2: 362–66. Bolton, Gary E., and Axel Ockenfels. 2000. “ERC: A Theory of Equity, Reciprocity, and Competition.” American Economic Review 90, no. 1: 166–93. Bowles, Samuel. 2003. Microeconomics: Behavior, Institutions, and Evolution. Princeton: Princeton University Press. Buchan, Nancy R., Rachel T. A. Croson, and Robyn M. Dawes. 2002. “Swift Neighbors and Persistent Strangers: A Cross-cultural Investigation of Trust and Reciprocity in Social Exchange.” American Journal of Sociology 108, no. 1: 168–206. Buskens, Vincent, and Werner Raub. 2013. “Rational Choice Research on Social Dilemmas.” In The Handbook of Rational Choice Social Research, edited by Rafael Wittek, Tom Snijders, and Victor Nee. Stanford, CA: Stanford University Press. Camerer, Colin F. 2003. Behavioral Game Theory. Princeton: Princeton University Press. ———. 2009. “Behavioral Game Theory and the Neural Basis of Strategic Choice.” In Neuroeconomics: Decision Making and the Brain, edited by Paul W. Glimcher, Colin F. Camerer, Ernst Fehr, and Russell A. Poldrack, 193–206. Amsterdam: Elsevier. Camerer, Colin F., George Loewenstein, and Matthew Rabin, eds. 2004. Advances in Behavioral Economics. Princeton: Princeton University Press. Cameron, Lisa. 1995. “Raising the Stakes in the Ultimatum Game: Experimental Evidence from Indonesia.” Economic Inquiry 37, no. 1: 47–59. Carpenter, Jeffrey. 2007. “The Demand for Punishment.” Journal of Economic Behavior and Organization 62, no. 4: 522–42. Carpenter, J. E., C. Connolly, and C. Knowles Myers. 2008. “Altruistic Behavior in a Representative Dictator Experiment.” Experimental Economics 11, no. 3: 282–98. Carpenter, Jeffrey P., Amrita G. Daniere, and Lois M. Takahashi. 2004. “Cooperation, Trust, and Social Capital in Southeast Asian Urban Slums.” Journal of Economic Behavior and Organization 55, no. 4: 533–51. Carpenter, Jeffrey P., and Erika Seki. 2011. “Do Social Preferences Increase Productivity? Field Experimental Evidence from Fishermen in Toyama Bay.” Economic Inquiry 49, no. 2: 612–30. Charness, Gary, and Matthew Rabin. 2002. “Understanding Social Preferences with Simple Tests.” Quarterly Journal of Economics 117, no. 3: 817–69. Chaudhuri, Ananish. 2009. Experiments in Economics. Playing Fair with Money. Oxon: Routledge. ———. 2011. “Sustaining Cooperation in Laboratory Public Goods Experiments: A Selective Survey of the Literature.” Experimental Economics 14, no. 1: 47–83. Clark, K., and M. Sefton. 2001. “The Sequential Prisoner’s Dilemma: Evidence on Reciprocation.” Economic Journal 111, no. 468: 51–68. Cochard, F., P.Van-Nguyen, and M. Willinger. 2004. “Trusting Behavior in a Repeated Investment Game.” Journal of Economic Behavior and Organization 55, no. 1: 31–44. Coleman, James. 1990. Foundations of Social Theory. Cambridge, MA: Belknap Press of Harvard University Press. Colman, Andrew M. 1999. Game Theory and Its Applications in the Social and Biological Sciences. London and New York: Routledge.

Rationality, Social Preferences, and Strategic Decision-making

67

Cookson, R. 2000. “Framing Effects in Public Goods Experiments.” Experimental Economics 3, no. 1: 55–79. Cooper, R., D. V. DeJong, R. Forsythe, and T. W. Ross. 1996. “Cooperation without Reputation: Experimental Evidence from Prisoner’s Dilemma Games.” Games and Economic Behavior 12, no. 2: 187–318. Cox, James C. 2004. “How to Identify Trust and Reciprocity.” Games and Economic Behavior 46, no. 2: 260–81. Croson, R., and S. Gächter. 2010. “The Science of Experimental Economics.” Journal of Economic Behavior and Organization 73, no. 1: 122–31. Dana, J., R. A.Weber, and J. Kuang. 2007.“Exploiting Moral Wiggle Room: Experiments Demonstrating an Illusory Preference for Fairness.” Economic Theory 33, no. 1: 67– 80. Davis, Douglas D., Glenn W. Harrison, and Arlington W. Williams. 1993. “Convergence to Nonstationary Competitive Equilibria: An Experimental Analysis.” Journal of Economic Behavior and Organization 22, no. 3: 305–26. De Quervain, Dominique J. F., Urs Fischbacher,Valerie Treyer, Melanie Schellhammer, Ulrich Schnyder, Alfred Buck, and Ernst Fehr. 2004. “The Neural Basis of Altruistic Punishment.” Science 305, no. 5688: 1254–58. Diamond, Peter, and Hannu Vartiainen, eds. 2007. Behavioral Economics and Its Applications. Princeton: Princeton University Press. Dohmen, Thomas, Armin Falk, David Huffman, and Uwe Sunde. 2009. “Homo Reciprocans: Survey Evidence on Behavioral Outcomes.” Economic Journal 119, no. 536: 592–612. Dufwenberg, M., and G. Kirchsteiger. 2004. “A Theory of Sequential Reciprocity.” Games and Economic Behavior 47, no. 2: 268–98. Eckel, Catherine C., and Philip J. Grossman. 1996. “The Relative Price of Fairness: Gender Differences in a Punishment Game.” Journal of Economic Behavior and Organization 30, no. 2: 143–58. Egas, Martijn, and Arno Riedl. 2008. “The Economics of Altruistic Punishment and the Maintenance of Cooperation.” Proceedings of the Royal Society B—Biological Sciences 275, no. 1637: 871–78. Elster, Jon. 2007. Explaining Social Behavior: More Nuts and Bolts for the Social Sciences. Cambridge: Cambridge University Press. Engel, C. 2011. “Dictator Games: A Meta Study.” Experimental Economics 14, no. 4: 583– 610. Engelmann, Dirk, and Martin Strobel. 2004. “Inequality Aversion, Efficiency, and Maximin Preferences in Simple Distribution Experiments.” American Economic Review 94, no. 4: 857–69. Ermisch, John, Diego Gambetta, Heather Laurie, Thomas Siedler, and S. C. Noah Uhrig. 2009. “Measuring People’s Trust.” Journal of the Royal Statistical Society: Series A (Statistics in Society) 172, no. 4: 749–69. Falk, A. 2007. “Gift Exchange in the Field.” Econometrica 75, no. 5: 1501–11. Falk, Armin, Ernst Fehr, and Urs Fischbacher. 2003. “On the Nature of Fair Behavior.” Economic Inquiry 4, no. 1: 20–26. ———. 2005. “Driving Forces behind Informal Sanctions.” Econometrica 73, no. 6: 2017–30. Falk, Armin, and Urs Fischbacher. 2006. “A Theory of Reciprocity.” Games and Economic Behavior 54, no. 2: 293–315. Falk, Armin, and James J. Heckman. 2009. “Lab Experiments Are a Major Source of Knowledge in the Social Sciences.” Science 326, no. 5952: 535–38. Falk, Armin, Stephan Meier, and Christian Zehnder. 2012. “Did We Overestimate the Role of Social Preferences? The Case of Self-selected Student Samples.” Journal of the European Economic Association (forthcoming).

68 Simon Gächter Falkinger, Josef, Ernst Fehr, Simon Gächter, and Rudolf Winter-Ebmer. 2000. “A Simple Mechanism for the Efficient Provision of Public Goods: Experimental Evidence.” American Economic Review 90, no. 1: 247–64. Fehr, Ernst. 2009a. “The Economics and Biology of Trust.” Journal of the European Economic Association 7, nos. 2–3: 235–66. ———. 2009b. “Social Preferences and the Brain.” In Neuroeconomics: Decision Making in the Brain, edited by Paul W. Glimcher, Colin F. Camerer, Ernst Fehr, and Russell A. Poldrack, 215–32. Amsterdam: Elsevier. Fehr, Ernst, Urs Fischbacher, Bernhard von Rosenbladt, Jürgen Schupp, and Gert G. Wagner. 2002. “A Nationwide Laboratory. Examining Trust and Trustworthiness by Integrating Behavioral Experiments into Representative Surveys.” Schmoller’s Jahrbuch 122, no. 1: 519–42. Fehr, Ernst, and Simon Gächter. 2000a. “Cooperation and Punishment in Public Goods Experiments.” American Economic Review 90, no. 4: 980–94. ———. 2000b. “Fairness and Retaliation: The Economics of Reciprocity.” Journal of Economic Perspectives 14, no. 3: 159–81. ———. 2002. “Altruistic Punishment in Humans.” Nature 415, no. 6868: 137–40. Fehr, Ernst, Simon Gächter, and Urs Fischbacher. 2002. “Strong Reciprocity, Human Cooperation, and the Enforcement of Social Norms.”Human Nature 13, no. 1: 1– 25. Fehr, Ernst, Erich Kirchler, Andreas Weichbold, and Simon Gächter. 1998. “When Social Norms Overpower Competition: Gift Exchange in Experimental Labor Markets.” Journal of Labor Economics 16, no. 2: 324–51. Fehr, Ernst, Georg Kirchsteiger, and Arno Riedl. 1993. “Does Fairness Prevent Market Clearing? An Experimental Investigation.” Quarterly Journal of Economics 108, no. 2: 437–59. Fehr, Ernst, and John A. List. 2004. “The Hidden Costs of Incentives—Trust and Trustworthiness among CEOs.” Journal of the European Economic Association 2, no. 5: 743–71. Fehr, Ernst, and Klaus M. Schmidt. 1999. “A Theory of Fairness, Competition, and Cooperation.” Quarterly Journal of Economics 114, no. 3: 817–68. ———. 2006. “The Economics of Fairness, Reciprocity and Altruism—Experimental Evidence and New Theories.” In Handbook of the Economics of Giving, Altruism and Reciprocity, edited by Serge-Christophe Kolm and Jean Mercier Ythier, 615–91. Amsterdam: Elsevier. Fischbacher, Urs, Christina M. Fong, and Ernst Fehr. 2009. “Fairness, Errors and the Power of Competition.” Journal of Economic Behavior and Organization 72, no. 1: 527– 45. Fischbacher, Urs, and Simon Gächter. 2010. “Social Preferences, Beliefs, and the Dynamics of Free Riding in Public Good Experiments.” American Economic Review 100, no. 1: 541–56. Fischbacher, Urs, Simon Gächter, and Ernst Fehr. 2001. “Are People Conditionally Cooperative? Evidence from a Public Goods Experiment.” Economics Letters 71, no. 3: 397–404. Forsythe, Robert, Joel Horowitz, N. Eugene Savin, and Martin Sefton. 1994. “Fairness in Simple Bargaining Experiments.” Games and Economic Behavior 6, no. 3: 347–69. Frey, Bruno S., and Stephan Meier. 2004. “Social Comparisons and Pro-social Behavior: Testing ‘Conditional Cooperation’ in a Field Experiment.” American Economic Review 94, no. 5: 1717–22. Friedman, Daniel, and Shyam Sunder. 1994. Experimental Methods: A Primer for Economists. Cambridge: Cambridge University Press. Gächter, S., and A. Falk. 2002. “Reputation and Reciprocity: Consequences for the Labour Relation.” Scandinavian Journal of Economics 104, no. 1: 1–26.

Rationality, Social Preferences, and Strategic Decision-making

69

Gächter, Simon, and Benedikt Herrmann. 2009. “Reciprocity, Culture, and Human Cooperation: Previous Insights and a New Cross-cultural Experiment.” Philosophical Transactions of the Royal Society B—Biological Sciences 364, no. 1518: 791–806. ———. 2011. “The Limits of Self-governance when Cooperators Get Punished: Experimental Evidence from Urban and Rural Russia.” European Economic Review 55, no. 2: 193–210. Gächter, Simon, and Christian Thöni. 2011. “Micromotives, Microstructure and Macrobehavior:The Case of Voluntary Cooperation.” Journal of Mathematical Sociology 35, no. 1: 26–65. Gigerenzer, Gerd, and Reinhard Selten. 2001. Bounded Rationality: The Adaptive Toolbox. Cambridge: MIT Press. Gintis, Herbert. 2000a. Game Theory Evolving. Princeton: Princeton University Press. ———. 2000b. “Strong Reciprocity and Human Sociality.” Journal of Theoretical Biology 206, no. 2: 169–79. ———. 2009. The Bounds of Reason. Game Theory and the Unification of the Behavioral Sciences. Princeton: Princeton University Press. Gintis, Herbert, Samuel Bowles, Robert T. Boyd, and Ernst Fehr, eds. 2005. Moral Sentiments and Material Interests: The Foundations of Cooperation in Economic Life. Cambridge: MIT Press. Gouldner, Alvin W. 1960. “The Norm of Reciprocity: A Preliminary Statement.” American Sociological Review 25, no. 2: 161–78. Guala, Francesco. 2005. The Methodology of Experimental Economics. Cambridge: Cambridge University Press. Güth, Werner, Carsten Schmidt, and Matthias Sutter. 2003. “Fairness in the Mail and Opportunism in the Internet—A Newspaper Experiment on Ultimatum Bargaining.” German Economic Review 4, no. 2: 243–65. Güth, Werner, Rolf Schmittberger, and Bernd Schwarze. 1982. “An Experimental Analysis of Ultimatum Bargaining.” Journal of Economic Behavior and Organization 3, no. 4: 367–88. Hagen, Edward H., and Peter Hammerstein. 2006. “Game Theory and Human Evolution: A Critique of Some Recent Interpretations of Experimental Games.” Theoretical Population Biology 69, no. 3: 339–48. Hechter, Michael, and Satoshi Kanazawa. 1997. “Sociological Rational Choice Theory.” Annual Review of Sociology 23: 191–213. Hedström, Peter. 2005. Dissecting the Social: On the Principles of Analytical Sociology. Cambridge: Cambridge University Press. Henrich, Joseph, Robert Boyd, Samuel Bowles, Colin F. Camerer, Ernst Fehr, and Herbert Gintis. 2004. Foundations of Human Sociality: Economic Experiments and Ethnographic Evidence from Fifteen Small-Scale Societies. Oxford: Oxford University Press. Herrmann, Benedikt, Christian Thöni, and Simon Gächter. 2008. “Antisocial Punishment across Societies.” Science 319, no. 5868: 1362–67. Hertwig, Ralf, and Andreas Ortmann. 2001. “Experimental Practices in Economics: A Methodological Challenge for Psychologists?” Behavioral and Brain Sciences 24, no. 3: 383–403. Hoffman, Elizabeth, Kevin McCabe, and Vernon L. Smith. 1996. “Social Distance and Other-regarding Behavior in Dictator Games.” American Economic Review 86, no. 3: 653–60. Huck, Steffen, Hans-Theo Normann, and Jörg Oechssler. 2004. “Two Are Few and Four Are Many: Number Effects in Experimental Oligopoly.” Journal of Economic Behavior and Organization 53: 435–46. Jensen, Keith, Josep Call, and Michael Tomasello. 2007. “Chimpanzees Are Rational Maximizers in an Ultimatum Game.” Science 318, no. 5847: 107–9.

70 Simon Gächter Johnson, Noel D., and Alexandra A. Mislin. 2011. “Trust Games: A Meta-analysis.” Journal of Economic Psychology 32, no. 5: 865–89. Kagel, John, and Alvin E. Roth. 1995. The Handbook of Experimental Economics. Princeton: Princeton University Press. Kahneman, Daniel, Jack L. Knetsch, and Richard H. Thaler. 1986. “Fairness as a Constraint on Profit Seeking—Entitlements in the Market.” American Economic Review 76, no. 4: 728–41. Kahneman, Daniel, and Amos Tversky, eds. 2000. Choices,Values, and Frames. Cambridge: Cambridge University Press. Karlan, Dean. 2005. “Using Experimental Economics to Measure Social Capital and Predict Financial Decisions.” American Economic Review 95, no. 5: 1688–99. Kreps, David, Paul Milgrom, John Roberts, and Robert Wilson. 1982. “Rational Cooperation in the Finitely Repeated Prisoners’ Dilemma.” Journal of Economic Theory 27, no. 2: 245–52. Lindenberg, Siegwart. 2008. Social Rationality and Well-being. Manuscript, University of Groningen, Department of Sociology. Loewenstein, George F. 2007. Exotic Preferences: Behavioral Economics and Human Motivation. New York: Oxford University Press. Loewenstein, George F., Ted O’Donoghue, and Matthew Rabin. 2003. “Projection Bias in Predicting Future Utility.” Quarterly Journal of Economics 118, no. 4: 1209–48. Loewenstein, George F., Daniel Read, and Roy F. Baumeister, eds. 2003. Time and Decision: Economic and Psychological Perspectives on Intertemporal Choice. New York: Russell Sage Foundation. Loewenstein, George F., Leigh Thompson, and Max H. Bazerman. 1989. “Social Utility and Decision Making in Interpersonal Contexts.” Journal of Personality and Social Psychology 57, no. 3: 426–41. Mas-Colell, Andreu, Michael D. Whinston, and Jerry R. Green. 1995. Microeconomic Theory. Oxford: Oxford University Press. Morton, Rebecca B., and Kenneth C.Williams. 2010. From Nature to the Lab: Experimental Political Science and the Study of Causality. Cambridge: Cambridge University Press. Nikiforakis, Nikos, and Hans-Theo Normann. 2008. “A Comparative Statics Analysis of Punishment in Public Goods Experiments.” Experimental Economics 11, no. 4: 358–69. Nowak, Martin A. 2006. “Five Rules for the Evolution of Cooperation.” Science 314, no. 5805: 1560–63. Oosterbeek, H., R. Sloof, and G. van de Kuilen. 2004. “Cultural Differences in Ultimatum Game Experiments: Evidence from a Meta-analysis.” Experimental Economics 7: 171–88. Ostrom, Elinor, and James M. Walker, eds. 2003. Trust and Reciprocity: Interdisciplinary Lessons from Experimental Research. New York: Russell Sage Foundation. Ostrom, Elinor, James M.Walker, and Roy Gardner. 1992. “Covenants with and without a Sword—Self-governance Is Possible.” American Political Science Review 86, no. 2: 404–17. Pillutlaa, Madan M., and J. Keith Murnighan. 1996. “Unfairness, Anger, and Spite: Emotional Rejections of Ultimatum Offers.” Organizational Behavior and Human Decision Processes 68, no. 3: 208–24. Rabin, Matthew. 1993. “Incorporating Fairness into Game-theory and Economics.” American Economic Review 83, no. 5: 1281–1302. Rustagi, Devesh, Stefanie Engel, and Michael Kosfeld. 2010. “Conditional Cooperation and Costly Monitoring Explain Success in Forest Commons Management.” Science 330, no. 6006: 961–65. Sanfey, Alan G., James K. Rilling, Jessica A. Aronson, Leigh E. Nystrom, and Jonathan D.

Rationality, Social Preferences, and Strategic Decision-making

71

Cohen. 2003. “The Neural Basis of Economic Decision-making in the Ultimatum Game.” Science 300, no. 5626: 1755–58. Schelling, Thomas. 1978. Micromotives and Macrobehavior. New York: W. W. Norton. Selten, Reinhard, and Rolf Stoecker. 1986. “End Behavior in Sequences of Finite Prisoners-Dilemma Supergames—A Learning-Theory Approach.” Journal of Economic Behavior and Organization 7, no. 1: 47–70. Sen, Amartya. 1977. “Rational Fools: A Critique of the Behavioral Foundations of Economic Theory.” Philosophy and Public Affairs 6, no. 4: 317–44. Shang, Jen, and Rachel Croson. 2009. “A Field Experiment in Charitable Contribution: The Impact of Social Information on the Voluntary Provision of Public Goods.” Economic Journal 119, no. 540: 1422–39. Simon, Herbert. 1982. Models of Bounded Rationality. Cambridge: MIT Press. Smith, Vernon L. 1962. “An Experimental Study of Competitive Market Behavior.” Journal of Political Economy 70, no. 2: 111–37. ———. 1986. “Experimental Methods in the Political-Economy of Exchange.” Science 234, no. 4773: 167–73. Snijders, Chris, and Gideon Keren. 2001. “Do You Trust? Whom Do You Trust? When Do You Trust?” In Advances in Group Processes 18, edited by Shane R. Thye, Edward J. Lawler, Michael W. Macy, and Henry A. Walker, 129–60. Amsterdam: JAI, Elsevier Science. Sobel, Joel. 2005. “Interdependent Preferences and Reciprocity.” Journal of Economic Literature 43, no. 2: 392–436. Stigler, George J., and Gary S. Becker. 1977. “De Gustibus Non Est Disputandum.” American Economic Review 67, no. 2: 76–90. Sutter, Matthias, and Martin Kocher. 2007. “Trust and Trustworthiness across Different Age Groups.” Games and Economic Behavior 59, no. 2: 364–82. Vieth, Manuela. 2009. Commitments and Reciprocity: Experimental Studies on Obligation, Indignation, and Self-Consistency. Utrecht: University of Utrecht. Vohs, Kathleen D., Roy F. Baumeister, and George Loewenstein, eds. 2007. Do Emotions Help or Hurt Decision Making? A Hedgefoxian Perspective. New York: Russell Sage Foundation. Yamagishi, T. 1986. “The Provision of a Sanctioning System as a Public Good.” Journal of Personality and Social Psychology 51, no. 1: 110–16.

chapter

Social Rationality, Self-Regulation, and Well-Being: The Regulatory Significance of Needs, Goals, and the Self

2

siegwart lindenberg

Introduction Clearly, humans influence their own well-being, by and large in a way that is positive for their well-being, albeit not always for their well-being in the long run. Another way of saying this is that humans are bent upon improving their condition in an agentic way; they are thus self-regulators, and maybe this is the root meaning of “rationality.” In fact, this paper suggests that when we speak of rationality we should actually refer to self-regulatory processes. But then, the important question is of course how humans self-regulate. It seems to be a complex combination of processes, and this paper is devoted to presenting them in some detail. Any theory that would answer the question how humans self-regulate has to deal with what self-regulation may actually do. What are the recurrent problems that impact the well-being of humans and about which they can do something by self-regulation? For example, do humans need the subjective experience of needs? Do they have to be able to pursue goals? Do they have to deal with possibly conflicting goals? Do they have to deal with regulating their emotions? Do they have to deal with uncertainty? Do they have to have a sense of self? Do they have to understand other minds? The more we simplify the list of problems, the simpler the theory can be. However, if we make the list too simple, we will miss out on important aspects of how people influence their own well-being and how the environment helps or hinders them in this regard. For example, there are versions of rational choice theory that would assume that there is no other goal than maximizing one’s own utility (which is fully defined by stable preferences and given constraints) and that there are no preferences that could create problems of inner conflict. The theory also assumes that even though there are emotions, uncertainty, and problems with understanding other minds, one can safely abstract from these issues. This kind of rational choice theory then boils down to a theory of self-regulation for which almost all the problems about influencing one’s own well-being are shifted to the constraints—that is, to dealing with external resources, given the

Social Rationality, Self-Regulation, and Well-Being

73

(shadow) prices of relevant goods, linked to a uniform self-regulatory capacity. The implication is that one can safely disregard differences in self-regulatory capacity. However, given such differences, the perceived constraints also differ. For example, for somebody with low self-regulatory capacity, changing longterm opportunities are not likely to have much effect.Worse, all constraints that affect self-regulatory capacity itself are ignored. Thus, for example, if somebody is surrounded by others with low self-control, his own self-regulatory ability will suffer (see Christakis and Flower 2007; Evans and Kutcher 2011). There is overwhelming evidence that self-regulatory capacity differs and that it matters for income, status, health, crime, and many other important outcomes of behavior that would possibly be covered by rational choice theory (see, for example, Moffitt et al. 2011). It differs as a trait but also by circumstance and by development over the life course. For social policy this may be a very important issue. As Moffit et al. (ibid.: 2697) put it: “Understanding the key ingredients in self-control and how best to enhance them with a good costbenefit ratio is a research priority.” We need a more complex theory of selfregulation, and I suggest that it is one that fully acknowledges the social roots of human rationality.

Social Rationality and Self-Regulation Maybe the best starting point to think about self-regulation is human evolution: due to evolutionary pressures, self-regulation and social regulation are intimately intertwined for human beings. The basic idea of a sociologically informed evolutionary theory is twofold: (a) there have been selective pressures concerning the individual with regard to self-regulation, but (b) for primates and especially for humans, living in groups has individual adaptive advantages if the groups can deliver collective goods. For primates, and humans in particular, the processes of getting offspring to be reproductive is much too complex to allow simple solutions to the collective good problem (such as protective advantages of swarms or the workings of chemical signals as in ant colonies; see Cacioppo et al. 2006; Hrdy 2009). This makes it likely that for humans, there was also group selection with regard to groups being able to make individuals contribute to collective goods (Wilson 2006). Dunbar (2003) has provided evidence for the fact that the human neocortex (which contains the frontal lobes with the “command post” of self-regulation [see Goldberg 2009]) has evolved mainly to allow higher primates to function in groups. For this reason Dunbar calls the neocortex “social brain.” Human neural, hormonal, cognitive, and motivational structures thus coevolved with the relational and group contexts, and there is a functional relationship between these individual and social structures in the sense that the human capacities developed for the sake of adaptive advantages that can be derived from the social context. In other words, much of human self-regulatory capacity is dedicated to making humans able to take care of themselves, to elicit the cooperation of others, and to be able to adequately cooperate with others. The functionality of self-regulation for one’s own well-being (one could say “one’s rationality”) is thus thoroughly social in nature. As a result of these evolutionary selective processes, there are many evolved

74 Siegwart Lindenberg and more or less automatic self-regulatory processes that make people want to do what is socially expected or socially adaptive. For example, bonding with one’s infant and mating are adaptively and socially very important processes that are to a considerable degree influenced by physiologically triggered shifts in preferences. Plasma oxytocin stimulates bonding behavior by the mother with her infant (Feldman et al. 2007). Oxytocin also affects interpersonal trust (Kosfeld et al. 2005). Various hormones thoroughly affect preferences when people fall in love (Fisher, Aron, and Brown 2006). Also, the brain works very much with cost/benefit calculations. Many animals, including humans, deal with scarcity of energy in such a way that exertion of energy is related to hardwired processes of cost-benefit analysis (via dopaminergic processes; see Denk et al. 2005; Niv 2007). In turn, what an individual considers cost and what benefit and the subjective expectations concerning cost and benefits are wide open to social influence. For example, in the evolutionary environment it was often adaptive to follow those who are visibly more successful than oneself. This leads to social imitation from the top down and to a probably hard-wired tendency to have one’s likes and dislikes be influenced by those one admires (status effect; see, for example, Cohen and Prinstein 2006; Galliani and Vianello 2012), by the group one identifies with (for example, Cohen 2003), and by how useful or thwarting things are for one’s goal pursuit (Ferguson and Bargh 2004). Similarly, the expectations about costs and benefits are heavily influenced by the stereotypes about social categories such as status, race, gender, and age groups; these stereotypes steer expectations about the competence of self and others, likelihood of success, trustworthiness, effort level, helping behavior, and so forth (Shelly 2001; Tiedens, Ellsworth, and Mesquita 2000). The development of the social brain went hand in hand with the development of quite a variety of self-regulatory processes. Self-regulation is thus not a uniform capacity. Rather, for each individual, there are multiple self-regulatory systems, subject to different degrees of volitional control. They may or may not act in harmony and answer different problems. For example, at one end of the continuum of volitional control, there is the autonomic nervous system, which regulates bodily functions such as heart rate but also sexual desire. Then there are processes like motivated cognition that serve selfenhancement and that are partially open to volitional control. At the other end there are highly controlled actions, such as suppressing a negative utterance in the face of the boss. In the following, I will try to sketch the architecture of selfregulatory processes. To simplify things, I will dichotomize the continuum and distinguish only between lower- and higher-order self-regulation. The lowerorder self-regulatory processes, such as self-enhancing biases, belong mostly to the “old” brain (for example, basal ganglia, thalamus). The higher-order regulatory processes, such as emotion regulation, belong mostly to the “new” brain (the neocortex, especially the frontal lobes) and govern (a) the situational appropriateness of lower-order self-regulation, (b) the balance between various lower-order self-regulatory processes, (c) resource-oriented behavior, and (d) norm-oriented behavior. The higher-order regulation steers processes in just about every region of the brain, while the lower-order regulation operates in a more localized manner. Even though they both steer self-regulation, there is an important difference: the degree to which context is taken into account

Social Rationality, Self-Regulation, and Well-Being Goal-related self-regulation

Need-related self-regulation

75

Self-related self-regulation

figure 2.1. Three inter-related sets of self-regulatory processes

(Goldberg 2009). For example, the more hungry people are, the more they will focus just on eating and the less they will consider matters of taste, ownership, etiquette, longer-term health effects, and the like. By contrast, higher-order processes push back this immediacy and singular focus in favor of attention to ownership, etiquette, health, and so on. The social brain is involved in a number of different sets of self-regulatory processes that are interrelated but should also be studied separately. There are features of human functioning that are not traditionally looked at as belonging to self-regulation but that in fact can fruitfully be treated as such. For example, needs themselves, not just need satisfaction, can be interpreted as part of selfregulation. I will discuss three sets (see Figure 2.1): need-related self-regulation, goal-related self-regulation, and finally self-regulation dedicated to stabilizing the self. In each set, we find lower- and higher-order self-regulatory processes that are of great importance for the work of sociologists because they strongly affect social and institutional problems and problem solutions.

Fundamental Needs as Self-Regulatory Mechanisms The concept of “need” (as different from “want”) is not very sharply defined, but it basically refers to the combination of finding something rewarding, being aroused to seek satisfaction, and experiencing pathological effects from deficits in satisfaction (with the last characteristic missing in wants; see Deci and Ryan 2000; Baumeister and Leary 1995). A person whose basic needs are unfulfilled will have a low score on well-being. This view is by now widely shared (Deci and Ryan 2000; Lyubomirsky, King, and Diener 2005; Maslow 1971; Baumeister and Leary 1995; Steverink, Lindenberg, and Slaets 2005). The self-regulatory aspect of needs is that they are cognitively represented and set in motion action toward their satisfaction. Thus having a certain need also means having a self-regulatory mechanism to do something about its satisfaction (not necessarily consciously; see Tiffany and Conklin 2000). Adaptation to a changing environment under selective pressures can thus lead to the development of new needs. These new needs make the organism find things rewarding that are adaptive in the new environment and to go after the new rewards. From the point of view of the social brain, the main innovation in needs for primates and especially for human beings is social needs, and needs concerning resources that are, to a large extent, tied to the social context. Thus we have to identify what these needs are. For this purpose, we also have to identify physical needs, because their satisfaction is likely to interact with the satisfaction of social and resource-related needs.

76 Siegwart Lindenberg The identification of needs from this evolutionary view has been worked out over the years under the name of social production function theory (SPF theory).1 In order to stress the important role of resources and of self-regulation concerning the acquisition and use of resources, the approach taken by SPF theory is that need satisfaction can best be viewed in terms of production functions (see also Stigler and Becker 1977). Human beings are producers of their own well-being in terms of need satisfaction. A particular level of need satisfaction (the output) is “produced” by a particular input. For this reason, we will first deal with production needs, before we come to discuss substantive needs. production needs: autonomy, competence, safety⁄security, and structure Self-regulation by needs first of all focuses on needs that are related to the quality of the production function (improvement of the input/output ratio in production functions). Because deficits in need satisfaction are damaging, effectiveness of given resources is of great importance and has found its way into self-regulation by needs. First of all, autonomy is essential for choosing inputs in production functions (according to their effectiveness), and Ryan and Deci claim that autonomy is a basic need even in collectivist cultures (see Chirkov et al. 2003; see also Leotti, Iyengar, and Ochsner 2010). Secondly, competence is important for resource use (see also Deci and Ryan 2000). To improve one’s competence is to improve the amount of output (need satisfaction) for a given amount of input. For example, improving one’s social skills allows one to gain social approval with less effort or fewer resources. This need to improve one’s competence may be strongest in childhood. White’s view of “mastery” (1959) and Bandura’s concept of “self-efficacy” (1997) refer to the result of the satisfaction of this need. In certain cultures, this need for improving one’s competence can become strongly emphasized and linked to identity formation, in which case it would cover what Maslow (1971) called “self-actualization” (see Lindenberg 1996). The quality of a production function also depends on stability conditions. For example, when the effectiveness of a given means changes in unpredictable ways, self-regulation becomes difficult or impossible. Highly insecure property rights or random violence paralyze self-regulation and even reduce cognitive abilities. For being able to invest in the future, predictability is especially important. Human beings need a certain amount of safety/security (including stability/predictability) with regard to their production functions and, being bent on improvement, they will do something about it (Heiner 1983; Maslow 1971; Mendes et al. 2007). This need can also be interpreted as a desire for the world to conform to expectations about possibilities for goal achievement. Related to safety/security is the need for structure (or order), which has been given a prominent place in the literature. It is also a production need in the sense that agentic need satisfaction depends on having or creating “meaningful” situations—that is, situations with discernible structure (Neuberg and Newsom 1993; Proulx, Heine, and Vohs 2010). This need may actually be in the service of aiding predictability. In short, striving for autonomy, competence, safety/security (with stability, predictability), and structure can be taken to be basic human “production

Social Rationality, Self-Regulation, and Well-Being

77

needs” that evolved through the adaptive advantage of individuals who are bent on improving their condition in an agentic way. substantive needs: physical and social well-being Already in Durkheim we find a strong emphasis on what he calls “the dual character” of humans: “man is double, that is, social man superimposes himself on physical man” (Durkheim 1951 [1897]: 213). Both sides have their own needs. This is also the view derived from the social brain hypothesis. Small wonder that in one way or another, this dichotomy keeps turning up.We find it in Maslow’s need hierarchy (1971) (physiological needs and love/belonging and esteem needs). We also find it in Ford’s taxonomy (Ford and Nichols 1987; and Ford 1992), which is based on a dichotomy of “within person” (for instance, arousal, physical well-being), and “person-environment” needs (for instance, superiority, belongingness). Physical needs What are the most plausible candidates for universal physical needs? There is, of course, a question about the level of abstraction. A great number of concrete needs can be taken together under the concept of comfort. “Comfort” refers to the degree to which a person is free of noxious stimuli (such as hunger pangs, thirst, pain, and so forth).There is a general need to increase one’s own comfort. This is not exclusively a human need, but some stimuli that create discomfort may be specific to humans (such as sympathetic reactions vis-à-vis suffering others), or even specific to cultures (such as reactions to being exposed to certain sounds). A second plausible candidate for a universal physical need is the opposite of tension reduction: seeking excitement, arousal, satisfying curiosity. Ford and Nichols (1987) list both arousal and exploration as basic needs. There is overwhelming evidence that individuals seek arousal, which led Hebb (1958), Berlyne (1960), and others to abandon the classical drive theories in favor of a theory of arousal, or electrical activity of the brain. Far from seeking either comfort or arousal, individuals often seek both at the same time. A need for arousal and exploration can be covered by the concept of the need for stimulation (see also Scitovsky 1976;Wippler 1990).These needs can be reliably measured (see Nieboer et al. 2005). Social needs The needs on which social well-being depends have been variously identified by philosophers, anthropologists, sociologists, and psychologists.There is a great deal of convergence among them, in the sense that everybody is agreed on the fact that human beings crave a positive opinion from other human beings. The universality of a need is best examined from the point of view of evolution. From this and, as we will see, also from a sociological point of view, it makes sense not to lump all forms of social approval into one. There are first the two quite distinct approval needs, variously identified in the literature as status (domination, prestige) and affection (love, attachment, intimacy). This distinction also links to Bakan’s concepts (1966) of agency and communion needs. However, for Bakan (ibid.) and his followers (such as McAdams et al. 1996), agentic needs covered both production needs (for example, autonomy) and substantive needs (such as

78 Siegwart Lindenberg status) under one concept. As already mentioned, from a self-regulatory point of view, it is useful to separate these needs (see Steverink and Lindenberg 2006). Let me begin with status. Status. The evolutionary view that status is a universal human need may have become established with Barkow (1989), who pointed to the importance of relative standing for preferential access to resources (for example, mating opportunities, food, allies) and the likelihood that primates have been selected to seek higher relative standing. It is important to observe that for humans, the basis for status differences has shifted in the course of evolution from domination toward prestige, with a related shift in strategies to achieve and maintain status (Gilbert 2003). In dominance hierarchies, status is related to coercion, threats, and inspiring fear, whereas in prestige hierarchies, status is related to the display of competence and talent and to eliciting positive affect (such as admiration; see also Gilbert and McGuire 1998, Galliani and Vianello 2012; van Vugt 2006). However, even among humans, domination hierarchies have not vanished and may even merge with prestige hierarchies (see Fessler 2004; Halevy et al. 2012). Besides the evolutionary research, empirical evidence for the ubiquity of status striving comes from various contemporaneous sources. For one, there is historical evidence, such as the research by Max Weber concerning the importance of “honor” in virtually every society. Second, Steverink and Lindenberg (2006) found that older people who don’t seem to care about status anymore, do care and profit for their own well-being from status pursuit, if the opportunity for such pursuit arises. Third, Frank (1985) provides much evidence for the ubiquity of status striving in contemporary society. Fourth, the claim that status as a goal in itself is universal has recently received experimental support by a study of Huberman, Loch, and Önçüler (2004). All in all, the evidence strongly speaks for the assumption that status striving is a universal need. However, there is also the other side. In order to get status that is not based on threat or fear, there must be people who grant status. Social needs must have evolved with their counterpart, so the need for acquiring status must have evolved with the need to evaluate others in terms of their valued skills and possible or actual contribution to the collective good, flanked by emotions (such as deference and admiration, superiority) (Ridgeway et al. 1998; Galliani and Vianello 2012; Halevy, Chou, and Galinksy 2011). Affection. The insight that affection is a need may seem obvious today (see review by Pendell 2002), but there were times when influential people were convinced of the opposite. Orphanages and charity for adults in the nineteenth and first half of the twentieth century were mainly focused on physical aid, and expert opinion even found affection a potential danger (see Blum 2002). In the early 1940s, Maslow (1943) presented a very different picture. He described love and affection as basic needs and maintained that “practically all theorists of psychopathology have stressed love needs as basic in the picture of maladjustment. Many clinical studies have therefore been made of this need and we know more about it perhaps than any of the other needs except physiological ones.” Despite this pronouncement, the need for affection became widely accepted only after the famous study of rhesus monkey infants by Harlow (1958), Maslow’s teacher. Roughly at the same time, Bowlby and Fry (1953) researched the importance

Social Rationality, Self-Regulation, and Well-Being

79

of secure attachment of children to certain adults. Even though attachment (as protection against threat) is not exactly the same as affection (see MacDonald 1992), the two concepts are often confounded, and Bowlby’s research helped to draw professional and public attention to the importance of affection. Many have since proclaimed a need for affection (Schultz 1958). Seen from an evolutionary point of view, the important role of affection (warmth) is connected to the social brain (Dunbar 2003). The relatively larger neocortex of human beings allows living in larger groups but also necessitates what MacDonald (1992) calls “high-investment parenting.” The long period of dependency of the infant requires prolonged parental care (especially motherchild) and support from the male partner, secured through pair bonding (see also Buss 1994), and from others (Hrdy 2009). Affection facilitates both pair bonding (Weisfeld 1999) and positive parenting (Russell 1997), including nurturance, empathy, and transfer of scarce resources, and it increases a child’s willingness to be influenced by the adult (Eberly and Montemayor 1999). Cultural transmission, so important for culling adaptive advantages from living in groups, is greatly aided by this willingness to be influenced (Euler, Hoier, and Rohde 2001). There is also considerable evidence that deprivation of affection leads to psychopathology and ill health. For example, Uchino, Cacioppo, and KiecoltGlaser (1996) and Cacioppo et al. (2000) review a great number of studies which show that affection (emotional support) has reliable positive effects on physiological functioning, and lack of affection has negative effects (impaired cardiovascular functioning, hormonal functioning, immune functioning). Behavioral confirmation. In evolutionary terms, the great importance of group membership for one’s own adaptive advantages makes it likely that there has been a selective pressure to be sensitive to signals about one’s standing in the group and that a need has developed to feel accepted by the group. Baumeister and Leary (1995: 497) had called this “need to belong” a “powerful, fundamental, and extremely pervasive motivation.”They too link this need to its evolutionary roots about the importance of belonging to a group. Because the “need to belong” is actually a mixture of the group-related need and affection, and because the two should be treated separately, I use a different term, “behavioral confirmation” (see Lindenberg 1996), even though that term is also used in the literature to indicate a specific aspect of the sensitivity to the opinion of others—namely, that people become what they believe others think of them (see Snyder and Klein 2005). The term “behavioral confirmation” as used here refers to the fact that in virtually everything they do (including uttering opinions and expressing feelings), people seek the confirmation of others. Status and affection are something else. Status is a relative standing within the group, and affection is social approval central to close relationships and generally more unconditional than behavioral confirmation (see Baldwin and Sinclair 1996). Feeling accepted and confirmed by the group (irrespective of one’s status position and irrespective of affection in close relationships within the group)2 is a separate social need. Failure to conform to group standards is often accompanied by negative emotions, prominent among them shame (Fessler 2004; de Hooge, Breugelmans, and Zeelenberg 2008). All three social needs can be reliably measured (see Nieboer et al. 2005).

80 Siegwart Lindenberg Lower-order and higher-order need-related self-regulation Lower-order self-regulation via needs consists most of all of the ability of need states to affect cognitive and motivational processes in the service of adaptive behavior.This is not a conscious process, and it can go wrong (Tiffany and Conklin 2000). For example, there is much research on people not knowing what they want (Ariely, Loewenstein, and Prelec 2006; Hsee and Hastie 2006), or people being subjectively wrong about thinking they want something (Gilbert and Wilson 2000). A related lower-order self-regulatory process has to do with a particular sensitivity to situational opportunities. It works via cues in the environment that trigger urges. The limbic system can amplify the incentive salience of reward cues in the environment, which makes a person temporarily change preferences as a result of the opportunity of getting a particular reward at that moment. Since lower-order self-regulatory processes disregard the wider context (such as longer term behavioral consequences), this cue sensitivity can also be the source of adaptive problems that have to be dealt with by higherorder self-regulation (see, for example, Bernheim and Rangel 2004). Social emotions are self-regulatory devices to help satisfy social needs and to indicate that behavior might have to be changed for better need satisfaction. For example, the negative affect produced by social disapproval helps alert people to the danger of being rejected by the group and to adapt their behavior in such a way that they are accepted. Leary (see Leary et al. 1995; Leary and Baumeister 2000) has called this mechanism “sociometer.” The sociometer effect, however, is limited. As mentioned above, it is one of the defining characteristics of a need that a deficit in need fulfillment will lead to pathological consequences. Thus, if for example behavioral confirmation is a need, then deprivation in this area can also lead to pathological consequences rather than to functional repair work. Once the deficit in behavioral confirmation is large (for example, when one is fully rejected by one’s ingroup), there is little impetus left to make amends. At best, if one is optimistic about success, one seeks new circles (Maner et al. 2007). A large deficit in behavioral confirmation may altogether reduce goal-directed behavior at restoring it.Thus one may lose agentic pursuit of need satisfaction (say, through becoming lethargic; see Twenge, Catanese, and Baumeister 2003), become hostile toward one’s own (former) ingroup (Twenge et al. 2001; Maner et al. 2007), and become less likely to act in a prosocial way, even to others outside one’s former group (Twenge et al. 2007). Higher-order need-related self-regulation An important higher-order form of self-regulation is the ability of the brain to deal with scarcity of resources, especially energy (see Denk et al. 2005; Niv 2007). Humans (and also other animals) deal with scarcity of energy in such a way that exertion of energy is related to hard-wired processes of cost-benefit analysis. However, these cost-benefit calculations are often not conscious. For humans, in any case, the inputs into these calculations, such as expectations and evaluations, are subject to influence from the environment. Other important higher-order need-relational forms of self-regulation concern the ability to be agentic and stay agentic in the face of failure and the ability to balance the satisfaction of needs (Steverink, Lindenberg, and Slaets 2005).

Social Rationality, Self-Regulation, and Well-Being

81

The ability to be and stay agentic refers to the capacity to gain and maintain a belief in personal competence, control, or self-efficacy (Bandura 1997). It also refers to the ability to keep a positive frame of mind, even in adversity. People with a positive frame of mind are less likely to be discouraged and to become paralyzed in the face of rejection (Pass, Lindenberg, and Park 2010) or failure (Taylor et al. 2003), whereas negativity can create emotional and behavioral disorders (Lengua and Long 2002). The ability to balance need satisfaction concerns the capacity to seek out synergetic satisfaction of various fundamental needs and the capacity to balance needs satisfaction now and in the future. A synergetic kind of need satisfaction (also called “multifunctionality”; see Steverink, Lindenberg, and Slaets 2005) leads to higher levels of well-being (Sheldon and Niemiec 2006) and makes resources that allow multifunctional satisfaction of physical and social needs (such as intimate partners) particularly important (Nieboer and Lindenberg 2002). The definition of a good intimate relationship is virtually identical with multifunctionality: the relationship is stimulating, physically comforting, socially affectionate, and it increases both a feeling of self-worth (status) and a sense of belonging (behavioral confirmation). The satisfaction of each of these needs can aid the satisfaction of the other needs (synergy). This implies that multifunctionality is particularly productive of subjective well-being and that loss of multifunctional relationships belongs to the most dramatic reductions in subjective well-being (see Nieboer, Lindenberg, and Ormel 1998–99; Lane 2000). Balancing short-term and longer-term need satisfaction has important positive effects on well-being (Prenda and Lachman 2001).When impaired, this kind of self-regulation may be the most damaging for well-being (see Moffitt et al. 2011), and it does not concern only the satisfaction of needs but also the balance of overarching goals (see below). Disturbances in need-related self-regulation (and thus deficits) can derive from highly asymmetric salience of one of the needs. For example, substance addiction will increase the need for comfort to a degree that it interferes with multifunctionality (Hirschman 1992). Also, a particularly high need for stimulation creates high risk taking and jeopardizes the satisfaction of the other needs (see Sijtsema et al. 2010). People with a particularly high need for status tend to dominate and thereby reduce the ability to realize both affection and behavioral confirmation (see Sijtsema, Veenstra, Lindenberg, and Salmivalli 2009).

Self-Regulation via Goals Self-regulation concerning needs overlaps with another set of self-regulatory processes that are distinctive enough to be treated separately: goals. Goals are mental representations of desired states, but they are at the same time elaborate processes of self-regulation. Lower-order goal-related self-regulatory processes contain (a) the ability to monitor the degree to which a goal that is presently focal has been achieved, to detect errors, and to react to this information in such a way that, when the goal is realized, one turns to another goal, or, when progress is not satisfying, to take action for improvement (see Carver

82 Siegwart Lindenberg and Scheier 1998). They also contain (b) emotional responses to success and failure in goal-pursuit that aid in a quick determination of the direction of action (approach or avoidance, see ibid.). Finally (c), they contain the ability to deal with goal conflicts by inhibiting incompatible goals. Goal pursuit is not necessarily conscious (see Bargh et al. 2001). overarching goals and goal-framing theory Other self-regulatory functions of goal-processes (including higher-order self-regulation) are particularly present in overarching goals. Overarching goals are all by themselves a form of self-regulatory devices because, when they are activated (“focal”), they coordinate a great number of cognitive and motivational processes. This allows the individual to be focused and prepared for action (all the way to the motoric level) at the same time. When such a goal is focal, it organizes cognitions and evaluations in a semimodular way and it selectively activates hardwired and learned modules. A focal high-level goal can thus be seen as a composite module, comprising a particular selection of semimodules and hardwired and learned submodules. In that sense, goals create domain specificity and selective sensitivity to specific inputs. For example, the highlevel goal “to act appropriately” is likely to make situationally relevant norms more cognitively accessible; make people particularly sensitive to information about what is expected; activate the modules to process information on gaze and on certain facial expressions of approval and disapproval; and activate response tendencies and habitual behavioral sequences concerning conformity to norms (such as facial expression, shaking hands, keeping a certain distance to the other person, helping in need, and so forth). It also activates expectations about how other people are likely to act and positive evaluations of the means to reach the goal (Ferguson and Bargh 2004). This power of overarching goals to coordinate a large number of cognitive and motivational processes is the basis for the goal-framing theory on overarching goals (Lindenberg 2001b, 2006, 2008; Lindenberg and Steg 2007).This theory specifies three goal-related self-regulatory processes: a. the ability to have goals that set the mind by coordinating cognitive and motivational processes (“overarching goals” called “goal-frames” when activated); b. specific overarching goals that regulate tasks that are instrumental but different from need satisfaction; c. processes that balance the relative strengths of overarching goals in favor of adaptive behavior. Three goal-frames Goal-frames are activated (that is, focal) overarching goals; they “frame” the mind. Need satisfaction refers to many different needs, but the overarching goal with regard to need satisfaction is to improve (or maintain, as the case may be) the way one feels at the moment. Thus the most basic goal-frame is what is called a hedonic goal-frame. It activates one or more subgoals that promise to improve the way one feels in a particular situation (such as avoiding effort, avoiding negative thoughts and events, avoiding direct uncertainty, seeking direct pleasure, seeking direct improvement in self-esteem, seeking excitement,

Social Rationality, Self-Regulation, and Well-Being

83

and the like). Its time horizon is very short and the criterion for goal realization is an improvement in the way one feels. People in a hedonic frame are also especially sensitive to what increases and what decreases their pleasure and affects their mood. For example, in a hedonic goal-frame, people are likely to react much more strongly to being made to feel bad, say, by being treated unfairly, than in the other two goal-frames. As overarching goal, the hedonic goal-frame is not purely a lower-order self-regulatory process but, surely in comparison with the other two overarching goals, it is strongly linked to lowerorder processes (McClure et al. 2004). The hedonic goal-frame is oriented toward the here and now. The ability to be concerned about resources in a focused way comes from another overarching goal, which, when activated, is called a gain goal-frame. It will make people very sensitive to changes in their personal resources. Its time horizon is middle- or long-term, and the criterion for goal realization is an improvement of (or prevention of decrease in) one’s resources or efficiency of resources. When this goal is focal, subgoals having to do with resources (such as saving money, increasing one’s income, dealing with threats to one’s financial security) will be easily activated; and subgoals, having to do with the way one feels and with normative behavior (see below), will be more or less inhibited—that is, they are pushed into the cognitive background. Even though both hedonic and gain goal-frames can be said to be linked to rewards, they are linked to different kinds of rewards and to different time perspectives, even in the neural systems (McClure et al. 2004). Note that norms can play an important role in a gain goal-frame to the degree that the individual is focused on costs for norm-conformity. For example, cheating is against the established norms, but in a gain goal-frame only the expected costs (say, in terms of a fine or reputational damage) of cheating will be considered, not any feeling of “obligation.” For people in gain goal-frame, when a particular good in the supermarket is more expensive than another of comparable quality, not much attention will be paid to the fact that one was produced in an environmentally friendly way while the other was not, even if the person values a sustainable environment. In short, in such a goalframe, norms play a role only as sources of constraints (such as disapproval or a fine). A gain goal-frame flanks need satisfaction in the sense that it focuses on the resources necessary for need satisfaction. However, the social brain also contains a goal-modus in which individuals focus on being members of a group. When activated, it is called a normative goal-frame, which covers all sorts of subgoals associated with appropriateness (such as behaving the right way, contributing to a joint project, showing exemplary behavior). It will make people especially sensitive to what they think one ought to do. When in a normative goalframe, the important aspects of a situation are normative, both in the sense that one is sensitive to “oughts” according to self or others and in the sense that one is sensitive to what one observes other people do (corresponding to the distinction made by Cialdini, Reno, and Kallgren [1990] on injunctive and descriptive norms). For example, a person in a normative goal-frame is not likely to throw a piece of trash on the street because that is inappropriate.When people are in a normative goal-frame, subgoals having to do with the way one

84 Siegwart Lindenberg feels and with personal resources are pushed into the cognitive background. Thus, for example, people who see a situation as a joint project (in a normative goal-frame) will contribute more to a collective good than people who see the situation as an “economic” one (in a gain goal-frame, see, for example, Pillutla and Chen 1999). A normative goal-frame is linked to feelings of “oughtness” about situationally relevant norms. This feeling contains at least three elements: the subjective importance of the norm; the tendency to react negatively to norm violations by others; and a tendency to feel obliged to follow the norm oneself. Like the other two goal-frames, the normative goal-frame is also linked to a particular neural system (Mendez 2009; Moll et al. 2005). Balance: background goals and the a priori strength of goal-frames There are two important additional points to be made about the goal-frames that have to do with higher-order self-regulation. The first point concerns the fact that the modularity of goal-frames is porous, that it is open to some influence from the background goals, an important reason for modularity to be “semi.” This makes possible the advantage of focus and coordination derived from overarching goals without the cost of complete neglect of the other two overarching goals. From the cognitive background, the other two overarching goals can strengthen or weaken the relative weight of the foreground (focal) goal. Motivations are thus rarely totally homogeneous, as we know from experimental evidence and daily experience. More often than not they are mixed, and it depends on the relative strength of the foreground and background goals to determine what the final effect will be. For example, the goal to eat (a hedonic goal) may be focal and the goal to remain healthy (a gain goal) may be in the background. Köpetz et al. (2011) showed that when the goal to eat is activated and its salience is boosted, then subjects, asked to choose between various kinds of foods, do not make much difference between high- or low-caloric foods; they eat almost everything that is equally tasty. By contrast, when the focal goal to eat is not boosted, then subjects are still focused on food, but the background goal (health) becomes relatively stronger and subjects become quite discriminating: they choose more low-caloric food. These effects of background goals imply that even in a dominant normative goal-frame considerations about gains are not completely gone. Conversely, experimental evidence shows that people rarely act completely egotistically, even if their main goal is gain. Rather, even then they seem to be somewhat restrained by normative concerns (see Camerer 2003; Ligthart and Lindenberg 1994). At any time, one goal is focal and influences cognitive process the most (that is, it is a goal-frame), while other goals are in the background and increase or decrease the strength of the focal goal to a greater or lesser degree. How does this work? Often the goal-frame and background goals will be in conflict. For example, if being cooperative is quite expensive, the normative goal-frame and the gain goal in the background are incompatible. This may not change the goal-frame from a normative one to a gain goal-frame—that is, the background motive may not affect the orientation (the ordering of alternatives is still in terms of appropriateness, not in terms of price), but it may lead to the choice of a less appropriate (but cheaper) alternative. In this case, price will affect the choice

Social Rationality, Self-Regulation, and Well-Being

85

but less than appropriateness. If goal-frame and background were reversed in this example, appropriateness concerns would affect the choice less than price (see examples below). The background goals do not necessarily weaken the workings of the goalframe. When they are compatible with the goal-frame, they strengthen it. This is particularly important for the normative goal-frame, which, as I will discuss in a moment, is a priori the weakest goal-frame that needs the most support (from compatible background goals) in order to withstand the weakening effect of conflicting background goals. What actually happens is that alternatives may serve both the focal and the background goal(s) to various degrees. For example, up to a point, community pharmacists may both advance their professionalism and commercial interests. But when making profit becomes more prominent, the two goals become conflictual (see Cancrinus et al. 1996). When there is a conflict, alternatives that serve background goals the best serve the focal goal quite badly, and vice versa. When there is compatibility, alternatives that serve the focal goal well will also serve the background goal well. As will be shown, this is the basis for balancing the overarching goals. Often, both compatible and incompatible background goals will be present. In sum, motives are mostly mixed in the sense that both foreground and background goals are operative. For example, the normative and gain motives often mix, but it makes a big difference whether people are in a gain goalframe and don’t go all out in the pursuit of gain, or whether they are in a normative goal-frame and cut corners because of the influence of gain motives. This difference lies in the fact that the focal goal, and not the goal(s) in the background, governs the selection and representation of preferences and constraints. Second, a priori, the three goal-frames are not likely to be equally strong.This asymmetry is a lower-order self-regulatory device that makes great evolutionary sense. The hedonic goal-frame, being directly related to need satisfaction and thus being the most basic, is very likely to be a priori the strongest of the three goal-frames. In other words, in order to displace the hedonic goal from the foreground, the gain and normative goals must have additional supports. Because, in evolutionary terms, the group is there for the adaptive advantage of the individual and not the other way around, the normative goal-frame is, a priori, the weakest. The gain goal-frames, being linked to one’s own resources, is in between. In order to withstand the onslaught of conflicting hedonic goals, gain and normative goal-frames need to be supported by compatible goals in the background. These supportive background goals are, in turn, often dependent on institutional arrangements. As Weber (1961) has shown, the gain goal-frame needs institutions (such as religion or secure property rights) that allow the individual to act on behalf of a reasonably well-established future self. The normative goal-frame is even more dependent on external support, be it through institutions and moralization (see Lindenberg 1983, 1992; Rozin 1999), or explicit disapproval for not following the norm (see Tangney and Dearling 2002). Just how precarious the normative goal-frame is can be demonstrated with an experiment we performed, concerning the norm of stealing (Keizer, Lindenberg, and Steg 2008). We placed a very noticeable envelop with a

86 Siegwart Lindenberg transparent window in a public mailbox, but we did it in such a way that it stuck out and people walking by could clearly see what was inside. What they could see was a five Euro bill peaking through the window of the envelope. The question was how many people who passed the mailbox would go so far as to take the envelope with them. The results showed that without graffiti 13 percent of all passersby took the envelope, and that with graffiti this percentage more than doubled (27 percent). Thus, if one lives in an environment with many indicators of low concern for acting appropriately, there is a risk that self-regulation will be impaired simply because of disorder in the social environment. Balance: changing the relative weight of gain and normative goal-frames The relative weakness of the gain and normative goal-frames compared with the normative goal-frame would mean a permanent dominance of the hedonic goal-frame were it not for the higher-order self-regulatory capacity to seek and or create extra support for the a priori weaker goal-frames. Goalframes cannot be directly chosen, but they can be made more likely or more stable by changing the environment, by distraction, or by conjuring up images (Mischel and Ebbesen 1970). In the literature, this is often referred to as “selfdiscipline”—that is, an effortful resistance to being tempted by hedonic goals at the expense of gain or normative goals (Baumeister and Vohs 2007). But because the gain goal is still relatively stronger than the normative goal, selfdiscipline is also applied to situations in which people are tempted by personal advantage not to act normatively (Keizer et al. 2008). Self-discipline is thus a particular kind of self-regulation that has to do with balancing the overarching goals by effortful control. When the weaker overarching goals have relatively little support, they can be strengthened somewhat by effortful control. However, this uses both physical and mental energy and leads to depletion, which in turn diminishes the relative weight of the normative goal-frame (DeWall et al. 2008) or of the gain goal-frame vis-à-vis the hedonic goal-frame by increasing risk-taking (Freeman and Muraven 2010). The ability to regulate one’s emotions also belongs to this realm of self-regulation. Emotions such as fear or anger make it difficult to sustain a gain or normative goal-frame, and they can be socially very disruptive. Emotion regulation is a crucial element in social competence (see Denham et al. 2003; Schultz et al. 2001), and lack of it can have severe longterm consequences in terms of occupational downward mobility, erratic work lives, and problematic partner relationships (see Caspi, Elder, and Bem 1987). The inability to self-regulate hedonic goal-frames also makes people smoke and eat more than they would like to, lowering their subjective well-being (see Stutzer and Frey 2007). The point is not that successful self-regulation does away with the hedonic goal-frame. Not showing emotions when it is called for (say, when your mother dies) is socially also inadequate. Some people have managed to stabilize their normative goal-frame to such an extent that they have to plan times for hedonic experiences (see Kivetz and Simonson 2002). Conversely, temptations are not always a danger for the stability of a normative goal-frame. It has been shown that being exposed to temptations can actually strengthen the normative

Social Rationality, Self-Regulation, and Well-Being

87

goal-frame (see Fishbach, Friedman, and Kruglanski 2003), so that normative goal-frames in very sheltered environments may be particularly susceptible to the rare intrusion of hedonic or gain opportunities (think of Dürrenmatt’s play “The Visit of the Old Lady”). There are supports for the weaker goal-frames that lower the needs for effortful control and thus also lower depletion effects.3 From what was said above about the power of cues that show deviant behavior, it follows that a very important source for weakening or strengthening the normative goal-frame is the behavior by others. In other words, very important social influences derive from the fact that the goal-frame of people in the surrounding also influences the stability of one’s own goal-frame (“goal-frame resonance,” see Lindenberg 2000; and “goal contagion,” see Aarts, Gollwitzer, and Hassin 2004). Hence, one of the most basic forms of self-regulation is to remove oneself from unwanted sources of influence, either physically or by shifting attention (Hoch and Loewenstein 1991; Mischel and Ebbesen 1970), or by removing the unwanted sources of influence themselves.Thus, for example, cleaning up physical disorder is likely to stabilize people’s self-regulation (by removing cues of other people’s transgressions; see Keizer et al. 2008). But there is also social disorder, and one may or may not be able to clean it up. For example, the behavior of high-status people showing disrespect for norms has a particularly strong effect on the normative goal-frame (Cohen and Prinstein 2006). This means that politicians and celebrities can have a considerable negative influence on people’s selfregulatory ability, an effect that is exacerbated by the fact that the powerful often use norm violations to demonstrate their power (Van Kleef et al. 2011). Peers may be individually less influential than high-status people, but then they mostly come in groups and thus exert considerable power over goalframes of the members. For example, being in a group of peers who seek fun and entertainment (a hedonic goal-frame), it is difficult to keep up a normative goal-frame (Sentse et al. 2010). If one wants to keep up a normative goal-frame, one is likely to proactively avoid the group, or if one is already in it, to leave it, if possible. Here, timing is of the essence. If one waits too long, the contagion will have progressed beyond the point at which self-control is likely to be strong enough to battle the effect of goal contagion and make one leave the group. For good or bad, the company one keeps may thus have a lot to do with the goals one pursues. The same effect has been observed with moods (Neumann and Strack 2000). Goals and expectations are influenced by one’s mood. If one is in a group of people who are in a bad mood, one’s own mood is likely to be negatively affected, and one would have to remove oneself early on from the group in order to escape this influence. James (1890), always attentive to issues of self-regulation, advises people to “accumulate all the possible circumstances that re-enforce the right motives” (123). Yet, some influences are difficult to escape. Children rarely can escape the influence of their parents, even if these parents beat them, are neglectful, and humiliate them. Even if they could escape, they often would not be able to improve their condition much by doing so, because they have a high chance of ending up in institutionalized care or no care at all. This does not only hamper the execution of self-regulation, but also the development of self-regulatory abilities. School contexts can help in this regard. Coleman and Hoffer (1987)

88 Siegwart Lindenberg found that in the United States children from disadvantaged family backgrounds do better when they are placed in school environments that require disciplined work (such as much homework, participation in academic programs, provided by Catholic schools compared with private and public schools). Another example is the finding that the adult time horizon for financial planning is strongly influenced by early parental influences on the child’s extension of the self into the future (Hershey, Henkens, and van Dalen 2010). A low extension hampers self-regulatory ability (see Nenkov, Inman, and Hulland 2008).Thus the inability to escape certain environments as a child can have long-term consequences. There are also path-dependent effects regarding opportunities and the ability to use them. If the environment is not conducive to the consideration of future consequences, be it because of goal contagion or of highly uncertain futures, people will be more frequently confronted with negative life events, at least a good deal of which are likely due to failures of self-regulation (see Brady and Matthews 2002). Research shows that people from lower income classes have more difficulty dealing with reasoning that is related to a gain-goal frame and necessary for handling economic decisions (such reasoning in terms of costs and benefits and ignoring sunk costs; see Larrick, Nisbett, and Morgan 1993). Such influences are not just concerning the future orientation (gain goalframe) but also the normative goal-frame. For example, lower class youth have been found to have much more trouble recognizing general social norms (see Parker and Fishhoff 2005). In part, the described effects can also be circular. For example, socioeconomic status can be interpreted as an indicator of influential environments with regard to health-related lifestyles (smoking, drinking, eating, sleep habits). In turn, negative health indicators (such as obesity) may contribute to locking people even more into a lower socioeconomic status (Mulatu and Schooler 2002). Probably even more important for self-regulation is the exemplary behavior of others as a cue that strengthens one’s own normative goal-frame. Thus exposing oneself selectively to influences on one’s overarching goals is perhaps the most important part of one’s ability to regulate oneself via goals (Lindenberg 2008; Dohmen and Falk 2011). Among these influences, probably the most prominent is having (and making oneself vulnerable to the influence of) significant others that represent normative claims. For this reason, I will dedicate a longer paragraph to that form of support. The important role of significant others. Not everybody is equally important for one’s self-regulatory abilities. In the course of their development, people acquire significant others (such as mother, partner, close friends, religious leaders) whose opinions and standards weigh heavily and who can be called upon especially to strengthen the normative goal-frame. Of course, some of the most important significant others are the direct socializers in early childhood, and especially the mother. They represent norms and standards, and in interaction with them the ability to stabilize the normative goal-frame is developed (see Gralinski and Kopp 1993; Kochanska 2002). However, the significant others are not just important in the formative years (and for the internalization of substantive norms), but also for the inner dialogue that keeps going on. They remain in the person as a private audience to which the self turns and virtually interacts

Social Rationality, Self-Regulation, and Well-Being

89

(see Baldwin and Holmes 1987).Thus a significant other does not even have to be physically present to influence one’s behavior. Experimental research shows that when certain significant others have been made salient in somebody’s mind, their norms will influence behavior quite strongly (see Baldwin, Carrell, and Lopez 1990; Baldwin and Holmes 1987; Fitzsimons and Bargh 2003; and Shah 2003a,b). Significant others also influence the readiness to follow social norms in general, not just specific norms. Thus having significant others on one’s mind helps stabilize the normative goal-frame. Shah (2003a) has shown that thinking of significant others can influence a person’s goals, in the sense that goals attributed to the significant other activate the same goals in the attached person, and also in the sense that goals the significant other is thought to disapprove of are inhibited. For example, Shah could show that individuals primed with father-related words were more committed to goals the father valued and also performed better at reaching those goals, the more so, the closer they felt to the father. Conversely, the closer they felt to the father, the more goals he disvalued were inhibited for them. Shah (2003b) also showed that the effect of significant others on one’s behavior does not run only via goal activation or inhibition, but also via self-appraisal and the emotional response to goal achievement or achievement failure. For example, if your mother thinks you can achieve a goal, then thinking of your mother will positively affect the appraisal of your own ability to achieve it. The converse holds for negative expectations. In addition, the more important the mother finds the goal, the more satisfied you will be by achieving it, and the more dissatisfied by a failure to achieve it (see also Baldwin, Carrel and Lopez 1990). Self-regulation thus involves a “psychological presence,” an inner meeting and dialogue of the self with significant others. Persons who have significant others who believe in their abilities and who find their goals important have a definite self-regulatory advantage in pursuing those goals. Yet having or not having significant others for self-regulation is often subject to self-regulation itself. For example, it has been found that people motivate themselves to achieve a valued goal by seeking out significant others that are successful at achieving the goal as role models (Lockwood, Jordan, and Kuna 2002). When people cannot attach to significant others (especially those that represent important social norms), the normative goal-frame cannot easily be strengthened by alternative means to the same degree. As a result, self-regulatory capacity is likely to be lower and depletion effects higher. For example, Gestsdottir and Lerner (2007) show how important self-regulation is for a positive development of youths.Yet children may have systematic disadvantages with regard to self-regulation. An important case in point is attachment problems in early childhood resulting from aggressive parenting. Children with attachment problems will grow up with a deficit in significant others and thus a deficit in self-regulation (see Calkins 2004). However, this problem is not randomly distributed but occurs more in low Socioeconomic Status (SES) families (see Pinderhughes et al. 2000; Raikes and Thompson 2005; Shaw et al. 2001). Thus children from low SES backgrounds run the risk of lower selfregulation capacity, as well as later in life, and bear the risk of lower well-being (see also Hart, Atkins, and Matsuba 2008). Note that self-regulatory ability has an impact on problem behavior and performance quite different from general

90 Siegwart Lindenberg intelligence, so that it is not simply a matter of a negative correlation between SES and intelligence (see Ayduk et al. 2007; Blair and Razza 2007; Moffitt et al. 2011). Lack of self-regulatory ability is also likely to affect status (see Bear and Rys 1994; Moffitt et al. 2011), so that, as already mentioned above, we get a vicious circle that may trap people in a low-status position. A similar effect can be expected for the children of immigrants. They are better integrated into the new culture than the parents, and are often confused about who their significant others are.This negatively affects their self-regulatory abilities, which makes it more difficult to break out of low-status positions. An important consequence of these phenomena is that paying attention to significant others will often work better than punishments or rewards. For example, Sampson, Laub, and Wimer (2006) show that being married (that is, having a close significant other who cares) has a considerable effect on reducing criminal activity. By contrast, it is by now well known that the interventions directed individually at problem youths (such as incarceration, probation, shocking youth by the experience of brief incarceration or by having criminals tell them about the horrors of prison [“scared straight”], courtordered school attendance) don’t work very well (see Kazdin and Weisz 1998; Lipsey and Wilson 1998; Sherman et al. 1997). Where the traditional rational choice models would assume that negative incentives (such as incarceration or shock experiences) steer behavior away from trouble, the social rationality approach, with a central place for self-regulation, would look first of all at the functioning of significant others for self-regulation capacity. Incarceration is likely to increase self-regulation problems because it reinforces the importance of delinquent peers and decreases the importance of adults in authority as significant others (see Huey et al. 2000). What is likely to help is to improve the positive role parents and teachers can play as significant others by focusing intervention on teacher and family functioning (Kazdin and Weisz 1998) and to coordinate the role teachers and parents can play as significant others (Eddy, Reid, and Fetrow 2000). This also involves communicating clear rules and expectations that emanate from the significant others (see Sherman et al. 1997: ch. 5). Conversely, changing one’s ways in order to become a better significant other for somebody else requires that, for example, a parent improve his or her own self-regulation by accepting therapists as significant others (Kazdin and Whitley 2006). Self-regulation is a socially embedded process and thus needs continuous social support. Thus it also helps to make the youths more susceptible to the influence of relevant significant others. For example, training in cognitive problem-solving skills (prominently including perceiving how others feel and anticipating the effects of one’s behavior on others) seems to be quite effective in reducing antisocial behavior in (pre)adolescents (see Kazdin and Weisz 1998). Note that this reasoning is not just based on the workings of social influence or “social integration” (as social control theory would have it; see Hirschi 1969). For example, the attempt to change youth violence by redirecting highrisk youth through enriching their recreational activities in the peer group context (for example, by midnight basketball games) did not work (Elliot and Tolan 1999; Patterson et al. 1998). Such interventions do not establish links to significant others who strengthen the normative goal-frame.

Social Rationality, Self-Regulation, and Well-Being

91

Meaningful need states. A more “radical” way to increase goal-related selfregulation is to change not just the environment but also the goal at the same time.When self-regulation is lacking, people can experience extreme frustration at not doing what they set out to do, or not knowing what to do, or feeling bereft of meaningful activity. By contrast, being eager to achieve a meaningful goal, knowing what steps to take, feeling efficacious to take those steps: jointly, these things give one a feeling of being in control, being part of something meaningful, and of being motivated to go forward. This is what may be called a “meaningful need state,” and it can be more or less actively sought as a way to increase the balance of the goal-frames. Often, but not always, it is also linked to significant others. The advantage of such a state is that when it is active, people feel that they have a purpose, they feel energized to pursue it, have feelings of deprivation if they cannot pursue it, are frustrated if they fail at pursuing it, and experience satisfaction by making progress in their pursuit. In short, meaningful need states provide both self-regulatory capacity (direction, planfulness, link to norms) as well as sources of satisfaction.When such a state is active, a normative goal-frame is strongly supported by a hedonic goal from the background, or (less advantageous for self-regulation, as we will see later), a hedonic goal-frame is strongly supported by a normative goal in the background. Bunderson and Thompson (2009) provide a graphic example of this combination: zookeepers, and how they experience both a deep moral duty to follow this “calling” and a passion for their calling. The concept of intrinsic motivation (as traditionally employed) is not suitable to cover meaningful need states because it lacks the normative element necessary for the meaningfulness besides enjoyment (see Lindenberg 2001b). It is probably difficult to overestimate the importance of the search for meaningful need states in society for both people’s attempts at increasing their self-regulatory capacity and also their sense of purposefulness and well-being. Very likely it is ubiquitous and much in need to be studied. How do people get into such a state? The basis of its working is the possibility that goals can acquire a needlike urgency. The assumption here is that when goals are considered meaningful and they are stated in such a way that they have a clear end state and clear steps that lead toward the end state, they can create an especially strong goal-gradient effect. The closer to the end state, the stronger the motivation to reach the end state (Kivetz, Urminsky, and Zhang 2006). Such goals open, as it were, a space for purposeful and worthy pursuit with the built-in driver of structured approximation toward a good end. A situation could trigger such a need state, but it is more likely that people, in search for sources of purpose and self-regulatory capacity, find themselves gradually in a situation in which they discover, welcome, and embrace the tug of a meaningful need state. Projects can be organized to grip people in such a way. For example, many people in the Netherlands work overtime. A survey (OSA 2003) reports that in the Netherlands, 25 percent of the Dutch labor force work paid overtime, and 27 percent put in unpaid overtime. Why do so many people put in unpaid overtime? It cannot be the lumpiness of labor supply in which the job with the exact preferred number of hours is unavailable. In the Netherlands a new law, the Working Hours Adjustment Act (Wet Aanpassing Arbeidsduur) was

92 Siegwart Lindenberg introduced in 2000 that gives employees the right to reduce or increase their contractual working hours. Generally, organizations comply with this law. We have looked into this question and found that it is also not the improved chances of promotion that drive unpaid overtime. Rather, our research shows that it is the project organization that does it (van Echtelt, Glebbeek, and Lindenberg 2006).The most likely circumstance leading to the decision to work overtime is that there are clear steps that have to be taken toward the completion of a project and that these steps do not get quite finished during regular time. Something presumably worthy has to be brought to a good end, and the steps in between may be such that they act as their own motivators. In other words, the unpaid overtime is very likely the result of the workings of meaningful need states. Even though organizations make strategic use of this effect (van Echtelt, Glebbeek, Lewis, and Lindenberg 2009), it can work only because people let themselves be swept into the self-motivating and self-constraining stream of a project because they profit from it in terms of self-regulatory capacity and satisfaction. Hobbies are another common source of meaningful need states. Hobbies are often quite socially regulated and organized in such a way that they provide purpose and at the same time embedded and often concatenated end states. For example, stamp collectors create worthiness of their pursuit by the networks and communities that exchange information and provide standards of competence and value, as well as opportunities to demonstrate (more or less competitively) expertise, stamina, and cunning with communal appreciation. In addition, the stamp collector community and the postal authorities that partially cater to this community create embedded meaningful end states by defining sets of stamps that belong together. There is in principle no end to this pursuit, because by new groupings, the end states are inexhaustible and, at least in part, also concatenated by sequences. Another example is bird watching, which is also socially organized and structured in terms of projects each one of which can be brought to a good end, only to be followed by another. Socially embedded hobbies are likely to be linked to particular significant others who represent the worthiness and normative standards for this form of improving one’s selfregulatory capacity. This means that both forms of support for self-regulation are mostly interlinked. The downside of meaningful need states. The positive side of the meaningful need states can also be their downside: they have a hold on people, especially if the significant others are lacking or not demanding with regard to social norms. Since the meaningful need states are actually activated goals, they tend to inhibit possibly competing goals (in case of overtime work, it is goals such as family obligations and leisure time that get sidelined; see Caruso et al. 2004; Dembe et al. 2005; and Dahlgren 2006). There can even be a reversal of hedonic and normative goals, such that the hedonic goals are in the foreground. For example, this might happen with computer games that are structured to create seemingly worthy, embedded, and concatenated end states. However, the emphasis is on embedded end states with relatively low standards for meaningfulness, which allows a preponderance of the hedonic goal. The significant others may be nonexisting or linked only to normative standards that are internal to the game structure and thus do not help with the self-regulation of social contacts. In

Social Rationality, Self-Regulation, and Well-Being

93

terms of self-regulatory capacity, such versions of meaningful need states may thus be counterproductive.

Self-Related Processes of Self-Regulation The very basis of top-down forms of self-regulation is that individuals are able and motivated to distinguish between self and nonself, to be selfreflective (that is, they can become an object of their own attention), and to be agentic—that is to say, feel themselves as the cause of changes in the world and changes in their own inner states. This process requires elaborate cognitive and motivational processing (Christoff et al. 2011; Jeannerod and Anquetil 2008; Lieberman 2007). The self has occupied a central place in both psychological and sociological theorizing, resulting in many different theories of the self; this is not the place to review that literature. However, there is considerable consensus about the fact that the development of the self is itself a social process, and so is the maintenance of the self. In fact, people are treated and evaluated as selves by others, and they evaluate themselves as selves, compare themselves with others, try to maintain a positive view of themselves, have ideas about their ideal selves, and strive to reduce the discrepancy between their present and their ideal selves. In all this, people are able to bridge the gap between self and social nonself with a theory of mind, perspective taking, and affective empathy—that is, with processes in which they experience others also as selves (Malle and Hodges 2005). Social interactions and evaluative processes (both of which are also relevant for satisfying fundamental needs) thus heavily lean on “selves” and their maintenance (see also Erikson 1964;Vygotsky 1978). From the foregoing, it is clear that the integration and maintenance of the self and the capacity to bridge the self and (social) nonself are important. But what exactly is so special about the self? What gives it such an important place? The main reason for this important place is the fact that individuals must be able to have internal representations of themselves and their own mental states (Frith and Frith 1999; Goldberg 2009; Greenwald and Breckler 1985). This is important for self-monitoring, for plans, and for reacting to one’s own prepotent impulses and the feedback from others. For adaptive behavior, individuals must be able to keep track of themselves, demand things of themselves, and respond to their own representations. Importantly, this response is also evaluative, creating more or less self-esteem. Individuals must also be able to project themselves into the future, make plans, and pursue interrelated goals. The quality of their decision-making depends to a large extent also on the way they feel about things. Inconsistencies or ambiguities in feelings will in many cases also negatively affect the quality of decision-making (see Bechara and Damasio 2005). In addition, there are constraints resulting from reaction to others. What feedback should an individual react to, and how? How to avoid impressions of arbitrariness or lack of direction? All these points are negatively affected when the self-nonself distinction is in flux, or when the self is experienced as fragmented or uncertain or unworthy. There is also yet another reason for the importance of the self. It is likely that perspective taking and empathy (putting oneself into the shoes of the other, both cognitively and affectively) depends on the ability to simulate what

94 Siegwart Lindenberg might go on in the other, which, in turn, requires a sense of self (Lieberman 2007; Uddin et al. 2007), all the more since the observer might have to imagine the situation of the other without having experienced it (Eisenberg and Sulik 2012). Schematically, the process of self-formation can be described as follows (see also Markus and Cross 1990). The child learns early on to demand things from the social environment. It is also genetically equipped with the ability to distinguish social from nonsocial objects and pays special attention to social objects (Wynn 2007). Later, it also learns to put itself into the shoes of significant others in the social environment, which presupposes some cognitive separation of self and others and the understanding of the other as intentional being (Malle and Hodges 2005). It also requires the ability to see itself through the eyes of others (Mead 1934). The child thereby also learns that things are demanded of it. The child develops self-representations and experiences itself as causal and agentic (Jeannerod 2003). It also develops shared representations with others (for example, joint attention; Tomasello and Carpenter 2007). In a further step, the ability to put oneself into the shoes of others is applied to oneself, as the child also learns to put itself into the shoes of its own future self, and also into the shoes of itself as member of a collective (dyad or group). This is also the basis for the overarching gain goal and normative goal, described earlier. Thus the child learns to be represented to itself in a need-related, resource-related, or collective guise. The ability to demand and respond to demands is present in these different selves. In this sense, the person is a social system. As the child begins to talk, it eventually also learns to talk to itself from different perspectives and to demand and suggest things to itself, in private speech (aloud but directed to itself) and later also in inner speech (Dias and Berk 1992), including nonverbal inner communication such as conjuring up images and auditions (Barkley 2004). For example, the child learns to demand from itself to block aggressive emotions from arising, and has the capacity to influence its own electrochemical processes in the brain (see Banks et al. 2007). In this development, the child will evaluate various aspects of itself (or of its various selves), and as the child grows up there is increasing internal and external pressure to integrate these aspects in a more or less harmonious way (Erikson 1964; Nowak et al. 2000). lower-order self-regulatory processes concerning the self There are at least two important lower-order self-regulatory processes that kick in when the self feels either invulnerable or threatened.With regard to the feeling of invulnerability, the self-regulatory tendency is to lower the degree to which one is socially influenced. In terms of evolution, this makes sense. The less dependent one is on others, the more adaptive it is not to use scarce resources on them. Yet high-order self-regulatory processes are needed to contextualize this tendency because it may lead to the negative consequences in the long run (say, because ignoring people may turn them against you and thus increase interdependency). This possible dysfunctional side of the lowerorder reaction to the feeling of invulnerability shows up in research on power. For example, powerful people tend to be worse at perspective-taking (Galinsky et al. 2006) and tend to be overconfident in their decision-making (Fast et al.

Social Rationality, Self-Regulation, and Well-Being

95

2012). Organizations can greatly suffer from these tendencies (Bunderson and Reagans 2011). Probably the two most important lower-order self-regulatory processes concerning threats to the self are (a) self-defense, in which a particular selfimage is defended against (mostly external) threats; and (b) cognitive egoenhancing biases. Self-defense is a regulatory process that involves both cognitive and affective aspects. Cognitively, it is linked to closing one’s mind to threatening information, or denying evidence (for example, denying that smoking is dangerous if smoking is part of one’s self-image). Affectively, self-defense is linked to avoidance or approach tendencies, and both can be mediated by a sense of shame (de Hooge, Zeelenberg, and Breugelmans 2010). There can be attempts to make amends, but, as people react defensively to a threatened selfimage, there can also be aggression and violence (Baumeister, Smart, and Boden 1996). For example, people may aggressively blame others in order to protect an unblemished self-image against threatening accusations. Negative feedback from others (such as disapproval) can be experienced as being directed against a particular kind of behavior (in which case it lowers satisfaction of the need for behavioral confirmation and may lead to repair behavior), or it may be experienced as being directed at the person, in which case it threatens the self and may lead to aggressive self-defense (see, for example, Bagozzi,Verbeke, and Gavino 2003). If the approach reaction is not feasible or too risky, then shame may lead to denial, withdrawal, or escape. Ego-enhancing biases consist of people’s tendency to see things such that it enhances their self-image, especially after it has been threatened (Dunning, Leuenberger, and Sherman 1995). Many such biases have been identified (Dunning 2002). One example is the better than average effect, which makes people assume that, regarding traits that are important to their self-image, they are better than the average other.This kind of assessment is not just for the sake of impressing others. It is a truly lower-order self-regulatory mechanism such that people believe their self-enhancing assessment (Williams and Gilovich 2007). In-group favoritism is another form of self-enhancement (Gramzow and Gaertner 2005). I will not list all the self-regulatory strategies to enhance one’s ego, especially after the ego is being threatened. However, all these forms share the basic mechanism of defending a positive self-image. Even though people with a negative self-image also tend to enhance their ego, they are also prone to seek validation of their low estimation of their self and thus have mixed self-regulatory ego-defensive strategies (Sherman and Cohen 2006). higher-order self-regulatory processes concerning the self For sociology, the higher-order self-regulatory processes concerning the self are even more important, because they are crucial for battling the negative side of the lower-order self-regulation and for a smoother running of social interaction in virtually all contexts. In the literature we find many different theories about high-order self-regulation concerning the self; however, the three sociologically most important groups of self-regulatory processes are likely to be the following:

96 Siegwart Lindenberg a. The search for clarity of core aspects of the self, b. The search for harmony among aspects of the self, c. The search for positivity concerning the self. I will briefly discuss them in order. Clarity. Even though there are many partial selves, there are aspects that are likely to belong to what Markus (1977) called “self-schema.”This core contains aspects that one considers particularly important about oneself. No matter what goal-frame is salient, persons will always be particularly sensitive to information about themselves. Aspects that belong to the self-schema are more accessible and thus come to mind more easily.They are also better remembered and better linked to other parts of one’s knowledge structure. For example, if somebody talks about your qualities as a father, it may make you particularly vigilant about what is said because your being a good father belongs to your self-schema. A poorly defined self-schema (that is, confusion about who you are) creates less focused and consistent guides to attention and retention (Campbell et al. 1996) and thus weaker tendencies to override lower-order self-regulatory processes. It also lowers self-esteem and well-being (Sheldon et al. 1997; Stinson,Wood, and Doxey 2008). For the establishment and maintenance of clarity of core aspects of the self, the search for and availability of significant others (including role models) have been identified as particularly important (Bosma and Kunnen 2001; DuBois et al. 2002; Meeus, Oosterwegel, and Vollebergh 2002). People search for support from significant others and look at role models in order to find out what may be the core aspects of their personal identity (Markus and Cross 1990). This holds for all phases of life, even though the significant others and role models may change. For example, for adolescents it may be parents, peers, teachers, or celebrities; for adults it may be partners, supervisors, colleagues, or friends; and for elderly people it may be partners, friends, or spiritual leaders (Carstensen 2006). Such search is intensified in times of negative life events (Thoits 1991). In short, the degree of identity clarity heavily depends on having supporting significant others and role models, especially when doubt about one’s identity is most salient (as in times of negative life events). People who have reduced access to such significant others are likely to have more problems with maintaining identity clarity. Another factor that has been identified as playing an important role is cultural clarity. This concept refers to clarity about one’s cultural group (whatever that group may be). Clarity in this area has a positive impact on the clarity of one’s personal self (Usborne and Taylor 2010). Thus cultural confusion is likely to lead to confusion about one’s personal identity. Self-regulation is thus likely to focus also on cultural clarity, for example by culture-relevant selectivity in interaction and negative out-group attributions. Harmony. Even with clear core aspects of the self, there are many partial selves (that also overlap with important roles), and they may be more or less in conflict with one another or in harmony. For example, one’s self as a mother and wife may be in conflict or harmonious, as is one’s present self in relation to one’s possible self (Markus and Nurius, 1986). Conflict may create difficulties

Social Rationality, Self-Regulation, and Well-Being

97

in making plans, in fulfilling obligations, in emotion regulation, and in selfesteem. Harmony refers to the compatible combinations (not necessarily to the integration) of partial selves. A higher degree of harmony has been associated with better resource use and higher well-being (Brook, Garcia, and Fleming 2008). Self-regulation with regard to harmony focuses on compatibility in terms of activities, circles of interaction, plans, and so forth. For example, creating a career is an important process for young adults, and thus it is important which aspects they can pursue without conflict (see, for example, Syed 2010). Part of their career planning will be dedicated to creating harmony among their partial selves, both in terms of identity-relevant roles and traits. Ideologies, narratives, and “self-portraits” are ways to help create and maintain harmony, and the higher the need for harmony the more likely that people will make use of these tools for self-regulation (see Habermas and Bluck 2000; Harter and Monsour 1992). Again, significant others are an important potential source of harmony (Chen, Boucher, and Tapias 2006). Positivity. Even though self-regulation focuses on clarity and harmony, this does not mean that an individual thereby automatically has a positive view of the own core self.Yet, since the core self is the basis of one’s agency, one’s plans, dreams, and hopes, one would like to be able to have a high regard for it (Jones 1973; Tesser 1988). This regard is not simply given. In comparison to the lowerorder self-defensive strategies, these higher-order strategies do not work via biased cognitions and aggressive outbursts. They involve strengthening a core sense of self in an agentic rather than reactive way, even when threats to the self are involved (Leary 2007).The strategies for this purpose include self-categorization into groups that can add positivity to the sense of self (Turner 1985). Then there is social comparison, which can be used to enhance one’s sense of self, both downward (Gibbons et al. 2002) and upward social comparison, depending on what works best for the positive sense of self. Two self-regulatory strategies are particularly important for overpowering the lower-level self-defensive strategies (and many of their negative consequences) and will therefore be discussed with a bit more detail. It is self-affirmation and self-presentation. When people reflect on what their personal values are and on how they lived up to those values, they strengthen their core self by affirming it (Steele 1988). They thereby reduce the often dysfunctional lower-level self-regulatory processes of self-defense and the concomitant closing of the mind in the process. Self-defensive strategies imply, for example, not changing one’s beliefs in the face of contrary evidence, not heeding negative health information, misperceiving one’s capacities, and jeopardizing one’s relationships with others (Sherman and Cohen 2006). Ironically, self-affirmation does not seem to work if the situational cues appeal to one’s being rational and pragmatic. Nor will the appeal to rationality and pragmatism prevent closing of the mind. Thus, for example in negotiations, appeals to the parties to be rational and pragmatic can be highly counterproductive (see Cohen et al. 2007). The explanation for this ironic result may be that making one’s rational identity salient implies lessening the focus on values and identity, thereby suppressing effects of affirming values. Threats to identity are then not buffered and therefore might trigger selfdefensive strategies.

98 Siegwart Lindenberg Self-regulation can stabilize the self-evaluation by influencing the evaluative feedback from others. Goffman has worked out this aspect of self-regulation in terms of what he called impression management (see Goffman 1959; Schlenker 2003). The presentation covers aspects that people deem important to their identity, possibly including physical appearance (dress, hair, posture) and facial composure, but also proper attention, orientation, and conversational focus, and especially clear signs of being a purposeful person. Take an example of appearing purposeful. Imagine that a person walks in the street and realizes that he has forgotten his bag in a store a block behind him. It is unlikely that he will just turn around and walk in the opposite direction, looking as if he had no purpose walking in either direction. Rather, his expressions and gestures will indicate that he has forgotten something before he turns back (for instance, he may briefly put his hand to his forehead, visibly shake his head, stop for a moment, and then reverse his direction). Other people expect to see purposefulness, and they react negatively when a person appears to lack that quality (see Gigerenzer 2002). The idea of impression management has long roots in sociology (see Cooley 1922), but in the past the literature focused less on social factors influencing self-related self-regulation. For example, social networks can be far more than sources of information or social support in times of trouble; they can be important stabilizers of one’s identity. Lacking social networks or having the wrong social networks can thus also impair one’s self-regulatory abilities via its impact on the self, not just via the impact on goal-contagion.

Conclusions When human rationality is seen as having evolved together with the problems that needed solving for adaptive behavior in changing environments, it basically comes down to a social kind of rationality that is linked to self-regulation adapted to the complex social interdependencies of human reproduction, socialization, and collective efforts.Viewed in this way, self-regulation, not consistency, should be the core meaning of human rationality. The human brain developed as a sophisticated social brain to be able to handle self-regulatory behavior in such contexts and to overrule, if need be, more basic (less context-dependent) forms of self-regulation. This led to three different evolved sets of self-regulatory processes that are interconnected but can be usefully distinguished and looked at by themselves: need-related, goal-related, and self-related self-regulation. These capacitates differ by person, but also by circumstance, and by lifelong development. This paper has discussed all three separately, but it is important not to forget their interrelation. For example, deficits in the satisfaction of fundamental needs lead to pathological behavior that negatively affects the other sets of self-regulatory processes. As we have seen, an insecure sense of self, in turn, negatively affects the balancing of overarching goals and thus leads to lower levels of self-discipline. Lower levels of self-discipline with regard to normative behavior lower in turn the satisfaction of fundamental social needs because these needs are best fulfilled obliquely, as a side effect of normative behavior (Lindenberg 1989; Sheldon 2004). When fundamental needs are being satisfied, the sense of self is strengthened (Luyckx et al. 2009). In addition to these interrelations, there are social events that disturb all three sets of self-regulation

Social Rationality, Self-Regulation, and Well-Being

99

at the same time. Lack or loss of significant others is such a factor. Another prominent example of this is social exclusion. A modest negative feedback is useful as a corrective for performance, especially if it supports one’s autonomy (Buckley,Winkel, and Leary 2004; Mouratidis, Lens, and Vansteenkiste 2010), but a feeling of being rejected by or excluded from the group seemingly leads to a decrease in the satisfaction of social needs, in the secure sense of self (provoking self-defensive behavior), and in self-discipline (Twenge, Catanese, and Baumeister 2003). It is important to realize that all three sets of self-regulatory processes heavily depend on social supports, be it in the form of informal relationships, organizational contexts or of formal institutions. Because these interrelated processes of self-regulation are rather complex, one may want to reduce the complexity by simplifying. This is perfectly legitimate, and one may concentrate on only one of the three sets (because they are interrelated), or focus only on higher-order processes because they often trump the lower-order processes. But, as sociologists, it is not advisable to simplify by focusing on situations in which self-regulation has been made trivial, in the sense that there are presumably no needs involved and no goal conflicts, there are presumably no threats to the self and no emotional reactions to success or failure. Standard models of rational choice are rudimentary theories of self-regulation in the sense that they assume agency and top-down regulation of behavior. However, in these models virtually all aspects of selfregulation are made trivial: no attention to needs or goal conflicts (thus an assumed chronic gain goal-frame), no changes in the level of rationality due to lower-order self-regulation, no emotional responses to success or failure (but instead undisturbable Bayesian updating). Such a rudimentary theory may be useful as ideal type (see Gächter, this volume), as a benchmark against which the actual complexity of self-regulation is being brought into profile, but by and large, this view of a restricted usefulness in terms of self-regulation has not yet received wide acceptance. Of course one has to simplify, but there is also a principle of sufficient complexity (Lindenberg 2001a). Many social situations and institutions are either a problem for or a solution to (or both) self-regulatory problems. In order to understand these situations and institutions, we would do well to make self-regulatory processes and their social embedding a core business in sociology.

Notes 1. See, for example, Lindenberg 1996; Ormel et al. 1999; Steverink, Lindenberg, and Ormel 1998; Nieboer and Lindenberg 2002; Nieboer, Lindenberg, Boomsma, and van Bruggen 2005; Steverink and Lindenberg 2006. 2. When the group concerned consists of just a close relationship, then behavioral confirmation and affection often blur, and so does the way they are talked about in the literature (see, for example, Baumeister and Leary 1995: 505; and Murray et al. 2003: 63). 3. For reasons of space, I will concentrate mainly on the supports of the normative goal-frame, even though equal space could be allotted to the strengthening of the gain goal-frame. With regard to the latter there is, however, less need of elaboration, as the work of Max Weber and contributions such as Williamson’s exposition (1985) of the institutional and organizational supports of capitalism provide important insights into the conditions that potentially strengthen the gain goal-frame.

100 Siegwart Lindenberg

References Aarts, H., Gollwitzer, P. M., and Hassin, R. R. 2004. “Goal Contagion: Perceiving Is for Pursuing.” Journal of Personality and Social Psychology 87: 23–37. Ariely, Dan, George Loewenstein, and Drazen Prelec. 2006. “Tom Sawyer and the Construction of Value.” Journal of Economic Behavior and Organization 60, no. 1: 1–10. Ayduk, Ozlem, Monica L. Rodriguez, Walter Mischel, Yuichi Shoda, and Jack Wright. 2007. “Verbal Intelligence and Self-regulatory Competencies: Joint Predictors of Boys’ Aggression.” Journal of Research in Personality 41, no. 2: 374–88. Bagozzi, Richard, Willem Verbeke, and Jacinto C. Gavino, Jr. 2003. “Culture Moderates the Self-Regulation of Shame and Its Effects on Performance: The Case of Salespersons in the Netherlands and the Philippines.” Journal of Applied Psychology 88, no. 2: 219–33. Bakan, David. 1966. The Duality of Human Existence: Isolation and Communion in Western Man. Boston: Beacon Press. Baldwin, Mark W., Suzanne E. Carrell, and David F. Lopez. 1990. “Priming Relationship Schemas—My Adviser and the Pope Are Watching Me from the Back of My Mind.” Journal of Experimental Social Psychology 26, no. 5: 435–54. Baldwin, Mark W., and John G. Holmes. 1987. “Salient Private Audiences and Awareness of the Self.” Journal of Personality and Social Psychology 52, no. 6: 1087–98. Baldwin, Mark W., and Lisa Sinclair. 1996. “Self-Esteem and ‘If-Then’ Contingencies of Interpersonal Acceptance.” Journal of Personality and Social Psychology 71, no. 6: 1130–41. Bandura, Albert. 1997. Self-Efficacy:The Exercise of Control. New York: Freeman. Banks, Sarah J., Kamryn T. Eddy, Mike Angstadt, Pradeep J. Nathan, and K. Luan Phan. 2007. “Amygdala-frontal Connectivity during Emotion Regulation.” Social Cognitive and Affective Neuroscience 2, no. 4: 303–12. Bargh, John A., Peter M. Gollwitzer, Annette Lee-Chai, Kimberly Barndollar, and Roman Trötschel. 2001. “Automated Will: Nonconscious Activation and Pursuit of Behavioral Goals.” Journal of Personality and Social Psychology 81, no. 6: 1014–27. Barkley, Russell A. 2004. “Attention-deficit/Hyperactivity Disorder and Self-regulation: Taking an Evolutionary Perspective on Executive Functioning.” In Handbook of Selfregulation: Research,Theory, and Applications, edited by Roy F. Baumeister and Kathleen D.Vohs, 301–23. New York: Guilford Press. Barkow, Jerome H. 1989. Darwin, Sex, and Status. Toronto: University of Toronto Press. Baumeister, Roy F., C. Nathan DeWall, Natalie J. Ciarocco, and Jean M. Twenge. 2005. “Social Exclusion Impairs Self-regulation.” Journal of Personality and Social Psychology 88, no. 4: 589–604. Baumeister, Roy F., and Mark R. Leary. 1995. “The Need to Belong: Desire for Interpersonal Attachment as a Fundamental Human Motivation.” Psychological Bulletin 117: 497–529. Baumeister, Roy F., Laura Smart, and Joseph M. Boden. 1996. “Relation of Threatened Egotism to Violence and Aggression: The Dark Side of High Self-Esteem.” Psychological Review 103, no. 1: 5–33. Baumeister, Roy F., and Kathleen D. Vohs. 2007. “Self-regulation, Ego Depletion, and Motivation.” Social and Personality Psychology Compass 1, no. 1: 115–28. Bear, George G., and Gail S. Rys. 1994. “Moral Reasoning, Classroom Behavior, and Sociometric Status among Elementary School Children.” Developmental Psychology 30, no. 5: 633–38. Bechara, A., and A. R. Damasio. 2005. “The Somatic Marker Hypothesis: A Neural Theory of Economic Decision Making.” Games and Economic Behavior 52, no. 2: 336–72.

Social Rationality, Self-Regulation, and Well-Being

101

Berlyne, D. E. 1960. Conflict, Arousal, and Curiosity. New York: McGraw-Hill. Bernheim, B. D., and A. Rangel. 2004. “Addiction and Cue-triggered Decision Processes.” American Economic Review 94: 1558–90. Blair, C., and R. P. Razza. 2007. “Relating Effortful Control, Executive Function, and False Belief Understanding to Emerging Math and Literary Ability in Kindergarten.” Child Development 78, no. 2: 647–63. Blum, D. 2002. Love at Goon Park: Harry Harlow and the Science of Affection. New York: Berkeley Publishing (Penguin). Bosma, Harke A., and E. Saskia Kunnen. 2001. “Determinants and Mechanisms in Ego Identity Development: A Review and Synthesis.” Developmental Review 21, no. 1: 39–66. Bowlby, J., and M. Fry. 1953. Child Care and the Growth of Love. London: Pelican Brady, S. S., and K. A. Matthews. 2002. “The Influence of Socioeconomic Status and Ethnicity on Adolescents’ Exposure to Stressful Life Events.” Journal of Pediatric Psychology 27: 575–83. Brook, A. T., J. Garcia, and M. Fleming. 2008. “The Effects of Multiple Identities on Psychological Well-being.” Personality and Social Psychology Bulletin 34, no. 12: 1588– 600. Buckley, Katherine E., Rachel E. Winkel, and Mark R. Leary. 2004. “Reactions to Acceptance and Rejection: Effects of Level and Sequence of Relational Evaluation.” Journal of Experimental Social Psychology 40, no. 1: 14–28. Bunderson, J. S., and R. E. Reagans. 2011.“Power, Status, and Learning in Organizations.” Organization Science 22: 1182–94. Bunderson, J. S., and J. A.Thompson. 2009. “The Call of the Wild: Zookeepers, Callings, and the Double-edged Sword of Deeply Meaningful Work.” Administrative Science Quarterly 54: 32–57. Buss, D. M. 1994. The Evolution of Desire: Strategies of Human Mating. New York: Basic Books. Cacioppo, J. T., G. G. Berntson, J. F. Sheridan, and M. K. McClintock. 2000. “Multilevel Integrative Analyses of Human Behavior: Social Neuroscience and the Complementing Nature of Social and Biological Approaches.” Psychological Bulletin 126, no. 6: 829–43. Cacioppo, John T., Louise C. Hawkley, John M. Ernst, Mary Burleson, Gary G. Berntson, Bita Nouriani, and David Spiegel. 2006. “Loneliness within a Nomological Net: An Evolutionary Perspective.” Journal of Research in Personality 40, no. 6: 1054–85. Calkins, Susan D. 2004. “Early Attachment Processes and the Development of Emotional Self-regulation.” In Handbook of Self-regulation: Research,Theory, and Applications, edited by Roy F. Baumeister and Kathleen D.Vohs, 324–39. New York: Guilford Press. Camerer, C. F. 2003. Behavioral Game Theory. New York: Russell Sage. Campbell, J. D., P. D. Trapnell, S. J. Heine, I. M. Katz, L. F. Lavallee, and D. R. Lehman. 1996. “Self-concept Clarity: Measurement, Personality Correlates, and Cultural Boundaries.” Journal of Personality and Social Psychology 70, no. 1: 141–56. Cancrinus-Matthijsse, A. M., S. Lindenberg, A. Bakker, and P. P. Groenewegen. 1996. “The Quality of Professional Practice of Community Pharmacists: What Can Still Be Improved in Europe.” Pharmacy World and Science 18, no. 6: 217–28. Carstensen, Laura L. 2006.“The Influence of a Sense of Time on Human Development.” Science 312, no. 5782: 1913–15. Carver, C. S., and M. F. Scheier. 1998. On the Self-regulation of Behavior. Cambridge: Cambridge University Press. Caruso, C. C., E. M. Hitchcock, R. B. Dick, J. M. Russo, and J. M. Schmit. 2004. Overtime and Extended Work Shifts: Recent Findings on Illnesses, Injuries and Health Behaviors. Cincinnati, OH: National Institute for Occupational Safety and Health (NIOSH).

102 Siegwart Lindenberg Caspi, A., G. H. Elder, Jr., and D. J. Bem. 1987. “Moving against the World: Life-course Patterns of Explosive Children.” Developmental Psychology 23, no. 2: 308–13. Chen, Serena, Helen C. Boucher, and Molly Parker Tapias. 2006. “The Relational Self Revealed: Integrative Conceptualization and Implications for Interpersonal Life.” Psychological Bulletin 132, no. 2: 151–79. Chirkov, Valery, Richard M. Ryan, Youngmee Kim, and Ulas Kaplan. 2003. “Differentiating Autonomy from Individualism and Independence: A Selfdetermination Theory Perspective on Internalization of Cultural Orientations and Well-being.” Journal of Personality and Social Psychology 84, no. 1: 97–110. Christakis, Nicholas A., and James H. Fowler. 2007. “The Spread of Obesity in a Large Social Network over 32 Years.” New England Journal of Medicine 357, no. 4: 370–79. Christoff, Kalina, Diego Cosmelli, Dorothée Legrand, and Evan Thompson. 2011a. “Specifying the Self for Cognitive Neuroscience.” Trends in Cognitive Sciences 15, no. 3: 104–12. ———. 2011b. “Clarifying the Self: Response to Northoff.” Trends in Cognitive Sciences 15, no. 5: 187–88. Cialdini, Robert B., Raymond R. Reno, and Carl R. Kallgren. 1990. “A Focus Theory of Normative Conduct: Recycling the Concept of Norms to Reduce Littering in Public Places.” Journal of Personality and Social Psychology 58, no. 6: 1015–26. Cohen, G. L. 2003. “Party over Policy: The Dominating Impact of Group Influence on Political Beliefs.” Journal of Personality and Social Psychology 85: 808–22. Cohen, G. L., J. Garcia,V. Purdie-Vaughns, N. Apfel, and P. Brzustoski. 2009. “Recursive Processes in Self-Affirmation: Intervening to Close the Minority Achievement Gap.” Science 324: 400–403. Cohen G. L., and M. J. Prinstein. 2006. “Peer Contagion of Aggression and Health Risk Behavior among Adolescent Males: An Experimental Investigation of Effects on Public Conduct and Private Attitudes.” Child Development 77: 967–83. Cohen, G. L., D. K. Sherman, A. Bastardi, L. Hsu, and M. McGoey. 2007. “Bridging the Partisan Divide: Self-Affirmation Reduces Ideological Closed-Mindedness and Inflexibility in Negotiation.” Journal of Personality and Social Psychology 93: 415–30. Coleman, J. S., and T. Hoffer. 1987. Public and Private High Schools: The Impact of Communities. New York: Basic Books Cooley, Charles Horton. 1922 [1902]. Human Nature and the Social Order. Revised edition by Charles Horton Cooley. New York: Charles Scribner’s Sons. Deci, E. L., and R. M. Ryan. 2000. “The ‘What’ and ‘Why’ of Goal Pursuits: Human Needs and the Self-determination of Behavior.” Psychological Inquiry 11: 227–68. de Hooge, I. E., S. M. Breugelmans, and M. Zeelenberg. 2008. “Not So Ugly after All: When Shame Acts as a Commitment Device.” Journal of Personality and Social Psychology 95: 933–43. de Hooge, I. E., M. Zeelenberg, and S. M. Breugelmans. 2010. “Restore and Protect Motivations following Shame.” Cognition and Emotion 24: 111–27. Dembe, A. E., J. B. Erickson, R. G. Delbos, and S. M. Banks. 2005. “The Impact of Overtime and Long Working Hours on Occupational Injuries and Illnesses: New Evidence from the United States.” Occupational and Environmental Medicine 62: 588– 97. Denham, S. A., K. A. Blair, E. DeMulder, J. Levitas, K. Sawyer, S. Auerbach-Major, and P. Queenan. 2003. “Preschool Emotional Competence: Pathway to Social Competence?” Child Development 74: 238–56. Denk, F., M. E. Walton, K. A. Jennings, T. Sharp, M. F. S. Rushworth, and D. M. Bannerman. 2005. “Differential Involvement of Serotonin and Dopamine Systems in Cost-benefit Decisions about Delay or Effort.” Psychopharmacology 179: 587–96. DeWall, C. N., R. F. Baumeister, M.T. Gailliot, and J. K. Manor. 2008. “Depletion Makes

Social Rationality, Self-Regulation, and Well-Being

103

the Heart Grow Less Helpful: Helping as a Function of Self-regulatory Energy and Genetic Relatedness.” Personality and Social Psychology Bulletin 34: 1653–62. Dias, Rafael M., and Laura E. Berk. 1992. Private Speech: From Social Interaction to SelfRegulation. Mahwah, NJ: Erlbaum. Dohmen, T., and A. Falk. 2011. “Performance Pay and Multi-dimensional Sorting— Productivity, Preferences and Gender.” American Economic Review 101: 556–90. DuBois, D. L., C. Burk-Braxton, L. P. Swenson, H. D.Tevendale, E. M. Lockerd, and B. L. Moran. 2002. “Getting By with a Little Help from Self and Others: Self-esteem and Social Support as Resources during Early Adolescence.” Developmental Psychology 36: 822–39. Dunbar, R. I. M. 2003. “The Social Brain: Mind, Language, and Society in Evolutionary Perspective.” Annual Review of Psychology 32: 163–81. Dunning, D. 2002. “The Zealous Self-affirmer: How and Why the Self Lurks so Pervasively behind Social Judgment.” In Motivated Social Perception: The Ontario Symposium,Vol. 9, edited by S. Fein and S. Spencer, 45–72. Mahwah, NJ: Erlbaum. Dunning, D., A. Leuenberger, and D. A. Sherman. 1995. “A New Look at Motivated Inference: Are Self-Serving Theories of Success a Product of Motivational Forces?” Journal of Personality and Social Psychology 69: 58–68. Durkheim, Emile. 1951 [1897]. Suicide: A Study in Sociology. Translated by John A. Spaulding and George Simpson, edited by George Simpson. Glencoe, IL: Free Press. Eberly, M. B., and R. Montemayor. 1999. “Adolescent Affection and Helpfulness toward Parents: A 2-year Follow-up.” Journal of Early Adolescence 19: 226–48. Eddy, J. M., J. B. Reid, and R. A. Fetrow. 2000. “An Elementary-school Based Prevention Program Targeting Modifiable Antecedents of Youth Delinquency and Violence: Linking the Interests of Families and Teachers (LIFT).” Journal of Emotional and Behavioral Disorders 8: 165–76. Eisenberg, N., and M. J. Sulik. 2012. “Is Self-other Overlap the Key to Understanding Empathy?” Emotion Review 4: 34–35. Elliott, D. S., and P. H. Tolan. 1999. “Youth Violence, Prevention, Intervention, and Social Policy.” In Youth Violence: Prevention, Intervention, and Social Policy, edited by D. J. Flannery and C. R. Huff, 3–46. Washington, DC: American Psychiatric Press. Erikson, E. H. 1964. Insight and Responsibility. New York: Norton. Euler, Harald A., Sabine Hoier, and Percy A. Rohde. 2001. “Relationship-specific Closeness of Intergenerational Family Ties: Findings from Evolutionary Psychology and Implications for Models of Cultural Transmission.” Journal of Cross-Cultural Psychology 32: 147–58. Evans, G. W., and R. Kutcher. 2011. “Loosening the Link between Childhood Poverty and Adolescent Smoking and Obesity: The Protective Effects of Social Capital.” Psychological Science 22: 3–7. Fast, N. J., N. Sivanathan, N. D. Mayer, and A. D. Galinsky. 2012. “Power and Overconfident Decision-making.” Organizational Behavior and Human Decision Processes 117: 249–60. Feldman, R., A. Weller, O. Zagoory-Sharon, and A. Levine. 2007. “Evidence for a Neuroendocrinological Foundation of Human Affiliation.” Psychological Science 18: 965–70. Ferguson, M. J., and J. A. Bargh. 2004. “Liking Is for Doing: The Effects of Goal Pursuit on Automatic Evaluation.” Journal of Personality and Social Psychology 87: 557–72. Fessler, D. M. T. 2004. “Shame in Two Cultures: Implications for Evolutionary Approaches.” Journal of Cognition and Culture 4: 207–62. Fishbach, A., R. S. Friedman, and A.W. Kruglanski. 2003.“Lead Us Not unto Temptation: Momentary Allurements Elicit Overriding Goal Activation.” Journal of Personality and Social Psychology 84: 296–309.

104 Siegwart Lindenberg Fisher, H. E., A. Aron, and L. C. Brown. 2006. “Romantic Love: A Mammalian Brain System for Mate Choice.” Philosophical Transactions of the Royal Society London B Biological Sciences 361: 2173–86. Fitzsimons, Gráinne M., and John A. Bargh. 2003. “Thinking of You: Nonconscious Pursuit of Interpersonal Goals associated with Relationship Partners.” Journal of Personality and Social Psychology 84: 148–64. Ford, M. E. 1992. Motivating Humans: Goals, Emotions, and Personal Agency Beliefs. Hillsdale, NJ: Erlbaum. Ford, M. E., and C.W. Nichols. 1987. “A Taxonomy of Human Goals and Some Possible Applications.” In Humans as Self-Constructing Systems: Putting the Framework to Work, edited by M. E. Ford and D. H. Ford. Hillsdale, NJ: Erlbaum. Frank, Robert H. 1985. Choosing the Right Pond: Human Behavior and the Quest for Status. New York: Oxford University Press. Freeman, Nicholas, and Mark Muraven. 2010. “Self-control Depletion Leads to Increased Risk Taking.” Social Psychological and Personality Science 1, no. 2: 175–81. Frith, C. D., and U. Frith. 1999. “Interacting Minds: A Biological Basis.” Science 286: 1692–95. Galinsky, A. D., Magee, J. C., Inesi, M. E., and Gruenfeld, D. H. 2006. “Power and Perspectives Not Taken.” Psychological Science 17: 1068–1074. Galliani, E. M., and M. Vianello. 2012. “The Emotion and Admiration Improves Employees’ Goal Orientation and Contexual Performance.” International Journal of Applied Psychology 2: 43–52. Gestsdottir, S., and R. M. Lerner. 2007. “Intentional Self-regulation and Positive Youth Development in Early Adolescence: Findings from the H-4 Study on Positive Youth Development.” Developmental Psychology 43: 508–21. Gibbons, F. X., D. J. Lane, M. Gerrard, M. Reis-Bergan, C. L. Lautrup, N. A. Pexa, and H. Blanton. 2002. “Comparison-Level Preferences after Performance: Is Downward Comparison Theory Still Useful?” Journal of Personality and Social Psychology 83: 865– 80. Gigerenzer, Gerd. 2002. “The Adaptive Toolbox.” In Bounded Rationality: The Adaptive Toolbox, edited by G. Gigerenzer and R. Selton, 31–50. Cambridge: MIT Press. Gilbert, D.T., and T. D.Wilson. 2000. “Miswanting: Some Problems in the Forecasting of Future Affective States.” In Thinking and Feeling: The Role of Affect in Social Cognition, edited by J. Forgas, 178–97. Cambridge: Cambridge University Press. Gilbert, P. 2003. “Evolution, Social Roles, and the Differences in Shame and Guilt.” Social Research 70: 1205–30. Gilbert, Paul, and Michael T. McGuire. 1998. “Shame, Social Roles and Status: The Psychobiological Continuum from Monkey to Human.” In Shame: Interpersonal Behavior, Psychopathology and Culture, edited by Paul Gilbert and Bernice Andrews, 99–125. New York: Oxford University Press. Goffman, E. 1959. The Presentation of Self in Everyday Life. Garden City, NY: Doubleday Anchor Books. Goldberg, E. 2009. The New Executive Brain. Oxford: Oxford University Press. Gralinski, H., and C. Kopp. 1993. “Everyday Rules for Behavior: Mothers’ Request to Young Children.” Developmental Psychology 29: 573–84. Gramzow, R. H., and L. Gaertner. 2005. “Self-Esteem and Favoritism toward Novel In-Groups: The Self as an Evaluative Base.” Journal of Personality and Social Psychology 88: 801–15. Greenwald, A. G., and S. J. Breckler. 1985. “To Whom Is the Self Presented?” In The Self and Social Life, edited by B. R. Schlenker, 126–45. New York: McGraw-Hill. Greenwald, A. G., J. E. Pickrell, and S. D. Farnham. 2002. “Implicit Partisanship: Taking Sides for No Reason.” Journal of Personality and Social Psychology 83: 367–79.

Social Rationality, Self-Regulation, and Well-Being

105

Gürerk, Özgür, Bernd Irlenbusch, and Bettina Rockenbach. 2006. “The Competitive Advantage of Sanctioning Institutions.” Science 312: 108–11. Habermas, T., and S. Bluck. 2000. “Getting a Life: The Emergence of the Life Story in Adolescence.” Psychological Bulletin 126: 248–69. Halevy, N., E. Chou, T. R. Cohen, and R. W. Livingston. 2012. “Status Conferral in Intergroup Social Dilemmas: Behavioral Antecedents and Consequences of Prestige and Dominance.” Journal of Personality and Social Psychology 102: 351–366. Halevy, N., E. Chou, and A. D. Galinsky. 2011. “A Functional Model of Hierarchy: Why, How and When Hierarchical Differentiation Enhances Group Performance.” Organizational Psychology Review 1: 32–52. Harlow, H. F. 1958. “The Nature of Love.” American Psychologist 13: 573–685. Hart, D., R. Atkins, and M. K. Matsuba. 2008. “The Association of Neighborhood Poverty with Personality Change in Childhood.” Journal of Personality and Social Psychology 94: 1048–61. Harter, S., and A. Monsour. 1992. “Developmental Analysis of Conflict Caused by Opposing Attributes in the Adolescent Self-portrait.” Developmental Psychology 28: 251–60. Hebb, D. O. 1958. A Textbook of Psychology. Philadelphia: W. B. Saunders. Heiner, R. A. 1983. “The Origins of Predictable Behavior.” American Economic Review 73: 560–95. Hershey, D. A., K. Henkens, and H. P. van Dalen. 2010. “Aging and Financial Planning for Retirement: Interdisciplinary Influences Viewed through a Cross-cultural Lens.” International Journal of Aging and Human Development 70: 1–38. Hirschi, Travis. 1969. Causes of Delinquency. Berkeley: University of California Press. Hirschman, E. C. 1992. “The Consciousness of Addiction: Toward a General Theory of Compulsive Consumption.” Journal of Consumer Research 19: 155–79. Hoch, S. J., and G. F. Loewenstein. 1991. “Time Inconsistent Preferences and Consumer Self-control.” Journal of Consumer Research 18: 492–507. Hrdy, S. B. 2009. Mothers and Others: The Evolutionary Origins of Mutual Understanding. Cambridge: Harvard University Press. Hsee, C. K., and R. Hastie. 2006. “Decision and Experience: Why Don’t We Choose What Makes Us Happy?” Trends in Cognitive Sciences 10: 31–37. Huberman, B. A., C. H. Loch, and A. Önçüler. 2004. “Status as a Valued Resource.” Social Psychology Quarterly 67: 103–14. Huey, S. J., Jr., S. W. Henggeler, M. J. Brondino, and S. G. Pickrel. 2000. “Mechanisms of Change in Multisystemic Therapy: Reducing Delinquent Behavior through Therapist Adherence and Improved Family and Peer Functioning.” Journal of Consulting and Clinical Psychology 68: 451–67. James, W. 1890. Principles of Psychology. New York: Macmillan Jeannerod, M. 2003. “The Mechanism of Self-recognition in Humans.” Behavioural Brain Research 142: 1–15. Jeannerod, M., and T. Anquetil. 2008. “Putting Oneself in the Perspective of the Other: A Framework for Self-other Differentiation.” Social Neuroscience 3: 356–67. Jones, S. C. 1973.“Self and Interpersonal Evaluation: Esteem Theories versus Consistency Theories.” Psychological Bulletin 79: 185–99. Kazdin, A. E., and J. R.Weisz. 1998. “Identifying and Developing Empirically Supported Child and Adolescent Treatment.” Journal of Consulting and Clinical Psychology 66: 19–36. Kazdin, A. E., and M. K. Whitley. 2006. “Comorbidity, Case Complexity, and Effects of Evidence-based Treatment for Children Referred for Disruptive Behavior.” Journal of Consulting and Clinical Psychology 74: 455–67. Keizer, K., S. Lindenberg, and L. Steg. 2008. “The Spreading of Disorder.” Science 322: 1681–85.

106 Siegwart Lindenberg Kivetz, R., and I. Simonson. 2002. “Self-control for the Righteous: Toward a Theory of Precommitment to Indulgence.” Journal of Consumer Research 29: 199–217. Kivetz, R., O. Urminsky, and Y. Zhang. 2006. “The Goal-Gradient Hypothesis Resurrected: Purchase Acceleration, Illusionary Goal Progress, and Customer Retention.” Journal of Marketing Research 43: 39–58. Kochanska, G. 2002. “Committed Compliance, Moral Self, and Internalization: A Mediational Model.” Developmental Psychology 38: 339–51. Köpetz, C., T. Faber, A. Fishbach, and A. Kruglanski. 2011. “Multifinality Constraints Effect: How Goal Multiplicity Narrows the Means Set to a Focal End.” Journal of Personality and Social Psychology 100: 810–26. Kosfeld, M., M. Heinrichs, P. J. Zack, U. Fischbacher, and E. Fehr. 2005. “Oxytocin Increases Trust in Humans.” Nature 435: 673–76. Lammers, J., Gordijn, E. H., and Otten, S. 2008. “Looking through the Eyes of the Powerful: The Effect of Power on Metastereotyping and the Mediating Role of Perspective Taking.” Journal of Experimental Social Psychology 44: 1229–38. Lane, Robert Edwards. 2000. The Loss of Happiness in Market Democracies. New Haven: Yale University Press. Larrick, R. P., R. E. Nisbett, and J. N. Morgan. 1993. “Who Uses the Cost-benefit Rules of Choice? Implications for the Normative Status of Microeconomic Theory.” Economic Behavior and Human Decision Processes 46: 331–47. Leary, M. R. 2007. “Motivational and Emotional Aspects of the Self.” Annual Review of Psychology 58: 317–44. Leary, M. R., and R. F. Baumeister. 2000. “The Nature and Function of Self-esteem: Sociometer Theory.” In Advances in Experimental Social Psychology, Vol. 32, edited by M. P. Zann, 2–51. San Diego: Academic Press. Leary, M. R., E. S. Tambor, S. K. Terdal, and D. L. Downs. 1995. “Self-esteem as an Interpersonal Monitor: The Sociometer Hypothesis.” Journal of Personality and Social Psychology 68: 518–30. Lengua, L. J., and A. C. Long. 2002. “The Role of Emotionality and Self-regulation in the Appraisal-coping Process: Tests of Direct and Moderating Effects.” Journal of Applied Developmental Psychology 23: 471–93. Leotti, L. A., S. S. Iyengar, and K. N. Ochsner. 2010. “Born to Choose: The Origins and Value of the Need for Control.” Trends in Cognitive Sciences 14: 457–63. Lieberman, M. D. 2007. “Social Cognitive Neuroscience: A Review of Core Processes.” Annual Review of Psychology 58: 259–89. Ligthart, P. A. M., and S. Lindenberg. 1994. “Solidarity and Gain Maximization in Economic Transactions: Framing Effects on Selling Prices.” In Ethics and Economic Affairs, edited by A. Lewis and K. E. Wärneryd, 215–30. London: Routledge. Lindenberg, S. 1983. “Utility and Morality.” Kyklos 36: 450–68. ———. 1989. “Choice and Culture: The Behavioral Basis of Cultural Impact on Transactions.” In Social Structure and Culture, edited by H. Haferkamp, 175–200. Berlin: De Gruyter. ———. 1992. “An Extended Theory of Institutions and Contractual Discipline.” Journal of Institutional and Theoretical Economics 148: 125–54. ———. 1996. “Continuities in the Theory of Social Production Functions.” In Verklarende Sociologie: Opstellen voor Reinhard Wippler, edited by H. Ganzeboom and S. Lindenberg, 169–84. Amsterdam: Thesis. ———. 2000. “The Extension of Rationality: Framing versus Cognitive Rationality.” In L’Acteur et ses Raisons. Mélanges en l’honneur de Raymond Boudon, edited by J. Baechler, F. Chazel, and R. Kamrane, 168–204. Paris: Presses Universitaires de France (PUF). ———. 2001a. “Social Rationality versus Rational Egoism.” In Handbook of Sociological Theory, edited by J. Turner, 635–68. New York: Kluwer Academic/Plenum.

Social Rationality, Self-Regulation, and Well-Being

107

———. 2001b. “Intrinsic Motivation in a New Light.” Kyklos 54: 317–42. ———. 2006. “Prosocial Behavior, Solidarity, and Framing Processes.” In Solidarity and Prosocial Behavior: An Integration of Sociological and Psychological Perspectives, edited by D. Fetchenhauer, A. Flache, A. P. Buunk, and S. Lindenberg, 23–44. New York: Springer. ———. 2008. “Social Rationality, Semi-Modularity and Goal-Framing: What Is It All About?” Analyse & Kritik 30: 669–87. Lindenberg, S., and B. S. Frey. 1993. “Alternatives, Frames, and Relative Prices: A Broader View of Rational Choice.” Acta Sociologica 36: 191–205. Lindenberg, S., and L. Steg. 2007. “Normative, Gain and Hedonic Goal Frames Guiding Environmental Behavior.” Journal of Social Issues 65: 117–37. Lipsey, M. W., and D. B. Wilson. 1998. “Effective Intervention for Serious Juvenile Offenders: A Synthesis of Research.” In Serious and Violent Juvenile Offenders: Risk Factors and Successful Interventions, edited by R. Loeber and D. P. Farrington, 313–45. Thousand Oaks, CA: Sage Publications. Lockwood, P., C. H. Jordan, and Z. Kunda. 2002. “Motivation by Positive or Negative Role Models: Regulatory Focus Determines Who Will Best Inspire Us.” Journal of Personality and Social Psychology 83: 854–64. Luyckx, K., M.Vansteenkiste, L. Goossens, and B. Duriez. 2009. “Basic Need Satisfaction and Identity Formation: Bridging Self-Determination Theory and Process-Oriented Identity Research.” Journal of Counseling Psychology 56: 276–88. Lyubomirsky, S., L. King, and E. Diener. 2005. “The Benefits of Frequent Positive Affect: Does Happiness Lead to Success?” Psychological Bulletin 131: 803–55. McAdams, D. P., B. J. Hoffman, E. D. Mansfield, and R. Day. 1996. “Themes of Agency and Communion in Significant Autobiographical Scenes.” Journal of Personality 63: 339–77. McClure, S. M., D. I. Laibson, G. Loewenstein, and J. D. Cohen. 2004. “Separate Neural Systems Value Immediate and Delayed Monetary Rewards.” Science 306, no. 5695: 503–7. Macdonald, K. 1992. “Warmth as a Developmental Construct: An Evolutionary Analysis.” Child Development 63: 753–73. Malle, B. F., and S. D. Hodges, eds. 2005. Other Minds: How Humans Bridge the Divide between Self and Others. New York: Guilford Press Maner, J. K., C. N. DeWall, M. Schaller, and R. F. Baumeister. 2007. “Does Social Exclusion Motivate Interpersonal Reconnection? Resolving the ‘Porcupine Problem.’” Journal of Personality and Social Psychology 92: 42–55. Markus, Hazel. 1977. “Self-schemata and Processing Information about the Self.” Journal of Personality and Social Psychology 35: 63–78. Markus, H., and S. Cross. 1990. “The Interpersonal Self.” In Handbook of Personality, edited by L. A. Pervin, 576–608. New York: Guilford Press. Markus, H., and P. Nurius. 1986. “Possible Selves.” American Psychologist 41: 954–69. Maslow, A. 1943. “A Theory of Human Motivation.” Psychological Review 50: 370–96. ———. 1971. The Farther Reaches of Human Nature. New York:Viking Press. Mead, G. H. 1934. Mind, Self and Society. Chicago: University of Chicago Press. Meeus,W., A. Oosterwegel, and W.Vollebergh. 2002. “Parental and Peer Attachment and Identity Development in Adolescence.” Journal of Adolescence 25: 93–106. Mendes,W. B., J. Blascovich, S. Hunter, B. Lickel, and J.T. Jost. 2007. “Threatened by the Unexpected: Physiological Responses during Social Interactions with Expectancyviolating Partners.” Journal of Personality and Social Psychology 92: 698–716. Mendez, M. F. 2009. “The Neurobiology of Moral Behavior: Review and Neuropsychiatric Implications.” CNS Spectrums 14: 608–20. Mischel, W., and E. B. Ebbesen. 1970. “Attention in Delay of Gratification.” Journal of Personality and Social Psychology 16: 329, 337.

108 Siegwart Lindenberg Moffitt, Terrie E., Louise Arseneault, Daniel Belsky, Nigel Dickson, Robert J. Hancox, HonaLee Harrington, Renate Houts, Richie Poulton, Brent W. Roberts, Stephen Ross, Malcolm R. Sears, W. Murray Thomson, and Avshalom Caspia. 2011. “A Gradient of Childhood Self-control Predicts Health, Wealth, and Public Safety.” PNAS 108, no. 7: 2693–98. Moll, J., R. Zahn, R. de Oliveira-Souza, F. Krueger, and J. Grafman. 2005. “The Neural Basis of Human Moral Cognition.” Nature Reviews Neuroscience 6: 799–809. Mouratidis, A., W. Lens, and M. Vansteenkiste. 2010. “How You Provide Corrective Feedback Makes a Difference: The Motivating Role of Communicating in an Autonomy-supporting Way.” Journal of Sport and Exercise Psychology 32: 619–37. Mulatu, M. S., and C. Schooler. 2002. “Causal Connections between Socio-economic Status and Health: Reciprocal Effects and Mediating Mechanisms.” Journal of Health and Social Behavior 43: 22–41. Murray, S. L., G. M. Bellavia, P. Rose, and D.W. Griffin. 2003. “Once Hurt Twice Hurtful: How Perceived Regard Regulates Daily Marital Interactions.” Journal of Personality and Social Psychology 84: 126–47. Nenkov, G.Y., J. J. Inman, and Hulland. 2008.“Considering the Future: Conceptualization and Measurement of Elaboration on Potential Outcomes.” Journal of Consumer Research 35: 126–41. Neuberg, S. L., and J. T. Newsom. 1993. “Personal Need for Structure: Individual Differences in the Desire for Simpler Structure.” Journal of Personality and Social Psychology 65: 113–31. Neumann, R., and F. Strack. 2000. “‘Mood Contagion’: The Automatic Transfer of Mood between Persons.” Journal of Personality and Social Psychology 79, no. 2: 211– 23. Nieboer, A., and S. Lindenberg. 2002. “Substitution, Buffers and Subjective Well-being: A Hierarchical Approach.” In The Universality of Subjective Well-Being Indicators, edited by E. Gullone and R. A. Cummins, 175–89. Social Indicators Research Book Series, Vol. 16. Dordrecht: Kluwer Academic Publishers. Nieboer, A., S. Lindenberg, A. Boomsma, and A. C. van Bruggen. 2005. “Dimensions of Well-being and Their Measurement: The SPF-IL Scale.” Social Indicators Research 73: 313–53. Nieboer, A. P., S. Lindenberg, and J. Ormel. 1998/1999. “Conjugal Bereavement and Wellbeing of Elderly Men and Women: A Preliminary Study.” Omega 38: 113–41. Niv, J. 2007. “Cost, Benefit, Tonic, Phasic: What Do Response Rates Tell Us about Dopamine and Motivation?” Annals of the New York Academy of Sciences 1104: 357– 76. Nowak, Andrzej, Robin R. Vallacher, Abraham Tesser, and Wojciech Borkowski. 2000. “Society of Self: The Emergence of Collective Properties in Self-structure.” Psychological Review 107, no. 1: 39–61. Ormel, J., S. Lindenberg, N. Steverink, and L. M. Verbrugge. 1999. “Subjective Well Being and Social Production Functions.” Social Indicator Research 46: 61–90. OSA. 2003. OSA-arbeidsaanbod panel 1985–2000. Retrieved from http: //pi0736/kub. nl. Parker, A. M., and B. Fishhoff. 2005. “Decision-making Competence: External Validation through an Individual-difference Approach.” Journal of Behavioral Decision-Making 18: 1–27. Pass, J., S. Lindenberg, and J. H. Park. 2010. “All You Need Is Love: Is the Sociometer Especially Sensitive to One’s Mating Capacity?” European Journal of Social Psychology 40: 221–34 Patterson, F. R., M. S. Forgatch, K. L.Yoerger, and M. Stoolmiller. 1998. “Variables that Initiate and Maintain an Early-onset Trajectory for Juvenile Offending.” Development and Psychopathology 10: 531–47.

Social Rationality, Self-Regulation, and Well-Being

109

Pendell, S. D. 2002. “Affection in Interpersonal Relationships: Not Just ‘A Fond or Tender Feeling.’” Communication Yearbook, Vol. 26 (Mahwah, NJ: Lawrence Earlbaum), 70–115. Pillutla, M. M., and X. P. Chen. 1999. “Social Norms and Cooperation in Social Dilemmas:The Effects of Context and Feedback.” Organizational Behavior and Human Decision Processes 78: 81–103. Pinderhughes, E. E., K. A. Dodge, J. E. Bates, and G. S. Petit. 2000. “Discipline Responses: Influences of Parents’ Socioeconomic Status, Ethnicity, Beliefs about Parenting, Stress, and Cognitive-emotional Processes.” Journal of Family Psychology 14: 380–400. Prenda, K. M., and M. E. Lachman. 2001. “Planning for the Future: A Life Management Strategy for Increasing Control and Life Satisfaction in Adulthood.” Psychology and Aging 16: 206–16. Proulx, T., S. J. Heine, and K. D. Vohs. 2010. “When Is the Unfamiliar the Uncanny? Meaning Affirmation after Exposure to Absurdist Literature, Humor, and Art.” Personality and Social Psychology Bulletin 36: 817–29. Raikes, H. A., and R. A.Thompson. 2005. “Link between Risk and Attachment Security: Models of Influence.” Journal of Applied Developmental Psychology 26: 440–55. Ridgeway, C. L., E. H. Boyle, K. Kuipers, and D. T. Robinson. 1998. “How Do Status Beliefs Develop? The Role of Resources and Interactional Experience.” American Sociological Review 63: 331–50. Rozin, P. 1999. “The Process of Moralization.” Psychological Sciences 10: 218–21. Russell, A. 1997. “Individual and Family Factors Contributing to Mothers’ and Fathers’ Positive Parenting.” International Journal of Behavioral Development 21: 111–32. Sampson, R. J., J. H. Laub, and C. Wimer. 2006. “Does Marriage Reduce Crime? A Counterfactual Approach to Within-individual Causal Effects.” Criminology 44: 465– 508. Schlenker, B. R. 2003. “Self-presentation.” In Handbook of Self and Identity, edited by M. R. Leary and J. P. Tangney, 492–518. New York: Guilford. Schultz, D., C. E. Izard, Brian P. Ackerman, and E. A. Youngstrom. 2001. “Emotion Knowledge in Economically Disadvantaged Children: Self-regulatory Antecedents and Relations to Social Difficulties and Withdrawal.” Development and Psychopathology 13: 53–67. Schultz, W. 1958. FIRO: A Three-dimensional Theory of Interpersonal Behavior. New York: Rinehart. Scitovsky, T. 1976. The Joyless Economy. Oxford: Oxford University Press. Sentse, M., J. K. Dijkstra, S. Lindenberg, J. Ormel, and R.Veenstra. 2010. “The Delicate Balance between Parental Protection, Unsupervised Wandering, and Adolescents’ Autonomy and Its Relation with Antisocial Behavior. The TRAILS Study.” International Journal of Behavioral Development 34: 159–67. Shah, J. 2003a. “Automatic for the People: How Representations of Significant Others Implicitly Affect Goal Pursuit.” Journal of Personality and Social Psychology 84: 661– 81. ———. 2003b. “The Motivational Looking Glass: How Significant Others Implicitly Affect Goal Appraisals.” Journal of Personality and Social Psychology 85: 424–39. Shaw, D., E. Owens, J. Giovannelli, and E. Winslow. 2001. “Infant and Toddler Pathways Leading to Early Externalizing Disorders.” Journal of the American Academy of Child and Adolescent Psychiatry 40: 36–43. Sheldon, K. M. 2004. “The Benefits of a ‘Sidelong’ Approach to Self-esteem Need Satisfaction: Comment on Crocker and Park.” Psychological Bulletin 130: 421–23. Sheldon, K. M., and T. Kasser. 1995. “Coherence and Congruence: Two Aspects of Personality Integration.” Journal of Personality and Social Psychology 68: 531–43. Sheldon, K. M., and C. Niemiec. 2006. “It’s Not Just the Amount That Counts: Balanced

110 Siegwart Lindenberg Need Satisfaction Also Affects Well-Being.” Journal of Personality and Social Psychology 91: 331–41. Sheldon, K. M., R. M. Ryan, L. J. Rawsthorne, and B. Ilardi. 1997. “Trait Self and True Self: Cross-role Variation in the Big-Five Personality Traits and Its Relations with Psychological Authenticity and Subjective Well-being.” Journal of Personality and Social Psychology 73: 1380–93. Shelly, R. 2001. “How Performance Expectations Arise from Sentiments.” Social Psychology Quarterly 64: 72–87. Sherman, D. K., and G. L. Cohen. 2006. “The Psychology of Self-defense: Selfaffirmation Theory.” In Advances in Experimental Social Psychology, Vol. 38, edited by M. P. Zanna, 183–242. San Diego: Academic Press. Sherman, L.W., D. C. Gottfredson, D. L. MacKenzie, J. Eck, P. Reuter, and S. D. Bushway. 1997. Preventing Crime: What Works, What Doesn’t, What’s Promising: A Report to the United States Congress (NCJ 171676). Washington, DC: U.S. Department of Justice, Office of Justice Programs. Sijtsema, J. J., R. Veenstra, S. Lindenberg, and C. Salmivalli. 2009. “Empirical Test of Bullies’ Status Goals: Assessing Direct Goals, Aggression, and Prestige.” Aggressive Behavior 35: 57–67. Online 2008: DOI: 10.1002/ab.20282. Sijtsema, J. J., R. Veenstra, S. Lindenberg, A. M. van Roon, F. C. Verhulst, J. Ormel, and H. Riese. 2010. “Mediation of Sensation Seeking and Behavioral Inhibition on the Longitudinal Relationship between Heart Rate and Antisocial Behavior: The TRAILS Study.” Journal of the American Academy of Child and Adolescent Psychiatry 49: 493–502. Snyder, M., and O. Klein. 2005. “Construing and Constructing Others: On the Reality and the Generality of the Behavioral Confirmation Scenario.” Interaction Studies 6: 53–67. Steele, C. M. 1988. “The Psychology of Self-affirmation: Sustaining the Integrity of the Self.” In Advances in Experimental Social Psychology, Vol. 21, edited by L. Berkowitz, 261–302. New York: Academic Press. Steverink, N., and S. Lindenberg. 2006. “Which Social Needs Are Important for Subjective Well-being? What Happens to Them with Aging?” Psychology and Aging 21: 281–90. ———. 2008. “Do Good Self-managers Have Less Physical and Social Deficits and More Well-being in Later Life?” European Journal of Ageing 5: 181–90. Steverink, N., S. Lindenberg, and J. Ormel. 1998. “Towards Understanding Successful Ageing: Patterned Change in Resources and Goals.” Ageing and Society 18: 441–67. Steverink N., S. Lindenberg, and J. P. J. Slaets. 2005. “How to Understand and Improve Older People’s Self-management of Wellbeing.” European Journal of Ageing 2: 235– 44. Stigler, G., and G. S. Becker. 1977. “De Gustibus Non Est Disputandum.” American Economic Review 67: 76–90. Stinson, D. A., J.V. Wood, and J. R. Doxey. 2008. “In Search of Clarity: Self-esteem and Domains of Confidence and Confusion.” Personality and Social Psychology Bulletin 34: 1541–55. Stutzer, A., and B. S. Frey. 2007. “What Happiness Research Can Tell Us about Selfcontrol Problems and Utility Misprediction.” In Economics and Psychology: A Promising New Cross-Disciplinary Field, edited by B. S. Frey and A. Stutzer, 169–95. Cambridge: MIT Press. Syed, M. 2010. “Developing an Integrated Self: Academic and Ethnic Identities among Ethnically Diverse College Students.” Developmental Psychology 46: 1590–604. Tangney, J. P., and R. L. Dearing. 2002. Shame and Guilt. New York: Guilford Press. Taylor, S. E., J. S. Lerner, D. K. Sherman, R. M. Sage, and N. K. McDowell. 2003.

Social Rationality, Self-Regulation, and Well-Being

111

“Are Self-enhancing Cognitions Associated with Healthy or Unhealthy Biological Profiles?” Journal of Personality and Social Psychology 85: 605–15. Tesser, A. 1988. “Towards a Self-evaluative Maintenance Model of Social Behavior.” In Advances in Experimental Social Psychology, Vol. 21, edited by L. Berkowitz, 181–227. Orlando, FL: Academic Press. Thoits, P. 1991. “On Merging Identity Theory and Stress Research.” Social Psychology Quarterly 54: 101–12. Tiedens, L. Z., P. C. Ellsworth, and B. Mesquita. 2000. “Sentimental Stereotypes: Emotional Expectations for High- and Low-status Group Members.” Personality and Social Psychology Bulletin 26: 560–75. Tiffany, S. T., and C. A. Conklin. 2000. “A Cognitive Processing Model of Alcohol Craving and Compulsive Alcohol Use.” Addiction 95: 145–54. Tomasello, M., and M. Carpenter. 2007. “Shared Intentionality.” Developmental Science 10: 121–25. Turner, J. C. 1985. “Social Categorization and the Self-concept: A Social Cognitive Theory of Group Behavior.” In Advances in Group Processes: Theory and Research, Vol. 2, edited by E. J. Lawler, 77–121. Greenwich, CT: JAI Press. Twenge, J. M., R. F. Baumeister, C. N. DeWall, N. J. Ciarocco, and J. M. Bartels. 2007. “Social Exclusion Decreases Prosocial Behavior.” Journal of Personality and Social Psychology 92: 56–66. Twenge, J. M., R. F. Baumeister, D. M. Tice, and T. S. Stucke. 2001. “If You Can’t Join Them, Beat Them: Effects of Social Exclusion on Aggressive Behavior.” Journal of Personality and Social Psychology 81: 1058–69. Twenge, J. M., K. R. Catanese, and R. F. Baumeister. 2003. “Social Exclusion and the Deconstructed State: Time Perception, Meaninglessness, Lethargy, Lack of Emotion, and Self-awareness.” Journal of Personality and Social Psychology 85: 409–23. Uchino, B. N., J.T. Cacioppo, and J. K. Kiecolt-Glaser. 1996. “The Relationship between Social Support and Physiological Processes: A Review with Emphasis on Underlying Mechanisms and Implications for Health.” Psychological Bulletin 119: 488–531. Uddin, L. Q., M. Iacoboni, C. Lange, and J. P. Keenan. 2007. “The Self and Social Cognition: The Role of Cortical Midline Structures and Mirror Neurons.” Trends in Cognitive Sciences 11: 153–57. Usborne, E., and D. M. Taylor. 2010. “The Role of Cultural Identity Clarity for Selfconcept Clarity, Self-esteem, and Subjective Well-being.” Personality and Social Psychology Bulletin 36: 883–97. van Echtelt, E. P., A. Glebbeek, S. Lewis, and S. Lindenberg, 2009. “Post-Fordist Work: A Man’s World?: Gender and Working Overtime in the Netherlands.” Gender and Society 23: 188–214. van Echtelt, E. P., A. C. Glebbeek, and S. Lindenberg. 2006. “The New Lumpiness of Work: Explaining the Mismatch between Actual and Preferred Working Hours.” Work, Employment and Society 20: 493–512. Van Kleef, G. A., A. C. Homan, C. Finkenauer, S. Gündemir, and E. Stamou. 2011. “Breaking the Rules to Rise to Power: How Norm Violators Gain Power in the Eyes of Others.” Social Psychological and Personality Science 2: 500–507. van Vugt, M. 2006. “Evolutionary Origins of Leadership and Followership.” Personality and Social Psychology Review 10: 354–71. Vygotsky, L. 1978. Mind in Society:The Development of Higher Mental Processes. Cambridge: Harvard University Press. Weber, M. 1961. General Economic History. New York: Collier Books. Weisfeld, G. E. 1999. Evolutionary Principles of Human Adolescence. New York: Basic Books. White, R. W. 1959. “Motivation Reconsidered: The Concept of Competence.” Psychological Review 66: 297–333.

112 Siegwart Lindenberg Williams, E. F., and T. Gilovich. 2007. “Do People Really Believe They Are Above Average?” Journal of Experimental Social Psychology 44: 1121–28. Williamson, O. E. 1985. The Economic Institutions of Capitalism. New York: Free Press. Wilson, D. S. 2006.“Human Groups as Adaptive Units:Towards a Permanent Consensus.” In The Innate Mind, Vol. 2, edited by P. Carruthers, S. Laurence, and S. Stich, 78–90. Oxford: Oxford University Press. Wippler, R. 1990. “Cultural Resources and Participation in High Culture.” In Social Institutions:Their Emergence, Maintenance, and Effects, edited by M. Hechter, K.-D. Opp, and R. Wippler, 187–204. New York: Aldine de Gruyter. Wynn, K. 2007. “Some Innate Foundations of Social and Moral Cognition.” In The Innate Mind, Vol. 3, edited by P. Carruthers, S. Laurence, and S. Stich, 330–47. Oxford: Oxford University Press.

chapter

Rational Choice Research on Social Dilemmas: Embeddedness Effects on Trust

3

vincent buskens and werner raub

Introduction social dilemmas by example: trust in economic exchange Consider economic exchange through the Internet. July 18, 2007, was the end date on which to purchase a copy of the first edition of Theory of Games and Economic Behavior by John von Neumann and Oskar Morgenstern at eBay from the seller “bibliomonster” for U.S.$1,900.00. The item had a fixed price listing (using eBay’s “Buy It Now” option) and could be purchased only without bidding in an auction. Assume that the seller did own the copy, that the description of the copy was accurate (“Bound in original publishers red cloth a bit rubbed at head of spine. Black (ink?) mark on top board. Minor shelf wear, else very good. Internally, clean and free of ink, marginalia and soiling. No dog-eared pages or tears. Includes the often missing corrigenda leaf. A nice, collectable copy.”), and that the accompanying photos were not misleading. Assume that a buyer existed who would have preferred purchasing the copy as described for the price mentioned. The buyer had to pay the price before the seller would ship the book. Thus, the buyer had to trust the seller that the copy would indeed be shipped. If the buyer decided not to pay, there would be no transaction. If the buyer decided to pay, the seller had to decide whether or not to ship the copy. As a benchmark scenario, imagine an “isolated encounter”—that is, a oneshot transaction in the sense that buyer and seller of the book have never done business with each other before, do not expect to do business with each other in the future, and that verifying the seller’s identity is prohibitively costly for the buyer. For the benchmark, imagine further that eBay would not maintain its feedback forum that allows buyers to evaluate sellers, with evaluations being publicly available and easy to access. As a second and less artificial scenario for buying antiquarian books, consider that websites of antiquarian booksellers typically offer indications of their identity, such as information on the physical location of their shop. Imagine, too, that the buyer has purchased antiquarian books from the seller in the past, and the seller expects that the buyer may also

114 Vincent Buskens and Werner Raub purchase such books in the future. Finally, consider a third scenario that takes a core feature of the eBay platform into account—namely, eBay’s feedback forum, which provides information on the seller from other buyers. On July 16, 2007, the seller of the first edition of Theory of Games and Economic Behavior had positive feedback from 386 other eBay members and negative feedback from 5 members, resulting in an eBay feedback score of 381, with 98.7 percent positive feedback. trust as a social dilemma Our example of the purchase of a first edition of Theory of Games and Economic Behavior represents a trust problem between the buyer and the seller in Coleman’s sense (1990: 97−99), with the buyer in the role of the trustor and the seller in the role of the trustee. Coleman emphasizes four points that characterize a trust problem: (1) Placing trust by the trustor allows the trustee to honor or abuse trust, while this alternative is not available for the trustee without placement of trust. In the example, if the buyer decides to buy, the seller can ship the copy of the first edition or can abstain from shipping. In addition, while Coleman does not mention this explicitly, it is important that the trustee has not only an opportunity but also an incentive to abuse trust. For example, the seller could change his or her virtual identity and offer the book once again through the Internet. (2) Compared to the situation with no trust placed, the trustor is better off if trust is placed and honored but is worse off if trust is abused. The buyer prefers to purchase the first edition to, say, owning only the 1980 Princeton paperback edition, while owning the paperback only is most likely still preferred by the buyer to paying U.S.$1,900.00 without receiving the first edition. Again in addition to Coleman’s characterization, the trustee is better off if trust is honored than if no trust is placed. Selling the book for U.S.$1,900.00 is profitable for the seller. (3) There is no “real commitment” (ibid.: 98) of the trustee to honor trust. Thus the trustor voluntarily places resources in the hands of the trustee. In the benchmark scenario, since the buyer cannot verify the identity of the seller, the buyer is not able to enforce shipment of the copy after payment. (4) There is a time lag between placement of trust by the trustor and the action of the trustee. The buyer first pays the price and, subsequently, the seller decides on whether or not to ship the book. In the benchmark scenario resembling a one-shot transaction, it seems intuitive that incentive-guided and goal-directed behavior of trustor and trustee implies that the trustee would indeed abuse trust, if trust is placed. Assuming that the trustor anticipates this, the trustor does not place trust in the first place. If trust is not placed, however, both trustor and trustee are worse off than when trust is placed and honored. Technically speaking, the no trust outcome is Pareto-suboptimal. As Rapoport (1974) aptly put it, individual rationality in the sense of incentive-guided and goal-directed behavior can lead to collective irrationality in the sense of Pareto-suboptimality. Such a “conflict” between individual and collective rationality is the core feature of a social dilemma, and trust relations are a paradigmatic example of a social dilemma involving two actors. While “social dilemma” is a label commonly used in social psychology and also sociology, such a situation is often referred to as a “problem of collective

Rational Choice Research on Social Dilemmas

115

action” or the “tragedy of the commons” in political science and as a “public goods problem” in economics (see Ledyard 1995: 122). Social dilemmas are intimately related to the problem of order in Parsons’s sense (1937). After all, in Hobbes’s “naturall condition of mankind” (1991 [1651]: ch. 13), actors are interdependent in a world of scarcity, while binding and externally enforced contracts are unfeasible. They may thus end up in the “warre of every man against every man.” In that situation, the life of man is “solitary, poore, nasty, brutish, and short,” and everybody is worse off than in a peaceful situation. This is a social dilemma among many actors. Parsons (1937: 89−94) posed the challenge to specify conditions such that individually rational actors solve the problem of order. He thus referred to the problem of order as “the most fundamental empirical difficulty of utilitarian thought” (ibid.: 91). In his meanwhile classic early contribution to rational choice social research, Coleman (1964: 166−67) clearly realized the challenge and formulated it even more radically than Parsons: “Hobbes took as problematic what most contemporary sociologists take as given: that a society can exist at all, despite the fact that individuals are born into it wholly self-concerned, and in fact remain largely self-concerned throughout their existence. Instead, sociologists have characteristically taken as their starting point a social system in which norms exist, and individuals are largely governed by those norms. Such a strategy views norms as the governors of social behavior, and thus neatly bypasses the difficult problem that Hobbes posed . . . . I will proceed in precisely the opposite fashion . . . . I will make an opposite error, but one which may prove more fruitful . . . . I will start with an image of man as wholly free: unsocialized, entirely self-interested, not constrained by norms of a system, but only rationally calculating to further his own self interest.” While it is part of the sociological folklore that Parsons’s challenge focuses on how rational choice social research can cope with social dilemmas, it is less well appreciated that Durkheim put forward a similar argument in his analysis of the division of labor in society (1973 [1893]: bk. I, ch. 7) that relates to the antiquarian book example. Durkheim’s point is that economic transactions often deviate from what is conventionally assumed in standard neoclassical models of spot exchange on perfect markets. Durkheim argued that the governance of transactions exclusively via bilateral contracts requires that the present and future rights and obligations of the partners involved in the transaction are specified explicitly for all circumstances and contingencies that might arise during and after the transaction. Anticipating much of the modern economic and game-theoretic literature on incomplete and implicit contracts, Durkheim pointed out that such purely contractual governance of economic transactions is problematic: typically, many unforeseen or unforeseeable contingencies could or actually do arise during or after a transaction. Negotiating a contract explicitly covering all these contingencies would be unfeasible or at least prohibitively costly. Likewise, renegotiations in the case that contingencies arise are also costly (for similar arguments on the limits of contractual governance, see Weber 1976 [1921]: 409 in his sociology of law). Such renegotiations characteristically offer incentives for opportunistic behavior, since an unexpected contingency will often strengthen the bargaining position of one partner while weakening the position of the other. Hence Durkheim argues that mutually beneficial

116 Vincent Buskens and Werner Raub economic exchange presupposes the solution of a trust problem and thus involves a social dilemma. focus of the chapter Game-theoretic models have emerged as a tool for the analysis of social dilemmas in rational choice social research. This is not accidental. Interdependence between actors is a core feature of a social dilemma. For example, the behavior of the trustor has effects for the trustee, and vice versa. Game theory is the branch of rational choice theory that models interdependent situations, providing concepts, assumptions, and theorems that allow specifying how rational actors behave in such situations. The theory assumes that actors behave as if they try to realize their preferences in decision situations with restrictions, taking their interdependencies as well as rational behavior of the other actors into account (see, for example, Harsanyi 1975: 89−117). It is therefore natural that applications of game-theoretic models figure prominently in rational choice social research on social dilemmas. Moreover, one should observe that interdependencies between actors and actors taking their interdependencies into account are likewise the core of Weber’s famous definition of social action (1947: 88, emphasis added): “Sociology . . . is a science which attempts the interpretive understanding of social action in order thereby to arrive at a causal explanation of its course and effects . . . . Action is social in so far as . . . it takes account of the behaviour of others and is thereby oriented in its course.”This is a reason why social dilemmas are a strategic research site not only for rational choice social research but also for sociology in general, and why game theory is an important tool, too, for sociological analyses in the spirit of Weber. Reviews of social dilemma research are readily available that highlight how social psychological theory and other approaches with a firm basis in methodological individualism can be used in this field that differ from rational choice assumptions (for example, Kollock 1998). Somewhat suprisingly, though, there is no systematic review of applications of game theory in this field with a focus on how sociologically informed hypotheses can be derived from these models, and how these hypotheses fare in empirical research. This chapter contributes to filling that gap. We do so by reconstructing research strategies that are often employed in applications of game theory for the analysis of social dilemmas, as well as reconstructing core assumptions and implications, including testable hypotheses. The chapter is analytical in nature, trying to structure the field, rather than providing a general overview of and comparison with alternative theories. Suggesting that theoretically and empirically informed middle range theory on social dilemmas has emerged from applications of game theory in this field, empirical insights generated by theoretical models are a core topic of the chapter. In the spirit of Goldthorpe’s plea (2000: ch. 5) for an alliance between rational action theory (RAT) and the quantitative analysis of data (QAD), we explore how research in this field can contribute to narrowing the gap between rational choice models and empirical research (Green and Shapiro 1994). The review emphasizes that relevant empirical research in the field includes survey designs, quasi-experimental designs, and more qualitative case studies in addition to experimental designs. Such a “multimethod”

Rational Choice Research on Social Dilemmas

117

perspective conceives QAD broadly and is particularly appropriate when it leads to testing similar hypotheses with complementary empirical designs, thus providing an indication of the robustness of empirical findings. This Handbook (see Introduction) focuses on two strategies through which rational choice theory has been extended beyond the highly stylized assumptions of neoclassical economics—namely, atomized interaction on perfect markets of rational and selfish actors with full information. One strategy involves making the assumptions on the actors more complex by relaxing the rationality assumption or the selfishness assumption (see Gächter’s chapter in this Handbook on how this strategy can be usefully employed for improving on standard models and applications of game theory in rational choice social research).The other strategy aims at using more complex and more appropriate assumptions on the social context by replacing the assumption of atomized interactions on perfect markets. This chapter highlights the second strategy.1 It does so by combining strong assumptions on individual rationality with assumptions on the “embeddedness” of action in ongoing relations and networks of relations, showing that embeddedness crucially affects behavior of rational actors in social dilemmas. This is in line not only with Coleman’s heuristic advice (1987) to combine robust assumptions on rational behavior with more complex assumptions on social structure. Also Granovetter (1985) advocated precisely such a combination of assumptions in his often cited programmatic sketch. Granovetter’s criticism of the shortcomings of the neoclassical model of perfect markets with atomized actors has often been taken to imply that one had better abandon rational choice models in favor of more “realistic,” socially inspired models of man. It has been widely overlooked, though, that Granovetter sharply opposes “psychological revisionism,” characterizing it as “an attempt to reform economic theory by abandoning an absolute assumption of rational decision making” (ibid.: 505). Rather, he suggests maintaining the rationality assumption: “[W]hile the assumption of rational action must always be problematic, it is a good working hypothesis that should not easily be abandoned. What looks to the analyst like nonrational behavior may be quite sensible when situational constraints, especially those of embeddedness are fully appreciated” (ibid.: 506). He argues that investments in tracing the effects of embeddedness are more promising than investments in the modification of the rationality assumption: “My claim is that however naive that psychology [of rational choice] may be, this is not where the main difficulty lies—it is rather in the neglect of social structure” (ibid.). This chapter explores the potential of such an approach for the analysis of social dilemmas. We use trust problems as a paradigmatic example of social dilemmas, sometimes indicating generalizations of results for other types of social dilemmas. Trust problems involve two actors. We thus largely neglect social dilemmas involving many actors (see Gächter’s chapter in this Handbook for some references, as well as the chapter by Kiser and Powers). We illustrate how game-theoretic tools can be used for modeling trust problems and other social dilemmas and sketch the logic of deriving testable hypotheses from game-theoretic models. We then turn to theory and hypotheses on how social structure—that is, embeddedness of a trust problem or, more generally, embeddedness of a social dilemma—affects behavior in such situations. The review of empirical research

118 Vincent Buskens and Werner Raub takes stock of evidence for and against hypotheses on embeddedness effects, with an emphasis on results obtained from complementary research designs and an emphasis on applications to the Internet economy and other economic exchange that resembles a social dilemma. Some directions for future research are also suggested.

Trust in Isolated Encounters the trust game Game theory provides tools for the analysis of situations with interdependence of two or more actors: choices of an actor affect the other actor(s), and vice versa. We sketch the analysis of trust situations in different social contexts, without elaborating on game-theoretic principles underlying the analysis.2 Consider the standard Trust Game (Camerer and Weigelt 1988; Dasgupta 1988; Kreps 1990; Snijders 1996: chs. 1−4; Buskens 2002: chs. 1−3), which models trust problems as outlined above. The game (see Figure 3.1) involves two actors, the trustor and the trustee. The game starts with a move by the trustor, who can choose between placing or not placing trust. If trust is not placed, the interaction ends and the trustor receives payoff P1, while the trustee receives payoff P2. If trust is placed, the trustee chooses between honoring and abusing trust. If the trustee honors trust, the payoffs for trustor and trustee are Ri > Pi, i = 1,2. If trust is abused, the payoff for the trustor is S1 < P1, while the trustee receives T2 > R2. The Trust Game models the benchmark scenario of a one-shot transaction between a buyer and a seller of an antiquarian book. The buyer is the trustor and chooses between placing trust by paying the price for the book and not placing trust by not buying. The seller is the trustee, who can honor trust by shipping the book or abuse trust by not shipping. Under standard gametheoretic assumptions, the payoffs in the game represent utilities for the actors, there is common knowledge implying that all actors know that all actors know all elements of the game, and the actors are rational in the sense that they maximize utility, given their expectations on the behavior of other actors. Using game-theoretic tools, one can then find a solution for the Trust Game. Such a solution implies equilibrium behavior—that is, each actor plays a best reply, given the behavior of the other actor(s) (Nash 1951). Under these assumptions, it is easy to see that the trustee would abuse trust if the trustor would place trust. The trustor, being able to anticipate this, will not place trust. This equilibrium is indicated by double lines in Figure 3.1. Trustor Trust

No trust

Trustee

figure 3.1. The Trust Game (S1 < P1 < R1, P2 < R2 < T2).

Abuse trust P1 P2

S1 T2

Honor trust R1 R2

Rational Choice Research on Social Dilemmas

119

A slightly more complex model of a trust problem is the Investment Game (Berg, Dickhaut, and McCabe 1995; Ortmann, Fitzgerald, and Boeing 2000; Barrera 2005). Again, the game is played by two actors. However, while in the Trust Game the actors make binary choices, in the Investment Game the trustor now chooses the degree to which the trustee will be trusted, and the trustee chooses the degree to which that trust will be honored. More precisely, the trustor has an endowment E1 and chooses an amount M1 to send to the trustee (0 ≤ M1 ≤ E1).This “investment” M1 is then multiplied by m > 1, and the trustee receives mM1. The parameter m can be seen as indicating the trustee’s returns resulting from the trustor’s investment. Subsequently, the trustee chooses an amount K2 to return to the trustor, with 0 ≤ K2 ≤ mM1. Afterward, the game ends with the trustor receiving V1 = E1 − M1 + K2 and the trustee receiving V2 = mM1 − K2. While M1 indicates how much the trustor trusts the trustee, K2 indicates how trustworthy the trustee is. Both the Trust Game and the Investment Game represent the time lag between placing trust and the trustee’s response. In both games trust is risky because the trustor regrets being trusting if the trustee turns out not to be trustworthy. The games thus model risks for the trustor in the sense of “opportunistic” or “strategic” behavior of the trustee, who has an incentive for abusing trust. deriving testable hypotheses from game-theoretic models How to use game-theoretic models for generating empirically testable hypotheses on social dilemmas? For answering this question, one can use Coleman’s scheme (1990: ch. 1; see also Raub, Buskens, and Van Assen 2011) for relating macro- and microlevel propositions in social science explanations (see Figure 3.2). First, the specification of the game includes social conditions in the sense of opportunities and restrictions for the actors’ behavior represented by the top-left node of Coleman’s scheme. The specification of the game also comprises assumptions on “independent variables” of rational choice theory, such as preferences of the actors that are represented by the actors’ payoffs.These are the microlevel assumptions related to the bottom-left node of Coleman’s scheme. Furthermore, the specification of the game includes assumptions on macro-micro transitions that are summarized by the vertical arrow 1 in Coleman’s scheme: note that the specification of the game models how an actor’s payoffs and information depend on social conditions. For example, an actor’s payoff function is a function of the possible choices of all actors—that is, a function of interdependencies. Thus, the specification of a game refers to empirical assumptions on the macrolevel of social conditions, on the microlevel of the actors’ preferences and information, and on macro-micro transitions. Rationality assumptions are microlevel assumptions that are summarized in Coleman’s scheme by arrow 2. These are assumptions such as that the solution has to be an equilibrium. Game-theoretic analysis then comprises deriving propositions on equilibria of the game and on properties of these equilibria. This allows one to derive implications concerning the behavior of rational actors. These implications are represented by the bottom-right node in Coleman’s scheme. In a final step, one can derive propositions on macrolevel effects—for example, on Paretooptimality or suboptimality of the outcomes that result from the behavior of

120 Vincent Buskens and Werner Raub Macro conditions

Macro outcomes

1

Micro conditions

3 2

Micro outcomes

figure 3.2. Coleman’s Scheme.

the actors.We illustrate below the derivation of hypotheses along these lines for trust situations under different social conditions. deriving hypotheses for trust in isolated encounters We briefly consider trust in isolated encounters.Two actors play the Trust Game once and only once. Neither the two actors, nor other actors, can condition behavior in future interactions on what happens in the Trust Game. Isolated encounters are hardly a standard feature of interactions in social and economic life. After all, eBay’s feedback forum implies that buyer and seller are not involved in an isolated encounter. Hence, isolated encounters are typically studied in the laboratory and are used to study nonstandard assumptions on preferences, because other factors such as the social embeddedness can be controlled in the laboratory. More precisely, assume that subjects play the Trust Game from Figure 3.1, with payoffs S1 < P1 = P2 < R1 = R2 < T2 in terms of monetary incentives or points converted into money at the end of the experiment. As explained above, the standard prediction is no trust, and, if trust were to be placed anyway, it would be abused. For the Investment Game, the analogous prediction is that the trustor sends nothing and, if the trustor were to send anything, the trustee would never return anything. Clearly, these are very strong predictions for the behavior of any subject in the laboratory. The predictions are clearly rejected (see Snijders 1996; Snijders and Keren 1999, 2001, on the Trust Game; Berg, Dickhaut, and McCabe 1995, on the Investment Game; and Camerer 2003: ch. 2.7, for an extensive review). Experiments show that substantial percentages of subjects trust in the Trust Game and send positive amounts in the Investment Game. Also, many subjects in the role of trustee honor trust and return substantial amounts. More generally, opportunism is not ubiquitous in isolated encounters resembling a social dilemma. Different approaches can be envisaged that account for such empirical regularities (see Fehr and Schmidt 2006 for an instructive overview). Each of these approaches involves making the assumptions on the actors more complex in one way or the other. First, one could relax the rationality assumption and employ a bounded rationality perspective. For example, one could assume that subjects are used to repeated interactions in life outside the laboratory. As we will see below, placing and honoring trust as well as other forms of cooperative, nonopportunistic behavior can be a result of equilibrium behavior in repeated interactions. The assumption then is that subjects erroneously apply rules in isolated encounters that are appropriate when interactions are repeated (see Binmore 1998 for a sophisticated discussion of such approaches). Second,

Rational Choice Research on Social Dilemmas

121

there are approaches that maintain the rationality assumption but modify the selfishness assumption. These approaches thus abandon the assumption that subjects care exclusively about their own material resources (“utility = own money”). Rather, it is assumed that subjects, or at least some subjects, have other-regarding preferences. It is quite often argued (see, for example, Fehr and Gintis 2007) that such preferences are the result of socialization processes and internalized social norms and values. Also, it is often assumed that subjects may differ with respect to their other-regarding preferences—there may be selfish subjects as well as subjects with other-regarding preferences—and that subjects are incompletely informed on the preferences of other subjects. To get the flavor of how assumptions on other-regarding preferences can be used to account for placing and honoring trust in a Trust Game as an isolated encounter, consider a simple version of a social preferences model—namely, Snijders’s guilt model (1996; see also Snijders and Keren 1999, 2001), a simplified version of the Fehr-Schmidt (1999) model of inequity aversion. Assume that actor i’s utility is given by Ui(xi,xj) = xi − βimax(xi − xj, 0) with monetary payoffs xi and xj for the actors i and j and βi ≥ 0 a parameter representing i’s guilt resulting from an inequitable allocation of monetary payoffs. Hence, in a Trust Game with payoffs in terms of money and P1 = P2 and R1 = R2, the trustee’s utility from abused trust would be T2 − β2(T2 − S1), while utilities correspond to own monetary payoffs in all other cases. Furthermore, assume actor heterogeneity with respect to the guilt parameter βi in the sense that there are actors with a large guilt parameter, while βi is small or even equals zero for other actors—namely, those with selfish preferences. Finally, denote by π a trustor’s belief of the probability that the trustee’s utility from abusing trust is smaller than his utility from honoring trust—that is, T2 − β2(T2 − S1) < R2. Equilibrium behavior now requires that a trustee with β2 > (T2 − R2)/(T2 − S1) honors trust, while a trustor places trust if π > (P1 − S1)/(R1 − S1). We can assume that placing trust becomes more likely when the condition π > (P1 − S1)/(R1 − S1) becomes less restrictive. Similarly, we can assume that honoring trust becomes more likely when the condition β2 > (T2 − R2)/(T2 − S1) becomes less restrictive. Furthermore, we can assume that π depends on the trustee’s incentives and hence decreases in (T2 − R2)/(T2 − S1). It follows from this model that the likelihood of placing trust decreases in the trustor’s risk (P1 − S1)/(R1 − S1), as well as in the trustee’s temptation (T2 − R2)/(T2 − S1), and that the likelihood of honoring trust decreases in the trustee’s temptation (T2 − R2)/(T2 − S1). These implications correspond with experimental evidence (see Snijders 1996; Snijders and Keren 1999, 2001). Obviously, assumptions on other-regarding preferences should be used with care (see Camerer 2003: 101; Fehr and Schmidt 2006: 618): (almost) all behavior can be “explained” by assuming the “right” preferences and adjusting the utility function.Thus, one would prefer first of all parsimonious assumptions on other-regarding preferences, adding as few new parameters as possible to the model. Second, when assumptions on other-regarding preferences are employed, one should aim at using the same set of assumptions for explaining behavior in a broad range of different experimental games. Third, one should not only account for well-known empirical regularities but also aim at deriving and testing new predictions. It is therefore important from a methodological

122 Vincent Buskens and Werner Raub perspective that the same set of assumptions on other-regarding preferences is consistent with empirical regularities of behavior not only in Trust Games but also in other social dilemmas, in games involving distribution problems such as the Ultimatum Game (Güth, Schmittberger, and Schwarze 1982; see Camerer 2003 for a survey), or the Dictator Game (Kahneman, Knetsch, and Thaler 1986; see Camerer 2003 for a survey), and in market games. We refer to Gächter’s chapter in this Handbook for a careful discussion of how to refine the model of rational and selfish actors (another useful overview for sociologists is Fehr and Gintis 2007). In the subsequent sections we return to employing standard assumptions on the level of individual behavior—namely, assumptions on game-theoretic rationality as well as basically selfish preferences. We now refine a standard assumption on the social context in neoclassical economics. Rather than assuming atomized interactions—in our case, “isolated encounters”—we explore the implications of “embeddedness” for rational and selfish behavior in social dilemmas.

Theory and Hypotheses on Effects of Social Embeddedness Roughly, embeddedness (Granovetter 1985) can mean that the actors involved in a focal Trust Game maintain an ongoing relation with prior and expected future interactions. We refer to this as dyadic embeddedness. An example is the second scenario for the purchase of the antiquarian book in which the buyer repeatedly purchases from the same antiquarian. Furthermore, a focal Trust Game can be related to interactions of trustor or trustee with third parties. We refer to this as network embeddedness. The buyer of the antiquarian book may happen to know others who purchase books from the antiquarian, or may obtain third-party information about the antiquarian’s behavior in the past through an institution that enhances network embeddedness. An example is our third scenario for purchasing the antiquarian book where eBay’s feedback forum provides network embeddedness for the transaction. We distinguish two mechanisms, control and learning, through which dyadic and network embeddedness may affect trust. Control refers to the case in which the trustee has short-term incentives for abusing trust, while some long-term consequences of his behavior in the focal Trust Game depend on behavior of the trustor. More precisely, if the trustee honors trust in the focal Trust Game, the trustor may be able to reward this by applying positive sanctions in the future. Conversely, if the trustee abuses trust in the focal Trust Game, the trustor may be able to punish this by applying negative sanctions. Given dyadic embeddedness, the trustee has to take into account that honoring trust in the focal Trust Game may affect whether or not the trustor places trust again in the future. Given network embeddedness, the trustee has to take into account that a trustor can inform third parties about the trustee’s behavior in the focal Trust Game, such as other trustors with whom the trustee may be involved in future Trust Games. Again, whether or not other trustors are willing to trust the trustee may depend on honoring or abusing trust in the focal Trust Game.Thus, the trustee has to trade off the short-term incentives to abuse trust against the long-term benefits of honoring trust and the long-term costs of abusing trust. This mechanism is also known as conditional cooperation (Taylor 1987) or reciprocity (Gouldner 1960; Blau 1996 [1964]; Diekmann 2004). Reciprocity

Rational Choice Research on Social Dilemmas

123

tab le 3 .1

Types of Embeddedness and Mechanisms through which Embeddedness Affects Trust Two types of embeddedness Two mechanisms

Control

Learning

Dyad

Sanctioning possibilities of the trustor without involving third parties. Information about the trustee from past experiences of the trustor.

Network

Sanctioning possibilities of the trustor that involve third parties. Information about the trustee from third parties.

in this sense (sometimes labeled “weak reciprocity”—see, for example, Fehr and Schmidt 2006: 620; Fehr and Gintis 2007) can be driven exclusively by long-term, “enlightened” self-interest of the actors. Thus reciprocity in this sense differs fundamentally from reciprocal behavior of the trustee in isolated encounters based on other-regarding preferences (“strong reciprocity”). Embeddedness may affect trust through a second mechanism—namely, learning. The trustor need not be completely informed on the behavioral alternatives and incentives of the trustee. Beliefs by the trustor on the trustee’s characteristics can be affected by information regarding past interactions. This information can be obtained from past interactions of trustor and trustee—that is, through dyadic embeddedness. Given network embeddedness, information can also be obtained from third parties who have interacted with the trustee in the past. If a trustee has been trustworthy in past interactions, a trustor might be more convinced that the trustee will be trustworthy again in the focal Trust Game than if information on untrustworthy behavior of the trustee in the past has been revealed. Table 3.1 summarizes our distinction between dyadic and network embeddedness, as well as between learning and control (Buskens and Raub 2002; see Yamagishi and Yamagishi 1994: 138−39 for a similar discussion of learning and control effects through network embeddedness). Our sketch indicates that embeddedness may help actors to mitigate social dilemmas such as trust problems (see Taylor 1987; Kollock 1998 for an overview of different ways in which social dilemmas can be mitigated). Note that embeddedness effects on social and economic interactions and exchange are a common theme of the sociological literature. However, clearly disentangling different types of embeddedness effects and the underlying mechanisms theoretically as well as empirically is often neglected.We now show how gametheoretic tools allow for modeling embeddedness and, more important, for deriving hypotheses on effects of embeddedness on trust. This can be done by “embedding” a focal Trust Game in a more complex game. Subsequently, one establishes conditions for an equilibrium of the more complex game such that trust is placed and honored in the focal Trust Game. Propositions on such conditions yield hypotheses on embeddedness effects on trust. a simple model of dyadic control: conditional trust in the indefinitely repeated trust game To see how trust can be a result of purely selfish rational actors who are “enlightened” in the sense that they take long-term effects of their behavior into account, we consider a model of control effects through dyadic

124 Vincent Buskens and Werner Raub embeddedness—namely, Kreps’s model (1990) of a repeated Trust Game (see also Gibbons 2001). In this model, the Trust Game is played repeatedly in rounds 1, 2,…, t,… By way of example, a buyer purchases repeatedly from the same seller of antiquarian books. More precisely, after each round t, another round t + 1 is played with probability w (0 < w < 1), while the repeated game ends after each round with probability 1 − w.The focal Trust Game is thus embedded in a more complex game in which the Trust Game is repeated indefinitely often. In each round, trustor and trustee observe each other’s behavior. In the repeated game, an actor’s behavior in each round t may depend on the behavior of both actors in the previous rounds. An actor’s expected payoff for the indefinitely repeated Trust Game is the discounted sum of the actor’s payoffs in each round, with the continuation probability w as discount parameter. For example, a trustor who places trust throughout the repeated game, with trust being honored throughout, receives payoff R1 + wR1 + … + wt−1R1 + … = R1/(1 − w). Thus, using the apt label coined by Axelrod (1984), the continuation probability w represents the “shadow of the future”: the larger w, the more an actor’s payoff from the repeated game depends on what the actor receives in future rounds. In the indefinitely repeated Trust Game, the trustor can exercise control by employing conditional behavior that rewards a trustee who honors trust in a focal Trust Game by placing trust again in future games. Conversely, conditional behavior of the trustor can imply punishing the trustee’s abuse of trust in the focal Trust Game by not placing trust in at least some future games. If the trustor uses weak reciprocity in the sense of implementing conditional behavior, the trustee can gain T2 rather than R2 in the current Trust Game by abusing trust. However, abusing trust will then be associated with obtaining only P2 in (some) future encounters with no trust placed by the trustor, while honoring trust will result in larger payoffs than P2 in those future encounters if the trustor goes on placing trust. Moreover, the larger the shadow of the future, the more important are the long-term effects of present behavior. Thus, anticipating that the trustor may employ conditional behavior, the trustee has to balance short-term (T2 − R2) and long-term (R2 − P2) incentives. It can be shown that weak reciprocity can be a basis for rational trust in the sense that the indefinitely repeated Trust Game has an equilibrium such that trust is placed and honored in each round. Consider conditional behavior of the trustor that is associated with the largest rewards for trustworthy behavior of the trustee and with the most severe sanctions for untrustworthy behavior. This is realized if trust is placed in the first round and also in future rounds, as long as trust has been placed and honored in all previous rounds. However, as soon as trust is not placed or abused in some round, the trustor refuses to place trust in any future round. Straightforward analysis shows that always honoring trust (and always abusing trust as soon as there has been any deviation from the pattern “place and honor trust”) is a best reply of the trustee against such conditional behavior of the trustor if and only if w ≥ (T2 − R2)/(T2 − P2).

(1)

This condition requires that the shadow of the future is large enough compared with (T2 − R2)/(T2 − P2), a convenient measure for a selfish trustee’s temptation to abuse trust. The condition refers exclusively to the incentives of

Rational Choice Research on Social Dilemmas

125

the trustee and not at all to the incentives of the trustor.3 This highlights that placing and honoring trust in the indefinitely repeated Trust Game is driven by the strategic interdependence of the actors. If condition (1) applies, the indefinitely repeated Trust Game has an equilibrium such that trust is always placed and honored. Enlightened selfinterest can thus be a basis for trust among rational actors in the sense of placing and honoring trust being equilibrium behavior.4 The equilibrium, however, is not unique. For example, never placing trust, while placed trust would always be abused, is always an equilibrium of the indefinitely repeated game. The “folk theorem” (see, for example, Fudenberg and Maskin 1986; Rasmusen 2007: ch. 5.2) for repeated games implies that the indefinitely repeated Trust Game has many other equilibria, too, for large enough w. Thus, an equilibrium selection problem emerges. A typical, though sometimes implicit, argument in the literature on equilibrium selection in this context is “payoff dominance.” An equilibrium is payoff dominated if there is another equilibrium that is associated with higher payoffs for at least some actor and is not associated with lower payoffs for any actor. In the indefinitely repeated Trust Game, an equilibrium that implies placed and honored trust throughout the game is evidently not payoff dominated by other equilibria, while the notrust-throughout equilibrium is payoff dominated. Rather than claiming that actors indeed follow the very strict behavioral rules described above, we can use condition (1) to derive more qualitative predictions about behavior in the indefinitely repeated Trust Game. If one proceeds from the observation that condition (1) is a necessary and sufficient condition for equilibria in the indefinitely repeated Trust Game such that trust is placed and honored throughout the game, one could then assume that placing and honoring trust becomes more likely when the condition becomes less restrictive. This leads directly to testable hypotheses on control effects through dyadic embeddedness. Specifically, one would expect that the likelihood of placing and honoring trust increases in the shadow of the future w and decreases in the temptation (T2 − R2)/(T2 − P2) for a selfish trustee. The results for the indefinitely repeated Trust Game can be generalized. For example, analogous results hold for an indefinitely repeated Investment Game. Friedman (1971, 1990) shows that analogous results apply to a broad class of indefinitely repeated 2- and n-person games. Roughly speaking, if a social dilemma is repeated indefinitely often and the shadow of the future is large enough relative to the short-term incentives of the actors, there exists an equilibrium of the indefinitely repeated game such that the actors cooperate: the equilibrium of the repeated game induces a Pareto-optimal outcome and a Pareto-improvement compared with the Pareto-suboptimal solution of the original dilemma. Of course, these generalizations should be interpreted with care. For example, conditional behavior requires the observability of the behavior of other actors. Hence, the underlying assumption that each actor receives reliable information on each other actor’s behavior in each round of the game is crucial, while such an assumption will often be rather problematic from an empirical perspective in games with many actors (see Bendor and Mookherjee 1987).

126 Vincent Buskens and Werner Raub network control Models of repeated Trust Games can be extended to account for control effects resulting from network embeddedness in addition to dyadic embeddedness. In these extended models, the trustee interacts with a set of trustors, while the trustors are connected through a network that allows for communication about the behavior of the trustee. The focal Trust Game is now embedded in a more complex game that comprises Trust Games of the trustee with different trustors. The important feature is that the trustor in a focal Trust Game can transmit information on the trustee’s behavior in that game to other trustors. Next to direct reciprocity exercised by the trustor who interacts personally with the trustee in the focal Trust Game, network embeddedness allows for indirect reciprocity exercised by other trustors. A trustee contemplating honoring or abusing trust in a focal Trust Game now has to consider future sanctions by the trustor with whom he interacts in the focal Trust Game, as well as sanctions that can be applied by other future trustors who receive information on the trustee’s behavior in the focal Trust Game and who may condition their future behavior on that information. In terms of our example of purchasing antiquarian books, we thus now consider variants of an eBay feedback forum. First, such network embeddedness can be a substitute for dyadic embeddedness (see Kreps 1990: 106−8). Assume that the trustee interacts with a different trustor in each round of the indefinitely repeated Trust Game in the previous section. Thus each trustor plays the Trust Game only once with the trustee. Dyadic embeddedness is then removed completely from the repeated game and has been replaced by network embeddedness. However, if the trustor in a given round is reliably informed of what has happened in previous rounds, all trustors can condition their behavior in a given round in the same way as a trustor who plays in each round: trust is placed if and only if there is no information that trust has not been honored before. Evidently, the trustee’s best reply against such behavior of the trustors is again to honor trust in each round if condition (1) is fulfilled. Conversely, placing trust is indeed then the best reply behavior also for the trustors. Hence, we see that network embeddedness can induce trust among rational and selfish actors. Dyadic as well as network embeddedness is included in more complex models (for example, Weesie, Buskens, and Raub 1998; Buskens and Weesie 2000a; Buskens 2002: ch. 3; see Raub and Weesie 1990 for a related model of network embeddedness for the Prisoner’s Dilemma). In these models, a trustee interacts with a trustor in an indefinitely repeated Trust Game. After the interaction with a given trustor ends, the trustee goes on playing an indefinitely repeated Trust Game with another trustor, while information on behavior in the Trust Games with the first trustor is communicated to the second trustor with some probability. Interactions with a third trustor start after the interactions with the second trustor have ended, and so forth.These models are relatively general and allow for quite some heterogeneity with respect to various features: the incentive T2 for abusing trust varies between games, the probability of starting interactions with the trustee as well as the continuation probability for these interactions varies between trustors, and the probability of information transmission varies between pairs of trustors. One can then

Rational Choice Research on Social Dilemmas

127

study equilibria such that trustors place trust if T2 is not “too large” and if they do not have information that trust has ever been abused. A nice feature of these models is that they account for the intuition that trust will not always be placed. In addition to hypotheses on how the likelihood of trust is affected by the shadow of the future and the short-term incentives of the trustee, such models allow for deriving hypotheses on effects of network characteristics. Specifically, the likelihood of placing and honoring trust in a focal Trust Game increases in the density of the network of trustors as well as in the trustor’s outdegree—that is, the probability that the trustor will transmit information to the next trustor who interacts with the trustee. This is intuitively plausible, since network density as well as outdegree increase the sanction possibilities of the trustor. Hence, if the trustee thinks in terms of long-term consequences, higher network density and outdegree allow for placing and honoring trust even if the trustee’s short-term incentive to abuse trust is fairly large. A problem of these models is that they assume that information is reliable and that incentive problems associated with the supply of information are neglected (see Lorenz 1988; Raub and Weesie 1990: 648; Williamson 1996: 153−55; Blumberg 1997: 208−10; Buskens 2002: 18−20). However, supplying information on the trustee’s behavior is a contribution to a public good— namely, enforcing trustworthy behavior of the trustee. Such contributions are problematic: after all, public good production is itself a social dilemma when contributions are costly (this feature is often discussed as a major problem of institutions such as eBay’s feedback forum; see Bolton and Ockenfels 2009). Moreover, information from third parties can be inconsistent with one’s own experiences. Also, information from third parties can be problematic because of misunderstanding or strategic misrepresentation: imagine that the trustors are competitors who purchase the same goods from the same seller. In a nutshell, one would expect that effects of network embeddedness are attenuated when such problems become more serious. Notice, too, that we have focused on the case of network control in the sense that other trustors can sanction the trustee in future interactions. This is control through “voice,” in Hirschman’s sense (1970). A different case of network control is that a trustor has access to alternative trustees and can exercise control through “exit”: whether or not the trustor interacts again with the trustee in the future depends on the trustee’s behavior in the focal Trust Game. Modeling network control through exit opportunities for the trustor is not trivial (see Hirshleifer and Rasmusen 1989; Schüßler 1989; Vanberg and Congleton 1992 for related models), but one would expect in general that the likelihood of placing and honoring trust increases in the trustor’s exit opportunities. game-theoretic models of trust based on learning and control The game-theoretic models of embeddedness effects on trust discussed above have been (repeated) games with complete information: roughly speaking, each actor is informed on the behavioral alternatives and incentives of all actors. Specifically, trustors are completely informed on the behavioral alternatives and the incentives of the trustee. Hence, there is no need—and no opportunity— for trustors to learn during the game about unobservable characteristics of

128 Vincent Buskens and Werner Raub the trustee. This means that these models do not yield hypotheses on learning effects of embeddedness.5 Hypotheses on control as well as learning effects can be derived from models of games with incomplete information. Typically, these are models of finitely repeated games. To get a flavor of these games, consider first of all a finitely repeated game with complete information. Assume that trustor and trustee play the Trust Game from Figure 3.1 repeatedly—namely, N times. Clearly, in the final round, equilibrium behavior requires that trust would be abused and no trust will be placed. This means that behavior in the last but one round cannot have effects on behavior in the final round. Hence, no trust will be placed in the last but one round and so forth, back to the first round. This backward induction argument shows that placing and honoring trust cannot be a result of rational and selfish behavior in a finitely repeated Trust Game with complete information. Things change by introducing incomplete information in the finitely repeated Trust Game (see Camerer and Weigelt 1988; Dasgupta 1988; Neral and Ochs 1992; Bower, Garber, and Watson 1997; Buskens 2003). Introducing incomplete information means relaxing another core assumption of the standard rational choice model. Assume that there is a positive ex-ante probability π that the trustee actually has no incentive to abuse trust—that is, his payoff from abusing trust is T ∗2 < R2 (an alternative assumption leading to essentially the same results would be to assume that the trustee has no opportunity to abuse trust with probability π). The trustor knows the probability π but cannot directly observe whether the trustee’s payoff from abusing trust is T2 or T ∗2 . Now, if the trustor places trust in some round of the repeated game that is not the final round, trust may be honored for one of two very different reasons. First, the trustee’s payoff could be T ∗2 < R2 so that there is no incentive at all for the trustee to abuse trust. Second, the trustee’s payoff could be T2 > R2 but the trustee follows an incentive for reputation building. The trustee knows that if he abuses trust, the trustor can infer for sure that the trustee’s payoff from abusing trust is T2 > R2 and will thus never place trust again in future rounds. On the other hand, if the trustee honors trust, the trustor remains uncertain about the trustee’s incentives and may place trust again in the future. Conversely, the trustor can anticipate such behavior of the trustee and may therefore indeed be inclined to place trust. In this game the trustor can control the trustee, in that placing trust in future rounds depends on honoring trust in the current round and the trustor can learn about the incentives of the trustee from the trustee’s behavior in previous rounds. The result is a subtle interplay of a trustor who tries to learn about and to control the trustee, taking the trustee’s incentives for reputation building into account, and a trustee who balances the long-term effects of his reputation and the short-term incentives for abusing trust, taking into account that the trustor anticipates on this balancing. It can be shown that the game has an equilibrium (Kreps and Wilson 1982) that does involve placing and honoring trust in some rounds of the repeated game. More precisely, in that equilibrium, the game starts with trust being placed and honored in a number of rounds. Afterward, a second phase follows in which the trustor and the trustee with T2 > R2 randomize their behavior until the trustor does not place trust or the trustee abuses trust. After trust has

Rational Choice Research on Social Dilemmas

129

not been placed or has been abused for the first time, the third and last phase starts, in which no trust is placed until the end of the game. A remarkable feature of the model is that quite some honored trust can be induced by equilibrium behavior even if the probability π that the trustee has no incentive to abuse trust is small. In the equilibrium, learning occurs—in the sense that the trustor updates her belief about the probability that she is playing with a trustee without an incentive to abuse trust—if trust is abused, and in the second phase as long as trust is honored. Learning is rational in the sense of Bayesian updating. The first phase of the game with trust being placed and honored is shorter, the higher the risk (P1 − S1)/(R1 − S1) for the trustor, the smaller the number of rounds of the repeated game, and the smaller the ex-ante probability π. While the risk for the trustor is a driving force of the model, the trustee’s temptation (T2 − R2)/(T2 − P2) only affects behavior in the second randomization phase of the repeated game. Quite counterintuitively, the probability that the trustor places trust in that phase increases (!) in the trustee’s temptation (see Buskens 2003: 239 for an explanation). Game-theoretic models with incomplete information such as the finitely repeated Trust Game are complex. They become even more complex by including learning resulting from network embeddedness. A shortcut linking learning effects of network embeddedness to such models would be to assume that the trustor’s ex-ante probability π of interacting with a trustee who would never abuse depends on information the trustor receives from third parties such as other trustors who played Trust Games previously with the trustee. Specifically, based on information diffusion models in networks of trustors (see, for example, Buskens 2002: ch. 4) and assuming that the information about the trustee is positive (it is information that the trustee has honored rather than abused trust), one would expect that the ex-ante probability π increases in the density of the network of trustors as well as in the extent to which the trustor in the focal Trust Game receives information about the trustee from other trustors—that is, increases in the trustor’s indegree. A more explicit game-theoretic model of network effects in games with incomplete information has been provided by Buskens (2003). In that model, the trustee plays Trust Games with two different trustors, A and B. With some probability, each trustor can inform the other trustor on the trustee’s previous behavior. We can conceive of the probability that trustor A transmits information to trustor B as A’s outdegree and B’s indegree (and vice versa). Thus, trustor A controls the trustee through her outdegree and learns from B about the trustee through her indegree. If each trustor transmits information to and receives information on the trustee from the other trustor with sufficiently high probability, the first phase of the repeated game such that trust is placed and honored becomes longer and in this sense network embeddedness increases trust. Counterintuitively, in the randomization phase, the probability of placing trust decreases (!) in the probabilities of information transmission. Summarizing and interpreting the results of the game-theoretic models for learning effects through dyadic and network embeddedness yields the hypotheses that the likelihood of placing and honoring trust decreases in the trustor’s risk (P1 − S1)/(R1 − S1) and increases if the trustor’s previous experiences with the trustee are positive (the trustee honored trust) rather

130 Vincent Buskens and Werner Raub than negative (the trustee abused trust). Furthermore, assuming that the trustor receives positive information about the trustee from other trustors, the likelihood of placing and honoring trust increases in the density of the network of trustors, and in the trustor’s indegree. These are relatively robust hypotheses that lend themselves also to empirical research outside the laboratory. The counterintuitive hypotheses on behavior in the randomization phase of the games are clearly best tested in lab experiments. Models for control and learning effects of embeddedness in games with incomplete information are not only problematic in that they use very strong assumptions on the actors’ rationality including rational (Bayesian) updating of beliefs.These models are also problematic in that they neglect learning on other features than unobservable characteristics of the trustee. For example, a trustor could try to use information received from other trustors for inferring how to reasonably cope with trust problems. Also, past interactions may give rise to other effects than exclusively learning. For example, actors may have pledged investments in their relation through past interactions, and these investments affect the incentives in the focal Trust Game. The attractive feature of game-theoretic models involving incomplete information is that control and learning can be analyzed simultaneously. The price tag attached to these models is a set of rather strong assumptions on the actors’ rationality. Alternatives are “pure” learning models in which actors adapt their behavior based on past experiences. Actors try to optimize short-term outcomes, while not (or “hardly”) looking ahead. This implies, too, that actors do not take other actors’ incentives into account (see Camerer 2003: ch. 6 for a useful overview of learning models; Macy and Flache 1995, 2002; Flache and Macy 2002 provide applications to social dilemmas; Buskens 2002: ch. 4 is an example of a model of learning in networks). Hence, these models neglect control effects. Typically (see Buskens and Raub 2002: 173−76), learning models yield hypotheses that the likelihood of placing trust decreases in the trustor’s risk (P1 − S1)/(R1 − S1). Also, the trustor’s estimation of the probability π that trust will be honored will typically increase with positive information about the trustee’s behavior in previous interactions, be it information from the trustor’s own previous interactions with the trustee or information from third parties.Therefore, one would again hypothesize that more positive information increases the likelihood of placing trust. Table 3.2 summarizes the hypotheses discussed in this section.

Empirical Research on Effects of Social Embeddedness We organize our overview of empirical evidence on effects of social embeddedness by type of research design, focusing on evidence closely related to the hypotheses summarized in Table 3.2. First, lab experiments are used for testing hypotheses on embeddedness effects (see Cook and Cooper 2003 for an overview of experimental studies on how other elements in the social context can affect trust). Experiments allow for control over the variation in important independent variables, and the causal relation between manipulations and outcome differences is mostly obvious.The disadvantage is that setups are often rather artificial. Subjects are typically students who are engaged in abstract

Rational Choice Research on Social Dilemmas

131

tab le 3 .2

Hypotheses on Effects of Social Embeddedness. Two types of embeddedness Two mechanisms

Control

Learning

Dyad

1. Trust and trustworthiness decrease with the temptation to abuse trust for the trustee and increase with the likelihood that an interaction is repeated. 2. Trust and trustworthiness decrease with the trustor’s risk and increase with positive experiences with a trustee.

Network

3. Trust and trustworthiness increase with the density of the trustor’s network, her outdegree, and the availability of institutions that provide information. 4. Trust and trustworthiness increase with the density of the trustor’s network, her indegree, and the availability of institutions that provide information (given that information about the trustee is predominantly positive).

interactions. This questions external validity. Therefore, evidence from settings beyond the lab is a complement to experimental evidence. In the second part, we thus review evidence from survey studies, with some emphasis on evidence from on-line transactions. It is typically difficult to disentangle learning and control effects of embeddedness in these studies. We conclude our overview with a brief sketch of two vignette experiments that were specifically designed to overcome this problem. Clearly, vignette experiments have their limitations, too. For example, incentives for subjects are problematic, since vignette designs involve hypothetical decisions in hypothetical situations. One may thus conclude that it makes sense to employ different and complementary research designs, each having specific strengths and shortcomings, for testing hypotheses on embeddedness effects in order to assess the robustness of empirical findings. laboratory experiments Effects of dyadic embeddedness Camerer and Weigelt (1988) initiated experimental research that aims at carefully testing hypotheses on behavior in finitely repeated Trust Games with incomplete information, with follow-up studies by Neral and Ochs (1992), Anderhub, Engelmann, and Güth (2002), and Brandts and Figueras (2003). See Camerer (2003: 446−53) for a more detailed overview of these experiments.6 While experiments on one-shot Trust Games focus on payoff effects and reveal that these effects, in particular effects of risk and temptation, are strong, experiments on repeated Trust Games focus on embeddedness effects. Experiments confirm that trust as well as trustworthiness are high in early rounds and decrease when the end of the repeated game approaches (dyadic control). Trust is almost absent after any abuse of trust, while trust remains relatively high as long as trust has been honored (dyadic learning). However, the trustor’s tendency to place trust as long as trust has been honored does not increase as the end of the game comes nearer.This is consistent with the theory, since trustors have to realize that trustees with an incentive to abuse trust also

132 Vincent Buskens and Werner Raub have an incentive to make trustors believe that they do not abuse trust, while these trustees will in fact abuse trust toward the end of the game. Brandts and Figueras (2003) also find that trust and trustworthiness increase with the probability that a trustee has no incentive to abuse trust. Summarizing, the equilibrium described above for the finitely repeated Trust Game predicts quite some global patterns of behavior reasonably well. However, the experiments of Neral and Ochs (1992), Anderhub, Engelmann, and Güth (2002), and Brandts and Figueras (2003) also show that behavior of subjects does not completely follow the predicted equilibria. For example, it is predicted that in the second phase of the game in which trustors and trustees with an incentive to abuse trust both randomize, the probability that trustors trust increases (!) with the temptation for the trustee to abuse trust. This implication is not only counterintuitive but also inconsistent with experimental findings. Results from some other experiments are quite in line with these findings. Gautschi (2000) reports findings for finitely repeated Trust Games that comprise two or three rounds of play. He finds that positive past experience matters (dyadic learning) and that the number of remaining rounds to be played also increases trust (dyadic control). For a more contextualized setting with buyers and sellers and an incentive structure similar to the Trust Game, Kollock (1994) finds similar evidence. Still, the studies by Gautschi and Kollock report quite some untrustworthy behavior by trustees very early in the games. This can be explained by the difference that subjects in the studies of Gautschi (2000) as well as Kollock (1994) play relatively few games, while subjects in the studies by Camerer and Weigelt (1988) and the related follow-up studies play very many games. Camerer and Weigelt (1988) as well as their followers analyze mainly the later games in the experiment. In this way, suboptimal behavior is minimized. For example, if trustees build up more experience, they “learn” that they actually earn less if they behave opportunistically too early. In the studies by Gautschi (2000) and Kollock (1994), subjects have much less opportunities for this type of learning. Engle-Warnick and Slonim (2004, 2006) compare finitely and indefinitely repeated games. In principle, the trustor’s opportunities to exercise control in an indefinitely repeated game with constant continuation probability are the same in round t and in round t + 1. Still, the authors find decreasing trust over time in such games. However, this decrease is much smoother than in the finitely repeated games. This can be understood in the sense that learning effects in terms of negative experiences reduce trust over time, and subsequently trust seems to be difficult to restore. On the other hand, trust remains reasonably high because control opportunities do not diminish over time and enable some pairs to continue to trust each other. An additional explanation for decreasing trust in indefinitely repeated games might be that subjects believe that after many repetitions of the game the probability increases that a specific round will be the last one, even if experimenters do their very best to make it apparent that the continuation after every round is constant (for example, by using a publicly thrown die). While there are many experiments on the Investment Game, only few use the finitely repeated Investment Game. The findings of Cochard, Nguyen Van, and Willinger (2004) are in line with empirical regularities that have been

Rational Choice Research on Social Dilemmas

133

found for the Trust Game. Subjects send more in the Investment Game if there is a longer future ahead (dyadic control), but if receivers do not return enough they stop sending (dyadic learning). In early rounds, trustors send more if the trustees return more. While Cochard, Nguyen Van, and Willinger refer to this finding as a reciprocity effect, it can also be interpreted as a learning effect. Again, there is a strong endgame effect, although it is observed very late in the games. Trustees start to return less in the last-but-one round. Trustors react on low return rates by sending less in the last round, but there is no significant evidence that trustors send less as a pure result of being in the last round.7 Effects of network embeddedness Experiments with Trust Games or Investment Games that include network embeddedness are still rare. Bolton, Katok, and Ockenfels (2004; see also Bolton and Ockenfels 2009) compare one-shot Trust Games that are isolated encounters in the strict sense, finitely repeated Trust Games (with the same partner), and a third treatment in which subjects play multiple one-shot Trust Games with different partners but obtain information about the past behavior of their partners in interactions with other subjects (for a similar setup and results, see Bohnet and Huck 2004). In the one-shot Trust Games, trust and trustworthiness decline quickly after subjects have some experience. Trust and trustworthiness remain high in the repeated Trust Games and collapse only in the last couple of rounds. This finding resonates with evidence on effects of dyadic embeddedness. In the third treatment, there is initially less trust and trustworthiness than in the finitely repeated Trust Game setting, but trustees apparently learn fast enough that they have a considerable problem if they do not honor trust. In this treatment, trust and trustworthiness stabilize for some time in the middle of the series of interactions, although at a somewhat lower level than in the repeated Trust Game setting. Again, trust collapses in the last few rounds. Bolton, Katok, and Ockenfels (2004) interpret their third treatment as an experimental implementation of an institutionalized reputation system that is common for on-line transactions.The treatment could also be interpreted as a complete network in which information diffusion is perfect. Below we will come back to this. Note that the Bolton, Katok, and Ockenfels reputation treatment involves opportunities for learning as well as control through third parties.While indicating that network embeddedness matters, it remains unclear to which mechanism—learning or control or both—trust can be attributed. Buskens, Raub, and Van der Veer (2010) introduce a network setting with subjects playing finitely repeated Trust Games in groups of three (see also Barrera and Buskens 2009 for a related study on the Investment Game). There are two trustors and one trustee. Both trustors play with the same trustee. The design varies whether or not trustors obtain information about transactions from the other trustor who plays with the trustee. It turns out that there is more trust in the condition in which trustors do know what happens in the games of the other trustor. Similar to findings from other experiments, within dyads, trustors send more after positive experiences with the trustee, and trust as well as trustworthiness collapse near the end of the game. This is once again evidence for dyadic control, as well as dyadic learning. Buskens (2003) implies that the decrease in trust and trustworthiness should start later with increasing

134 Vincent Buskens and Werner Raub network embeddedness because of the network control effect. Buskens, Raub, and Van der Veer (2010) do not find evidence for this network control effect on the trust of the trustor. Still, they do find evidence for network control on trustworthiness of the trustee. They offer a bounded rationality argument why network control has an effect only for trustees and not for trustors: the trustee needs to anticipate third-party sanctions, while the trustor needs a further step of strategic reasoning—namely, anticipating that the trustee anticipates the third-party sanctions. survey studies There is quite some qualitative evidence that dyadic embeddedness (for example, Uzzi 1996) and network embeddedness (for example, Wechsberg 1966) affect trust. We focus on more quantitative evidence from surveys. As we will see, although many surveys offer evidence for effects of embeddedness, it is hardly ever the case that we can determine whether the effects are the result of learning, control, or a combination of the two mechanisms. Effects of dyadic embeddedness Gulati (1995a, 1995b) employs data on strategic alliances between business firms. Such alliances typically involve incentives for opportunism and trust problems between the partners. He finds that the probability that firms will form alliances is larger if they have been previously involved in alliances with the same partner. Gulati interprets this as an indication that previous and presumably positive experiences enlarge trust among partners. Moreover, the probability that partners in alliances use equity as a formal governance mechanism and commitment device decreases with the number of previous alliances between the partners. Using other data, Gulati and Wang (2003) show that joint ventures with a longer positive relation generate more value than joint ventures with less dyadic embeddedness. This is another indication that dyadic embeddedness helps to reach more efficient solutions in the social dilemmas the partners face. More precisely, it is tempting to assume learning effects resulting from dyadic embeddedness as an underlying mechanism. Baker, Faulkner, and Fisher (1998) find that interorganizational ties between advertising agencies and their clients have a smaller probability of being dissolved if they have existed for a longer period. Although these findings are interpreted in terms of learning— positive past experience increases trust, while trust enlarges the probability of staying together—a control interpretation seems likewise plausible. After all, the increased probability of staying together improves control opportunities. Some studies on trust in interfirm relations are noteworthy for addressing the effects of embeddedness at the dyadic level on the investment in formal arrangements such as investments in contracting. We consider investments in formal contractual arrangements as an indication for lack of trust among partners, since such arrangements provide, for example, compensation for the trustor in case of untrustworthy behavior by the trustee. This can also be interpreted as that there exist substitution effects between contracting and embeddedness in facilitating transactions that involve trust problems. In a study on seventy-two subcontracting relationships, Lyons (1994) finds that the probability for arranging the relationship with a formal contract decreases with

Rational Choice Research on Social Dilemmas

135

the number of years subcontractors have been trading with their most important customers. Similarly, Corts and Singh (2004) find that repeated interactions between oil companies and off-shore drillers reduce the probability that they choose fixed-price contracts to arrange the transaction. Blumberg (1997: ch. 4.2) uses the investment in formal arrangements and the extensiveness of the contract as measures for distrust in R&D-relations. He finds that both measures decrease and, thus, that trust increases with the extent to which the partners had transactions in the past. These results support the learning hypothesis that positive experiences increase trust. Blumberg actually distinguishes between the effect of past transactions and transactions expected with the partner in the future, but he does not find an effect of the transactions partners expect in the future. Batenburg, Raub, and Snijders (2003) study relations between buyers and suppliers of IT products. Their dependent variable is a combination of time and money spent in partner search, negotiating with the partner, and the extensiveness of the contract.This dependent variable represents investments in the ex-ante planning of transactions. They find that these investments decrease if the partners had transactions in the past. Furthermore, they find that the investments decrease even more if the partners already had past transactions and expect more transactions in the future. They do not find an effect of expected future transactions if the partners had no previous transactions.Their explanation employs two arguments. First, costly investments in ex-ante planning are less necessary if more future transactions are expected because of the sanction opportunities from subsequent transactions. This is a control effect based on the expectation of future transactions. Second, however, it is worthwhile to invest more in formal arrangements if more future transactions are expected, because these investments can be used again in subsequent transactions. This is an investment effect brought about by the expectation of future transactions. The driving force of this effect is that relation-specific investments associated with a focal transaction affect the incentive structure of future transactions. Combining the arguments on control and on the investment effect, it is unclear what the total effect of future transactions is. However, a negative interaction effect between past and future on ex-ante planning is indeed expected, since the investment effect will be larger in initial transactions than in later transactions. Another explanation for such an interaction effect could be that control plays a role only if the partners have sufficient information about each other and that uncertainties about an unknown partner are simply too large to allow for reliance on control through future sanctions already in the first transaction. Effects of network embeddedness Buskens, Raub, and Weesie (2000) use the same data as Batenburg, Raub, and Snijders (2003). They address the effects of network embeddedness on trust.They find that there are fewer issues addressed in the contract if the buyer and supplier are located closer to each other. An interpretation of this finding is that buyers and suppliers who are located closer to each other are probably embedded in a denser network. Although alternative explanations might be possible, this is a first indication that network embeddedness increases trust. Obviously, being located close to one another improves learning as well as

136 Vincent Buskens and Werner Raub control opportunities, so that it is unclear whether this effect is the result of learning or control. Rooks, Raub, and Tazelaar (2006) also use the same data to study trustworthiness of the supplier by analyzing how many ex-post problems arise in an IT-transaction.They find evidence that fewer ex-post problems arise in a transaction if the buyer knows more other buyers of the supplier, if the supplier is more visible, and if the buyer has more alternative suppliers to choose from. While the first and the second effect can be interpreted as learning as well as control effects of network embeddedness, the third effect is a network control effect on trustworthiness of the supplier. This pattern of findings is consistent with the experiment of Buskens, Raub, and Van der Veer (2010), showing that the control effect of network embeddedness is more apparent for the trustee’s than for the trustor’s behavior. Gulati (1995b) argues that social networks help firms to obtain information about facilities and the abilities of potential partners. He indeed finds that alliances occur more often among partners who have more common ties with third parties. Gulati and Gargiulo (1999) is one of the studies which shows that specific network properties—in this case, centrality—are related to the likelihood of alliance formation. These findings can be interpreted as a result of learning as well as control effects of network embeddedness, because central actors potentially receive more information and they can also transmit information more widely in the network. Using data on strategic alliances in the biotechnology sector, Robinson and Stuart (2007) try to disentangle the effects of different mechanisms based on network embeddedness. Their dependent variable is equity participation, which is a measurement for mistrust, because it reduces incentive problems and increases formal control opportunities. Their important independent variables related to our study are the centrality of the trustor (“client”) and trustee (“agent” or “target”) in the network; third parties trustor and trustee share in the network; and past alliances among the two partners. Robinson and Stuart provide strong evidence for effects of dyadic embeddedness and network embeddedness on trust by explaining the use of informal network management mechanisms rather than equity participation for partners with repeated interactions, partners that are more central in the network, and partners that have more other partners in common. Again, however, it is impossible to distinguish clearly between learning and control effects, because Robinson and Stuart use centrality measures such that the ties considered are symmetric and can be used for sending as well as receiving information. Moreover, the evidence in these network studies is based on the assumption that the network structure related to actual alliances corresponds largely with the network structure of communication among the relevant firms. If this assumption does not hold, learning and control can be the result of ties other than the alliance ties. Thus these studies provide at best indirect evidence for the mechanisms implied by the theory, because there is no information on how actual information about behavior of the firms is spread among other firms. The Internet economy confronts exchange partners with trust problems and illustrates how actors try to solve trust problems through institutionalized information exchange that improves network embeddedness in a setting in which direct face-to-face contact is not sufficient. An important advantage of

Rational Choice Research on Social Dilemmas

137

studies on the Internet economy is that for transactions at eBay and other platforms, researchers know which information the buyers have about sellers. In addition, the large number of transactions provides good opportunities for quantitative analyses. Both selling probability and selling price can be interpreted as measures of trust. First, if trust is too low, the transaction will not take place. Second, a good reputation of the seller increases the buyer’s trust and the expected value of the product for the buyer, implying that she is willing to pay more. In one of the earliest studies, Kollock (1999) claims preliminary evidence that reputation scores have an effect on price. Resnick and Zeckhauser (2002) conclude from an overview of studies on eBay auctions that good reputation scores increase the likelihood that a product is sold, but that there is no evidence that reputation scores affect price. In subsequent studies, various statistical pitfalls in the analyses of Internet auctions are addressed. Resnick et al. (2006) develop a field experiment in which they use an experienced seller who sells under different identities and with different reputation scores. In this way, they keep constant many confounding factors—for example, seller’s experience. Indeed, they find a large price premium related to the better reputation score. Also, in their extensive review of existing empirical work, Resnick et al. show that the price effect becomes more apparent in the more sophisticated studies. In addition, Snijders and Zijdeman (2004) show that unobserved heterogeneity can obscure the effect of reputation on price (see also Diekmann, Jann, and Wyder 2009). Lucking-Reiley et al. (2007) find evidence that looking only at the net reputation score—that is, positive minus negative evaluations—is not sufficient, because negative evaluations have a much larger impact on the price than positive evaluations. Most data collected from eBay and other standard auction sites have another problem—namely, that they include information about completed deals, while there is no, or at best limited, information about the alternatives buyers had. This implies that the actual choice problem of buyers cannot be properly evaluated, which might lead to serious selection problems. Snijders and Weesie (2009) solve this problem by looking at a different type of site at which producers of software can offer their services for specific demands by the buyer. At the end of the auction, the buyer chooses a producer. Snijders and Weesie (ibid.) again find that better reputation scores have a positive effect on price. The evidence on Internet auctions shows that institutionalized information exchange indeed improves trust of buyers in sellers. Clearly, this can be interpreted as a learning effect: positive past information convinces buyers that a seller will act in a trustworthy way. Still, institutionalized information exchange is likewise related to control through network embeddedness. Given that positive reputations have a price premium, buyers can damage reputations of sellers if sellers do not perform well. However, the evidence on how negative evaluations should be weighted against positive ones and whether negative evaluations have a larger negative impact on price for sellers with good reputations than for sellers with less well developed reputations is not so clear yet. Note that we do not address the question of why people provide feedback at all. Clearly, this involves a collective good dilemma in itself, since providing feedback is costly, while there is no direct benefit for oneself in

tab le 3.3

Key References on Evidence about Embeddedness Effects;Type of Data in Brackets. Research designs and mechanisms

Experiments Control

Learning

Surveys Control

Control and/or learning

Learning

Two types of embeddedness Dyad

Network

Camerer and Weigelt 1988; Neral and Ochs 1992; Anderhub et al. 2002; Brandts and Figueras 2003 (finitely repeated Trust Game) Engle-Warnick and Slonim 2004, 2006 (indefinitely as well as finitely repeated Trust Game) Cochard et al. 2004 (finitely repeated Investment Game)

Bolton et al. 2004; Bolton and Ockenfels 2009; Bohnet and Huck 2004 (one-shot Trust Game finitely repeated within a group, while new partners obtain information about everyone’s behavior in the past) Buskens et al. 2010; Barrera 2005; Barrera and Buskens 2009 (finitely repeated Trust Game, trustors also obtain information about behavior of their trustee with another trustor) Bolton et al. 2004; Bolton and Ockenfels 2009; Bohnet and Huck 2004 (one-shot Trust Game finitely repeated within a group, while new partners obtain information about everyone’s behavior in the past) Buskens et al. 2010; Barrera 2005; Barrera and Buskens 2009 (finitely repeated Trust Game, trustors also obtain information about behavior of their trustee with another trustor)

Camerer and Weigelt 1988; Neral and Ochs 1992; Anderhub et al. 2002; Gautschi 2000; Brandts and Figueras 2003 (finitely repeated Trust Game) Kollock 1994 (experimental buyer-seller market) Engle-Warnick and Slonim 2004, 2006 (indefinitely as well as finitely repeated Trust Game) Cochard et al. 2004 (finitely repeated Investment Game) Blumberg 1997 (R&D alliances) Batenburg et al. 2003 (IT transactions) Gulati 1995a, 1995b (strategic alliances) Gulati and Wang 2003 (joint ventures) Baker et al. 1998 (advertising agencies and their clients) Lyons 1994 (subcontractors) Corts and Singh 2004 (off-shore drillers and oil companies) Robinson and Stuart 2007 (strategic alliances)

Blumberg 1997 (R&D alliances) Batenburg et al. 2003 (IT transactions)

Buskens et al. 2000; Rooks et al. 2006 (IT transactions, proximity, third-party relations, exit opportunities) Gulati 1995b; Gulati and Gargiulo 1999 (R&D alliances, third-party relations, centrality) Robinson and Stuart 2007 (strategic alliances, centrality, proximity) Resnick and Zeckhauser 2002; Diekmann et al. 2009; Snijders and Zijdeman 2004; Resnick et al. 2006; Lucking-Reiley et al. 2007; Snijders and Weesie 2009 (Internet auctions)

Rational Choice Research on Social Dilemmas

139

providing feedback. Solving this dilemma would bring us into the literature on self-organizing institutions (see Ostrom 1990; and Janssen 2006 for a specific model related to Internet auctions). The evidence on learning and control effects on trust through social networks or more formal institutions that facilitate information exchange is still far from conclusive. While there is evidence that trust can emerge in dense social networks, it remains unclear what drives the emergence of trust. Is it learning, or is control through the promise of positive and the threat of negative sanctions more important? Presumably, the empirical evidence is also limited because of scarce theoretical explanations that can guide the search for empirical evidence. Researchers have focused primarily on establishing the relationship between network embeddedness and trust considering at most one mechanism that drives this relationship. The distinction between learning and control effects of embeddedness proposed here and the fact that embeddedness facilitates learning as well as control, however, asks for an integrated approach that allows for disentangling these two mechanisms.Table 3.3 offers a summary of key references to empirical research on embeddedness effects. vignette experiments for disentangling control and learning effects of embeddedness Distinguishing empirically between control and learning effects of embeddedness is a complex task. While laboratory experiments mostly allow variation in only a small number of variables, survey research often lacks the necessary control on causes and consequences. As a complement to lab experiments and survey studies, we discuss two vignette experiments in which subjects are presented with hypothetical economic transactions that involve trust problems. The subjects answer questions about their behavior related to these transactions (see Rossi and Nock 1982 on vignette experiments). Vignette experiments are useful in providing more control over the variation of somewhat more key variables, as well as over what the causes of changes in the dependent variable are. In addition, in surveys and experiments, actors are engaged in series of transactions in which opportunities for learning and control often co-occur, while in vignette experiments it is more straightforward to vary opportunities for learning and control independently. These advantages of vignette designs, however, come with two disadvantages. First, the choices are purely hypothetical, which implies that the choices do not have actual consequences for the decision-makers. This questions the actual incentives to choose one or the other option. Second, given that the decision situations are hypothetical, it can be a problem that the decision situation is rather artificial for the decision-maker, which compromises the validity of the decisions. In the vignette experiments, subjects have to imagine themselves in the role of buyers in economic transactions. The description of the situation makes it plausible that buyers face a trust problem by indicating that the transaction partner might have an incentive to behave opportunistically in the transaction. In the first experiment, purchase managers of Dutch companies are asked to answer questions about hypothetical transactions with suppliers (see Rooks et al. 2000). The description of the transactions comprises information about transaction characteristics such as price and specific investments of the buyer associated with the transaction, but also about the relationship of the buyer

140 Vincent Buskens and Werner Raub with the supplier. Four variables are varied that are related to embeddedness: (1) The extent to which the buyer did business with the same supplier in the past (dyadic learning); (2) The extent to which the buyer expects to do business with the supplier in the future (dyadic control); (3) The extent to which the buyer and supplier have common business partners (network embeddedness that provides opportunities for learning as well as control); and (4) The availability of alternative suppliers for the buyer (network control). The dependent variable is lack of trust of the buyer, measured by the extent to which the buyer wants to invest in safeguards (for example, contractual agreements) before the transaction takes place. Results reveal a strong effect of embeddedness on trust resulting from learning within a dyadic relation. Positive past experiences reduce the investment in safeguards. Although there is no main effect of expected interactions in the future, there is indeed a negative interaction effect of past transactions and expected future transactions, indicating that the use of control opportunities is contingent on some previous learning opportunities. This finding is in line with the results of the survey on IT transactions of Batenburg, Raub, and Snijders (2003) discussed above, notably employing a very different research design. Concerning third-party effects, results show that knowing other business partners of the supplier increases trust. It is unclear whether this effect is the result of learning or control, since these third parties can be used to obtain information on previous behavior of the supplier, but they also can be informed on behavior of the supplier in the focal transaction, thus extending control opportunities for the buyer. There is indeed a negative effect of the availability of alternative trading partners on investments of the buyer in safeguards for the transaction. This supports the interpretation that purchase managers realize that alternative suppliers provide them with sanction opportunities, implying that suppliers are less likely to act in an untrustworthy way if they have more competitors. In another vignette experiment, students are asked to compare situations for buying a used car (see Buskens and Weesie 2000b for more details). Students are offered pairs of vignettes describing such a transaction, and they are asked which one they prefer. Five embeddedness variables are varied at the vignettes: (1) Whether the buyer has bought a car from the dealer before and was satisfied, or never bought a car from the dealer (dyadic learning). (2) Whether or not the buyer expects to move to the other side of the country soon (dyadic control). The probability that the buyer has future transactions with the dealer is smaller if the buyer moves. Hence, control is more difficult for a buyer who moves. Theoretically, the effect of expected future transactions is based on the sanctions of the buyer anticipated by the dealer. Therefore, strictly speaking, expected future transactions can be expected to affect the behavior of buyer and dealer only if the dealer is informed about the buyer’s plans to move. (3) Whether the dealer is or is not well known in the neighborhood of the buyer (network density). Again, learning as well as control of a well-known garage through the network of customers can be more effective than learning about or control of a garage that is not well known. (4) Whether or not the buyer has information from friends about transactions of these friends with the garage (network learning), with a focus on the difference between no information and positive information. (5) Whether or not both the buyer and the dealer are members of the same sports team (network control). This measures network

Rational Choice Research on Social Dilemmas

141

control because the number of acquaintances the buyer and dealer have in common is expected to be larger if the buyer and dealer are members of the same sports team. Common membership provides the buyer with possibilities of controlling the dealer through positive or negative reputational sanctions both in business and as a team member. An advantage of this operationalization is that the theoretical assumption of “common knowledge about the network” is unlikely to be violated because the buyer and the dealer both know that they are members of the sports team. Results show that all five embeddedness variables have positive effects on the likelihood that subjects prefer a vignette that includes the respective type of embeddedness over the one in which that type of embeddedness is not available. The strongest effects seem again to be those of dyadic and network learning variables. Positive information clearly enhances trust. Dyadic control, density, and network control likewise have positive effects on trust, which implies that control is important at the dyadic as well as at the network level. The evidence for a control mechanism is somewhat problematic, since these variables are subject to alternative explanations. There might be other disadvantages of buying a car just before you move, for example, because you need to find another garage if there is any problem with the car in the future. Although our arguments and results for network control are in accordance with DiMaggio and Louch (1998), we cannot exclude that actors prefer to trust well-known others over unknown others for other reasons than the control reasons advocated here. Summarizing the evidence from studies employing different research designs, we have quite unambiguous support for hypotheses on learning effects at the dyadic and network level. Also, hypotheses on effects of control opportunities at the dyadic level are quite consistently supported, as can be seen from endgame effects in finitely repeated Trust Games, surveys on transactions as well as the vignette experiments. Network control is less well studied and the evidence is more ambiguous. We would expect that these results might generalize to other kinds of social dilemmas, but we do not know about systematic studies that have compared embeddedness effects for other social dilemmas, including experimental as well as survey studies and distinguishing between different types of embeddedness effects.

Conclusions and Directions for Further Research We have provided a survey of rational choice research on social dilemmas by focusing on how game theory can be used to model social dilemmas, how testable hypotheses can be generated from game-theoretic models, and what empirical evidence tells us about those hypotheses. Trust problems have been our paradigm case of a social dilemma. In terms of the strategies for refining the model of atomized interactions on perfect markets of rational and selfish actors with full information, we have focused on models that retain the rationality assumption. In fact, game-theoretic models often employ particularly strong rationality assumptions. We have briefly considered how relaxing selfishness assumptions by including other-regarding preferences can help in accounting for behavior in social dilemmas and specifically in dilemmas that are isolated encounters. Our main focus, however, has been on effects

142 Vincent Buskens and Werner Raub of social embeddedness. We have concentrated on game-theoretic models that allow for an analysis of how embeddedness affects behavior in social dilemmas. Hence, the bulk of the models surveyed in this chapter relax the assumption of atomized interactions and often also the assumption of full information. We have stressed that game-theoretic models allow us to systematically distinguish different kinds of embeddedness and also to distinguish different mechanisms such as control and learning through which behavior in social dilemmas depends on embeddedness. From the empirical end, we have stressed the need for research designs that make possible discrimination between different types of embeddedness, as well as disentangling control and learning effects. We have also argued for using complementary research designs such as experiments, surveys, vignette studies, and the like as a strategy for establishing the robustness of empirical findings (see Levitt and List 2007 for a thorough discussion of this issue, as well as Falk and Heckman 2009 and Gächter and Thöni 2011). Our overview of studies shows that there is quite some empirical evidence for embeddedness effects. When research designs are employed that do allow for disentangling different kinds of embeddedness effects and mechanisms through which embeddedness works, hypotheses based on game-theoretic models often succeed in predicting the signs of coefficients (see Grofman 1993; Green and Shapiro 1994 for related discussions on the merits and problems of qualitative predictions on changes “at the margin” using comparative statics versus quantitative point predictions from rational choice models). Nevertheless, there is clearly much room for improvement in the predictions of game-theoretic models on behavior in social dilemmas. Roughly speaking, the overall impression is that assuming game-theoretic rationality as well as selfish actors (“utility = own money”) cannot account for quite some nonopportunistic behavior in social dilemmas that are isolated encounters, while it also often predicts “too much” cooperative behavior in repeated social dilemmas (see Bolton and Ockenfels 2009 for a similar point in the context of research on reputation systems in the Internet economy). Developing game-theoretic models on the interplay of social embeddedness and other-regarding preferences may be useful in this respect (see Gintis 2000: ch. 11 for related arguments). With respect to research on social embeddedness, theoretical as well as empirical work reviewed in this chapter assumed embeddedness characteristics as exogenously given. Using a notion that has become popular, embeddedness can be conceived as social capital of actors (see, for example, Coleman 1990). We have focused on the returns on social capital: embeddedness allows for overcoming Pareto-suboptimal outcomes in social dilemmas. What we have neglected are actors’ investments in their social capital (see Flap 2004 and the chapter by Flap and Völker in this Handbook for the distinction between returns on and investments in social capital). However, given the returns on embeddedness in social dilemmas, actors do have an incentive to invest in their embeddedness by strategically establishing, maintaining, or deleting ties to others, including search for potential interaction partners. One would thus like to endogenize embeddedness characteristics. Research on strategic network formation based on game-theoretic models is rather novel but meanwhile rapidly developing (see the textbooks Goyal 2007; Jackson 2008). Such work

Rational Choice Research on Social Dilemmas

143

on returns on and investments in social capital can likewise benefit from the development of “actor-driven” statistical models for the dynamics (“coevolution”) of networks and behavior (see Snijders 2001; Snijders’s chapter in this Handbook). How embeddedness can contribute to trust and cooperation has been a core topic of our chapter. We have thus highlighted the beneficial effects of embeddedness for the actors involved in a social dilemma. Beneficial effects for the actors directly involved can have negative effects for others. From the perspective of third parties or from a societal perspective, undermining rather than fostering cooperation is often the aim. It should be noted, too, that embeddedness can also have adverse effects for the actors who are themselves directly involved in social dilemmas. Focusing on learning and information diffusion rather than game-theoretic rationality as a driving force of behavior, Burt and Knez (1995) have shown that dense networks can amplify trust as well as distrust. The core argument is that because of the homogeneity of opinions in a dense network, actors become convinced about some information because they receive the information disproportionately often. Coethnics may be able to solve trust problems in economic exchange by transacting with each other, but this may lead to entrapment and missing opportunities from outside networks (see, for example, Portes 1998). Flache (2002) offers a game-theoretic model of how informal social ties between the members of a team can undermine cooperation between the members of the team since they have to trade off the benefits of sanctioning team members who do not cooperate against the costs of deteriorating informal social ties through negative sanctions. While there is quite some empirical research on adverse effects of embeddedness, more systematic theoretical modeling of such effects is needed. With respect to theoretical modeling, relaxing strong game-theoretic rationality assumptions or showing that and when equilibrium behavior in accordance with such assumptions is a—possibly long-term—result of bounded rationality and evolutionary or learning processes (for example, Fudenberg and Levine 1998; Gintis 2000) may have useful implications for research on social dilemmas, too. We would like to conclude, though, with a more specific suggestion.We have seen that, in principle, games with incomplete information are a tool for analyzing embeddedness effects in social dilemmas through control and learning in an integrated way. However, as we have also seen, it is difficult to strike a fruitful balance of analytic tractability and realistic assumptions about what information actors have and how they use relevant sources of information. On the one hand, models with more realistic informational assumptions are often difficult to analyze. On the other, knowledge about what realistic informational assumptions would be is limited, because most empirical research has not succeeded in clearly disentangling the effects of learning and control mechanisms. Disentangling the effects could provide some evidence on the relative importance of these mechanisms and evidence about changes in the importance of different effects related to different circumstances. To overcome these limitations, we propose a two-step empirical and theoretical approach. More experimental research is necessary to obtain better insights in the relative importance of the different mechanisms. For example, experiments should be designed such that subjects are involved in abstract Trust

144 Vincent Buskens and Werner Raub Games embedded in a social context that allows for communication among trustors. The experiments should explicitly provide insights in how subjects use information they obtain from other subjects in the network and whether or not they try to sanction by informing other trustors in the network. In this way, the experiments make possible the development of new models built on assumptions, for example, about information exchange that have an empirical basis rather than on assumptions chosen exclusively on the basis of introspection of researchers and mathematical tractability. Moreover, the experiments can be used to obtain initial insights in circumstances that affect the importance of control versus learning. The results of such experiments can inspire new theoretical models on the relative effects of learning versus control. Based on these models, survey designs can then be developed that allow for variations in learning and control variables such that the predicted effects can be distinguished.

Notes The order of authorship is alphabetical. Stimulating comments of and discussions with Jeroen Weesie and other members of our Utrecht group, Cooperation in Social and Economic Relations, are gratefully acknowledged. We also acknowledge helpful comments from participants of the Rational Choice Social Research Workshop and specifically from our discussant, Simon Gächter. Financial support for Buskens was provided by the Royal Netherlands Academy of Arts and Sciences (KNAW) for the project Third-Party Effects in Cooperation Problems, and by Utrecht University for the High Potential-program Dynamics of Cooperation, Networks, and Institutions. Financial support for Raub was provided by the Netherlands Organization for Scientific Research under grants S 96-168 and PGS 50-370 for the PIONIER-program The Management of Matches, and under grant 400-05-089 for the project Commitments and Reciprocity. 1. See Ostrom (2003) for first steps toward a theoretical framework combining both strategies. 2. For a textbook accessible to readers with modest training in formal theoretical model building and no prior exposure to game theory, see Rasmusen (2007). 3. In contrast, the expression derived for the temptation in isolated encounters using guilt incorporates also the trustor’s payoff S1. 4. Coleman, in his meanwhile classic sketch, clearly intuited this result when he argued that an important feature of socialization is “coming to see the long-term consequences to oneself of particular strategies of action,” rather than the internalization of norms (1964: 180).Voss (1982) seems to be the first sociologist who realized explicitly that the theory of repeated games has important implications for the problem of order and cooperation in social dilemmas. 5. One might argue that learning is still possible in these models, since there are many equilibria and it is not clear why actors should choose the same equilibrium to start with. We disregard this issue, assuming that actors coordinate instantly on the same equilibrium (see Fudenberg and Levine 1998: 20). 6. There are sizable parallel literatures on experiments with repeated social dilemmas like the Prisoner’s Dilemma (see Dawes 1980; Colman 1982: ch. 7; Sally 1995 for overviews), Public Goods Games (overview: Ledyard 1995), and still other strategic interactions (overview: Camerer 2003). 7. Experiments employing the closely related Gift-Exchange Game (Van der Heijden et al. 2001; Gächter and Falk 2002) likewise show that repeated play increases the efficiency of outcomes and that endgame effects occur in a similar manner.

Rational Choice Research on Social Dilemmas

145

References Anderhub, Vital, Dirk Engelmann, and Werner Güth. 2002. “An Experimental Study of the Repeated Trust Game with Incomplete Information.” Journal of Economic Behavior and Organization 48: 197−216. Axelrod, Robert. 1984. The Evolution of Cooperation. New York: Basic Books. Baker,Wayne E., Robert R. Faulkner, and Gene A. Fisher. 1998. “Hazards of the Market: The Continuity and Dissolution of Interorganizational Market Relationships.” American Sociological Review 63: 147−77. Barrera, Davide. 2005. Trust in Embedded Settings.Veenendaal: Universal Press. Barrera, Davide, and Vincent Buskens. 2009. “Third-Party Effects.” In eTrust: Forming Relationships in the Online World, edited by Karen S. Cook, Chris Snijders, Vincent Buskens, and Coye Cheshire, 37–72. New York: Russell Sage. Batenburg, Ronald S.,Werner Raub, and Chris Snijders. 2003. “Contacts and Contracts: Temporal Embeddedness and the Contractual Behavior of Firms.” Research in the Sociology of Organizations 20: 135−88. Bendor, Jonathan, and Dilip Mookherjee. 1987. “Institutional Structure and the Logic of Ongoing Collective Action.” American Political Science Review 81: 129−54. Berg, Joyce, John Dickhaut, and Kevin McCabe. 1995. “Trust, Reciprocity, and Social History.” Games and Economic Behavior 10: 122−42. Binmore, Ken. 1998. Game Theory and the Social Contract. Vol. 2: Just Playing. Cambridge, MA: MIT Press. Blau, Peter M. 1996 [1964]. Exchange and Power in Social Life. New Brunswick, NJ: Transaction Publishers. Blumberg, Boris F. 1997. Das Management von Technologiekooperationen: Partnersuche und Verhandlungen mit dem Partner aus empirisch-theoretischer Perspektive. Amsterdam: Thesis Publishers. Bohnet, Iris, and Steffen Huck. 2004. “Repetition and Reputation: Implications for Trust and Trustworthiness When Institutions Change.” American Economic Review (Papers and Proceedings) 94: 362−66. Bolton, Gary E., Elena Katok, and Axel Ockenfels. 2004. “How Effective Are Online Reputation Mechanisms? An Experimental Study.” Management Science 50: 1587−602. Bolton, Gary E., and Axel Ockenfels. 2009. “The Limits of Trust in Economic Transactions: Investigations of Perfect Reputation Systems.” In eTrust: Forming Relationships in the Online World, edited by Karen S. Cook, Chris Snijders, Vincent Buskens, and Coye Cheshire, 15–36. New York: Russell Sage. Bower, Anthony, Steven Garber, and Joel C.Watson. 1997. “Learning about a Population of Agents and the Evolution of Trust and Cooperation.” International Journal of Industrial Organization 15: 165−90. Brandts, Jordi, and Neus Figueras. 2003. “An Exploration of Reputation Formation in Experimental Games.” Journal of Economic Behavior and Organization 50: 89−115. Burt, Ronald S., and Marc Knez. 1995. “Kinds of Third-Party Effects on Trust.” Rationality and Society 7: 255−92. Buskens,Vincent. 2002. Social Networks and Trust. Boston: Kluwer. ———. 2003. “Trust in Triads: Effect of Exit, Control, and Learning.” Games and Economic Behavior 42: 235−52. Buskens, Vincent, and Werner Raub. 2002. “Embedded Trust: Control and Learning.” Advances in Group Processes 19: 167−202. Buskens, Vincent, Werner Raub, and Joris van der Veer. 2010. “Trust in Triads: An Experimental Study.” Social Networks 32: 301–12. Buskens, Vincent, Werner Raub, and Jeroen Weesie. 2000. “Networks and Contracting

146 Vincent Buskens and Werner Raub in Information Technology Transactions.” In The Management of Durable Relations: Theoretical and Empirical Models for Organizations and Households, edited by Werner Raub and Jeroen Weesie, 77–81. Amsterdam: Thela Thesis. Buskens, Vincent, and Jeroen Weesie. 2000a. “Cooperation via Networks.” Analyse & Kritik 22: 44−74. ———. 2000b. “An Experiment on the Effects of Embeddedness in Trust Situations: Buying a Used Car.” Rationality and Society 12: 227−53. Camerer, Colin F. 2003. Behavioral Game Theory: Experiments in Strategic Interaction. New York: Russell Sage. Camerer, Colin F., and Keith Weigelt. 1988. “Experimental Tests of a Sequential Equilibrium Reputation Model.” Econometrica 56: 1−36. Cochard, Francois, Phu Nguyen Van, and Marc Willinger. 2004. “Trusting Behavior in a Repeated Investment Game.” Journal of Economic Behavior and Organization 55: 31−44. Coleman, James S. 1964. “Collective Decisions.” Sociological Inquiry 34: 166−81. ———. 1987. “Psychological Structure and Social Structure in Economic Models.” In Rational Choice, edited by Robin M. Hogarth and Melvin W. Reder, 181−85. Chicago: University of Chicago Press. ———. 1990. Foundations of Social Theory. Cambridge, MA: Belknap Press of Harvard University Press. Colman, Andrew M. 1982. Game Theory and Experimental Games: The Study of Strategic Interactions. Oxford: Pergamon Press. Cook, Karen S., and Robin M. Cooper. 2003. “Experimental Studies of Cooperation, Trust, and Social Exchange.” In Trust and Reciprocity: Interdisciplinary Lessons from Experimental Research, edited by Elinor Ostrom and James Walker, 209–44. New York: Russell Sage. Corts, Kenneth S., and Jasjit Singh. 2004. “The Effect of Repeated Interaction on Contract Choice: Evidence from Offshore Drilling.” Journal of Law, Economics and Organization 20: 230−60. Dasgupta, Partha. 1988.“Trust as a Commodity.” In Trust: Making and Breaking Cooperative Relations, edited by Diego Gambetta, 49–72. Oxford: Blackwell. Dawes, Robert M. 1980. “Social Dilemmas.” Annual Review of Psychology 31: 169−93. Diekmann, Andreas. 2004. “The Power of Reciprocity.” Journal of Conflict Resolution 48: 487−505. Diekmann, Andreas, Ben Jann, and David Wyder. 2009. “Trust and Reputation in Internet Auctions.” In eTrust: Forming Relationships in the Online World, edited by Karen S. Cook, Chris Snijders, Vincent Buskens, and Coye Cheshire, 139–65. New York: Russell Sage. DiMaggio, Paul, and Hugh Louch. 1998. “Socially Embedded Consumer Transactions: For What Kinds of Purchases Do People Most Often Use Networks?” American Sociological Review 63: 619−37. Durkheim, Emile. 1973 [1893]. De la Division du Travail Social. 9th ed. Paris: PUF. Engle-Warnick, Jim, and Robert L. Slonim. 2004. “The Evolution of Strategies in a Repeated Trust Game.” Journal of Economic Behavior and Organization 55: 553−73. ———. 2006. “Learning to Trust in Indefinitely Repeated Games.” Games and Economic Behavior 54: 95−114. Falk, Armin, and James J. Heckman. 2009. “Lab Experiments Are a Major Source of Knowledge in the Social Sciences.” Science 326: 535–38. Fehr, Ernst, and Herbert Gintis. 2007. “Human Motivation and Social Cooperation: Experimental and Analytical Foundations.” Annual Review of Sociology 33: 43−64. Fehr, Ernst, and Klaus M. Schmidt. 1999. “A Theory of Fairness, Competition, and Cooperation.” Quarterly Journal of Economics 114: 817−68.

Rational Choice Research on Social Dilemmas

147

———. 2006. “The Economics of Fairness, Reciprocity and Altruism—Experimental Evidence and New Theories.” In Handbook of the Economics of Giving, Altruism and Reciprocity, edited by Serge-Christophe Kolm and Jean Mercier Ythier, 615–91. Amsterdam: Elsevier. Flache, Andreas. 2002. “The Rational Weakness of Strong Ties.” Journal of Mathematical Sociology 26: 189−216. Flache, Andreas, and Michael W. Macy. 2002. “Stochastic Collusion and the Power Law of Learning: A General Reinforcement Learning Model of Cooperation.” Journal of Conflict Resolution 46: 629−53. Flap, Henk. 2004. “Creation and Returns of Social Capital.” In Creation and Returns of Social Capital, edited by Henk Flap and Beate Völker, 3–23. London: Routledge. Flap, Henk, and Beate Völker. “Social Capital.” Chapter in this Handbook. Friedman, James W. 1971. “A Non-Cooperative Equilibrium for Supergames.” Review of Economic Studies 38: 1–12. ———. 1990. Game Theory with Applications to Economics. 2nd ed. New York: Oxford University Press. Fudenberg, Drew, and David K. Levine. 1998. The Theory of Learning in Games. Cambridge, MA: MIT Press. Fudenberg, Drew, and Eric Maskin. 1986. “The Folk Theorem in Repeated Games with Discounting or with Incomplete Information.” Econometrica 54: 533−54. Gächter, Simon. “Rationality, Social Preferences, and Strategic Decision-making from a Behavioral Economics Perspective.” Chapter in this Handbook. Gächter, Simon, and Armin Falk. 2002. “Reputation and Reciprocity: Consequences for the Labour Relation.” Scandinavian Journal of Economics 104: 1−26. Gächter, Simon, and Christian Thöni. 2011. “Micromotives, Microstructure, and Macrobehavior:The Case of Voluntary Cooperation.” Journal of Mathematical Sociology 35: 26−65. Gautschi, Thomas. 2000. “History Effects in Social Dilemma Situations.” Rationality and Society 12: 131−62. Gibbons, Robert. 2001. “Trust in Social Structures: Hobbes and Coase Meet Repeated Games.” In Trust in Society, edited by Karen S. Cook, 332–53. New York: Russell Sage. Gintis, Herbert. 2000. Game Theory Evolving. Princeton: Princeton University Press. Goldthorpe, John H. 2000. On Sociology: Numbers, Narratives, and the Integration of Research and Theory. Oxford: Oxford University Press. Gouldner, Alvin W. 1960. “The Norm of Reciprocity.” American Sociological Review 25: 161−78. Goyal, Sanjeev. 2007. Connections: An Introduction to the Economics of Networks. Princeton: Princeton University Press. Granovetter, Mark S. 1985. “Economic Action and Social Structure: The Problem of Embeddedness.” American Journal of Sociology 91: 481−510. Green, Donald P., and Ian Shapiro. 1994. Pathologies of Rational Choice Theory: A Critique of Applications in Political Science. New Haven, CT:Yale University Press. Grofman, Bernard. 1993. “Is Turnout the Paradox That Ate Rational Choice Theory?” In Information, Participation, and Choice, edited by Bernard Grofman, 93–103. Ann Arbor: University of Michigan Press. Gulati, Ranjay. 1995a. “Does Familiarity Breed Trust? The Implications of Repeated Ties for Contractual Choice in Alliances.” Academy of Management Journal 38: 85−112. ———. 1995b. “Social Structure and Alliance Formation Patterns: A Longitudinal Study.” Administrative Science Quarterly 40: 619−52. Gulati, Ranjay, and Martin Gargiulo. 1999. “Where Do Interorganizational Networks Come From?” American Journal of Sociology 104: 1439−93.

148 Vincent Buskens and Werner Raub Gulati, Ranjay, and Lihua Wang. 2003. “Size of the Pie and Share of the Pie: Implications of Structural Embeddedness for Value Creation and Value Appropriation in Joint Ventures.” Research in the Sociology of Organizations 20: 209−42. Güth, Werner, Rolf Schmittberger, and Bernd Schwarze. 1982. “An Experimental Analysis of Ultimatum Bargaining.” Journal of Economic Behavior and Organization 3: 367−88. Harsanyi, John C. 1975. Essays on Ethics, Social Behavior, and Scientific Explanation. Dordrecht: Reidel. Hirschman, Albert O. 1970. Exit, Voice, and Loyalty: Responses to Decline in Firms, Organizations, and States. Cambridge, MA: Harvard University Press. Hirshleifer, David, and Eric Rasmusen. 1989. “Cooperation in a Repeated Prisoner’s Dilemma with Ostracism.” Journal of Economic Behavior and Organization 12: 87−106. Hobbes, Thomas. 1991 [1651]. Leviathan. Cambridge: Cambridge University Press. Jackson, Matthew O. 2008. Social and Economic Networks. Princeton: Princeton University Press. Janssen, Marco. 2006. “Evolution of Cooperation When Feedback to Reputation Scores Is Voluntary.” Journal of Artificial Societies and Social Simulation 9: . Kahneman, Daniel, Jack L. Knetsch, and Richard Thaler. 1986. “Fairness as a Constraint on Profit Seeking: Entitlements in the Market.” American Economic Review 76: 728−41. Kiser, Edgar, and Erin Powers. “Rational Choice Approaches to State-Making.” Chapter in this Handbook. Kollock, Peter. 1994. “The Emergence of Exchange Structures: An Experimental Study of Uncertainty, Commitment, and Trust.” American Journal of Sociology 100: 313−45. ———. 1998. “Social Dilemmas: The Anatomy of Cooperation.” Annual Review of Sociology 24: 183−214. ———. 1999. “The Production of Trust in Online Markets.” Advances in Group Processes 16: 99−123. Kreps, David M. 1990. “Corporate Culture and Economic Theory.” In Perspectives on Positive Political Economy, edited by James E. Alt and Kenneth A. Shepsle, 90–143. Cambridge: Cambridge University Press. Kreps, David M., and Robert Wilson. 1982. “Sequential Equilibria,” Econometrica 50: 863−94. Ledyard, John O. 1995. “Public Goods: A Survey of Experimental Research.” In The Handbook of Experimental Economics, edited by John H. Kagel and Alvin E. Roth, 11–94. Princeton: Princeton University Press. Levitt, Steven D., and John A. List. 2007. “What Do Laboratory Experiments Measuring Social Preference Reveal about the Real World?” Journal of Economic Perspectives 21: 153−74. Lorenz, Edward H. 1988. “Neither Friends Nor Strangers: Informal Networks of Subcontracting in French Industry.” In Trust: Making and Breaking Cooperative Relations, edited by Diego Gambetta, 94–107. Oxford: Blackwell. Lucking-Reiley, David, Doug Bryan, Naghi Prasad, and Daniel Reeves. 2007. “Pennies from eBay: The Determinants of Price in Online Auctions.” Journal of Industrial Economics 55: 223−33. Lyons, Bruce R. 1994. “Contracts and Specific Investments: An Empirical Test of Transaction Cost Theory.” Journal of Economics and Management Strategy 3: 257−78. Macy, Michael W., and Andreas Flache. 1995. “Beyond Rationality in Models of Choice.” Annual Review of Sociology 21: 73−91. ———. 2002. “Learning Dynamics in Social Dilemmas.” Proceedings of the National Academy of Sciences U.S.A. 99: 7229−36.

Rational Choice Research on Social Dilemmas

149

Nash, John F. 1951. “Non-Cooperative Games.” Annals of Mathematics 54: 286−95. Neral, John, and Jack Ochs. 1992. “The Sequential Equilibrium Theory of Reputation Building: A Further Test.” Econometrica 60: 1151−69. Ortmann, Andreas, John Fitzgerald, and Carl Boeing. 2000. “Trust, Reciprocity, and Social History: A Re-examination.” Experimental Economics 3: 81−100. Ostrom, Elinor. 1990. Governing the Commons: The Evolution of Institutions for Collective Action. Cambridge: Cambridge University Press. ———. 2003. “Toward a Behavioral Theory Linking Trust, Reciprocity, and Reputation.” In Trust and Reciprocity: Interdisciplinary Lessons from Experimental Research, edited by Elinor Ostrom and James Walker, 19–79. New York: Russell Sage. Parsons, Talcott. 1937. The Structure of Social Action. New York: Free Press. Portes, Alejandro. 1998. “Social Capital: Its Origins and Applications in Modern Sociology.” Annual Review of Sociology 24: 1−24. Rapoport, Anatol. 1974. “Prisoner’s Dilemma—Recollections and Observations.” In Game Theory as a Theory of Conflict Resolution, edited by Anatol Rapoport, 18–34. Dordrecht: Reidel. Rasmusen, Eric. 2007. Games and Information: An Introduction to Game Theory. 4th ed. Oxford: Blackwell. Raub, Werner,Vincent Buskens, and Marcel van Assen. 2011. “Micro-Macro Links and Microfoundations in Sociology.” Journal of Mathematical Sociology 35: 1−25. Raub, Werner, and Jeroen Weesie. 1990. “Reputation and Efficiency in Social Interactions: An Example of Network Effects.” American Journal of Sociology 96: 626−54. Resnick, Paul, and Richard Zeckhauser. 2002. “Trust among Strangers in Internet Transactions: Empirical Analysis of eBay’s Reputation System.” Advances in Applied Microeconomics 11: 127−57. Resnick, Paul, Richard Zeckhauser, John Swanson, and Kate Lockwood. 2006. “The Value of Reputation on eBay: A Controlled Experiment.” Experimental Economics 9: 79−101. Robinson, David T., and Toby E. Stuart. 2007. “Network Effects in the Governance of Strategic Alliances.” Journal of Law, Economics and Organization 23: 242−73. Rooks, Gerrit, Werner Raub, Robert Selten, and Frits Tazelaar. 2000. “Cooperation between Buyer and Supplier: Effects of Social Embeddedness on Negotiation Effort.” Acta Sociologica 43: 123−37. Rooks, Gerrit, Werner Raub, and Frits Tazelaar. 2006. “Ex Post Problems in BuyerSupplier Transactions: Effects of Transactional Characteristics, Social Embeddedness, and Contractual Governance.” Journal of Management and Governance 10: 239–76. Rossi, Peter H., and Steven L. Nock, eds. 1982. Measuring Social Judgments: The Factorial Survey Approach. Beverly Hills, CA: Sage. Sally, David. 1995. “Conversation and Cooperation in Social Dilemmas: A MetaAnalysis of Experiments from 1958 to 1992.” Rationality and Society 7: 58−92. Schüßler, Rudolf A. 1989. “Exit Threats and Cooperation under Anonymity.” Journal of Conflict Resolution 33: 728−49. Snijders, Chris. 1996. Trust and Commitments. Amsterdam: Thesis Publishers. Snijders, Chris, and Gideon Keren. 1999. “Determinants of Trust.” In Games and Human Behavior, edited by David V. Budescu, Ido Erev, and Rami Zwick, 355–85. Mahwah, NJ: Lawrence Erlbaum. ———. 2001. “Do You Trust? Whom Do You Trust? When Do You Trust?” Advances in Group Processes 18: 129−60. Snijders, Chris, and Jeroen Weesie. 2009. “Online Programming Markets.” In eTrust: Forming Relationships in the Online World, edited by Karen S. Cook, Chris Snijders, Vincent Buskens, and Coye Cheshire, 166–85. New York: Russell Sage.

150 Vincent Buskens and Werner Raub Snijders, Chris, and Richard Zijdeman. 2004. “Reputation and Internet Auctions: eBay and Beyond. Analyse & Kritik 26: 158−84. Snijders, Tom A. B. 2001. “The Statistical Evaluation of Social Network Dynamics.” Sociological Methodology 29: 361−95. Snijders, Tom A. B. “Network Dynamics.” Chapter in this Handbook. Taylor, Michael. 1987. The Possibility of Cooperation. Cambridge: Cambridge University Press. Rev. ed. of Anarchy and Cooperation. London: Wiley 1976. Uzzi, Brian. 1996. “The Sources and Consequences of Embeddedness for the Economic Performance of Organizations:The Network Effect.” American Sociological Review 61: 674−98. Van der Heijden, C. M. Eline, Jan H. M. Nelissen, Jan J. M. Potters, and Harrie A. A. Verbon. 2001. “Simple and Complex Gift Exchange in the Laboratory.” Economic Inquiry 39: 280−97. Vanberg, Viktor J., and Roger D. Congleton. 1992. “Rationality, Morality, and Exit.” American Political Science Review 86: 418−31. Voss, Thomas. 1982. “Rational Actors and Social Institutions: The Case of the Organic Emergence of Norms.” In Theoretical Models and Empirical Analyses: Contributions to the Explanation of Individual Actions and Collective Phenomena, edited by Werner Raub, 76−100. Utrecht: ESP. Weber, Max. 1976 [1921]. Wirtschaft und Gesellschaft. 5th ed. Tübingen: Mohr. ———. 1947. The Theory of Social and Economic Organization. New York: Free Press. Wechsberg, Joseph. 1966. The Merchant Bankers. New York: Bedminster Press. Weesie, Jeroen, Vincent Buskens, and Werner Raub. 1998. “The Management of Trust Relations via Institutional and Structural Embeddedness.” In The Problem of Solidarity:Theories and Models, edited by Patrick Doreian and Thomas Fararo, 113–38. Amsterdam: Gordon and Breach. Williamson, Oliver E. 1996. The Mechanisms of Governance. New York: Oxford University Press. Wittek, Rafael, Tom A. B. Snijders, and Victor Nee. Introduction in this Handbook. Yamagishi, Toshio, and Midori Yamagishi. 1994. “Trust and Commitment in the United States and Japan.” Motivation and Emotion 18: 129−66.

chapter

Modeling Collective Decision-making

4

frans n. stokman, jelle van der knoop, and reinier c. h. van oosten

What kind of mental switch is flipped that changes people’s attitudes? And what kind of switch turns chimpanzee group mates into each other’s deadliest foes? I suspect the switches operate similarly in humans and apes and are controlled by the perception of shared versus competing interests. So long as individuals feel a common purpose, they suppress negative feelings. But as soon as the common purpose is gone, tensions rise to the surface. —Frans De Waal, Our Inner Ape (London: Granta Books, 2005), 135–36.

From Power Studies to Modeling Fundamental Processes When in the 1950s empirical studies on local power started, their main topic was the distribution of power in local societies. It gave rise to a huge debate about the concepts and measurement of power and influence. Ideological, theoretical, and measurement issues colored the empirical results and made them incomparable. The debate concentrated around two empirical local studies in that period, the study of Hunter (1953) in Atlanta and the New Haven study of Dahl (1958, 1961). Hunter represents the group of scholars who believed that the United States was ruled by a power elite, a group that was strongly inspired by the work of C. Wright Mills, best known by his later book The Power Elite (Mills 1956). Dahl represents the group of scholars who strongly believed that the United States is ruled by a plurality of groups. Starting from these opposite perspectives in the so-called elitist-pluralist debate, their definitions and measures of power could hardly do anything else than confirm their view of American politics. But the main benefit of the debate is that it revealed the necessity to reflect about different dimensions in the concepts of power and influence and how these dimensions have to be represented in measures. The first important distinction concerns the question of whether power and influence have to be defined as capacities (Hunter 1953) or as actual effectuation (Dahl 1958, 1961). As effectuation of power and influence depends heavily on the amount of perceived interest of the stakeholder in the problem and the issues involved, it is important to differentiate between the two and to define and measure power and influence as capabilities. The perceived interest

152 Stokman,Van der Knoop, and Van Oosten of an actor in the problem can then be characterized as the percentage or fraction of potential resources that a stakeholder will mobilize. Moreover, there are other restrictions that might hinder the actual effectuation of a stakeholder’s influence. Second, are the concepts of power and influence interchangeable, or do they refer to different phenomena? In more complex contexts, collective outcomes become binding through institutional arrangements. The formal aspects of decision-making consist of the identification of the actors who are legally or otherwise formally charged with taking the decision. This is particularly obvious in political decision-making. Such formal procedures often mean that stakeholders who have no formal right to codetermine the decision outcomes have very high stakes in those decisions. In Western democracies final decisionmaking is allocated to parliaments, composed of elected representatives who take the final decisions. Some political theories, such as the one of Schumpeter (1943), identify democratic decision-making with democratically taken decisions. Other theories stress that democratic procedures are only a necessary, but not a sufficient, condition for democratic decision-making (Bachrach and Baratz 1962; Lukes 1974). They stress that content and quality of decisions should be part of the evaluation of the democratic character of decisions. In their view, a decision should be based on a “balanced” weighing of different interests in a society. To arrive at such a balanced weighing, democracies recognize the right of assembly and free expression of opinion and often require certain consultations and hearings as part of the decision-making process. Particularly within this normative frame, we expect that authorities receive social approval when they weigh the intensity of interests and relative influence of different societal actors properly. Errors, particularly frequent errors, will result in serious social conflicts and poor implementation, which will reduce the likelihood of the authorities being re-elected. The power of actors in social systems is consequently not based solely on their voting power in the final decisionmaking stage, but also on actors’ ability to have their interests reflected in the final decisions. The latter we denote influence (Mokken and Stokman 1976). Third, should influence be measured as a relational variable, or as a characteristic of the stakeholder, or as a combination of the two? Influence is strongly determined by direct or indirect access to authorities, those actors who are formally empowered to take decisions. The increasing analytic possibilities of social network analysis gave rise to a large number of network studies to investigate power centers among large corporations (Mintz and Schwartz 1984; Stokman et al. 1985; Mizruchi 1982; Heemskerk and Fennema 2009; Windolf 2009), intellectual groups (Kadushin 1968), and between large corporations and government agencies (Mokken and Stokman 1979). On the other hand, influence also depends on resources of actors they can mobilize, resources to persuade authorities or to force them to take certain interests into account. One essential resource is information, particularly very specialized information. Numbers might also matter, such as the number of people a stakeholder, such as a trade union for example, can mobilize. The importance of different resources depends on the context in which the collective decision is taken. For example, a country’s military resources are unlikely to be relevant when international banking regulations are being debated. The three essential elements of power

Modeling Collective Decision-making 153

and influence in collective decision-making are, therefore, voting power in final decision-making, timely access, and resources. Voting power, access, and resources determine the potential power and influence of actors. In the bargaining stage, the actual mobilization of an actor’s potential influence depends on three other elements. First of all, it depends on how strongly the decision affects important goals of an actor, the issue salience of the decision for the actor. The other two elements are: (a) the degree to which actors expect the outcome will deviate from their preferred outcome, and (b) whether their participation is expected to have a positive effect through the mobilization of their resources (Zelditch and Ford 1994; Stokman and Stokman 1995). This implies that theories of collective decision-making cannot be based only on the three power elements of the actors, but also have to take into account their issue salience and their preference regarding the outcome. Power becomes visible only if actors have diverging preferences regarding decisions of sufficiently high salience to them. Similarly, if the status quo reflects the interests of the powerful, they are likely to prevent decision-making rather than exercise voting power and influence in the decision-making process. This phenomenon is called “non–decision making,” now better known as “agenda setting” (Cobb and Elder 1972; Kingdon 1995 [1984]; Ordeshook 1992; Tsebelis 1994; and many others). The distinction between power and influence is strongly related to the common conception of collective decision-making in many political systems, consisting of an influence stage followed by a voting stage. Achen (2006a: 86) notes that this general conception has been shared by a broad range of studies, including the work of Bentley (1967). Stokman and Van den Bos (1992) formalized this conception in their two-stage model of policy-making. At the bargaining stage, actors attempt to win support for the decision outcomes they favor most (denoted their policy positions). During this bargaining stage, actors employ a range of strategies in pursuit of this goal. As a consequence of bargaining, actors may end up supporting policy positions other than those they originally took. We refer to these new positions as actors’ voting positions. In the second stage, the voting stage, the process consists of the transformation of the voting positions into one outcome that is binding for all. This implies that the processes in the two stages are fundamentally different. In the bargaining stage policy positions are transformed into voting positions; in the voting stage voting positions are transformed into binding decisions. In complex systems, a final outcome may well be based on a repeated chain of these two stages, such as a decision-making process at three levels in the government and in two chambers of Parliament. Collective decision-making is necessary in any situation in which people wish to achieve things that can often only be achieved, or can be achieved more efficiently, with the contributions of others. This is referred to as joint production (Lindenberg and Foss 2011). Joint production requires collective decisions about what actions should be taken to realize shared interests: who should deliver which contributions, and how should the added value of the joint production be divided. But collective decision-making itself is also a special case of joint production, because individuals involved in such decisions are mutually dependent in making the required decisions. The joint product in

154 Stokman,Van der Knoop, and Van Oosten collective decision-making is a collective decision that is binding for all actors in the social system. Consider the wide range of situations in which people take collective decisions. Families take collective decisions about how to spend and save, where to live, and about the distribution of household tasks. Management boards of businesses and nonprofit organizations take collective decisions about what strategies to implement. Public policies in democracies are collective decisions taken by groups of elected representatives, often after consultations with affected stakeholders. In all these contexts, collective decision-making is the process in which stakeholders have to transform their different preferences into a single collective decision that binds all actors within the social system. In doing so, all actors try to influence the decision outcome, including efforts of some of them to prevent decision-making for the preservation of the status quo. Seen from this perspective, not power or influence but interest alignment is the key to understanding collective decision-making: how diverging preferences for collective outcomes nevertheless result in one collective outcome that is binding for all. Such an analysis requires a focus on and specification of fundamental processes by which interest alignment takes place, even when we realize that actors have different capabilities to do so and differ in their perceptions of how much of their interests are at stake. Joint production inevitably involves both shared and conflicting interests in the perceptions of the stakeholders (Stokman and Vieth 2005). Shared interests result from the perceived added value of the joint product; conflicting interests from the perceptions of the division of the added value and the division of the individual contributions to the joint production. We will show that the perceptions of the relative weight of shared and conflicting interests strongly affect the type of process we expect to emerge in different collective decisionmaking settings. This perception also determines the intensity with which people try to influence the collective decision outcome in line with their own position, versus their willingness to compromise in order to arrive at a broadly supported common position. Interest alignment implies coalition building. The dynamics in collective decision-making processes result from the simultaneous efforts of stakeholders with different policy positions to build as large coalitions as possible around their own positions. This implies that stakeholders are willing or forced to support other positions than they started with. Most studies study just one such process, without specifying the conditions under which that process is likely to take place. Here we specify three such processes: persuasion, logrolling, and enforcement. Each of them is associated with a specific type of network as well. We argue that in any decision-making context all three processes and associated networks take place simultaneously, but that only one of them is dominant. We specify the conditions under which each is likely to be dominant and under which conditions the logrolling and enforcement processes are likely to support or undermine the persuasion process. Characteristic of earlier studies on coalition processes is that they study only one and do not specify the conditions under which that process is likely to dominate. So-called contagion models (Friedkin and Johnsen 1990, 1997, 1999;

Modeling Collective Decision-making 155

Marsden and Friedkin 1993; Leenders 1995, 2002) assume that actors’ opinions and attitudes in a social system depend only partially on individual characteristics, and that these opinions and attitudes are also shaped by social influence. Social influence is represented in the form of an influence network, reflecting the dyadic influence of actors on each other. Technically, spatial autocorrelation algorithms are used to capture such processes. In the social influence part of the model, a person’s opinions or attitudes are modeled as the weighted mean of the opinions or attitudes of the people who have an influence relationship with that person. In the literature, a large variety of weights have been proposed, as Leenders (2002) has shown. Dynamic network models combine influence and selection effects in networks and investigate the relative impacts of the two effects (Stokman and Zeggelink 1996a; Steglich, Snijders, and Pearson 2010). Stokman and Zeggelink (1996b) and Stokman and Berveling (1998) connect these types of models with the fundamental step of aggregating the individual opinions to a collective outcome, a step the other models do not make. Moreover, they do not specify the conditions under which these processes are expected to occur. If we consider other approaches to social exchange in the literature, we find two main alternatives. The first consists of extensions of Coleman’s exchange model that incorporate networks (1972, 1990). Coleman assumed that actors have an interest in some events and control over others. By exchanging control over events in which they are less interested for control over events in which they are more interested, mutually beneficial outcomes can be achieved. The main mechanism in this model is that of a market. The model is able to predict the division of control among the actors in equilibrium. Power (and value of the events) is derived from the model, rather than being introduced in an ad hoc fashion. While the original Coleman model assumed that exchange possibilities are unrestricted, later models introduced the concept of unequal exchange opportunities by connecting Coleman’s exchange model to networks (Marsden and Laumann 1977; Laumann, Knoke, and Kim 1987; Knoke et al. 1996; König 1997; Pappi and Henning 1998). In these models, structural constraints force actors to exchange with particular other actors. Moreover, the models were adapted to predict outcomes on issues on which there are only two policy alternatives (such as yes or no). Coleman’s model thus became extended to outcomes of collective decision-making processes. The second approach to exchange consists of Network Exchange Models (see, for example, Bienenstock and Bonacich 1992; Cook and Yamagishi 1992; Friedkin 1992; Markovsky, Willer, and Patton 1988; Skvoretz and Willer 1993; Willer 1999).Whereas Coleman’s model is based on global equilibria, Network Exchange Models focus on network effects on exchange rates between pairs of actors. Actors’ power derives primarily from the possibilities they have to exclude others from exchange. This power is defined in terms of shifts of exchange rates to an actor’s own advantage. A difference between our work and work in Network Exchange Theory is that the latter deals mainly with exchanges of private goods (for an exception, see Dijkstra and Van Assen 2006, 2008a,b,c). Private goods are also the starting point for Coleman’s models, and generalizations to public goods are not straightforward (Stokman and Van Oosten 1994). We, however, investigate exchanges of voting positions. Changes

156 Stokman,Van der Knoop, and Van Oosten in voting positions affect all stakeholders in collective decision-making and have externalities for other actors (Van Assen, Stokman, and Van Oosten 2003; Dijkstra,Van Assen, and Stokman 2008). Another important difference between our approach and studies informed by Network Exchange Theory is that the latter study given and static networks (see Willer and Willer 2000 for an exception). In contrast, we derive exchange networks from the distribution of positions and saliences of the stakeholders on the issues (see Figure 4.3). As the processes in the bargaining stage and the voting stage are fundamentally different, game-theoretical models of the two stages are expected to be fundamentally different. Models of the bargaining stage formalize different views on the nature of the bargaining process, which results in shifts in actors’ positions. Models of the voting stage formalize different views on the way in which the procedural rules of decision-making affect decision outcomes. There are cooperative and noncooperative models of both the bargaining stage and the voting stage. The most important difference between cooperative and noncooperative models is that agreements between stakeholders are considered to be more or less binding in cooperative models, whereas they are not in noncooperative models. As agreements in noncooperative models are not binding, actors continually evaluate proposals against the status quo and other proposals in terms of the utility they provide them. As a consequence, the status quo plays a much more dominant role in noncooperative models than in cooperative ones as a reference point for evaluating support for different proposals. Table 4.1 summarizes the main classes of models of each of the two stages. Cooperative models of the voting stage were first dominant, starting with onedimensional coalition models (Axelrod 1970; De Swaan 1973). These models were later extended to multidimensional models (Schofield 1976; Laver and Schofield 1990). More recently, noncooperative models for the voting stage have become dominant in the new institutional approach (Baron and Ferejohn 1989; Austen-Smith and Banks 2005). A winset is a set of possible decision outcomes that improve the utility of a required majority of decision-makers relative to the status quo. If winsets are based on a careful analysis of procedural rules, the models are known as procedural models (see Steunenberg and Selck 2006 for an overview). As procedural rules are particularly complex in the European Union, the European Union is an attractive field of application for these models (ibid.; Tsebelis 1994; Moser 1996; Steunenberg 1994; Garret and Tsebelis 1999a,b; Hosli 1993, 1997; Lane and Maeland 1995; Widgrén 1994, 1995; Tsebelis 1996; Tsebelis and Garrett 1996, 1997; Pajala 2002). These models assume that outcomes of decisions are determined by the combination of preferences of the formally empowered decision-makers and the formal institutional rules that determine the voting weights and rules of the decisionmakers and the sequence of moves they can make. ta bl e 4 . 1 Classification of Main Types of Collective Decision-Making Models

Voting stage Bargaining stage

Cooperative models

Noncooperative models

Coalition PersuasionExchange

WinsetsProcedural models Enforcement/Challenge

Modeling Collective Decision-making 157

Scholars have also developed both noncooperative and cooperative models of the bargaining stage. The best known noncooperative model of the bargaining stage is the Expected Utility model of Bueno de Mesquita (Bueno de Mesquita, Newman, and Rabushka 1985; Bueno de Mesquita 1994). Despite having some similarities, cooperative models of the bargaining stage differ in important respects from coalition models that have been developed in the context of the voting stage. In the latter context, coalition models’ primary aim is to predict the composition of coalitions based on voting weights and decision rules. Voting weights and rules are less important at the bargaining stage than at the voting stage. Cooperative models of the bargaining stage, therefore, do not focus primarily on the composition of majority coalitions, but on the prediction of decision outcomes. These predictions depend on which of the bargaining processes dominate. Persuasion models are based on the assumption that the outcome is the one supported by all stakeholders often after being persuaded to do so with convincing arguments, whereas exchange models are based on the assumption that stakeholders try to build larger coalitions through bilateral exchanges between stakeholders or groups of stakeholders. The focus of the present article is on cooperative and noncooperative bargaining models and their empirical applicability. There are three reasons for this limitation. The first and most important reason for concentrating on bargaining models is that models of the voting stage in our view neglect one of the fundamental features of collective decision-making. As mentioned above, one of these fundamental features is that collective decision-making involves interdependencies among actors involved in joint production. Models of the voting stage ignore such interdependencies, and instead assume that decision outcomes are determined by formal decision-making procedures together with actors’ preferences. In many real world decision situations, however, formal rules provide relatively scant indications of how decisions are actually taken. For instance, there are many political systems in which a decision might formally be allowed if only a simple majority of decision-makers agree, but where in practice the support of a broader majority of decision-makers is required for policy change. Such an informal unanimity norm can be observed in many organizations (among them the Council of Ministers of the European Union). In such contexts, formal procedures do not appear to determine behavior or decision outcomes, but do set the boundaries within which action takes place. Institutional rules then work like legal contracts. They provide safeguards to actors in case fundamental problems arise or other actors misbehave. Such safeguards cannot, however, always loom large in consensual decision-making systems. If actors often have to fall back on the safeguards of formal decision rules, the more the norm of consensus building will be put under pressure. The frequent use of formal rules is likely to undermine the perception that shared interests are so salient that the actors will compromise for the sake of consensus. The mere existence of the rules should give sufficient constraint to enforce agreement and compliance. This implies that the voting stage formalizes primarily the result of the bargaining process and ensures that external sanctions can be used to enforce implementation and compliance. The result of the decision outcome, then, is not determined by

158 Stokman,Van der Knoop, and Van Oosten the combination of preferences of the authorities and the formal institutional rules, as the procedural models assume, but by the dominant bargaining process that precedes the voting stage. The second reason to limit the present discussion to bargaining models is that the models of the voting stage are limited to those decision situations in which there are well-developed formal rules. As described in the opening paragraph, collective decision-making also takes place in very informal settings in which there are few formal rules. Moreover, even in situations, such as national politics, in which there are formal rules, stakeholders who are not formally empowered to take decisions have huge influence. Models of the voting stage do not integrate these stakeholders into the analysis. A third and final reason for limiting the present discussion to bargaining models is that procedural models are generally poor at predicting decision outcomes. One of the largest and most comprehensive comparative tests of the performance of bargaining and procedural models was recently conducted on decision-making in the European Union (see particularly Achen 2006b) and confirmed in Thomson’s replication in the extended European Union of the twenty-five to twenty-seven members (Thomson 2011). Both studies found that bargaining models were generally much better at predicting decision outcomes across a large range of issues than were procedural models. In the European Union, the poor performance of procedural models results from the prevailing norm of consensus seeking, despite the formal possibility of supermajority voting.

Summary of Data Collection and an Empirical Example Applying different bargaining models to complex decision-making processes requires that we describe these decision situations in a systematic and stylized way. This section gives details of what these descriptions entail, illustrated with an example. The above overview of different approaches toward collective decision-making has made clear that a full-fledged model of collective decision-making needs to contain the following elements. All models require a specification of the set of issues and the relevant set of stakeholders and authorities with their policy positions on the issues. We must also obtain estimates of concepts that are specific to certain models, such as the position of the status quo, the relative saliences of the issues for the stakeholders, the relative power of the stakeholders, which higher-ordered goals are at stake, and which priorities stakeholders attach to these goals. More details of the bargaining models referred to above are then provided.There we will introduce the three fundamental bargaining processes in collective decision-making and the models that represent them, how they are related to each other, and under which conditions they will dominate. Subsequently we will discuss and illustrate how the models are applied in practice. As an illustration, we take some data from a recent study on the 2009 Copenhagen climate negotiations. An example. The fifteenth Conference of Parties (COP) meeting took place in Copenhagen from December 7 to 15, 2009. The aim of the conference was an agreement on measures to be taken against climate change caused by our fossil-based economy. The climate treaty of Kyoto ended in 2012. Obligations

Modeling Collective Decision-making 159

after 2012 had to be agreed upon either in an extension of the treaty period or in a new treaty, but COP 16 and COP 17 were also unable to do so. Two experts from the Stockholm Environment Institute specified seven issues as the main ones at stake in Copenhagen. Subsequently, they specified which countries and country groups have to be distinguished and their potential influence and salience for an overall consensus. Finally, they provided the data for each stakeholder on each issue: its position, issue salience, and potential influence. The two experts were interviewed on October 27 and 28, 2009. Table 4.2 presents the Party Groups the experts identified and the abbreviations we use in the remainder. Stakeholders are defined as individuals or groups that have both sufficient power resources potentially to exert influence in the decision-making process and sufficient stakes in the issues to exercise their influence, directly or indirectly (for example, by anticipation of others). If a stakeholder is a group, then the members of that group agree on the desired outcome of the decision and on the importance of the issue. Furthermore, the members of that group are seen to act collectively. Developing countries coordinate their positions within the Group of 77 (G77). At the establishment of this group in the 1960s, seventy-seven developing countries participated. The name of the group remained the same over the years, even though many new developing countries emerged and joined. Since the G77 countries are very diverse, the experts identified several subgroups within the G77 and provided data for each of the subgroups rather than for the whole G77. Table 4.2 also presents estimates of the relative influence of Party Groups during the informal negotiation process preceding the final vote. Above, we have seen that there are fundamentally different ways in which power and influence have been measured in the literature. We usually use voting power measures, such as the Shapley-Shubik index (Shapley 1953; Shapley and Shubik 1954; Pajala 2002) or the Banzaf index (Banzaf 1965), when our analysis is confined to formally empowered authorities (as in the studies of decision-making in the European Union by Thomson et al. [2006] and Thomson [2011]). Influence reputation measures are usually applied when other stakeholders are included as well. To reach agreement in Copenhagen, the vote should be unanimous; consequently, voting power is equal for all Parties, but the differences in influence of Parties and Party Groups in the preceding negotiations are very large. We deal with many different resources—such as exclusive information, financial resources, number of persons an organization represents, superior access to authorities or other stakeholders—that are difficult to weigh in a combined measure of overall influence. Expectation status theory (Berger et al. 1977; Berger, Rosenholz, and Zelditch 1980; Berger and Zelditch 1985) has demonstrated, both in experiments and in field studies, that the ascription of status differences makes them real, that ascription is linked with performance differences. This is the strongest argument for the use of reputation-based influence measures in collective decision-making studies. As a consequence, we follow the approach of Bueno de Mesquita, Newman, and Rabushka (1985) and use expert evaluations of relative influence as our measure of influence in collective decision-making processes.

160 Stokman,Van der Knoop, and Van Oosten ta bl e 4 . 2 Party Groups with Their Relative Influence and the Importance They Attach to Reaching an Overall Agreement Party groups

Abbreviation

Relative influence

Importance attached to reaching agreement

United States of America Canada Australia European Union Japan Russia China and India Brazil Least Developed Countries Alliance of Small Island States G77 minus LDC, AOSIS, China, India, and Brazil.

USA Canada Australia EU Japan Russia China India Brazil

100 15 10 60 20 5 95 10

10 40 50 90 60 10 70 60

LDC

30

85

AOSIS

30

90

Other G77

10

65

To reach agreement, the vote should be unanimous, but Party Groups differ in the importance they attach to reach an overall agreement (group salience). The more importance they attach to an overall agreement, the more they are willing to compromise.We asked the experts to score this on a scale from 0 (not important) to 100 (the Party Group will try to reach agreement with all means at its disposal). The expert ratings are given in the rightmost column of Table 4.2. The United States is estimated to have the greatest influence; however, it is also very little inclined to make concessions to come to a unanimous agreement. In contrast, the EU is willing to promote unanimity very strongly. The first step involved in collecting data is the specification of the problem to be analyzed in terms of a limited number of issues. This is often one of the most challenging stages in the data collection procedure. Each of the issues specified must be described in two ways: first, in terms of a specific policy question on which a collective decision must be taken; and, second, in terms of a scale or continuum on which the alternative possible outcomes of this decision can be placed. Each of the issue continua is assumed to be unidimensional, and each actor involved in the decision who has an interest in the issue can be placed on a point on the continuum to represent his policy position on that issue. We assume that actors have single peaked preference functions. Thus each actor evaluates points on the continuum that lie further away from his position more negatively. This means that each actor expects to receive most value from the realization of his own position on the continuum compared with other positions, and less from alternatives located further from his own position.The two extreme positions, or endpoints, on each issue continuum are usually occupied by the most extreme positions favored by any of the actors. Intermediate positions represent more moderate positions and also possible compromise outcomes. The specification of problems in terms of a limited number of issues provides a conceptual structure in which the positions of all actors can be represented. The specification of the issues should be comprehensive, in the

Modeling Collective Decision-making 161

sense that the decisions taken on these issues should determine the main contours of the solution to the collective decision problem. The number of issues that are necessary varies from one decision-making situation to another. Usually, one to five issues are sufficient to represent all combinations of possible outcomes in even highly complex decision-making processes, but up to twenty issues have been used in some applications. The requirement of specifying a limited number of issues is often a useful exercise in itself, because it compels analysts to distinguish the main points from subordinate ones. Ill-defined issue specifications mean that analyses are based on incorrect representations of the political problems. Poor issue specification will result in bad model predictions. In practice, the issues are specified using a combination of content analysis of documentation and interviews with key informants (also referred to as “subject area specialists” or “experts”). The estimation of the policy positions of each of the stakeholders depends on the specification of the issues as described above. If the issue does not represent a quantitative outcome (such as budget or time), the most extreme positions are usually placed at points 0 and 100 of the issue continuum. In the analysis, all issues are rescaled to a 0 to 100 continuum, based on the positions of the most extreme stakeholders. It is vital that the stakeholders are placed on the issue continua to reflect the political distances between the alternative decision outcomes they support. The next variable that informants are asked to estimate for each stakeholder is the level of salience that each stakeholder attaches to an outcome close to its policy position on the issue (the stakeholder’s issue salience). Note that obtaining an outcome close to the policy position on a given issue may be more important for one stakeholder than for another. In addition, any given stakeholder may attach more salience to one issue than to another. The more utility loss a stakeholder experiences resulting from a difference between its policy position and the outcome, the more likely that stakeholder will put into effect its potential power to obtain an outcome close to its policy position. The variable salience can therefore also be interpreted as a measure of the extent to which a stakeholder is willing to put into effect its potential power if the issue is brought up during interaction with other stakeholders. The level of salience each stakeholder attaches to each issue is usually expressed on a scale from 0 to 1. A score of 0 indicates that the issue is of no interest whatsoever to the actor. In fact, if an actor attaches zero salience to an issue, it is not considered to be a stakeholder on that issue. A score of 1 indicates that an actor will devote all of its potential power to this issue if the issue is brought up during the course of interaction with other stakeholders. A score of 0.5 indicates that the issue is neither important nor unimportant. Taken together, the potential influence times the salience of a stakeholder determines its effective influence with regard to a certain issue. In the literature and in practice it is often neglected that the combination of position and salience determines the behavior of stakeholders. These combinations are generated by the incentive structures of stakeholders. If one of the two is overlooked (in practice, often salience), serious miscalculations are inevitable.

Outcome NBS (61) Outcome after concessions to USA and China India (85)

Outcome after exchange (57)

A Collection of Decisions

10

A New Treaty

20

Extension of Kyoto

50

90

100

80 Russia (70)

Canada (80)

EU (40)

USA (90)

Japan (60) AOSIS (90) Australia (40 )

Brazil (50) China India (90) LDC (90) Other G77 (50)

figure 4.1. New Decisions vs. Extension of Kyoto (issue 1)

Outcome after concessions to USA and China India (82)

Outcome NBS (53)

Outcome after exchange (46) Low

High

10

15

China India (100)

30

Russia (10)

Brazil (90) Other G77 (80)

40

50

70

AOSIS (30) LDC (30)

80

90

Australia (60) EU (50) Canada (50) Japan (50)

figure 4.2. MRV CO2 Reduction in Developing Countries (issue 2)

USA (100)

Modeling Collective Decision-making 163

One of the major controversial questions in Copenhagen was whether the Kyoto Treaty had to be extended, or whether the decisions in Copenhagen had to result in a new treaty. This issue is particularly controversial, as the United States had not signed the Kyoto Treaty. The second most controversial issue concerned the measurable and verifiable contributions of China, India, and Brazil. These countries have no obligations under Kyoto, but their economic growth is now so high that they can be expected to contribute to the worldwide CO2 emission reduction. Most of the other five issues are related to mitigation and adaptation. Mitigation concerns the reduction of greenhouse gas emissions, such as CO2; adaptation concerns measures to circumvent or diminish damage caused by climate change. As an illustration, we will concentrate here on the two most controversial ones and report over the others only globally. For the full report, delivered one month before the conference, we refer to Stokman 2009. Figure 4.1 contains the data on the status of the Kyoto Treaty. Is the outcome of the Copenhagen COP an extension of the Kyoto Treaty (position 100 on the scale), a new treaty (position 50), or just a collection of decisions (position 0)? Saliences of Party Groups are given in parentheses after their acronym, ranging from 0 to 100. The salience is also represented by shades of gray. Party Groups in dark gray attach a salience of between 80 and 100 to the issue; in middle gray between 50 and 80; and in light gray below 50. Both ends of the scale are covered with dark gray Party Groups, indicating the highly controversial nature of the issue. Above the scale the outcomes we expect under different assumptions are presented. They will be discussed in more detail later. Figure 4.2 contains the data on the MRV CO2 Emission Reduction in Developing Countries. MRV CO2 emission reduction refers to reductions that are “Measurable, Reportable, and Verifiable” (MRV). These criteria are applied to ensure measurable CO2 emission reductions. Whereas in rich countries MRVs concern reductions in the total amount of emissions by 2020, developing countries are still allowed to increase their total emissions in order to obtain a higher welfare. MRV CO2 emission reductions in developing countries aim to increase the CO2-emission-free proportion in their growth, especially in sectors involving high emissions such as heavy industries, electricity, and transport.The MRV issue concerns, therefore, the commitments of developing countries to create a more sustainable economy.

Bargaining As described above, the dynamics in decision-making processes result from the fact that each of the stakeholders attempts to realize the policy position it favors as the outcome. The complexity of such processes derives from the fact that stakeholders often take quite different positions, have different levels of potential to influence the decision outcome, and differ from each other with respect to the intensity of their preferences. Stakeholders may attempt to build a coalition as large as possible in support of the policy positions they favor. By building such coalitions, stakeholders hope to affect the positions of the final decision-makers, the authorities, which will in turn lead to a collective outcome that reflects their interests as much as possible. Consequently, the dynamics

164 Stokman,Van der Knoop, and Van Oosten tabl e 4.3 Fundamental Processes, Dominant Networks, Approaches, Conditions for Processes to Dominate Fundamental Processes

Dominant Networks

Present Not-integrated Approaches

Persuasion

Information and Trust Networks

1. Contagion Models Reciprocal 2. Exchange Networks

Cooperative Nash Bargaining Solution for all relevant stakeholders

1. Reversion point very unattractive 2. Overall coalition possible/ Subcoalitions difficult to form Risk averse stakeholders

Logrolling

Negotiated Exchange Networks

3. Coleman Exchange Model 4. Network Exchange Theory

Voting Position Exchange Models (Cooperative solutions for subsets of stakeholders with positive and/or negative externalities for others)

Opposite positions and complementary interests

Enforcement

Hierarchical/ Power Networks

5. Noncooperative Models

(Noncooperative) Challenge Model

Opposite positions and noncomplementary interests

Integrated Approach

Conditions for process to dominate

of decision-making processes are based primarily on processes through which other stakeholders are willing or forced to change their positions. Three fundamental processes can result in such shifts in positions: persuasion, logrolling, and enforcement. Udehn (1996) derives these three fundamental processes from the literature in his sociological critique of economic models of politics. Each of these is associated with its own specific interdependencies.Table 4.3 gives an overview of these three processes, the types of networks associated with these processes, which approaches in the literature are associated with which process, and the conditions under which each of the processes is expected to dominate collective decision-making.We then elaborate on each process and the different elements contained in Table 4.3. For the logrolling and enforcement processes, we will also specify the conditions under which they strengthen persuasion processes and under which conditions they undermine persuasion processes. Persuasion Through persuasion, stakeholders aim to change other stakeholders’ initial positions, or preferences, and the levels of salience they attach to the issues that must be decided on (Stokman et al. 2000). When a stakeholder changes its position or alters the level of salience it attaches to an issue as a result of persuasion, this change constitutes a fundamental internal switch on the part of the stakeholder. Persuasion is achieved through the provision of convincing information. Persuasion strategies are particularly likely to dominate when collective decision-making based on unanimity is a strong formal or informal norm (that is, if the group consensus salience is high and includes all stakeholders). The Nash Bargaining Solution (NBS) (Nash 1950) provides an approach with which to model persuasion as a dominant mode of interaction. One of the central conditions conducive to persuasion is that stakeholders perceive shared

Modeling Collective Decision-making 165

interests to greatly outweigh their individual interests. When stakeholders have a strong shared interest in reaching a collective decision, failure to do so is highly undesirable, and far less desirable than any of the decision outcomes advocated by any of the stakeholders involved. This facilitates the feasibility of grand coalitions of all stakeholders, particularly when smaller coalitions are difficult to form. Under these conditions and assuming quadratic loss functions on the issue continua (implying risk-averse stakeholders), Achen (2006a) shows that the compromise model becomes a first-order approximation of the Nash Bargaining Solution. This compromise model prediction is simply the average of the stakeholders’ initial policy positions, weighted by the product of each stakeholder’s influence and salience. Conditions that are conducive to persuasion can exist only when stakeholders are embedded in dense trust networks or are severely punished when they deviate from shared interests. Stakeholders need to be confident that the information they receive is sincere and not strategically manipulated. Pursuing one’s own personal gains is permitted as long as this does not inflict harm on others, and as long as personal gains are compatible with shared interests.Within this context, stakeholders can be confident that the concessions they make to stakeholders who have strong interests in present issues will be compensated in future situations when their own interests are stronger. Reciprocal and generalized exchanges (Molm 1997) are therefore an integral part of decisionmaking by persuasion, and not of decision-making by logrolling as the name might suggest. Stakeholders who provide information will be trusted if they have proven to be reliable in the past and if they would experience future negative consequences from providing distorted or incomplete information. This “shadow of the future” (Axelrod 1984) is more effective if providers of distorted information lose reputation, not only with respect to the recipient stakeholder but also with respect to others (Raub and Weesie 1990; Buskens 1999; Panchanathan and Boyd 2004; Nowak and Sigmund 2005).Trust will also be greater if the information is less related to the provider’s central interests. These conditions for trust emerge more readily among like-minded stakeholders and among stakeholders who also meet each other in other contexts, than among stakeholders with conflicting interests. Stakeholders also tend to assign more weight to the opinion of powerful stakeholders, whereas powerful stakeholders tend to listen more to one another than to less powerful ones (Molm 1997; Stokman and Zeggelink 1996b). Large power differences, however, make it less likely that persuasion strategies will be successful. The same holds for highly polarized issues. In contrast to persuasion, logrolling and enforcement processes typically do not affect stakeholders’ initial positions or the levels of salience they attach to issues. Logrolling is a process of negotiated exchanges. The result is that stakeholders are willing to support another position on an issue that is of relatively less importance to them in exchange for support of another stakeholder on an issue that is relatively more important to them. Similarly, when enforcement is the dominant mode of interaction, stakeholders can feel forced to support another position under pressure from more powerful stakeholders or coalitions. Logrolling and enforcement are most likely if stakeholders’ initial positions fundamentally differ because of the different weights they attach to

166 Stokman,Van der Knoop, and Van Oosten different higher-ordered goals. In such situations, arguments do not help to bring initial positions closer together. Therefore, coalitions can be built only through processes that affect the final or voting positions of stakeholders. We will consider these two processes in the next two sections. Logrolling Whereas information and trust networks define persuasion, negotiated exchange networks define stakeholders’ exchange possibilities under logrolling. When stakeholders shift their policy positions as a result of logrolling, these shifts lead to changes in the expected outcomes on the issues involved in the exchange. Consequently, stakeholders experience gains and losses when the expected outcomes on issues move closer to or further from their initial positions. Stakeholders from two groups with opposing positions can profit from position exchange if the relative salience of the two issues for each of them is different (see Figure 4.3; Stokman and Van Oosten 1994). A position exchange is then profitable for both stakeholders. The model of logrolling bargaining processes assumes that each stakeholder has complete knowledge of the positions, saliences, and capabilities of all other stakeholders. We further assume that all stakeholders share a common view on what the collectively optimal outcome would be on each issue when considered separately. This collectively optimum outcome is assumed to be the Nash Bargaining Solution, approximated by the average of the stakeholders’ initial policy positions, weighted by the product of each stakeholder’s influence and salience (see above in the section on persuasion). Position exchanges link pairs of issues and provide pairs of stakeholders opportunities for bilateral winwin situations above the NBS solutions of the issues in isolation (see Figure 4.3). They can be seen as bilateral active optimizations of the NBS. Each stakeholder potentially has a number of possible exchanges. Each stakeholder has to choose which of these potential exchanges to realize. A potential exchange is realized only if both stakeholders agree to realize it. This will happen only if neither of them has a better alternative exchange. When an exchange is realized, both stakeholders are no longer able to change position on the issue on which they have moved their position. This of course limits future exchange possibilities in the bargaining process. In other words, when stakeholders realize an exchange they enter into a binding commitment, which is what makes the logrolling model a cooperative bargaining model. Modeling position exchanges requires careful consideration of the nature of these exchanges. In particular, a choice has to be made about which exchange rate to use. Utility gains and losses result from outcome shifts on the two issues because of their position shifts and depend on the size and direction of the outcome shifts and the issue saliences of the stakeholders. The exchange rate determines the extent to which each stakeholder shifts its position. Stokman and Van Oosten (1994) use an equal utility gain for both exchange partners. This has the advantage that exchanges have the same utility for both partners, and that the exchanges can be ordered in terms of their relative attractiveness to both exchange partners. The disadvantage of the equal utility gain assumption is that it involves an intersubjective comparison of utility, which is theoretically problematic. Two alternative solutions for the exchange rate have been derived

Modeling Collective Decision-making 167 Issue 1 A

D

B

C O1 (Nash Bargaining Solution as expected outcome)

Issue 2 A

D

C

B O2 (Nash Bargaining Solution as expected outcome)

figure 4.3. Effects of an exchange between stakeholders of type A and type D. The arrows mean that actor D shifts his position on issue 1 in A’s direction, while actor A shifts his position on issue 2 in the direction of actor D.

that are independent of the utility scale: the Nash solution (Achterkamp 1999; Van Assen 2001), and the Raiffa-Kalai-Smorodinsky (RKS) solution (Van Assen 2001).1 With the exception of the equal gain exchange rate, all exchange rates face the same problem of deadlock, whereby no two stakeholders prefer, and therefore realize, the same exchange.2 Bilateral exchanges also have important side effects or externalities with respect to other stakeholders’ utility. Externalities arise when stakeholders who are not involved in an exchange are either positively or negatively affected by it. This can clearly be seen in Figure 4.3. Assume that a stakeholder of type D attaches relatively more salience to issue 2 than to issue 1 if we compare its saliences with those of a stakeholder of type A. Then, issue 2 is D’s demand issue and A’s supply issue. Position exchange between A and D implies that A is willing to shift its position on issue 2 in the direction of D, while D does the same on issue 1. If they do, they both shift away from C in the direction of B on both issues. In that case, C is punished doubly and B rewarded doubly, while neither of the two is directly involved in the exchange (Van Assen, Stokman, and Van Oosten 2003). Positive and negative externalities also emerge within the A and D groups if A and/or D consists of more stakeholders. An exchange of two stakeholders from the A and D groups will have positive externalities for other members in the A and/or D group if the relative saliences within each group are relatively homogeneous. Otherwise, such an exchange may well have negative externalities within the A and D groups. In the most extreme case, one A member may want to use issue 1 as its supply issue whereas another A member may want to use that issue as its demand issue. Here, we return to the Copenhagen example. In Table 4.4, the Party Groups are ordered on the basis of their relative salience for the two most controversial issues—that is, the salience for the issue on the status of the new treaty (issue 1) divided by the salience for the issue on the MRV CO2 reduction in developing countries (issue 2). In the last column of Table 4.4, the Party Groups are allocated to the four cells in Figure 4.3 on the basis of their positions: their placement depends on whether they are located to the left or to the right of

168 Stokman,Van der Knoop, and Van Oosten ta bl e 4 . 4 Relative Saliences of Copenhagen Party Groups for the Two Most Controversial Issues and Their Positions Relative to the Expected Outcomes on the Two Issues Party group

Canada USA Japan China India LDC AOSIS Russia Australia Brazil Other G77 EU

Relative salience for Issue 1/Issue 2

Cell

1.33 1.20 1.00 1.00 1.00 0.95 0.78 0.67 0.67 0.63 0.50

A A A C D D C A C D B

note: Party Groups in cell A have positions left of the expected outcomes on both issues; Party Groups in cell B left on issue 1 and right on issue 2; Party Groups in cell C right on issue 1 and left on issue 2; Party Groups in cell D right on both issues.

the expected outcome (the Nash Bargaining Solution) on each issue. As in Figure 4.3, only Party Groups with opposite positions on both issues can make exchanges—that is, Party Groups in cell A can make exchanges with those in cell D, and Party Groups in cell B with those in cell C. Cell A consists of four Party Groups and cell D of three. Cell C consists of three Party Groups, cell B of one. This results in twelve potential exchanges between the A’s and D’s and three between the B’s and C’s, making a total of fifteen potential exchanges of voting positions. Three members of cell A attach relatively more salience to issue 1 than do all Party Groups in cell D. For these A’s, the first issue is the demand issue (as in Figure 4.3), and all potential exchanges go in the same direction: toward the initial position of the Party Groups in cell B. The fourth Party Group in the A cell (Australia), however, can make an exchange with Party Groups in cell D in both directions. With the Party Group Other G77 again its demand issue is issue 1, but with the Least Developing Countries (LDC) and the Alliance of Small Island States (AOSIS) Groups issue 2 is the demand issue, as the latter have a higher relative salience for issue 1 (1 resp. 0.95) than Australia (0.67). Whereas most potential exchanges between the Party Groups in cells A and D will result in a better outcome for the Party Group in cell B (the EU) and a worse outcome for the three Party Groups in cell C, the EU has potential exchanges with three Party Groups in cell C. If one or more are realized, the positions of the EU and the three Party Groups in cell C (China, India, Russia, Brazil) will all shift in the direction of the Party Groups in cell D, resulting in worse outcomes for the Party Groups in cell A. In other words, all potential bilateral exchanges will have negative externalities for at least some Party Groups, making an overall unanimous outcome less likely. We can now specify the conditions under which logrolling based on

Modeling Collective Decision-making 169

bilateral exchanges is compatible with cognitive interdependencies that support persuasion and consensus building.The following three conditions should hold simultaneously:3 1. One of the four groups should be empty.Without loss of generality, let us assume that group C is empty. 2. The relative saliences of the two groups that can exchange is such that the exchange shifts the decision outcomes in the direction of the positions of the stakeholders in the nonempty group. 3. The relative saliences of the two groups that can exchange are such that there are no negative externalities within each of these groups. This occurs under the following condition. Without loss of generality, let us assume that a stakeholder in the A group attaches the highest relative salience to issue 1 compared with all other stakeholders in the A-D group. Under the assumption of linear decreasing utility functions around the policy positions of stakeholders, no negative externalities occur if the exchange rate is lower than the relative salience of the A stakeholder with the lowest relative salience for issue 1, and is higher than the relative salience of the D stakeholder with the highest relative salience for issue 1. This implies that negative externalities within an exchange group are unavoidable when stakeholders of one group embed some stakeholders of the other group in the ordering of the relative salience they attach to the issues.4 If these conditions are not met, bilateral exchanges over pairs of issues produce negative externalities for other stakeholders. Such negative externalities will harm consensus building, unless the stakeholders who experience negative externalities are compensated in other respects.5 Certain institutional conditions may discourage stakeholders from realizing exchanges with negative externalities and encourage them to realize exchanges with positive externalities. Some decision-making rules stipulate that outcomes must be supported unanimously. In other contexts, informal norms stipulate that unanimous support should be sought, although outcomes could formally be taken by majority voting. In both these contexts, we expect stakeholders to avoid voting position exchanges with negative externalities and to realize only exchanges with positive externalities. Exchanges with positive externalities facilitate overall consensus, as the interests of the exchanging Parties are in harmony with those of the others. Under the condition that stakeholders avoid exchanges with negative externalities, linking issues can potentially improve the overall Nash Bargaining Solution for all stakeholders. Dijkstra, Van Assen, and Stokman (2008) incorporated this idea in a new version of the exchange model, denoted the Externality Exchange Model (EEM), and tested their model against the original model in the context of the European Union. A nonparametric test for which of the two models more often gives the best prediction shows no significant difference between the two. Relative to the original logrolling model, the predictions of the EEM model show a more substantial improvement, albeit not significant in the nonparametric test.This is a weak indication that stakeholders avoid exchanges with negative externalities in contexts in which an overall consensus is normatively promoted.

170 Stokman,Van der Knoop, and Van Oosten Enforcement When collective decision-making is driven by power processes, enforcement, not persuasion, is the dominant mode of interaction among the stakeholders (see Table 4.3). Stakeholders try to build as large a coalition as possible behind their own policy position by showing that they have sufficient power to enforce a decision and/or to block other alternatives. Solutions to substantive problems are not sought by arguments but by showing that there is sufficient support to enforce the decision on the basis of the formal procedures and/or informal power arguments. When enforcement is the dominant mode of interaction, stakeholders may shift their positions because they feel compelled to do so, not because they are convinced to do so. To the extent that a stakeholder’s issue salience is lower than that of stakeholders who support another position, and the power of those other stakeholders is greater, that stakeholder may be inclined to give up its initial position. Stakeholders may avoid costs by conceding on an issue that is only marginally related to their own interests. When enforcement occurs, decision outcomes can be seen as the result of a noncooperative game in which no binding agreements are made (Bueno de Mesquita, Newman, and Rabushka 1985; Bueno de Mesquita 1994, 2002). In his computer simulation model, a challenge to a stakeholder’s position is more likely to be successful if the stakeholder to whom the challenge is directed attaches less salience to the issue than does the challenger, and if the support for the challenged stakeholder’s position is lower than the support for the challenging stakeholder’s position. In the model, these two aspects dominate the determination of which stakeholders will challenge which other stakeholders. Each stakeholder makes this choice in relation to each of the other stakeholders. Based on the challenges made, each stakeholder has a set of cards in its hands that represent the challenges made and received. If a stakeholder received challenges, that stakeholder has to draw the one that is best. The result is either conflict (if the stakeholder made a challenge to the other as well) or a forced position change. These position changes create a new decision-making setting (iteration in the computer simulation model). In that new setting, the stakeholders repeat the choice process. This continues until none of the stakeholders move (substantially) or until all stakeholders take the same position. If enforcement dominates decision-making about organizational policies, hierarchy dominates over arguments also in the preparatory stage of decisionmaking. In such a situation, the goals of the organization are likely not primarily seen as shared goals, but as the goals of and set by the top of the organization. Such a setting leads to a cognitive interdependence model in which personal relationships are seen primarily in the light of their hierarchical place and ordering. In other words, power networks dominate the outcomes of collective decision-making processes. Again, as bilateral negotiated exchanges may well be compatible with consensus or even enhance consensus building (in the presence of large positive and the absence of negative externalities), it is unlikely that persuasion on the basis of high shared interests will long survive without clear institutional rules and clear responsibilities that are derived from them.6 They connect joint production with external sanction (legal) systems to enforce cooperation,

Modeling Collective Decision-making 171

resulting in sufficient trust that noncooperative individuals can effectively be sanctioned or even fired. Enforcement of cooperation is also important for the timely and correct implementation of collective decision-making. Recent EU studies have investigated such effects on implementation of distances between decision outcomes and policy positions of Member States and the European Commission and of consensus among Member States in the Council (see, among others, Falkner et al. 2005; Zhelyazkova and Torenvlied 2009; König and Luetgert 2009; Steunenberg 2010; Thomson 2009, 2010). If cognitive interdependencies are linked to norms that decisions should be based on consensus, institutional rules work like legal contracts. As mentioned before, they provide safeguards to stakeholders in case fundamental problems arise or other stakeholders misbehave. However, the more often you have to fall back on them, the more the norm of consensus building will be under pressure. Building sufficient support for a specific outcome may lead to a preferable outcome, but it may also lead to disturbed relations. Some stakeholders may not be interested in a specific outcome, but in any outcome as long as it is supported by all stakeholders. Other stakeholders may be interested solely in an outcome close to their policy position, even when it implies a lot of opposition and turmoil. From this perspective, each stakeholder can be perceived to have at least two objectives while intervening in decision-making. The first objective is to minimize the distance between the outcome and the policy position of the stakeholder on the issue. The second objective is to minimize the variance of the positions of all stakeholders or the subgroup of stakeholders with whom the stakeholder is associated. Earlier, we denoted the first issue salience and the second group consensus salience. The two objectives can be modeled by using an aggregate utility function in which both objectives are combined. This can be realized by applying the Cobb-Douglas function with two weights, one being the issue salience and the second the group consensus salience. The three bargaining processes: transitions and testing dominance It is interesting to study and model transitions from one dominant process of decision-making to another. This is the subject of future research in which both Lindenberg’s theory about frame switches (Lindenberg and Frey 1993; Lindenberg 1998, 2000) can be helpful, as well as Esser’s model building on shifts in the definition of situations (Esser 1997, 2000). The dominance of the three types of networks (persuasion, logrolling, enforcement) in the context of the European Union was evaluated on the basis of the accuracy of the three corresponding models. To determine the dominance of the three types of processes (persuasion, logrolling, enforcement) in the context of the European Union, the accuracy of each model is determined by the distance between the model-predicted outcomes and the actual outcomes on the issue scales (for the EU 2001 extension, see Stokman and Thomson 2004: 19; for the EU after the extension, see Thomson 2011). Models based on cooperative solutions that include the positions of all EU decision-makers give the best predictions. Unanimity, wherever possible, is a very strong norm in the EU, even when decision outcomes

172 Stokman,Van der Knoop, and Van Oosten supported by only a qualified majority of actors are possible (see also Mattila and Lane 2001). Decision outcomes in the EU tend to take into account actors’ essential interests, wherever possible, and actors avoid harming the essential interests of others (Schneider, Finke, and Bailer 2010). This implies that persuasion networks dominate in the European context. Negotiated exchange networks do not often support consensus building in the European Union because of the high negative externalities involved. Given the dominant norm of consensus building, this type of network is not dominant in the European context, as shown by its worse predictions than the persuasion model. Dijkstra, Van Assen, and Stokman (2008) show, however, that negotiated exchanges that avoid negative externalities indeed contribute to overall consensus building in the European Union. Power networks do not dominate European Union decision-making either: noncooperative procedural and bargaining models do even worse. We therefore conclude that, also in the European context, procedures do not determine behavior, but set the boundaries within which action takes place. The reader should be aware that the inferences about European Union decision-making can be made only by a comparative analysis of the three processes and corresponding networks.

Strategic Intervention in Decision-making The methodology of data collection and the dynamic analysis of the bargaining processes through computer simulation have not only been validated in scientific research but are also applied in commercial projects as a successful tool for strategic intervention (see Stokman et al. 2000 for two examples). Whereas in scientific applications the main aim is the prediction of outcomes and the determination of the dominant process, applied projects usually aim either to arrive at decisions close to the client’s position with sufficient support to be viable, or to arrive at a common position in stakeholder dialogues and mediation. The approach can be applied fruitfully both in contexts where organizational strategies have to be determined and where organizational strategies have to be implemented. Stokman et al. (ibid.) elaborates strategic moves for all three bargaining processes. Here we will illustrate just one such move in the context of our example of the Copenhagen climate conference in December 2009. The question of whether consensus could be reached in Copenhagen depended on two perceptions of the Party Groups.The first perception concerns the severity of the expected climate changes as a consequence of greenhouse gas emissions owing to currently unsustainable industrial production. The second is evaluating the importance of a worldwide agreement between the Parties in order to realize the transition to a more sustainable production. If both perceptions are strong and can be shared by all Party Groups, failing to reach a unanimous agreement will be seen as highly undesirable, even less desirable than a weak compromise. If this were the case, unanimity was expected to be reached in the end, even when the Party Groups fundamentally disagree on a number of issues. For each issue, the expected outcome will then be close to the mean of the Party positions on the scale, weighted by their influence and

Modeling Collective Decision-making 173

salience, the approximation of the Nash Bargaining Solution, as we have seen earlier.Table 4.2 shows, however, that certain Party Groups do not attach much importance to reaching an agreement. The EU, the least developed countries, and AOSIS want to reach an agreement, but others such as the United States and Russia do not.This implies that the NBS is unlikely to be a good predictor for the outcomes on this issue. The expected outcomes, but also the variation of the positions can change fundamentally if Party Groups exchange voting positions by linking the issues with each other. The degree of agreement after the exchange process increased substantially for five of the seven issues, but remains low for the two most controversial issues—namely, the state of the decisions in Copenhagen as new or as an extension of Kyoto (Issue 1; Figure 4.1), and the size of MRV CO2 reduction in advanced developing countries, such as China, India, and Brazil (Issue 2; Figure 4.2). In other words, the basis for agreement improves fundamentally, but two issues will continue to cause problems. Another reason why it is not expected that this exchange process will result in an overall agreement is that, over all simulated exchanges between Party Groups, the positive externalities are greater than the negative ones only for the EU, Russia, and some developing country groups. All other Party Groups perceive higher negative externalities than positive ones, which is the second reason for the main conclusion that the interests of the Party Groups are not sufficiently aligned to arrive at an overall agreement by simply exchanging positions. Two issues remain controversial and require another solution. There are simply not enough complementarities between interests to reach an overall agreement. The next question is then whether there are instruments to increase the complementarities of interests of the Party Groups in Copenhagen in such a way that an overall agreement can be achieved. A strategy for such an outcome is based on two small changes in the data on the basis of solid reasoning. Issue 1 is a problem mainly for the United States, which never ratified the Kyoto Treaty. If the new decisions are classified as an extension of the Kyoto Treaty, the U.S. House and Senate ratification of the Copenhagen agreement implies a ratification of the Kyoto Treaty. Moreover, after eight years of the Bush administration, the United States cannot easily catch up. Consequently, the U.S. will not likely sign a treaty that implies ratification of the Kyoto Treaty. On the other hand, China and India have high stakes in having a Copenhagen agreement as an extension of the Kyoto Treaty, as rich countries can realize their emission reduction obligations with projects in their countries.The MRV CO2 free reduction in the growth (Issue 2) is especially important to China and India, as they are willing to realize such a component in their growth but are not willing to make binding agreements to do so. A possible solution could be to accept nonobligatory intentions in both cases, but to put the realizations of CO2 reduction of these countries in the Copenhagen Treaty. Such a double arrangement considerably reduces the salience of the United States in Issue 1 and the salience of China and India in Issue 2, which can be investigated by a considerable reduction of the two saliences in the data. The salience of the U.S. on Issue 1 is reduced from 90 to an arbitrarily chosen value of 70 or lower, such as 50. Simultaneously, the salience of 100 of China and India on Issue 2 is also reduced to 50.

174 Stokman,Van der Knoop, and Van Oosten These changes reduce the variances of positions on all seven issues substantially enough to expect overall consensus. The results are stable as long as the salience of the United States is reduced to 70 or lower for Issue 1 and that of China and India to a value of 90 or lower on Issue 2. Doing so provides us with very stable results. Now, after bilateral exchanges, sufficient agreement is realized on all issues to arrive at a complete agreement. This prediction was made in November, one month before the start of the Copenhagen Conference and turned out to be the sole solution for something like an agreement in Copenhagen (see Stokman 2009 for details).

Conclusions Starting from the idea that collective decision-making is a special case of joint production, required in any situation in which individuals are mutually outcome dependent, we hope to have shown that the topic is of much wider importance than in simply the political sphere. Collective decision-making is at the heart of any collaboration, whether that is in small informal groups or in complex organizations or in political systems. Placing collective decisionmaking in this perspective, one’s attention is immediately drawn to the relative salience of the shared versus the conflicting interests that is of such importance in any joint production. We hope also to have shown convincingly that in the domain of collective decision-making this ratio strongly determines the type of dominant process and the likelihood of arriving at common positions, even when formal institutions do not require reaching them. The chosen perspective also shifts the attention from formal rules toward informal rules, without underestimating the importance of the formal rules for the evolution and effectiveness of the informal rules. If, then, outcomes cannot be seen as the result of the interplay of formal institutions and preferences, we almost automatically have to shift our focus toward the informal processes preceding the final vote. Notwithstanding the elegance of neoinstitutional models and the extra insights they have generated, they seem not to be able to predict outcomes of decision-making processes in reality. Models that represent informal bargaining processes seem to do much better in this respect. This does not imply that formal institutions are neglected in such models. On the contrary, formal institutions and the voting rights and rules that are based on them are, first of all, required to make outcomes binding for social systems. In addition, they codetermine power and influence distributions in social systems, codetermining which stakeholders have to be included in the analysis. Finally, they connect collective decision-making processes to external sanction systems, without which informal processes are likely soon to degenerate, as stakeholders will not have formal sanctions to enforce cooperation and norm-conforming behavior. In this article, we have specified three main bargaining processes and the conditions that each of them is likely to dominate in decision-making processes. Moreover, we have specified under which conditions logrolling and enforcement processes are likely to support or undermine persuasion processes. Finally, we have tried to develop an integrative set of models for all processes, enabling both a comparative analysis of the expected outcomes under each

Modeling Collective Decision-making 175

of the processes and a strategic analysis of how to align the different processes in such a way that they support each other. It is this combination that makes the approach valuable both for scientific analysis and strategic intervention in decision processes. The largest project in which this approach is applied is the Decision-Making in the European Union (DEU) project (Thomson et al. 2006; Thomson 2011). In the Forum Section of European Union Politics, Mattila (2012: 459) concluded independently: “In many respects the DEU project has led EU studies to a new level. It was the first project of its scale to analyse the EU’s decision-making system with a systematic rational choice approach.” If we have valid and reliable estimates of the main issues at stake and the positions, saliences, and influence of the relevant stakeholders, rational models are able to provide far-reaching insights and conclusions. This does not imply, however, that the positions and saliences of the stakeholders are based on rational considerations only, focused on an optimal outcome for each stakeholder. A theoretical derivation of these data requires a more complex model of man (see, for example, Lindenberg and Steg 2007). Nevertheless, there is more to do. We are presently particularly working on two lines of further development. First, the models can be improved by incorporating not only the issue salience of the stakeholders but also their group consensus salience. Stakeholders aim not only at outcomes close to their policy position but also at outcomes that receive support from either all stakeholders involved or the stakeholders they want to align with. We have indicated that the Cobb-Douglas function can be used for the simultaneous optimalization of these two goals, taking into account the relative saliences for both goals. Second, a further elaboration of persuasion models requires a further elaboration of the relationships between instrumental (issues) and higher ordered goals. In such an elaboration, both differences in priority of the higher ordered goals and differences in cognitive perceptions of the relationships between issues and goals have to be integrated with the collective decisionmaking models as treated in this article.

Notes The ideas in this article were developed in close collaboration with many persons, of whom we would particularly like to mention Siegwart Lindenberg, Robert Mokken, and Marcel Van Assen. We thank Vincent Buskens, Jacob Dijkstra, Beth Levy, Siegwart Lindenberg, Robert Thomson, Timo Septer, Hanne Van der Iest, David Willer, and Rafael Wittek for their comments on earlier drafts. 1. There is a strong link between our work and Network Exchange Theory because the exchange-resistance solution used in Network Exchange Theory (see, for example, Willer, Markovsky, and Patton 1989; Skvoretz and Willer 1993; Szmatka and Willer 1995) is also derived from the game theoretical RKS solution (Heckathorn 1980). 2. Stokman and Van Oosten’s exchange model has been tested in several contexts, ranging from complex negotiations between employers’ organizations and trade unions (Rojer 1996; Akkerman 2000), urban politics (Berveling 1994), European Union decision-making (Arregui, Stokman, and Thomson 2006), and the international climate conference in Copenhagen (Stokman 2009). We have published the outcomes for the Copenhagen study not only in advance but also at four other occasions at the beginning

176 Stokman,Van der Knoop, and Van Oosten of the negotiations. The first time, in 1996, Rojer and Stokman gave the predictions of outcomes on sixteen issues in a forthcoming negotiation process in the Dutch metal industry to a lawyer. They announced publicly that these predictions would be revealed at a press conference at the end of the negotiation process. Three times, in 2002, 2007, and 2009, predictions were published at the start of coalition negotiations between Dutch political parties. In all three cases, more than 80 percent of the outcomes were correctly predicted within strict, previously specified boundaries. (See http: //www. stokman.org/news-related%20activities.htm.) 3. The proof is given in Dijkstra,Van Assen, and Stokman 2008. 4. If such an exchange is possible, another condition may limit the size of the exchange. For none of the actors, the distance between the expected outcome and their position on the demand issue should be larger after exchange than before (Dijkstra,Van Assen, and Stokman 2008). 5. Van Assen, Stokman, and Van Oosten (2003) define measures with which the positive and negative externalities of bilateral exchanges for other stakeholders can be computed. In their approach, exchange is considered as a cooperative two-person game. That is, in the derivation and the calculation of the measures it is assumed that actors not involved in the exchange do not affect the exchange rate of the exchange under consideration. 6. For the interaction between informal cooperation and sanctioning systems, see, among others, Yamagishi 1986; Ostrom, Walker, and Gardner 1992; Fehr and Gächter 2002; and Rockenbach and Milinski 2006.

References Achen, Christopher H. 2006a. “Institutional Realism and Bargaining Models.” In The European Union Decides, edited by R. Thomson, F. N. Stokman, C. Achen, and T. König, 86–123. Cambridge: Cambridge University Press. ———. 2006b. “Evaluating Political Decision-making Models.” In The European Union Decides, edited by R. Thomson, F. N. Stokman, C. Achen, and T. König, 239–63. Cambridge: Cambridge University Press. Achterkamp, Marjolein C. 1999. Influence Strategies in Collective Decision Making: A Comparison of Two Models. Amsterdam: Thela Thesis. Akkerman, Agnes. 2000. Verdeelde Vakbeweging en Stakingen: Concurrentie om Leden. Amsterdam: Thela Thesis. Arregui, Javier, Frans N. Stokman, and Robert Thomson. 2006. “Exchange and Compromise in the European Union.” In The European Union Decides, edited by R. Thomson, F. N. Stokman, C. Achen, and T. König, 124–52. Cambridge: Cambridge University Press. Austen-Smith, David, and Jeffrey S. Banks. 2005. Positive Political Theory II: Strategies and Structures. Ann Harbor: University of Michigan Press. Axelrod, Robert. 1970. Conflict of Interest: A Theory of Divergent Goals with Applications to Politics. Chicago: Markham. ———. 1984. The Evolution of Cooperation. New York: Basic Books. Bachrach, Peter, and Morton S. Baratz. 1962. “The Two Faces of Power.” American Political Science Review 57: 947–52. Banzaf, John F. 1965. “Weighted Voting Does Not Work: A Mathematical Analysis.” Rutgers Law Review 35: 317–43. Baron, David, and John A. Ferejohn. 1989. “Bargaining in Legislatures.” American Political Science Review 83: 1181–206. Bentley, Arthur F. 1967 [1908]. The Process of Government. Cambridge: Harvard University Press. Berger, Joseph, Bernard P. Cohen, and Morris Zelditch, Jr. 1972. “Status Characteristics and Expectation States.” American Sociological Review 37: 241–55.

Modeling Collective Decision-making 177 Berger, Joseph, Hamit M. Fisek, Robert Z. Norman, and Morris Zelditch, Jr. 1977. Status Characteristics and Social Interaction: An Expectation States Approach. New York: Elsevier. Berger, Joseph, Susan J. Rosenholz, and Morris Zelditch, Jr. 1980. “Status Organizing Processes.” Annual Review of Sociology 6: 470–508. Berger, Joseph, and Morris Zelditch, Jr., eds. 1985. Status, Rewards, and Influence. San Francisco: Jossey-Bass. Berveling, Jaco. 1994. Het Stempel op de Besluitvorming: Macht, invloed en besluitvorming op twee Amsterdamse beleidsterreinen. Amsterdam: Thela Thesis. Bienenstock, Elisa J., and Phillip Bonacich. 1992. “The Core as a Solution to Exclusionary Networks.” Social Networks 14: 231–44. Black, Duncan. 1958. Theory of Committees and Elections. Cambridge: Cambridge University Press. Bueno de Mesquita, Bruce. 1994. “Political Forecasting: An Expected Utility Method.” In European Community Decision Making: Models, Comparisons, and Applications, edited by B. Bueno de Mesquita and F. N. Stokman, 71–104. New Haven:Yale University Press. ———. Predicting Politics. 2002. Columbus: Ohio State University Press. Bueno de Mesquita, Bruce, David Newman, and Alwin Rabushka. 1985. Forecasting Political Events:The Future of Hong Kong. New Haven:Yale University Press. Buskens,Vincent W. 1999. Social Networks and Trust. Amsterdam: Thela Thesis. Cobb, Roger W., and Charles D. Elder. 1972. Participation in American Politics: The Dynamics of Agenda-building. Boston: Allyn and Bacon. Coleman, James S. 1972. “Systems of Social Exchange.” Journal of Mathematical Sociology 2: 145–63. ———. 1990. Foundations of Social Theory. Cambridge: Belknap Press. Cook, Karen S., and Toshio Yamagishi. 1992. “Power in Exchange Networks: A PowerDependence Formulation.” Social Networks 14: 245–65. Dahl, Robert A. 1958. “A Critique of the Ruling Elite Model.” American Political Science Review 52: 463–69. ———. 1961. Who Governs? Democracy and Power in an American City. New Haven:Yale University Press. De Swaan, Abram. 1973. Coalition Theories and Cabinet Formations: A Study of Formal Theories of Coalition Formation Applied to Nine European Parliaments after 1918. Amsterdam: Elsevier. De Waal, Frans. 2005. Our Inner Ape. London: Granta Books. Dijkstra, Jacob, and Marcel A. L. M. Van Assen. 2006. “Externalities in Exchange Networks: An Exploration.” Sociological Theory and Methods 21, no. 2: 279–94. ———. 2008a. “Transferring Goods or Splitting a Resource:Testing the Consequences of the Violation of a Basic Assumption in Exchange Research.” Social Psychology Quarterly 71: 17–36. ———. 2008b. “The Comparison of Four Types of Everyday Interdependencies: Externalities in Exchange Networks.” Rationality and Society 20: 115–43. ———. 2008c. “Effects of Externalities on Patterns of Exchange.” Sociological Theory and Methods 43, no. 23: 91–110. Dijkstra, Jacob, Marcel A. L. M.Van Assen, and Frans N. Stokman. 2008. “Outcomes of Collective Decisions with Externalities Predicted.” Journal of Theoretical Politics 20: 415–41. Esser, Hartmut. 1997. “Die ‘Definition’ der Situation und die Rationalität des Handelns.” In Soziologie im Konzert der Wissenschaften, edited by T. Melegny, H. J. Niederzu, M. Preglau, F. Traxler, and B. Scheikal, 69–90. Opladen: Westdeutscher Verlag, 1997. ———. 2000. “Normen als Frames: Das Problem der ‘Unbedingtheit’ des normativen Handelns.” In Normen und Institutionen: Entstehung und Wirkungen, edited by R. Metze, K. Mühler, and K. D. Opp, 137–55. Leipzig: Leipziger Universitätsverlag.

178 Stokman,Van der Knoop, and Van Oosten Falkner, Gerda, Oliver Treib, Miriam Hartlapp, and Simone Leiber. 2005. Complying with Europe: EU Harmonization and Soft Law in the Member States. Cambridge: Cambridge University Press. Fehr, Ernst, and Simon Gächter. 2002. “Altruistic Punishment in Humans.” Nature 415: 137–40. Friedkin, Noah E. 1992. “An Expected Value Model of Social Power: Predictions for Selected Exchange Networks.” Social Networks 14: 213–29. Friedkin, Noah E., and Eugene C. Johnsen. 1990. “Social Influence and Opinions.” Journal of Mathematical Sociology 15: 193–205. ———. 1997. “Social Positions in Influence Networks.” Social Networks 19: 209–22. ———. 1999. “Social Influence Networks and Opinion Change.” Advances in Group Processes 16: 1–29. Garrett, Geoffrey, and George Tsebelis. 1999a. “Why Resist the Temptation to Apply Power Indices to the European Union?” Journal of Theoretical Politics 11: 291–308. ———. 1999b. “More Reasons to Resist the Temptation of Power Indices in the European Union.” Journal of Theoretical Politics 11: 331–38. Heckathorn, Douglas D. 1980. “A Unified Model for Bargaining and Conflict.” Behavioral Science 25: 261–84. Heemskerk, Eelke M., and Meindert Fennema. 2009. “Network Dynamics of the Dutch Business Elite.” International Sociology 24: 807–32. Hosli, Madeleine O. 1993. “Admission of European Free Trade Association States to the European Community: Effects on Voting Power in the European Community Council of Ministers.” International Organization 47: 629–43. ———. 1997. “Voting Strength in the European Parliament: The Influence of National and of Partisan Actors.” European Journal of Political Research 31: 351–66. Hunter, Floyd. 1953. Community Power Structure: A Study of Decision Makers. Chapel Hill: University of North Carolina Press. Kadushin, Charles. 1968. “Power, Influence, and Social Circles: A New Methodology for Studying Opinion Makers.” American Sociological Review 33: 685–99. Kingdon, John W. 1995 [1984]. Agendas, Alternatives, and Public Policies. New York: Harper Collins; Boston: Little, Brown. Knoke, David, Franz U. Pappi, Jeffrey Broadbent, and Yutaka Tsujinaka. 1996. Comparing Policy Networks: Labor Politics in the U.S., Germany, and Japan. Cambridge: Cambridge University Press. König,Thomas. 1997. Europa auf dem Weg zum Mehrheitssystem: Gründe und Konsequenzen nationaler und parlamentarischer Integration. Opladen: Westdeutscher Verlag. König, Thomas, and Brooke Luetgert. 2009. “Troubles with Transposition? Explaining Trends in Member-State Notification and the Delayed Transposition of EU Directives.” British Journal of Political Science 39: 163–94. Lane, Jan-Erik, and Reinert Maeland. 1995. “Voting Power under the EU Constitution.” Journal of Theoretical Politics 7: 223–30. Laumann, Edward O., David Knoke, and Yong-Hak Kim. 1987. “Event Outcomes.” In The Organizational State: Social Choice in National Policy Domains, edited by E. O. Laumann and D. Knoke, 343–73. Madison: University of Wisconsin Press. Laver, Michael, and Norman Schofield. 1990. Multiparty Government. Oxford: Oxford University Press. Leenders, Roger Th. A. J. 1995. Structure and Influence: Statistical Models for the Dynamics of Actor Attributes, Network Structure and Their Interdependence. Amsterdam: Thela Thesis. ———. 2002. “Modeling Social Influence through Network Autocorrelation: Constructing the Weight Matrix.” Social Networks 24: 21–48. Lindenberg, Siegwart M. 1998. “Solidarity: Its Microfoundations and Macro Dependence: A Framing Approach.” In The Problem of Solidarity:Theories and Models, edited by P. Doreian and T. J. Fararo, 61–112. Amsterdam: Gordon and Breach.

Modeling Collective Decision-making 179 ———. 2000. “It Takes Both Trust and Lack of Mistrust: The Working of Solidarity and Relational Signaling in Contractual Relationships.” Journal of Management and Governance 4: 11–33. Lindenberg, Siegwart, and Nicolai J. Foss. 2011.“Managing Joint Production Motivation: The Role of Goal-framing and Governance Mechanisms.” Academy of Management Review 36: 500–525. Lindenberg, Siegwart M., and Bruno S. Frey. 1993. “Alternatives, Frames, and Relative Prices: A Broader View of Rational Choice.” Acta Sociologica 36: 191–205. Lindenberg, Siegwart, and Linda Steg. 2007. “Normative, Gain and Hedonic Goal Frames Guiding Environmental Behavior.” Journal of Social Issues 63: 117–37. Lukes, Stephen. 1974. Power: A Radical View. London: Macmillan. Markovsky, Barry, David Willer, and Travis Patton. 1988. “Power Relations in Exchange Networks.” American Sociological Review 53: 220–36. Marsden, Peter V., and Noah E. Friedkin. 1993. “Network Studies of Social Influence.” Sociological Methods and Research 22: 127–51. Marsden, Peter V., and Edward O. Laumann. 1977. “Collective Action in a Community Elite: Exchange, Influence, Resources and Issue Resolution.” In Power, Paradigms, and Community Research, edited by R. J. Liebert and A. W. Imershein, 199–250. London: Sage. Mattila, Mikko. 2012. “Resolving Controversies with DEU Data.” European Union Politics 13: 451–61. Mattila, Mikko, and Jan-Erik Lane. 2001. “Why Unanimity in the Council? A Roll Call Analysis of Council Voting.” European Union Politics 2: 31–52. Mills, C. Wright. 1956. The Power Elite. Oxford: Oxford University Press. Mintz, Beth, and Michael Schwartz. 1984. The Structure of Power of the American Corporate System. Chicago: University of Chicago Press. Mizruchi, Mark S. 1982. The American Corporate Network, 1904–1974. Beverly Hills, CA: Sage Publications. Mokken, Robert J., and Frans N. Stokman. 1976. “Power and Influence as Political Phenomena.” In Power and Political Theory: Some European Perspectives, edited by B. M. Barry, 33–54. London: John Wiley. ———. 1979. “Corporate-Governmental Networks in The Netherlands.” Social Networks 1: 333–58. Molm, Linda D. 1997. Coercive Power in Social Exchange. Cambridge: Cambridge University Press. Moser, Peter. 1996. “The European Parliament as a Conditional Agenda Setter: What Are the Conditions? A Critique of Tsebelis (1994).” American Science Review 90: 834–38. Nash, John. 1950. “The Bargaining Problem.” Econometrica 18: 155–62. Nowak, Martin A., and Karl Sigmund. 2005. “Evolution of Indirect Reciprocity.” Nature 437: 1291–97. Ordeshook, Peter C. 1992. A Political Theory Primer. New York: Routledge. Ostrom, Elinor, James Walker, and Roy Gardner. 1992. “Covenants with and without a Sword: Self-governance Is Possible.” American Political Science Review 86: 404–17. Pajala, Antti. 2002. Expected Power and Success in Coalitions and Space: Empirical Voting Power in 17 European Parliaments and the Council of the EU. Turku: Turun Yliopisto. Panchanathan, Karthik, and Robert Boyd. 2004. “Indirect Reciprocity Can Stabilize Cooperation without the Second-order Free Rider Problem.” Nature 432: 499– 502. Pappi, Franz U., and Christian H. C. A. Henning. 1998. “Policy Networks: More than a Metaphor?” Journal of Theoretical Politics 10: 553–75. Raub, Werner, and Jeroen Weesie. 1990. “Reputation and Efficiency in Social Interactions: An Example of Network Effects.” American Journal of Sociology 96: 626– 54.

180 Stokman,Van der Knoop, and Van Oosten Raunio, Tapio. 1997. The European Perspective: Transnational Party Groups in the 1989– 1994 European Parliament. Ashgate: Brookfield. Rockenbach, Bettina, and Manfred Milinski. 2006. “The Efficient Interaction of Indirect Reciprocity and Costly Punishment.” Nature 444: 718–23. Rojer, Maurice. 1996. CAO-Onderhandelingen: Een voorspelbaar, logisch en rationeel proces? Amsterdam: Thesis Publishers. Schalk, Jelmer, René Torenvlied, Jeroen Weesie, and Frans Stokman. 2007. “The Power of the Presidency in EU Council Decision-Making.” European Union Politics 8: 229– 50. Schneider, Gerald, Daniel Finke, and Stefanie Bailer. 2010. “Bargaining Power in the European Union: An Evaluation of Competing Game-theoretic Models.” Political Studies 58: 85–103. Schofield, Norman. 1976. “Instability of Simple Dynamic Games.” Review of Economic Studies 45: 575–94. Schumpeter, Joseph A. 1943. Capitalism, Socialism and Democracy. London: Allen and Unwin. Shapley, Lloyd S. 1953. “A Value for n-Person Games.” In Contributions to the Theory of Games II, edited by H. W. Kuhn and A. W. Tucker, 307–17. Princeton: Princeton University Press. Shapley, Lloyd S., and Martin Shubik. 1954. “A Method of Evaluating the Distribution of Power in a Committee System.” American Political Science Review 48: 787–92. Skvoretz, John, and David Willer. 1993. “Exclusion and Power: A Test of Four Theories of Power in Exchange Networks.” American Sociological Review 58: 801–18. Steglich, Christian, Tom A. B. Snijders, and Mike Pearson 2010. “Dynamic Networks and Behavior: Separating Selection from Influence.” Sociological Methodology 40: 329– 93. Steunenberg, Bernard. 1994. “Decision-Making under Different Institutional Arrangements: Legislation by the European Community.” Journal of Institutional and Theoretical Economics: Zeitschrift fur die Gesamte Staatswissenschaft 150: 642–69. ———. 2010. “Is Big Brother Watching? Commission Oversight of the National Implementation of EU Directives.” European Union Politics 11: 359–80. Steunenberg, Bernard, and Torsten J. Selck. 2006. “Testing Procedural Models of EU Legislative Decision-Making.” In The European Union Decides, edited by R.Thomson, F. N. Stokman, C. Achen, and T. König, 54–85. Cambridge: Cambridge University Press. Stokman, Frans. 2009. Is a Copenhagen Climate Treaty Still Possible? Scientific Analysis Provides New Insights for Agreement and a Better Treaty for the Planet. http: //www. stokman.org/artikel/09Stok.Copenhagenstudy.pdf. Stokman, Frans N., and Jaco Berveling. 1998. “Dynamic Modeling of Policy Networks in Amsterdam.” Journal of Theoretical Politics 10: 577–601. Stokman, Frans N., and Jasper V. Stokman. 1995. “Strategic Control and Interests: Its Effects on Decision Outcomes.” Journal of Mathematical Sociology 20: 289–317. Stokman, Frans N., and Robert Thomson. 2004. “Winners and Losers in the European Union.” European Union Politics 5: 5–23. Stokman, Frans N., Marcel A. L. M. Van Assen, Jelle Van der Knoop, and Reinier C. H. Van Oosten. 2000. “Strategic Decision making.” Advances in Group processes 17: 131–53. Stokman, Frans N., and Jan M. M. Van den Bos. 1992. “A Two-Stage Model of Policy Making: With an Empirical Test in the U.S. Energy Policy Domain.” In The Political Consequences of Social Networks. Vol. 4: Research and Society, edited by G. Moore and A. J. Whitt, 219–53. Greenwich, CT: JAI Press. Stokman, Frans N., and Reinier C. H. Van Oosten. 1994. “The Exchange of Voting

Modeling Collective Decision-making 181 Positions: An Object-Oriented Model of Policy Networks.” In European Community Decision Making: Models, Comparisons, and Applications, edited by B. Bueno de Mesquita and F. N. Stokman, 105–27. New Haven:Yale University Press. Stokman, Frans N., and Manuela Vieth. 2005. “Was verbindet uns wann mit wem? Inhalt und Struktur in der Analyse sozialer Netzwerke.” Kölner Zeitschrift für Soziologie 44: 1–19. Stokman, Frans N., and Evelien P. H. Zeggelink. 1996a. “Self-Organizing Friendship Networks.” In Frontiers in Social Dilemmas Research, edited by W. B. G. Liebrand and D. M. Messick, 385–418. Berlin: Springer. ———. 1996b.“Is Politics Power or Policy Driven? A Comparative Analysis of Dynamic Access Models in Policy Networks.” Journal of Mathematical Sociology 21: 77–111. Stokman, Frans N., Rolf Ziegler, and John Scott, eds. 1985. Networks of Corporate Power: A Comparative Analysis of Ten Countries. Cambridge: Polity Press. Szmatka, Jacek, and David Willer. 1995. “Exclusion, Inclusion, and Compound Connection in Exchange Networks.” Social Psychology Quarterly 58: 123–32. Thomson, Robert. 2009. “Same Effects in Different Worlds: The Transposition of EU Directives.” Journal of European Public Policy 16: 987–1008. ———. 2010. “Opposition through the Back Door in the Transposition of EU Directives.” European Union Politics 11: 577–96. ———. 2011. Resolving Controversy in the European Union: Legislative Decision-Making before and after Enlargement. Cambridge: Cambridge University Press. Thomson, Robert, Frans N. Stokman, Christopher Achen, and Thomas König, eds. 2006. The European Union Decides. Cambridge: Cambridge University Press. Tsebelis, George. 1994. “The Power of the European Parliament as a Conditional Agenda Setter.” American Political Science Review 88: 128–42. ———. 1996. “More of the European Parliament as a Conditional Agenda Setter: Response to Moser.” American Political Science Review 90: 839–44. Tsebelis, George, and Geoffrey Garrett. 1996. “Agenda Setting Power, Power Indices, and Decision Making in the European Union.” International Review of Law and Economics 16: 345–61. ———. 1997. “Agenda Setting, Vetoes and the European Union’s Co-decision Procedure.” Journal of Legislative Studies 3: 74–92. Udehn, Lars. 1996. The Limits of Public Choice: A Sociological Critique of the Economic Theory of Politics. London: Routledge. Van Assen, Marcel A. L. M. 2001. Essays on Actor Models in Exchange Networks and Social Dilemmas. Amsterdam: Thela Thesis. Van Assen, Marcel A. L. M., Frans N. Stokman, and Reinier C. H. Van Oosten. 2003. “Conflict Measures in Cooperative Exchange Models of Collective Decision Making.” Rationality and Society 15: 85–112. Widgrén, Mika. 1994.“Voting Power in the EC Decision-Making and the Consequences of Two Different Enlargements.” European Economic Review 38: 1153–70. ———. 1995. “Probabilistic Voting Power in the EU Council: The Cases of TradePolicy and Social Regulation.” Scandinavian Journal of Economics 97: 345–56. Willer, David, ed. 1999. Network Exchange Theory. Westport, CT: Spraeger. Willer, David, Barry Markovsky, and Travis Patton. 1989. “Power Structures: Derivations and Applications of Elementary Theory.” In Sociological Theories in Progress: New Formulations, edited by J. Berger, M. Zelditch, Jr., and B. Anderson, 313–53. Newbury Park, Calif.: Sage Publications. Willer, Robb, and David Willer. 2000. “Exploring Dynamic Networks: Hypotheses and Conjectures.” Social Networks 22: 251–272. Windolf, Paul. 2009. “Coordination and Control in Corporate Networks: United States and Germany in Comparison, 1896–1938.” European Sociological Review 25: 443–57.

182 Stokman,Van der Knoop, and Van Oosten Yamagishi, Toshio. 1986. “The Provision of a Sanctioning System as a Public Good.” Journal of Personality and Social Psychology 51: 110–16. Zelditch, Morris, Jr., and Joan B. Ford. 1994. “Uncertainty, Potential Power, and Nondecisions.” Social Psychology Quarterly 57: 64–73. Zhelyazkova, Asya, and René Torenvlied. 2009. “The Time-Dependent Effect of Conflict in the Council on Delays in the Transposition of EU Directives.” European Union Politics 10: 35–62.

chapter

Social Exchange, Power, and Inequality in Networks

5

karen s. cook and coye cheshire

Introduction There is an interesting tension in the writings about social exchange, especially in the early work of Homans and Blau as well as in many of the subsequent writings about exchange. For some authors, exchange, especially social exchange, entails the transfer of valued resources or the performance of mutually rewarding actions by actors who are relatively equal. Exchange is voluntary and exit is easy. This image also fits the basic model of market exchange in economics if we add the condition that information is fully and freely available. Even our received image of the Kula Ring exchange system (Malinowski 1922), the classic example of generalized exchange, is viewed primarily as an exchange among equals that builds solidarity at the communal level. Levi-Strauss (1969: 266) argues that generalized exchange “presumes equality.” Ekeh (1974: 63) also expresses an egalitarian bias: “[T]he significance of social exchange for social dynamics lies in its integration of society not in its differentiation of society.” But exchange is often not among equals, and it frequently forms the basis for differentiation and inequality. It is no accident that Blau (1964) named his most famous book Exchange and Power in Social Life.1 Power inequality is an inevitable outcome of differentiation in resources and structural position. Over time some actors gain positions of advantage in their exchange relations (or networks of exchange relations) and thus have the capacity to exploit this advantage. However, distinct forms of exchange have different implications for social differentiation, the emergence of power inequalities, and social solidarity. We discuss the emergence of such inequalities in exchange networks and their implications for the analysis of power and solidarity. Differentiation and a kind of benign form of inequality are central to basic processes of social exchange. Without differentiation in preferences and endowments there would be little reason for exchange. As Blau (1964: 170) notes, “[T]he reason two men engage in a voluntary exchange transaction is that both benefit from it. Both can benefit only if they have divergent attitudes

186 Karen S. Cook and Coye Cheshire (preferences or endowments).” Specialization, he argues, provides each “man [sic]” with more of some resources than he can use and fewer of others than he needs. Some differentiation is thus presumed to exist prior to the formation of social exchange relations, but it is differentiation in endowments and/or preferences that are not necessarily imbued with wider social significance in terms of differential social status, power, and influence. These differences in endowments and preferences breed interdependence, not socially significant distinctions. Once inequality emerges, however, those with more resources are often viewed as having higher status and more social influence (Blau 1977). Such distinctions based on rewards may produce beliefs that support the development of differentiated status conceptions (see, for example, Ridgeway, Berger, and Smith 1985), which may eventually attain broader social significance.

Social Exchange, Power Inequalities, and Solidarity For Blau the source of power is the “one-sided dependence” of one actor on another. This notion is also the key to Emerson’s power-dependence formulation (1962, 1964) of power in exchange relations, though Emerson focuses on mutual dependence. As Blau argues (1964: 118–19), the conditions that Emerson defines as “power-balancing mechanisms” can be viewed as conditions that establish power imbalance itself. This schema, he goes on to suggest, can be used to identify the requirements of power, the conditions of social independence, and the bases of power conflicts and their structural implications. Even among those who are initially equal, power differences are produced, Blau (ibid.: 140) contends, by the imbalances in obligations that are incurred in social transactions across time. This is perhaps the most basic form of inequality that emerges in social exchange relations.These imbalances are built up through social exchange in which one party ends up providing services of greater value than another, inducing a kind of social debt that is difficult to discharge fully except in the return of benefit whenever possible. Such imbalances in obligations naturally occur in most social exchanges, eventually generating inequalities in the dependencies of the actors in the exchange relations that in turn produce power differentials. In discussing the “tension” in systems of exchange, Blau (1994: 158) comments: “A paradox of social exchange is that it gives rise to social bonds between peers and differentiation of status.”2 Blau’s influential analysis, however, focused primarily on the emergence of social differentiation. In his classic treatment of the exchange of advice for status in a work group, he provides an example of a common pattern in groups and presents it as a general principle of social differentiation. He subsequently discusses the role of legitimation and organization as they relate to the emergence and sustainability of differentiation.3 Blau (2002) identified his study of consultations in a work group as the original source of his idea of social exchange. For Blau (1955: 108), “A consultation can be considered an exchange of values; both participants gain something and both have to pay a price.” His version of exchange theory was quite different from Homans’s conception of exchange largely as a result of their different microlevel assumptions about the determinants of behavior. Blau’s microlevel theory was based loosely on an extension of principles of microeconomics,

Social Exchange, Power, and Inequality in Networks 187

whereas Homans’s was based on principles of behaviorism influential at the time he wrote his paper on social behavior as exchange (see also Cook and Gerbasi 2006). In this contest of ideas we see both the early rational choice roots of the individual level model behind some versions of exchange theory and the behaviorism or principles of learning and reinforcement that were reflected not only in Homans’s work but also in the early work of Emerson (1969, 1972a) and subsequently Molm (1985). Competition for scarce resources, Blau (1964: 141) argued, leads to power differentiation. “In informal groups the initial competition is for participation time, which is scarce, and which is needed to obtain any social reward from group membership” (ibid.). This insight is also supported by a large literature concerning the emergence of inequality in small groups both in status and influence (see the work of Bales, Berger, and his colleagues). In communities, Blau goes on to argue, the primitive competition is for “scarce means of livelihood.” He concludes that “as status differences emerge in consequence of differential success in the initial competition, the object of competition changes, and exchange relations become differentiated from competitive ones” (ibid.).4 Here we see the separation of cooperative exchange relations from relations of strict competition and conflict.5 Subsequently, this distinction was superseded by Emerson’s classification of exchange relations into more or less cooperative or competitive. Negatively connected exchange relations involve competition for access to resources. Positively connected exchange relations tend toward cooperative modes of interaction since there is no resource competition in relations that are positively connected. Like Blau (1964), Emerson (1972b) viewed the fundamental task of social exchange theory to be the building of a framework in which the primary dependent variables were social structure and structural changes. He argued that structural arrangements among actors who controlled valued resources created differences in potential power. Cook and Emerson (1978) and their colleagues subsequently demonstrated experimentally that power was a function of relative dependence, which was determined by the shape of the network of interconnected exchange relations and the positions each actor occupied in the network. While Cook and Emerson (ibid.) examined other exchange outcomes such as commitment formation between actors in this seminal piece, the connection between the structure of social networks and the use of power became the central focus of research for the next two decades. Return to the study of commitment and other indicators of solidarity, such as trust, did not occur until the 1990s (with a few exceptions). In networks composed of exchange relations, differentials in dependence generate structurally embedded power inequalities. Actors with greater access to valued resources by virtue of their location in a network of exchange opportunities can often enhance their structural advantage, while those with less access to the resources they value as a result of their network location often lose power with the inability to negotiate more favorable rates of exchange and with few alternatives and limited exit options. In this way, structurally embedded power inequalities in exchange networks often increase over time. Operating against this tendency are the power-balancing mechanisms identified by Emerson (1972b).

188 Karen S. Cook and Coye Cheshire Emerson’s original power-dependence formulation conceived of power as determined by the mutual dependence of one actor on another in a social relation (later applied to exchange relations).This formulation was influential in Blau’s development of his theoretical conceptualization of power and exchange as noted above. For Emerson (1962, 1964, 1972a,b) power in a two-party relation is a function of dependence (Pab = Dba),6 such that the power of actor A over actor B (Pab) is a direct function of B’s dependence on A. More formally, in an exchange relation between two actors, A and B, the power of actor A over B in the Ax:By exchange relation (where x and y represent resources of value) increases as a direct function of the value of y to A and decreases proportional to the degree of availability of y to A from alternative sources (other than B). These two factors—resource value and resource availability—determine the level of B’s dependence on A and thus A’s power over B. An exchange relation in which power (and dependence) are unequal is referred to as “imbalanced” in Emerson’s formulation (1972a,b). An exchange relation is balanced if Dab = Dba; that is, if both parties are equally dependent. The concept “balance” is important because it sets the stage for understanding the “balancing operations” Emerson developed to explain change in exchange relations and networks. He identified four possible “balancing” operations, or processes, that alter or mitigate the power differences between actors in exchange relations characterized by power inequality. If A is more powerful than B—that is, Pab > Pba and Dba > Dab in a two-party exchange relation— four generic options exist to alter the balance of power.These options are based on the two variables that determine levels of dependence: value of the resources at stake and the degree of availability of these resources from alternate sources. To alter the balance of power in the Ax:By exchange relation: (1) actor B can reduce the level of motivational investment in goals mediated by A or reduce the perceived value of the resources at stake7 (that is, a form of “withdrawal” from the relationship); (2) B can come up with alternative sources (for example, actor C) for those goals mediated by A or resources of value provided by A (referred to as “network extension”); (3) B can attempt to increase A’s motivational investment in the goals B mediates (for example, through “status-giving” or other efforts to increase the value of the resources provided to A); and/or (4) B can work to eliminate A’s alternative sources for the goals or resources B mediates access to (for example, by engaging in coalition formation with other actors, in particular suppliers of key resources). We discuss some examples of these processes in the section on power dynamics. The key assumptions of exchange theory, summarized by Molm and Cook (1995: 210), include: (1) behavior is motivated by the desire to increase gain and to avoid loss (or to increase outcomes that are positively valued and to decrease outcomes that are negatively valued); (2) exchange relations develop in structures of mutual dependence (both parties have some reason to engage in exchange to obtain resources of value or there would be no need to form an exchange relation); (3) actors engage in recurrent exchanges with specific partners over time (that is, they are not engaged in simple one-shot transactions); and (4) valued outcomes obey the economic law of diminishing marginal utility (or the psychological principle of satiation). Based on these core assumptions various predictions can be made about the

Social Exchange, Power, and Inequality in Networks 189

behavior of actors engaged in exchange and the effects of different factors on the outcomes of exchange. The power-dependence principle, in addition, is the basis for predictions concerning the effects of increasing the value of the resources in the exchange and their availability from alternate sources, referred to by Emerson as “power-balancing” mechanisms, as noted above. Exchange relations, however, are rarely isolated. Actors are typically embedded in networks of social relations, which provide alternative access to the resources they value. Since access to alternatives is one of the main determinants of differentials in power and dependence, it was natural for Emerson to extend his conception of relational power to the analysis of exchange networks representing linked relations of exchange. To make this link Emerson (1972b) defined several types of network connections that have different implications for the development of power differences within networks composed of exchange relations connected in various ways (see also Cook and Emerson 1978; Cook et al. 1983; and Yamagishi, Gillmore, and Cook 1988). The exchange network is the network of connected exchange relations in which the A:B exchange relation between any two parties is embedded. The network is the social structure created by the distribution of exchange opportunities, and power differences in the network are determined by the value of the available resources to each actor in the exchange and by the nature of their connections to others, which provide access to these valued resources (Cook and Whitmeyer 1992). This social structure of exchange opportunities forms the basis for Emerson’s structural theory of power. Power in this context is potential power and the use of this power is a variable to be explained. Networks are composed of exchange relations that are connected to the extent that exchange in one relation affects or is affected by the nature of the exchange in another relation. By this definition exchange relations are connected, and the actors are linked by virtue of their involvement in these exchange relations. The connection can be either positive or negative.8 A negative connection means that exchange in one relation reduces the amount or frequency of exchange in another exchange relation involving one of the same parties (for example, the A-B and B-C exchange relations are negatively connected at B if exchange in the A-B relation reduces the frequency or amount of exchange in the B-C relation).9 An example is a typical dating network. A connection is positive if the amount or frequency of exchange in one relation increases the amount or frequency of exchange in an exchange relation involving at least one of the parties to both exchanges (for example, the A-B relation is positively connected to the B-C relation if exchange in the A-B relation increases the frequency or amount of exchange in the B-C relation). A communication network is typically positively connected. Exchange in more complicated networks, often called mixed networks (Yamagishi, Gillmore, and Cook 1988), involves both positive and negative connections. Subsequently,Willer and his collaborators in developing Network Exchange Theory (NET) made a distinction between types of connections, similar to Emerson’s conceptualization. Positive connections typically result in what Willer and colleagues call “inclusion” (that is, the need to cooperate), and negative connections result in what they call “exclusion” as a result of competition for access to valued resources. The relationship between the specific structure of

190 Karen S. Cook and Coye Cheshire these social exchange networks (sets of connected exchange relations) and the distribution of power and power use in the network became the central focus of empirical research in the social exchange tradition for several decades. To test power-dependence theory experimentally social interaction was defined as the exchange of resources of value that could be manipulated in a laboratory setting. While the theory applies to the exchange of any resources (or services) of value, it was possible to create a setting for exchange by giving actors access to resources for exchange in the laboratory and giving them monetary value (even though the theory applies to anything of value; see Molm 1997). Emerson published his first empirical tests supporting powerdependence theory applied to dyadic relations in 1964. Subsequent research focused on exchange in networks of connected exchange relations, and the structural determinants of power and power use. We have already introduced the concepts of power, dependence, exchange relations, positive and negative connections, networks, and power-balancing mechanisms. Other key concepts include reciprocity and relational cohesion (linked subsequently to issues of social solidarity). For Emerson, reciprocity was central to all social exchange. Without it there would be no mutual social exchange; thus he conceived of reciprocity as part of the definition of exchange. Norms of obligation emerge to reinforce reciprocity, but reciprocity itself cannot be viewed as an explanation of the continuation of exchange. In Emerson’s earliest formulation (1972a), reinforcement principles provided sufficient explanation for the initiation and the continuity or extinction of exchange relations. Molm and her associates’ recent work (2004, 2007b, 2010), however, treats reciprocity as a variable feature of the different types of exchange. Cohesion represents the “strength” of the exchange relation or its propensity to “survive conflict” (Emerson 1972a). Relational cohesion is the average dependence of the two actors in the relation. Subsequently, Molm (1985) and others (for example, Lawler, Ford, and Blegen 1988) referred to this concept as average total power: the higher the mutual dependence, the greater the total power in the relation. It represents how much is actually at stake in the relation (not the relative power of each actor within the exchange relation, another important determinant of commitment, power use, and cohesion). From this perspective high mutual dependence is positively related to relational cohesion.

Network Determinants of Power Cook, Emerson, and their colleagues (see, for example, Cook and Emerson 1978; Cook et al. 1983; and others) published the results of their early tests of the application of power-dependence principles to the analysis of exchange in networks connected in various ways.The focus was the test of basic propositions concerning the structural effects of networks and specific types of exchange connections on power use, as measured by the distribution of exchange profits in the networks over time. This work led to subsequent studies of network exchange including experiments that challenged the power-dependence formulation (see our discussion of alternative theoretical formulations). Based on power-dependence reasoning, actors who have more alternatives for

Social Exchange, Power, and Inequality in Networks 191

obtaining resources they value are less dependent and thus have more power in the network. In a negatively connected network, actors with more partners with whom they can engage in exchange thus have more power. Access to alternatives increases the availability of the resources of value (assuming relatively constant valuation during an experimental session). Building on the initial study of Stolte and Emerson (1977), the first general test of the theory focused on negatively connected exchange networks. Assuming that to have power is to use it (Emerson 1972), this general proposition could be tested by measuring power use in terms of the extent to which one party could negotiate more favorable terms of exchange than others in a network of linked exchange relations. Exchange was operationalized by having participants negotiate the terms of trade for resources of value, converted to monetary payoffs at the end of the experiment. The interaction was computer-mediated to increase experimental control and to reduce experimenter effects. Assuming actors would use their potential power, the more powerful actor in an exchange relation will typically obtain a larger share of the valuable resources to be exchanged—that is, to receive more points— than the actor’s partner. The two experiments reported in Cook and Emerson (1978) involved four-actor, fully linked networks. The networks were either power-balanced or power-imbalanced. In the unbalanced networks, not only did the more powerful actors gain significantly more points than their partners, but they also gained significantly more points than any of the actors in equivalent positions in the balanced network. Other factors investigated in this study were commitment and equity effects. Cook et al. (1983) developed an extension of the theory to apply to larger networks. In this study, variation in network structures was the key variable of interest. The network studied consisted of five actors linked in a chain so each actor had only two potential trading partners. Ignoring the low-profitability relation (which connected F1 and F2 the two end-points of the chain) results in a line (later called “Line 5”) of five actors involving four exchange relations, F1 – E1 – D1 – E2 – F2.10 Previous theory on social networks had supposed that positional centrality in a network confers the most power, and thus that D1 would be most powerful. However, in this study power-dependence theory was used to predict that if this network is negatively connected, actors E1 and E2 will emerge as the most powerful actors, as a result of the increased availability of their alternatives (F1) and (F2). The experimental evidence supported the predictions concerning relative positional power.11 Cook and Yamagishi (1992) developed an algorithm consistent with powerdependence theory that they referred to as the “equi-dependence” formulation. It specifies equilibrium points in the network at which the dependence between exchange partners reaches “balance.”They argue that social exchanges in a network proceed toward an equilibrium point at which partners depend equally on each other for valued resources. This “equi-dependence” principle has implications for partner selection as well as for exchange outcomes. Three different types of relations that can emerge from a network of potential exchange relations (which they refer to as an exchange opportunity structure) were identified. Exchange relations are those in which exchanges routinely

192 Karen S. Cook and Coye Cheshire occur. Nonrelations are potential partnerships within the network that are never used, and that if removed from the network do not affect the predicted distribution of power. Finally, latent relations are potential exchange relations that also remain unused but that if removed affect the subsequent predicted distribution of power across positions in the network. The existence of latent relations is important because they may be activated at any time as an alternative source of valued resources. An example is the reactivation of a relationship with a former partner. When activated they modify the distribution of power in the network. This principle was supported by simulation results and was consistent with some of the empirical investigations of exchange in power-imbalanced networks of exchange, but the algorithm was limited in its applicability to large-scale networks and thus was superseded by the work of Markovsky, Willer, and their collaborators, who developed the GPI index and later, ESL, an “exchange-seek” algorithm for determining power distributions in networks of exchange relations. One of the most consistent findings in the experimental research on social exchange that emerged over time in various research settings is that relative position in a network of exchange relations produces differences in the relative use of power, typically manifested in the unequal distribution of rewards across positions in a social network (Bienenstock and Bonacich 1992; Cook and Emerson 1978; Cook et al. 1983; Friedkin 1992; Markovsky, Willer, and Patton 1988; Markovsky et al. 1993; Molm and Cook 1995; Skvoretz and Willer 1993). While a number of competing microtheories connecting network structure and power-use emerged in the intervening decades, these competing perspectives generally converge on the prediction that power differentials relate directly to actors’ network positions (Skvoretz and Willer 1993: 803). The theories differ to some extent, however, in the causal mechanisms at work in converting these differentials in network position into actual power differences and often in the nature of the experimental paradigms employed to test predictions. Network Exchange Theory (NET) derived from Willer’s “elementary theory” of social relations is the best-developed and most extensively tested alternative formulation (see below). power dynamics Inherent in Emerson’s treatment of power were the seeds of a theory of social change based on principles relating to changes in the distribution of power in the network of connected exchange relations. In his view the important feature of power inequality is that it creates “strains”12 in exchange relations and provides an impetus toward structural changes, creating problems of collective action unique to exchange contexts (Emerson 1972b; Cook and Gillmore 1984; Lawler and Yoon 1998). Power imbalances were viewed as a temporary state, generating strains in exchange relations. The four distinct “balancing” operations would stabilize relationships, he argued. One such mechanism was coalition formation as a power-gaining strategy. The other mechanisms, listed earlier, included network extension, status giving, and exit. The early work on coalition formation in exchange theory (Cook and Gillmore 1984) empirically demonstrated that power imbalances do promote the formation of coalitions.13 In the simplest hierarchical network structure

Social Exchange, Power, and Inequality in Networks 193

in which one power-advantaged actor exchanges with a number of powerdisadvantaged actors, a coalition of all power-disadvantaged actors against the powerful will lead to a balance of power in the network (ibid.). Coalitions that do not include all disadvantaged actors will not result in power-balance because the powerful actor still possesses alternatives to the coalition, which can be used to advantage. Moreover, coalitions including all of the power-disadvantaged actors tend to be more stable over time (Emerson 1972b) than coalitions that do not. However, the formation of large coalitions is problematic in part as a result of simple coordination issues. A different power-balancing mechanism is network extension. Powerdisadvantaged actors, rather than forming coalitions to balance power, may seek out new exchange relations, reducing their dependence on any given actor for valued resources. Leik (1992) proposed a theory of network extension and contraction based on principles derived from the Graph-theoretic Power Index (GPI) and related assumptions developed within Network Exchange Theory tradition (for example,Willer and Anderson 1981; Markovsky,Willer, and Patton 1988; Markovsky et al. 1993; see below for further details). Leik (1992) argued that as long as actors are assumed to be trying to maximize their power vis-àvis their partners, power-advantaged actors would attempt to reduce linkages between partners to consolidate their power, while power-disadvantaged actors would attempt to create new linkages to increase their power. Such a theory requires that actors have information and strategic capacity: “Without sufficient information and the savvy to utilize it, neither the weak nor the strong will be able to perceive the advantage of linkage changes” (ibid.: 316). In addition to laboratory research on network connections, there have been several attempts to apply these concepts outside the laboratory. For example, Anderson, Håkansson, and Johanson (1994) analyzed business networks, defined as two or more connected relations between businesses conceived as collective actors. A key proposition is that each firm will develop a network identity, which has three dimensions: an orientation toward other actors, competence, and power. Power is a function of an actor’s resources and its network context, following Emerson, Cook, and colleagues. Anderson, Håkansson, and Johanson (1994) report contrasting effects of positive and negative exchange connections in two case studies and point out that connections may actually switch between positive and negative through time, or may be simultaneously positive and negative (Whitmeyer 1997b). In addition, Anderson, Håkansson, and Johanson (1994) identify mechanisms, typically involving network identity that result in change over time in relationships and connections in business networks. Power-dependence theory has also been applied in the study of organizations (resource dependence) and in economic sociology with investigations of network embeddedness and its effects on various types of economic relations (for example, in the manufacturing industry and the like) in part because power-balancing operations are difficult to examine in the laboratory.The relatively short duration of a typical experimental session (often between one to two hours) makes it hard to study change over time. Thus field studies are often more appropriate for investigating power dynamics, as indicated by some of the early applications of power-dependence reasoning in the literature on organizations. One of the first applications in the field

194 Karen S. Cook and Coye Cheshire of organizations was the development of resource dependence theory (Pfeffer and Salancik 1978). Resource dependence creates power differences between organizations and subunits within organizations. According to Scott (1992: 115), a major contribution of this perspective is the specification of strategies that organizations use to manage their strategic dependencies—strategies “ranging from buffering to diversification and merger.” For Emerson (1972b; Cook et al. 1983), many of these strategic responses are examples of powerbalancing or power-gaining tactics14 (see also Thompson 1967; Cook 1977; Cook and Emerson 1984). Other strategies that reflect responses to strategic dependencies include vertical integration and specialization strategies designed to reduce competition and collective reactions that alter power relations such as joint ventures, long-term contracting, and consolidation, similar to coalition formation strategies in the research on individual actors within exchange networks characterized by power differences. Factors other than network structure can affect power use, including behavioral constraints created by fairness considerations (for example, Cook and Emerson 1978) or concerns over the ultimate impact of power use, such as the potential for termination of a relationship of value, especially in the face of strong interpersonal commitment. Analysis of the effects of uncertainty and risk on commitment between parties to the exchange and the structure of social exchange in networks (including the work of Molm and Lawler and their separate collaborators) builds directly on the early theoretical work of Emerson. In fact, Cook and Emerson (ibid.) included examination of the emergence of commitment between parties to an exchange relationship over time, arguing that commitment was more likely to occur under uncertainty, especially when risks were involved (see also Cook and Emerson 1984), especially the risk of failing to locate an exchange partner. Research focused on other factors related to the emergence of cohesive exchange partnerships, including exchange frequency. For Lawler the key factors leading to commitment or relational cohesion were frequent positive exchange and the positive emotions they generate (Lawler and Yoon 1996). Facing uncertain environments (including uncertainty about resource quality), actors involved in exchange were more likely to seek to form committed exchange relations (Cook and Emerson 1978; Kollock 1994) or networks of trusted exchange partners (Cook 2005). A significant effect of the emergence of commitment in many networks is that it reduces the extent to which actors seek exchange with alternative partners and thus serves to reduce power inequalities both within the exchange relation and within the network in which the relation is embedded (Rice 2002). Kollock’s work (1994) demonstrates that uncertainty not only results in commitment formation as a means of reducing uncertainty but also tends to be correlated with perceptions of trustworthiness of the actors involved in the exchange relations and thus trust (see also Cook et. al. 2005).15 Yamagishi, Cook, and Watabe (1998) similarly report that trust emerges in exchange relations under conditions of high uncertainty when actors form commitments to exclusive exchange relations, and it helps to avoid the possibility of exploitation by unknown actors who enter the exchange opportunity structure. Given low uncertainty, actors are much more likely to continue to “play the market” and to avoid forming commitments to

Social Exchange, Power, and Inequality in Networks 195

specific partners to maximize access to valued resources. (Uncertainty in these experiments refers to the likelihood of being exploited by a new partner in a network of exchange opportunities that changes over time.) In research on trust, uncertainty and vulnerability to exploitation are often defined as two of the key elements of situations in which trust considerations are paramount (see, for example, Heimer 2001). Cook, Rice, and Gerbasi (2004) identify the types of economic uncertainty that lead to the formation of trust networks for exchange. Under high political and economic uncertainty such as that characteristic of eastern European countries and Russia post-1989, trust networks, if they become closed, retard the transition to more open market economies. Commitment processes are often also reflected in business relations, even in those involving power differentials. For example, Seabright, Levinthal, and Fichman (1992) in a study of the dissolution of auditor-client relationships in which there is a clear power difference found that commitments (or attachments) mitigated the tendency for such relations to end. And Cook, Kramer, et al. (2004) found that in professional relationships between physicians and their patients (characterized by a power difference) the commitment that resulted from a long-term relationship increased trust and often provided the basis for better communication, higher levels of perceived patient compliance, and greater patient and physician satisfaction. Other relational outcomes including fairness also became the focus of subsequent investigation.

Inequality, Fairness, and Normative Constraints on Social Exchange Processes Homan’s distributive justice proposition is often viewed as central to some of the later developments in social exchange theory, since it clearly introduced a normative component, something he had ironically assiduously tried to avoid in his development of a theory of “sub-institutional” behavior. Homans (1961: 144) argues that justice is a “curious mixture of equality within inequality,” reflecting the tension that subsequent theorists wrote about in their treatment of exchange and power. The introduction of norms of fairness in Homans’s work and later Blau’s on social exchange provides a sense of how norms emerge in groups to regulate exchange (see also Ekeh 1974: 148). In addition, as Ekeh argues, distributive justice, in the hands of Homans, “leads to the justification of status inequalities in the distribution of rewards,” and he was quite critical of this fact. Without the conception of distributive justice the normative component would be missing from Homans’s basic theory of exchange. Cook and Emerson (1978) found that concerns over equity appeared to be more pronounced among those who were power-disadvantaged, not surprisingly, and it was the reduction in willingness to accept offers on their part that mitigated the use of power among those in power-advantaged positions, once the inequality in outcomes became common knowledge. Experimental research by Molm, Takahashi, and Peterson (2003) shows that the effects of fairness considerations in exchange situations are nuanced. They depend not only on the power of the actor involved but also on the type of exchange (for example, negotiated, reciprocal). Stolte (1987) identifies the conditions under

196 Karen S. Cook and Coye Cheshire which fairness considerations come into play in systems of productive exchange in which coordinated action is required. Even in Blau’s early work (1964) some consideration was given to the general role of social norms in the analysis of social exchange. Blau was much less oriented to the study of “sub-institutional” forms of behavior and viewed exchange relations as central to the analysis of associations, organizations, and institutions. In fact, he theorized about the role of such relations in institutional change and transformation. According to Blau (ibid.: 255): “Normative standards that restrict the range of permissible conduct are essential for social life. Although social exchange serves as a self-regulating mechanism to a considerable extent, since each party advances his own interests by promoting those of others, it must be protected against antisocial practices that would interfere with this very process.” He goes on to state: “Without social norms prohibiting force and fraud, the trust required for social exchange would be jeopardized, and social exchange could not serve as a self-regulating mechanism within the limits of these norms.” Moreover, superior powers and resources, which often are the result of competitive advantages gained in exchange transactions, make it possible to exploit others. This creates a need for social norms that prohibit at least those forms of exploitation that conflict with fundamental cultural values. In addition, Blau (ibid.: 257) argued that norms were needed to prohibit actions through which individuals could gain advantages at the expense of the “common interests of the collectivity.” Such norms serve to foster collective action and inhibit selfish acts that undermine collective well-being (see also Ostrom and Walker 2005 for a review). But the maintenance of such norms through monitoring and sanctioning is clearly problematic. We discuss some of the research in the exchange tradition that focuses on these issues of collective action as they relate to power dynamics. The specific role of norms as they emerge to regulate exchange in various settings has not been the focus of much empirical investigation (except for fairness considerations).

Alternative Approaches to the Analysis of Power in Exchange Networks A number of alternative models were developed to specify the exact nature of the link between structural features of networks and the distribution or location of power in exchange networks.These most notably included the work of David Willer and his collaborators (see Willer 1999). Willer differentiates his approach from Emerson’s by defining it as being structural rather than behavioral. This particular characterization does not accurately reflect Emerson’s primary purpose in extending his theory of exchange relations (Emerson 1972a) to the study of networks (Emerson 1972b). In fact, Emerson (1972a: 41) states that his purpose is to “address social structure and structural change within the framework of exchange theory,” even though the microlevel model of individual behavior he developed in the late 1960s was derived from behaviorism (as was Molm’s earliest work on exchange). We provide a brief overview of these alternative formulations (see also reviews by Molm 2000; Cook and Rice 2003; Willer and Emanuelson 2006). Subsequently we return to Emerson’s work and the theorists that have extended his program of

Social Exchange, Power, and Inequality in Networks 197

research—in particular, Edward Lawler and his colleagues, and Linda Molm and her colleagues. network exchange⁄resistance theory In the late 1980s David Willer, Barry Markovsky, John Skvoretz, and their collaborators developed a theory of power in networks that deviated from the earlier works of Blau (1964), Homans (1974), Emerson (1976), and Cook and her colleagues (1978, 1983). In response to what the researchers felt were omissions of scope in the earlier theories of power and social exchange, Markovsky, Willer, and Patton (1988) established a method for computing relative power among positions in different social networks and across different exchange conditions. This network exchange resistance approach defines power as a position’s structurally determined potential, which is indirectly measured through power use (manifested in resource distributions). Predictions about relative power differences are then made using graph theory to analyze structurally contingent interactions. Markovsky, Willer, and Patton (ibid.) formulated predictions about power differences using the Graph-theoretic Power Index (GPI) and then tested these hypotheses in a series of laboratory experiments.16 The GPI predicts levels of actors’ resource acquisition based on the probabilities of particular partnerships being formed. As a net result of actor positions and the lack or presence of ties between them, different structures have more of an impetus toward the exclusion of some parties from exchange than do others. Comparable to powerdependence reasoning, positions in a network that have more alternatives for exchange have more power, as long as the alternatives themselves have few or no options. The GPI method calculates power for each position in a network and then predicts which positions will exchange with others based on the resulting index. Paths with even numbers of links are weighed negatively, while those with odd numbers of links are weighed positively.The “power” of a given point i is computed by taking the sum of all weights connected to that point (Markovsky, Willer, and Patton ibid.). In addition to the use of the GPI, a key component of the network exchange resistance approach is what Willer and his colleagues refer to as elementary theory (Willer and Anderson 1981; Willer, Markovsky, and Patton 1989; Willer and Markovsky 1993). Willer and Anderson (1981) describe elementary theory as a fully general theory, designed to account for social aspects of human interaction as well as cognitive and biophysical factors. According to Willer and his colleagues, elementary theory is a response to what they believe is unnecessary reductionism in earlier psychological theories of operant exchange. Elementary theory disregards learned reinforcement in favor of structural contingencies between actors in social networks (Willer and Anderson ibid.) as the primary focus of study and basis for the analysis of power differences in networks. From a methodological standpoint, experiments originally conducted to test elementary theory differed fairly substantially from the earlier experiments of Emerson, Cook, and their collaborators. Researchers in the power-dependence tradition used computer-mediated systems that deliberately regulated the information available to the participants (for example, relative position in a network, total size of resource pool or profit overlap, and so forth). On the

198 Karen S. Cook and Coye Cheshire other hand, experiments derived from elementary theory used face-to-face interaction situations in which individuals were allowed to overhear other offers, the size of the resource pool was known, and locations within a given network structure were fully disclosed (see Willer and Emanuelson 2006 for a full review of the theory and other work in this tradition). forms of power and coalitions from the perspective of network exchange theory The network exchange approach has been extended in a number of ways since Markovsky, Willer, and Patton’s initial experiments (1988). Skvoretz and Willer (1991) examined additional network structures in which they varied the number of exchanges available to each position in the network. In addition, Markovsky et al. (1993) used the network exchange approach to examine networks that differed in what they called “weak” and “strong” power.The main distinction between the two is that strong-power networks include positions that can exclude particular partners without affecting their own relative power or benefit levels. Densely connected systems (that is, friendship networks) exemplify weak-power networks, while strong-power networks appear in sparsely connected systems such as monopolies and hierarchies. Using computer simulations, Markovsky et al. (ibid.) demonstrated that higher connectivity provides additional opportunities for weaker positions to lessen the structural advantages of the stronger positions. Coalition formation and the distribution of resources resulting from coalitions are also major topic areas for Network Exchange Theory researchers. Following the initial work by Cook and Gillmore (1984), Willer (1987) investigated coalitions in coercive structures in which “the coercer has negative sanctions that are costly for coercees to receive. Coercees seek to retain the positive sanctions they hold” (Borch and Willer 2006: 81). Willer’s laboratory experiments (1987) showed that individuals with low structural power could earn more points by forming a complete coalition with others in the same structural position. Following their earlier work, Willer and Skvoretz (1997) began to use strategic analysis to extract the game-theoretic situations embedded in various structures for the purpose of investigating power use and coalition formation across different network forms. As the researchers argue, individuals in lowpower positions engage in prisoner’s dilemma (PD) type interactions.17 More broadly, all strong power structures appear to include PD games, at least for some range of offers (ibid.: 18). Borch and Willer (2006) applied the same strategic analysis method as Willer and Skvoretz (1997) to examine the game structures in bipartite networks (that is, networks where nodes can be separated into two subsets and all nodes in one group connect to all nodes in the other group). Borch and Willer (2006) found that coalitions can transform the social dilemma for low-power actors from a simple PD to a Privileged game (that is, the dominant strategy equilibrium is Pareto optimal; no social dilemma exists), yet coalitions do not change the initial Privileged game for high-power actors. In this case, coalitions among the less powerful allowed them to gain power, as Emerson had originally predicted.

Social Exchange, Power, and Inequality in Networks 199

other directions in network exchange theory A relatively recent development in this line of research is Willer’s application (2003) of Network Exchange Theory to power use beyond a single relation (called “power-at-a-distance”).Willer (ibid.) argues that power can be exercised through many different levels such as generals who pass orders down a chain of command or in horizontal networks that are typified by competitive seller markets. In horizontal power extensions, an organization’s power is observed through competition and market position. In each of these examples, power can operate indirectly—not just as an exact measure of all face-to-face, direct relationships. In Willer’s conception of indirect power, “For A’s favorable outcome to be at B’s expense, it is not necessary that the two be adjacent. . . . [T]his understanding means that local power exercise is to be distinguished from power-at-a-distance” (ibid.: 1298). Willer allows for network exchange systems that can centralize power such that resources flow from distant actors and intermediaries, potentially giving one or more actors power centralization, defined in terms of power concentration and closeness centrality (ibid.: 1323). Using the experimental framework of Network Exchange Theory,Willer (ibid.) finds support for the predicted effects of power centralization and the ability for central actors to exercise power over indirectly connected actors. Unlike similar studies which limit payoff information to one’s own direct relations, in this study actors have access to information on the payoffs not only to self but also between others in the network. In an empirical extension of the Network Exchange framework, Thye, Willer, and Markovsky (2006) link status and power in negotiated exchange systems by integrating Network Exchange Theory with status characteristics theory. Status characteristics theory argues that in collectively oriented groups, culturally valued status characteristics (that is, gender, age) can generate expectations about performance. Thus, those with the highly valued state of a given characteristic are granted higher status and prestige relative to those with a less valued state of the characteristic (Berger, Cohen, and Zelditch 1972).Thye, Willer, and Markovsky (2006) examine the effect of status ascriptions on power through the theoretical models of status value (value derived from possessing a characteristic) and status influence (ability to change one’s beliefs such that it affects one’s behavior). The researchers show that high status individuals earn more points and have higher perceived competence and influence than lower status individuals in the same network location. As the authors note, status characteristics theory is limited “to the circumstances under which individuals initially select exchange partners and sample information” (ibid. 2006: 1487), while structural and relational theories such as Network Exchange Theory encompass the behaviors that occur within exchange. In general, linking theories to better understand the antecedents and consequences of power, exchange, and status deserves continued attention. In sum, Network Exchange Theory has led to a number of theoretical and empirical developments since the 1980s. During its nascent period, much of the early work led to lively discussion of social exchange networks, methods, and processes (for example, Markovsky, Willer, and Patton 1988; Willer, Markovsky, and Patton 1989;Yamagishi and Cook 1990). With more than three decades of research based on NET, power-dependence, and other frameworks, it is evident

200 Karen S. Cook and Coye Cheshire that the cumulative result is a varied yet empirically grounded body of work that unites on the importance of structural determinants of social network outcomes. probability and game-theoretic approaches to power in exchange networks In contrast to the equi-dependence algorithm based on power-dependence reasoning and network exchange theories of power, two very different approaches to power come from applying principles of game theory and probability to social exchange networks. Friedkin’s expected value theory (1992) builds on underlying assumptions about the probability of interactions in the various theories of social exchange. Friedkin (1995: 213) notes that the fundamental notion of power “as a relation that defines opportunities for interpersonal events, such as exchange transactions . . . or interpersonal influences” was already present in the early theoretical formulations of power (that is, French 1956; Emerson 1962). Friedkin (1992, 1993) argues that the outcomes from a given structure can be viewed in terms of the expected value of the possible exchanges weighted by their actual likelihood of occurrence. As with earlier approaches to power, the expected-value model predominantly focuses on dependencies between actors in a given network structure. Formally, the dependency between any two actors i and j is defined in terms of a joint probability that “i is excluded from an exchange and i does not exchange with j” (Skvoretz and Willer 1993: 816). Thus the relative dependence of one actor on another in a system determines the magnitude of offers. This principle is extended to all possible relations, which ultimately predicts total earnings. The expected-value model receives the most empirical support when the actor’s dependencies are created from observed relative frequencies (Skvoretz and Willer 1993). A subsequent extension of the expected-value model explains how social exchange outcomes shape emergent social structures, while simultaneously addressing some prediction anomalies in the original formulation of expected-value theory (Friedkin 1995).18 Bienenstock and Bonacich (1992, 1993, 1997) developed a theory for analyzing network exchange in terms of N-person cooperative game theory with transferable utility. Bienenstock and Bonacich (1993) demonstrated that a pure rational-choice game theoretic model accurately predicts the frequency of interaction and the distribution of valued outcomes in certain exchange networks based on the likelihood of coalition formation. In their formulation, the distribution of valued outcomes is the only real consequence of power. Thus they argue that power becomes an “unnecessary” intervening variable when analyzing network exchange structures (ibid.: 118). Despite their reservations about power as a variable in exchange networks, Bienenstock and Bonacich’s solution predicts relative power differences using the core, or a set of all outcomes such that “no group of players will accept an outcome if by forming a coalition they can do better” (1992: 125). Forming a coalition equates with engaging in exchange in this model. In game-theoretic terms, the core is the set of all outcomes that have individual rationality (individuals will not accept less in a coalition than could be earned alone), coalition rationality (a set of individuals will not accept less than what

Social Exchange, Power, and Inequality in Networks 201

could be earned in a coalition), and group rationality (all actors combined will maximize their overall reward). Many network structures have a core, which leads to a stable exchange environment. However, if there is no solution that an individual (or set of individuals) can improve on, then a core may not exist and the network is strategically undetermined. Such networks are unstable because no single position or set of positions can consistently exercise power over others (Bienenstock and Bonacich 1993).19 One alternative theoretical approach referred to as general equilibrium analysis (a fundamental tool of microeconomics) was adapted in various ways for application to exchange networks by a number of researchers (for example, Marsden 1983; Whitmeyer 1994, 1997b; Yamaguchi 1996), following early discussion of the subject (Blau 1964; Heath 1976; Cook and Emerson 1978). Unlike game theory, which applies to situations involving few actors who can act strategically, assumptions of general equilibrium analysis are appropriate for market situations, involving many actors, all of whom have competitors (Whitmeyer 1997a). For analyzing exchange networks, it has the merit of yielding a single power distribution, which lies within the range of power distributions identified by game theory.20

Other Forms of Social Exchange moving beyond reward power and negotiated exchange Extending the work of Emerson, Molm and her collaborators (Molm 1990, 1994a, 1994b, 1997; Molm, Peterson, and Takahashi 1999) investigated reward-based versus punishment-based power and affective outcomes such as commitment and trust in different types of exchange. In her early work, Molm (1990, 1994b) examined reciprocal exchange rather than negotiated exchange, the focus of most of the research on exchange and exchange networks by other scholars at that time. In reciprocal exchange, actors do not bargain or negotiate the exchange of resources but participate in reciprocal (and contingent) acts of giving and receiving resources of value. The failure of reciprocity in this context results in infrequent exchange or termination of the relationship. Much of Molm’s research focuses on the distinction between negotiated and reciprocal direct exchanges. The primary distinction between these two forms of exchange is the nature of the interaction involved as noted above. Both are instances of mutual dependence. In negotiated exchange two parties typically bargain over the terms of trade through a process of offers and counteroffers before reaching a mutual agreement (if they do). In reciprocal exchange there is no negotiation, only the contingent exchange of beneficial actions (or resources of value) over time. For Molm, and for Blau, this type of exchange was central to social exchange. Examples include neighbors or friends doing favors for one another or even engaging in reciprocal acts of kindness (for example, mowing each other’s lawns, inviting each other over for dinner on separate occasions, or exchanging baby-sitting).21 Such exchange systems are also common in underground and informal economies in which individuals engage in valued activities for one another in place of purchasing such services in the market. An exchange relationship develops as actors calibrate their acts of giving and receiving to maintain the relationship. Actors run the risk in

202 Karen S. Cook and Coye Cheshire such exchanges of not receiving in return. Obligations emerge over time as parties to the developing exchange relation evaluate their benefits and costs and determine the value of the continuation or termination of the relationship. Molm investigated coercion or punishment power in addition to reward power, noting that power is not based only on the legitimate use of authority (that is, authorized reward power). Her work is in direct contrast to the view that power is the use of one’s structural advantage in a network of exchange relations to exclude other actors from exchange to drive up profits. Molm analyzes the use of sanctions or the imposition of negative outcomes on others as a form of exchange power (arguing that actors have control over both rewards and sanctions in the typical exchange relation).22 The threat or practice of exclusion is most effective in networks in which there is a large power difference between the actors. In addition, actors who are most dependent (least powerful) are most likely to be excluded from exchange in certain networks (for example, networks in which there is a monopoly structure). Molm’s work demonstrates that not all power use is structurally motivated (Molm 1990, 1994a, 1997). Punishment power is not used unwittingly in the same way in which exclusion can produce the unconscious use of reward power in negotiated exchange contexts (Molm 1990). Power use can also have strategic motivations. Punishment power may be used much less frequently, but when it is used it is most likely to be employed purposively to influence the future actions of an exchange partner (Molm ibid., 1994a). The less frequent use of punishment power results from the risk that the target of punishment will simply withdraw from the relationship. Molm’s research demonstrates how coercive power is also constrained by structures of dependence. The primary force in exchange relations is the dependence on rewards, which motivates both the use of punishment as well as reward power (Molm 1990). Since those involved in ongoing exchange relations frequently control both rewards and punishments, Molm’s research facilitates the investigation of more complex exchange situations. In addition, Molm specifies the nature of the precise mechanisms that relate structural determinants of power with the actual use of power by those in positions of power. Norms of fairness or justice and attitudes toward risk play a central role in this analysis. Conceptions of fairness constrain the use of power under some conditions, especially the use of coercive power, and risk aversion makes some actors unwilling to use the structural power at their disposal for fear of loss (see also Kahneman and Tversky 2000). Molm, Takahashi, and Peterson (2003) and Molm (2003a) analyze the relationship between different forms of social exchange (for example, negotiated versus reciprocal exchange) as a key factor in predicting exchange outcomes. The relative importance of fairness, risk aversion, and the strategic use of power varies depending on whether the exchange is negotiated directly between the parties involved or is reciprocal, non-negotiated exchange. Molm’s (Molm, Peterson, and Takahashi 1999, 2001; Molm et al. 2000, 2003b, 2004, 2006, 2007) recent work focuses on explaining these differences in the outcomes of exchange based on the nature of the differences in these forms of exchange, rather than on the structure of the network in which the exchange relations are embedded.23 In addition, she focuses on the integrative (for

Social Exchange, Power, and Inequality in Networks 203

example, commitment, positive affect and trust) rather than the differentiating outcomes (for example, conflict and inequality). She analyzes three general forms of exchange: negotiated and reciprocal direct exchange relations, and indirect exchange referred to as generalized exchange. The study of integrative outcomes of exchange became the focus of the work of both Molm and Lawler and their collaborators, following Blau’s insight that exchanges involve acts that integrate the actors engaged in exchange as well as differentiate them, though the balance of these forces is determined in part by the form of the exchange, as Molm’s work has clearly demonstrated empirically. In particular, the differentiating outcomes are central in negotiated exchange, and under certain conditions integrative outcomes are more salient in reciprocal and generalized exchange. Integrative outcomes include those that reflect the emergence of bonds of attachment: trust, commitment, affect, and cohesion or solidarity. For Lawler the focus is the relationship itself rather than the individual actors engaged in exchange as he attempts to explain the development of group attachment (see below). Molm’s theoretical work specifies the key dimensions along which the basic forms of exchange differ, which she argues account for the differential effects of the form of exchange on both differentiating and integrative outcomes. These dimensions of variation include nature of the risk involved in the exchange, the salience of the potential for conflict, and the significance of the “voluntary” acts of reciprocity, which can occur in reciprocal exchange but not negotiated exchange. Findings to date suggest that many integrative outcomes are more positive in reciprocal exchange, stronger positive affect and perceived fairness, commitment and trust, as well as lower levels of perceived conflict. In part the risk involved allows actors to demonstrate their trustworthiness in reciprocal exchange, but not in negotiated exchange, in which agreements are typically binding (Rice 2002; Molm 2003b). More recent findings (Molm, Collett, and Schaefer 2007; Molm, Schaefer, and Collett 2007) extending this work to include generalized exchange that involves indirect reciprocity (not direct reciprocity as in reciprocal exchange) indicate that even stronger positive regard, affect, trust, and solidarity emerge in this type of exchange setting, reflecting some of the early anthropologists’ claims (for example, Mauss 1954; Malinowski 1922) about generalized exchange systems. Molm (2003b; Molm, Collett, and Schaefer 2007) demonstrates that two features of reciprocal exchange—the indirect form of the reciprocity it entails and the unilateral as opposed to the bilateral flow of resources—are key to explaining the different effects of these forms of exchange on integrative outcomes.When established, reciprocal exchange leads to greater positive affect, higher levels of trust and solidarity, than negotiated exchange that involves direct reciprocity and the bilateral flow of resources typically not inducing much trust. The main reasons behind these findings are developed in her updated theory of reciprocity (Molm 2010). They include the increased risk of the failure of reciprocity in reciprocal (rather than negotiated) exchange, the increased expressive value of reciprocating, and the decreased salience of the conflictual aspects of exchange, which are much stronger in negotiated exchange, as noted above. Lawler’s line of research on solidarity and integrative

204 Karen S. Cook and Coye Cheshire outcomes of exchange is similar, but he proposes a different set of theoretical explanations for their emergence in different types of exchange. relational cohesion and solidarity more on integrative outcomes Lawler and his collaborators (for example, Lawler and Yoon 1993) developed a theory of relational cohesion and subsequently relational exchange. This formulation is an extension of some of the principles derived from Emerson’s initial theory of social exchange (1972a,b). This research examines the conditions under which social exchange relationships emerge out of opportunities for exchange and lead to the emergence of positive emotions about the exchange relation itself. Such positive emotions subsequently lead to relational cohesion, commitment, and solidarity. They develop in response to positive evaluations of exchange outcomes and the frequency of exchange. Low frequency and unfavorable (or less favorable) outcomes are less likely to lead to commitment, positive feelings about the exchange, and bonds of cohesiveness or solidarity (that is, what Lawler summarized in the phrase “wefeeling”). This line of research, along with Molm’s work and related work on trust (Cook, Hardin, and Levi 2005; Cook 2005), addresses the nature of the links between exchange and solidarity. It also expands the scope of exchange theory to include the emotional bases of exchange, commitment, and cohesion, undertheorized in Emerson’s work and in the research that followed on power in exchange networks, including NET (Network Exchange Theory). Lawler and Yoon (1998) conducted experiments on equal and unequal power relations in different network structures to test their theory of relational cohesion. Their results generally supported their theoretical predictions. When networks create equal power relations and actors repeatedly engage in positive exchanges, relatively cohesive partnerships develop. Relational cohesion is high partly as a result of successful exchange and the associated positive feelings. Frequent exchanges in unequal power relations, however, do not have this effect, as predicted. In unequal power relations, low-power and high-power actors have quite different emotional responses to successful exchange. High-power actors are more satisfied and positively oriented to continued exchange than are low-power actors who expressed less satisfaction with successful exchange and perceived less cohesion. Group identity reduced the differentiation of exchange frequencies in unequal power networks by evoking egalitarian norms. Making salient an overarching group identity reduced powerful actors’ efforts to exploit their structural advantage. New work on generalized exchange by both Lawler and Molm also explores not only the underlying collective action issues but also solidarity and trust.24 inequality, generalized exchange, and social solidarity Different types of exchange systems yield different levels of inequality and different degrees of solidarity, as Molm’s work has demonstrated. Differences in inequality and solidarity exist between different forms of direct exchange (for example, reciprocal and negotiated), but also between the broader forms of direct and indirect exchange. When individuals exchange valued resources indirectly and without explicit agreement, it can create generalized exchanges, or

Social Exchange, Power, and Inequality in Networks 205

systems in which the rewards that an individual receives from others do not depend on the resources provided by that individual (Ekeh 1974; Emerson 1976;Yamagishi and Cook 1993). Under typical rules of generalized exchange, reciprocity is not mutual; it is “univocal” or one-directional. Anthropologists first observed and described systems of sustained generalized exchange in the Kula ring among the Trobriand Islanders of Papua New Guinea (Mauss 1954; Malinowski 1922). The Kula ring was a stable exchange of decorative items with no real monetary value in the local economies, but substantial symbolic value among the members of the society. Mauss (1954) described this and other similar systems as gift economies to contrast them with more common systems of bartering and negotiation. In his polemical discussion of social exchange, Peter Ekeh (1974) describes several different types of generalized exchange.Although there are many different varieties and subcategories of indirect exchange in Ekeh’s original formulation, Yamagishi and Cook (1993) label the two major classes as network-generalized and group-generalized exchange. Chains or cyclic patterns of exchange, such as A→B→C→A, characterize the former. In addition to the aforementioned Kula ring in Papua New Guinea (Malinowski 1922), Bearman’s observation (1997) of aboriginal tribes that exchange women between families also constitutes a form of network-generalized exchange. Unlike network-generalized systems, group-generalized exchanges do not typically involve any connections between individual actors. Instead, actors contribute to a collective pool, and valued resources are redistributed from this pool. Examples include citizens working together to build a community bridge, or housemates sharing cleaning responsibilities to maintain a communal kitchen (Yamagishi and Cook 1993). Other contemporary examples of groupgeneralized exchange include peer-to-peer file-sharing networks and other forms of information pooling and redistribution on the Internet (Cheshire 2007; Cheshire and Antin 2009). Generalized exchange, like coalition formation, presents a standard collective action problem. Individuals do better by not giving to others, but if all refuse to give, they all do worse since there is no gain for anyone in the network. Thus the typical structure of a generalized exchange system entails the classic incentives of a social dilemma (see Yamagishi 1995). Since a generalized exchange system requires a minimum of three actors, coordination problems are also likely, especially as the size of the network increases. Actors are not trading within dyads, so they must rely on the goodwill of a third party, over whom they have no direct control. The unilateral gift giving makes systems of generalized exchange vulnerable to free riders. Without guarantees of reciprocity or mutually contingent exchanges, actors can shirk reaping rewards by receiving and refusing to reward others to which they are connected. Giving and receiving resources in a unilateral fashion opens the door to exploitation by others (Bearman 1997; Takahashi 2000). Given the temptation to free ride in generalized exchange systems, it may seem surprising that such systems often emerge and persist. However, the social dilemmas inherent in generalized exchange are partially mitigated by group cohesion, solidarity, and normative constraints on behavior. For example, in the Kula ring (Malinowski 1922), tribes living on different islands in Papua New Guinea traded bracelets

206 Karen S. Cook and Coye Cheshire and necklaces in overlapping, counterdirectional cyclic chains. There was no economic value attached to the bracelets and necklaces, and no one enforced the trading practice. Yet the groups involved in the Kula exchange knew that they were part of a larger system that connected the various islands. Mauss (1954) took some issue with Malinowski’s original interpretation of the Kula ring, noting that the exchange was not so much about individual gifts between individual groups or actors, but was instead a series of exchanges that could not be separated from one another; it must be viewed as a whole (compare Ekeh 1974). In other words, to truly understand generalized exchange systems such as the Kula ring, we must look at the larger network (system) level, not just the dyadic level. Importantly, the individuals in a generalized exchange system may not actually perceive a specific fixed, network-generalized structure. Individuals might help one another (such as in the case of the stranded motorist), essentially choosing when or if to give resources to a recipient. Takahashi and Yamagishi (1996) and Takahashi (2000) refer to this as a pure-generalized exchange system. When viewed as a system rather than a series of isolated dyads, aggregate effects become apparent. Such effects include the reinforcement and transmission of norms of fairness or the development of social solidarity. Takahashi (ibid.) demonstrates that generalized exchange can emerge in a simulation environment, so long as individuals employ a fairness-based selectivegiving strategy. That is, he assumes that individuals will give more often to individuals that have higher ratios of giving over receiving. His solution, like many solutions to the problem of the evolution of cooperation in systems of repeated prisoner’s dilemmas, relies on the existence of network structures that provide some sort of localized information and accountability (for example, Axelrod 1984; Macy and Skvoretz 1998). In such a structure, norms regarding contribution can emerge and persist through reputation systems as well as informal monitoring and sanctioning. The power of Takahashi’s analysis (2000) is that it demonstrates how a normative base can create a system of generalized exchange within a community.25 For Blau, generalized exchange is supported at the group level by a social norm of reciprocity—since “doing favors for others is socially expected.” This is in effect a group norm, and the behavior, he argues, can more simply be explained as conformity to this norm: “Conformity with this norm is the reason all group members receive favors in the long run and solidarity is strengthened” (1994: 156). But there is no explanation of the emergence of the norm of reciprocity, which is key to Blau’s explanation of the emergence of generalized exchange systems. Subsequent work on social dilemmas made it clear that these norms that require monitoring and sanctioning represent what has been called the “second-order” social dilemma (and a collective action problem of the same type as the first-order social dilemma). That is, individuals can free ride on others to monitor and sanction the failure to conform to such norms (see Yamagishi 1995). For Durkheim, the “organic” solidarity built through exchange among equals is compromised when inequality exists. Levi-Strauss (1969) similarly viewed inequality as a barrier to building greater social solidarity through generalized exchange among equals. However, Bearman (1997) argues that generalized exchange tends to block tendencies toward subgroup cleavages

Social Exchange, Power, and Inequality in Networks 207

and thus is more stable, and generates greater communal solidarity. This is in contrast to restricted dyadic exchanges (Ekeh 1974;Yamagishi and Cook 1983), in which exploitation within dyads can occur, potentially weakening groups by allowing inequalities to emerge between pairs of exchange partners. In addition, he argues, dyads may isolate themselves from the group, creating their own exchange system. This type of subgroup cleavage has negative consequences for solidarity at the group or communal level. Dyadic isolation is not likely in generalized exchange systems because connections are composed of indirect, univocal reciprocity rather than direct reciprocity. Bearman’s analysis of the kinship data from Groote Eylandt identifies a clear case in which a generalized exchange system emerged to regulate the transfer of women as wives. “This system ensures solidarity by binding all members into a chain of univocal prestations, embedding each block in a network of debt and obligation” (1997: 1406). Bearman (ibid.: 1413) makes an interesting argument (similar to the arguments of Ekeh 1974) concerning the different solidarity implications of distinct forms of social exchange. “All exchange systems yield solidarity as their by-product, as they embed actors in chains of mutual obligation and debt. But different systems provide different levels of solidarity. In direct dyadic exchange, exploitation can occur[,]” increasing the inequality within and potentially between pairs. “Skilled actors build on ambiguity over valuation in exchange and thereby profit from within the exchange relation. Those exploited by reciprocity norms may appeal to the group for redress, but other actors may be indifferent to exchange outcomes among other pairs and therefore fail to sanction the exploiter.” And only embedded dyadic exchange relations could be affected by monitoring and sanctioning. Thus Bearman goes on to argue: “The inherent tendency in restricted exchange systems is toward increased inequality and differentiation between and within exchange pairs.” This insight is supported by the early experimental work on power and inequality in negotiated exchange networks. Furthermore, he emphasizes the point that “the structure of a society bound together by dyadic exchange is at risk of subgroup cleavages.” This work supports Ekeh’s original argument (1974) that restricted dyadic exchange was more brittle and less likely to lead to the type of organic solidarity and integrative bonds now argued to be a by-product of reciprocal exchange (and relations involving indirect reciprocity as is characteristic of generalized exchanges). Generalized exchange may produce a more secure form of solidarity than other forms of exchange. According to Bearman (1997: 1413),“Equals exchange, and only a violation of reciprocity norms allows exchangers to obtain more value.” Such violations may affect all parties equally, thus normative responses are likely to restrict exploitation. Restricted exchange systems change because of endogenous pressures, while generalized exchange systems are more stable and more likely to be changed by exogenous shocks.This is Bearman’s argument for the observed stability of generalized exchange systems—in particular, the system of cyclical exchange in marriage across blocks defined by kin relations. The structure of marriage across blocks is not one of dyadic, direct reciprocal exchange. Rather it is a systematic cycle from one block to another in a circle of generalized exchange—the origin is as sisters in one block or subgroup to their destination as wives in another subgroup or block.

208 Karen S. Cook and Coye Cheshire Beyond the early anthropological work and Bearman’s research, there have been few applications of generalized exchange conceptions outside the laboratory (see current work of Lawler, Thye, and Yoon 2008; Molm et al. 2007). In one such effort, Cook and Donnelly (1996) applied the concept of generalized exchange to intergenerational relations. Relations between generations can be examined as implicit exchange relations in which each generation must determine how to allocate its resources to the next generation and on what basis. Reciprocity, trust, dependence, power, fairness, and asymmetry in exchange benefits all play a significant role in these determinations. These dynamics are important within families and relate to long-term care, childcare, elder abuse, health care, and the transfer of wealth. Many of these issues also arise at the aggregate level for the society at large in terms of the nature of the relations between the generations with implications for property law, taxation, welfare policy, social and health services, and education. Finally, another important area where ongoing generalized exchange exists outside of the laboratory is on the Internet. Many different websites and services allow individuals to engage in generalized exchange of goods and services. For example, on websites such as NetCycler.com and Freecycle.org, individuals give their unwanted or superfluous items to others who have a matching need. In systems such as Freecycle.org, direct negotiation or payment is explicitly discouraged in order to sustain the gift economy. Researchers have begun to examine group identity, cohesion, and solidarity among individuals who use these generalized exchange systems (Willer, Flynn, and Zak 2010; Suhonen et al. 2010). Consistent with earlier anthropological and sociological work on generalized exchange, this line of work shows that online generalized exchange systems can foster solidarity and community among participants over time. However, online social interactions without negotiation, agreements, or sanctions also create high levels of uncertainty that can be difficult to overcome for potential exchange participants (Suhonen et al. 2010). These topics and others require further investigation.

Future Directions Throughout this chapter, we have detailed many theoretical avenues in the sociological study of social exchange. This body of work has laid a stable foundation on which researchers may further build theoretical and empirical work in this tradition. In this section we propose new directions for research based on common themes in existing research, especially on topics that have not been fully explored. As we have shown, previous work in direct-exchange systems (for example, reciprocal and negotiated exchange) has extensively explored the determinants of power and power use in exchange networks. However, little research has explored the role of power in generalized exchange systems. The lack of work in this area is likely because there are few if any alternative sources of resources or exchange partners in both network and group-generalized exchange systems. Often those who analyze these systems simply assume equality of the actors involved. Power in direct-exchange social networks is often accomplished with ties that grant access to additional resources or to alternative exchange opportunities, so generalized exchange

Social Exchange, Power, and Inequality in Networks 209

systems create new and interesting problems for the study of power relationships in indirect-exchange systems. Even when alternatives are available, such as in the pure-generalized exchange systems studied by Takahashi and Yamagishi (1996) and Takahashi (2000), the unilateral nature of exchange may hamper the development of relational power differences. Although relational power within generalized exchanges may initially seem difficult to formulate, the aggregate effects of generalized exchange systems may significantly connect to larger power dynamics within a given community. As Willer and Skvoretz point out, “[C]ollective action countervails power” (1997: 19). Since group-generalized exchange systems are synonymous with collective action efforts, it is reasonable to imagine group-generalized exchange networks acting as power-balancing exchange structures within a larger system of social exchange. For example, weak-power individuals who are able to work together within group-generalized exchange systems can pool resources to counterbalance alternative resource-rich actors. This collective action through group-generalized exchange is like the use of coalition formation to limit the use of power by others (that is, it is a power-balancing mechanism at the network or group level). This example of generalized exchange networks embedded in a broader system of exchange hints at some of the larger implications of overlapping exchange structures within a social system. While the focus of much empirical and theoretical work in social exchange and power has dealt with the separate forms of exchange, one of the key directions for future research is the study of multiplex relations, or overlapping forms of exchange among the same set(s) of actors. In the real world, social relations frequently involve different intersecting modes of exchange. The same group of actors might exchange different goods and services within positively connected, negatively connected, or mixed types of exchange networks. Furthermore, distinct sets of actors may participate in different systems of exchange, leading to power structures that apply in one domain yet not in another. Social exchange theory is moving increasingly close to analyzing the complexity of real-world dynamics, such as when an exchange relationship in one system affects other actors in distant (and seemingly unrelated) exchange networks. Future research needs to articulate the mechanisms behind these effects (for example, extending work on weak or bridging ties, brokerage, reputations, and so forth). Another crucial direction for research on social exchange processes and power dynamics is the use of real-world data. Experimental research is advancing toward analyses of complex exchange systems that approximate real-life social networks, and the future is profoundly contingent on access to data that can bridge the gap between rigid experimental control and ecological validity. One of the most promising new directions for maximizing both concerns is the study of computer-mediated interactions such as those that emerge on the Internet.This includes empirical research on digital information exchange (see, for example, Cheshire 2007; Cheshire and Antin 2009; Shah and Levine 2003) and gift economies of goods, services, and favors managed through websites and online services (for example, Suhonen et al. 2010; Willer, Flynn, and Zak 2010). A second direction is the investigation of pure economic exchange on the

210 Karen S. Cook and Coye Cheshire Internet. For example, nonbinding and binding negotiated exchange systems such as online auctions continue to be significant topics of investigation for both sociologists and economists. In particular, one of the central focal areas in direct-exchange via the Internet involves emergent reputation systems in online auction systems (for example, Kollock 1999; Houser and Wooders 2006; Yamagishi et al. 2009; Resnick et al. 2000). Reputation systems are critical to the perceived legitimacy of computer-mediated forms of social and economic exchange (Cook et al. 2009). Real-world systems of social exchange are often complex and dynamic, taking different forms and structures over time. Furthermore, individuals frequently join and leave systems of exchange, thereby changing the composition of nodes in a relational network. For example, many relationships start out as formal, contractual arrangements. As the individuals, parties, or organizations learn more about one another and build a history of interactions, they may shift into less formal exchange relationships. The converse is also possible, such as relationships that begin with informal or loose reciprocity (that is, small favors such as alternating rides to the airport). Eventually the experiences from the previous exchanges may carry over or evolve to include more formal, direct negotiations over new types of goods or services. While real-world systems provide an abundance of examples of transitions between forms of social exchange, current theoretical and empirical research is only beginning to tackle such complex social exchange arrangements. We believe that a primary direction for the future of social exchange theory involves the specification of the determinants of power and other relational outcomes within intersecting and multifaceted exchange systems that transition between forms over time. If a system of exchange shifts from one mode to another, this change should have attitudinal and behavioral consequences for the actors involved (Cheshire, Gerbasi, and Cook 2010). As current experimental work shows, a shift between two structural arrangements (for example, reciprocal exchange and negotiated exchange) may influence the attributions actors make about one another (Cook, Gerbasi, and Cheshire 2006). Thus shifts in modes of exchange and other temporal transitions may lead to changes in perceived fairness, trustworthiness, and locus of control, levels of commitment, cohesion, and solidarity. Alterations in the mode of exchange may also have consequences for rates of successfully completed exchanges (for example, cooperation) at the group or network level. As we have previously argued, one of the key differences between the social exchange perspective and related microeconomic theories is that the latter deals with efficiency while the former focuses on the emergence and maintenance of structures through processes such as power. In focusing on the structural determinants of power and the effects of power use, social exchange theory and research openly attempt to bridge microlevel interactions with increasingly larger social structures (for example, groups, networks, organizations, institutions). This is not to say that larger macro structures are simply reduced to constituent exchange structures; rather, the social exchange perspective shows how exchange processes link to observed social structures (Cook 1991) and how they emerge. A vital future direction for social exchange theory and the study of power

Social Exchange, Power, and Inequality in Networks 211

is to confront directly the micro-macro link by integrating interpersonal exchange processes with larger network dynamics. This task continues to elude sociologists, yet there is much agreement about the potential of social exchange theory to meet this challenge (see Lind 1986; Stolte 1988; Singelmann 1972; Cook 1991). Exchange processes exist through social interactions, yet other pre-existing macro structures clearly influence and constrain them. One of the key features of the social exchange perspective is the focus on the relational tie between actors rather than on the actors themselves. Thus the same exchange processes that help explain interpersonal power dynamics can be applied to interorganizational systems (Cook 1977). The challenge is connecting the different levels of analysis through the same theoretical tools and procedures, with a focus on shared processes rather than instrumental reductionism.

Notes 1. Homans (1961, 1974), on the other hand, includes almost no reference to the concept, power, in his book on elementary forms of social behavior in which he introduces his conception of social interaction as exchange (see also Cook and Gerbasi 2004). 2. Blau (1964: 328) argues: “Exchange processes, therefore, lead to the emergence of bonds of intrinsic attraction and social integration, on the one hand, and of unilateral services and social differentiation, on the other.” 3. Blau (1964: 333) extrapolated this argument to the group level, stating: “Transactions among organized collectivities, then, may give rise to social ties that unite them, just as social exchange among individuals tends to produce integrative bonds. These transactions also differentiate competing organizations and may result in the elimination or absorption of competitors and the dominance of one or a few organizations . . . just as unilateral transactions and competition among individuals generate hierarchical differentiation and may result in the dominance of one or a few leaders in a group.” 4. Blau comments on this competition stating: “Those successful in the earlier stages of competition tend to compete later for dominant positions and, in communities, for movement into higher social classes, while the unsuccessful ones cannot compete with them for dominance but become their exchange partners, who receive instrumental benefits in exchange for subordination and status support.” 5. Willer also classifies exchange as distinct from relations of conflict and competition. 6. In Emerson’s original words, the dependence of B on A in turn is a positive function of the “motivational investment” of B in “goals mediated by” A and a negative function of the “availability of those goals” to B outside the A-B relation Emerson (1962: 32). 7. An extreme form of devaluation of the resources provided by actor A to B or of the goals important to B mediated by A is complete withdrawal from the relationship— which came to be known as the “exit” option. 8. In Network Exchange Theory (see Willer 1999), a null connection is said to exist in networks for completeness. In Emerson’s theory if there is no connection between two exchange relations they are not linked in a network sense, though clearly the absence of a connection may have implications for network power. 9. An example is the situation in which A and C are alternative dating partners for B. 10. In addition to the empirical results, computer simulation results for four

212 Karen S. Cook and Coye Cheshire networks: the Line 5, and networks with seven, ten, and thirteen actors were also presented to extend the theory. An early algorithm for determining the distribution of power in a negatively connected exchange network directly from the network structure was also developed by Cook et al. (1983). This algorithm was grounded loosely in power-dependence theory, but its application involved only analysis of the network structure and did not use power-dependence theory or models of actor behavior explicitly, as did subsequent models. 11. The experiments reported in the 1978 and 1983 articles were designed to test theory. Thus many features of these experiments are not theoretically crucial. They are operationalizations of theoretical concepts in a laboratory setting. Exchange was operationalized as coming to agreement over the terms of the trade of resources (or profit points), and negative connection was operationalized by allowing each actor only one exchange per round. (Allowing more than one exchange would have resulted in a positive connection in Emerson’s terminology, giving actors the chance to exchange with each of their alternatives, thus eliminating any competition or reason for exclusion.) While some of these aspects of experimental design are not common in natural situations (for example, one exchange per round), they do instantiate the theoretical concepts in ways easy to control and measure and therefore permit clear tests of theory. Thus tested and supported, the theory can then be applied to more complicated naturally occurring situations involving exchange and exchange networks, the focus of much of the subsequent research. 12. In this work Emerson was clearly influenced by Heider’s “balance” theory, though in his case the strains were not cognitive (though they could be in the case of an individual in a power-imbalanced relationship), but they were defined as structural, the result of structural pressures for change. 13. A much larger literature on coalitions based in part on game theory also demonstrates this fact. 14. As Scott (1992: 193) puts it: “Unequal exchange relations can generate power and dependency differences among organizations, causing them to enter into exchange relations cautiously and to pursue strategies that will enhance their own bargaining position.” 15. Outside the laboratory exchange theory was applied to the study of interpersonal relationships. For example, in the study of dating couples, partners, and married people various authors have applied exchange concepts to the analysis of the longevity and quality of such relationships despite the argument that an exchange “logic” does not work in close, personal relations. Michaels, Edwards, and Acock (1984), for example, find that exchange outcomes are a more important predictor of relationship satisfaction than are equity concerns. In addition, Sprecher’s research (1988, 2001) indicates that relationship commitment is affected more by the level of rewards available to partners in alternative relations than by fairness or equity considerations. 16. The GPI is also a central component of Network Exchange Theory, which was developed by Willer and his colleagues as an alternative to other social exchange theories (for example, Willer 1987; Willer and Patton 1987). 17. Coalitions between low-power actors can also produce game situations like the Chicken game, in which the outcome of mutual defection is the worst possible outcome, yet the best possible outcome occurs if one defects and the partner cooperates (Simpson and Macy 2001). 18. Skvoretz and Willer (1993) found that the expected-value model couldn’t account for the observed advantage of the central position in the Kite network structure. Friedkin’s extension of expected-value theory (1995: 214) resolves this problem by considering the probability of an exchange at time t based on the value of an exchange at time t -1.

Social Exchange, Power, and Inequality in Networks 213 19. Bienenstock and Bonacich’s core solution demonstrates how game theory can successfully predict exchange frequency and resource allocation in negatively connected social exchange networks. While the other major sociological and social psychological theories of power are all based on some rational behavior assumptions (that is, individuals maximize gains over losses), core theory is a pure rational-choice model with no additional explanatory factors. As a consequence of this restricted approach, core theory cannot always make specific point predictions for all positions in some networks. As Skvoretz and Willer note, “Because no specific social psychological principle is assumed, rationality considerations alone cannot always single out a particular outcome” (1993: 804). Despite this limitation, the predictions and experimental results of core theory are consistent with other major sociological theories of power (Bienenstock and Bonacich 1993, 1997; Skvoretz and Willer 1993). 20. For application of general equilibrium analysis to the analysis of power in positively and negatively connected networks, see Yamaguchi (1996). Through estimation of a parameter s which changes sign when the connection changes sign, the model can approximate results from experimental networks, both positive and negative. 21. As Molm (1994b) notes, the distinction between negotiated and reciprocal exchange parallels game theorists’ distinction between cooperative and noncooperative games. “In cooperative games and negotiated exchanges, strictly binding agreements are made jointly by players who can communicate; in noncooperative games and reciprocal exchanges, actors make choices independently, without knowledge of others’ choices” (Heckathorn 1985). 22. For Willer, coercion is treated as distinct from exchange. 23. Molm typically controls for network size and number of alternatives to determine the independent effects of type of exchange. 24. Molm’s focus on reciprocal exchange has been a major impetus for this extension of exchange theory, building on the work of Blau (1964), Ekeh (1974), Emerson (1972a,b), and others. 25. As Takahashi (2000) points out, his simulation analysis is restricted to puregeneralized exchange systems where individuals choose their recipients.

References Anderson, James C., Håkan Håkansson, and Jan Johanson. 1994. “Dyadic Business Relationships within a Business Network Context.” Journal of Marketing 58: 1–15. Axelrod, Robert. 1984. The Evolution of Cooperation. New York: Basic Books. Bearman, Peter. 1997. “Generalized Exchange.” American Journal of Sociology 102: 1383– 415. Berger, Joseph, Bernard Cohen, and Morris Zelditch, Jr. 1972. “Status Characteristics and Social Interaction.” American Sociological Review 37: 241–55. Bienenstock, Elisa Jayne, and Phillip Bonacich. 1992. “The Core as a Solution to Negatively Connected Exchange Networks.” Social Networks 14: 231–43. ———. 1993. “Game-Theory Models for Exchange Networks: Experimental Results.” Sociological Perspectives 36: 117–35. ———. 1997. “Network Exchange as a Cooperative Game.” Rationality and Society 9: 37–65. Blau, Peter M. 1955. The Dynamics of Bureaucracy. Chicago: University of Chicago Press. ———. 1977. Inequality and Heterogeneity. New York: Free Press. ———. 1986 [1964]. Exchange and Power in Social Life. New York: John Wiley and Sons. ———. 1994. Structural Contexts of Opportunities. Chicago: University of Chicago Press.

214 Karen S. Cook and Coye Cheshire ———. 2002. “Reflections on a Career as a Theorist.” In New Directions in Contemporary Sociological Theory, edited by Joseph Berger and Morris Zelditch, Jr., 345–57. Lanham, MD: Rowman and Littlefield. Borch, Casey, and David Willer. 2006. “Power, Embedded Games and Coalition Formation.” Journal of Mathematical Sociology 30: 77–111. Cheshire, Coye. 2007. “Selective Incentives and Generalized Information Exchange.” Social Psychology Quarterly 70: 82–100. Cheshire, Coye, and Judd Antin. 2009. “Order, Coordination and Uncertainty.” In eTrust: Forming Relationships in the Online World, edited by Karen S. Cook, Chris Snijders, Vincent Buskins, and Coye Cheshire, 266–91. New York: Russell Sage Foundation. Cheshire, Coye, Alexandra Gerbasi, and Karen S. Cook. 2010. “Trust and Transitions in Modes of Social Exchange.” Social Psychology Quarterly 73, no. 2: 176–95. Cook, Karen S. 1977. “Exchange and Power in Networks of Interorganizational Relations.” Sociological Quarterly 18: 62–82. ———. 1991. “The Microfoundations of Social Structure: An Exchange Perspective.” In Macro-Micro Linkages in Sociology, edited by Joan Huber, 29–45. Newbury Park, CA: Sage. ———. 2005. “Networks, Norms and Trust: The Social Psychology of Social Capital.” Social Psychology Quarterly 68, no. 1: 4–14. Cook, Karen S., and Shawn Donnelly. 1996. “Intergenerational Exchange Relations and Social Justice.” In Current Societal Concerns about Justice, edited by Leo Montada and Melvin J. Lerner, 67–83. New York: Plenum Press. Cook, Karen S., and Richard M. Emerson. 1978. “Power, Equity and Commitment in Exchange Networks.” American Sociological Review 43: 721–39. ———. 1984. “Exchange Networks and the Analysis of Complex Organizations.” Research in the Sociology of Organizations 3: 1–30. Cook, K. S., R. M. Emerson, M. R. Gillmore, and T.Yamagishi. 1983. “The Distribution of Power in Exchange Networks: Theory and Experimental Results.” American Journal of Sociology 89: 275–305. Cook, Karen S., and Alexandra Gerbasi. 2006. “Homans and Emerson on Power: Out of the Skinner Box.” In George C. Homans: History, Theory and Method, edited by A. Javier Trevino, 157–76. New York: Paradigm Publishers. Cook, Karen S., Alexandra Gerbasi, and Coye Cheshire. 2006. “Exchange Network Transitions: Uncertainty, Risk and Shifts in Mode of Exchange.” Conference Paper. American Sociological Association Annual Meeting, Montreal, Quebec, Canada. Cook, Karen S., and M. R. Gillmore. 1984. “Power, Dependence and Coalitions.” In Advances in Group Processes. Vol. 1, edited by Edward J. Lawler, 27–58. Greenwich, CT: JAI Press. Cook, Karen S., Russell Hardin, and Margaret Levi. 2005. Cooperation without Trust? New York: Russell Sage Foundation. Cook, Karen S., Roderick Kramer, David Thom, Stephanie Bailey, Robin Cooper, and Irena Stepanikova. 2004. “Trust and Distrust in Patient-Physician Relationships: Perceived Determinants of High- and Low-Trust Relationships in Managed Care Settings.” In Trust and Distrust in Organizations: Dilemmas and Approaches, edited by Roderick M. Kramer and Karen S. Cook, 65–98. New York: Russell Sage Foundation. Cook, Karen S., and Eric R. W. Rice. 2003. “Social Exchange Theory.” In Handbook of Social Psychology, edited by John Delamater, 53–76. New York: Kluwer Academic/ Plenum Publishers. Cook, Karen S., Eric R. W. Rice, and Alexandra Gerbasi. 2004. “The Emergence of Trust Networks under Uncertainty: The Case of Transitional Economies—Insights from Social Psychological Research.” In Creating Social Trust in Post-Socialist Transitions,

Social Exchange, Power, and Inequality in Networks 215 edited by Susan Rose-Ackerman, Bo Rothstein, and Janos Kornai, 193–212. NY: Palgrave Macmillan. Cook, Karen S., Chris Snijders,Vincent Buskens, and Coye Cheshire, eds. 2009. eTrust: Forming Relationships in the Online World. New York: Russell Sage Foundation. Cook, Karen S., and Joseph W. Whitmeyer. 1992. “Two Approaches to Social Structure: Exchange Theory and Network Analysis.” In Annual Review of Sociology. Vol., 18, edited by Judith Blake and John Hagan, 109–27. Palo Alto, CA: Annual Reviews. Cook, Karen S., and Toshio Yamagishi. 1992. “Power in Exchange Networks: A PowerDependence Formulation.” Social Networks 14: 245–66. Cook, Karen S., Toshio Yamagishi, Coye Cheshire, Robin Cooper, Masafumi Matsuda, and Rie Mashima. 2005. “Trust Building via Risk Taking: A Cross-Societal Experiment.” Social Psychology Quarterly 68, no. 2: 121–42. Ekeh, Peter. 1974. Social Exchange Theory: The Two Traditions. Cambridge: Harvard University Press. Emerson, Richard M. 1962. “Power-Dependence Relations.” American Sociological Review 27: 31–41. ———. 1964. “Power-Dependence Relations: Two Experiments.” Sociometry 27: 282– 98. ———. 1969. “Operant Psychology and Exchange Theory.” In Behavioral Sociology, edited by R. Burgess and D. Bushell, 378–405. New York: Columbia University Press. ———. 1972a, 1972b. “Exchange Theory, Part I: A Psychological Basis for Social Exchange.” And “Exchange Theory, Part II: Exchange Relations and Network Structures.” In Sociological Theories in Progress. Vol. 2, edited by J. Berger, M. Zelditch, and B. Anderson, 38–87. Boston: Houghton Mifflin. ———. 1976. “Social Exchange Theory.” Annual Review of Sociology 2: 335–62. French, J. R. P., Jr. 1956. “A Formal Theory of Social Power.” Psychological Review 63: 181–94. Friedkin, Noah E. 1992. “An Expected Value Model of Social Power: Predictions for Selected Exchange Networks.” Social Networks 14: 213–29. ———. 1993. “An Expected Value Model of Social Exchange Outcomes.” In Advances in Group Processes, edited by Edward J. Lawler, Barry Markovsky, Karen Heimer, and Jodi O’Brien, 163–93. Greenwich, CT: JAI Press. ———. 1995. “The Incidence of Exchange Networks.” Social Psychology Quarterly 58: 213–22. Gillmore, Mary R. 1987. “Implications of Generalized versus Restricted Exchange.” In Social Exchange Theory, edited by Karen S. Cook, 170–89. Newbury Park, CA: Sage Publications. Gould, Roger. 2002. “The Origins of Status Hierarchies: A Formal Theory and Empirical Test.” American Journal of Sociology 107, no. 5: 1143–78. Heath, Anthony F. 1976. Rational Choice and Social Exchange: A Critique of Exchange Theory. Cambridge: Cambridge University Press. Heckathorn, Douglas D. 1985. “Power and Trust in Social Exchange.” In Advances in Group Processes, edited by Edward J. Lawler, 143–67. Greenwich, CT: JAI Press. Heimer, Carol. 2001. “Solving the Problem of Trust.” In Trust in Society, edited by Karen Cook, 40–88. New York: Russell Sage. Homans, George C. 1961. Social Behavior: Its Elementary Forms. New York: Harcourt, Brace and World. ———. 1974. Social Behaviour: Its Elementary Forms. Rev. ed. New York: Harcourt Brace Jovanovich. Houser, Daniel, and John Wooders. 2006. “Reputation in Auctions: Theory, and Evidence from eBay.” Journal of Economics and Management Strategy 15, no. 2: 353–69.

216 Karen S. Cook and Coye Cheshire Kahneman, D., and A. Tversky, eds. 2000. Choices, Values and Frames. New York: Cambridge University Press. Knoke, David. 1990. Political Networks:The Structural Perspective. Cambridge: Cambridge University Press. Kollock, Peter. 1994. “The Emergence of Exchange Structures: An Experimental Study of Uncertainty, Commitment and Trust.” American Journal of Sociology 100: 313–45. ———. 1999. “The Production of Trust in Online Markets.” In Advances in Group Processes, edited by Edward J. Lawler, 99–123. Greenwich, CT: JAI Press. Lawler, Edward J., Rebecca Ford, and Mary A. Blegen. 1988. “Coercive Capability in Conflict: A Test of Bilateral Deterrence versus Conflict Spiral Theory.” Social Psychology Quarterly 51: 93–107. Lawler, Edward J., Shane R. Thye, and Jeongkoo Yoon. 2008. “Social Exchange and Micro Social Order.” American Sociological Review 73, no. 4: 519–42. Lawler, Edward J., and Jeongkoo Yoon. 1993. “Power and the Emergence of Commitment Behavior in Negotiated Exchange.” American Sociological Review 58: 465–481. ———. 1996. “Commitment in Exchange Relations: A Test of a Theory of Relational Cohesion.” American Sociological Review 61: 89–108. ———. 1998. “Network Structure and Emotion in Exchange Relations.” American Sociological Review 63: 871–94. Leik, Robert K. 1992. “New Directions for Network Exchange Theory: Strategic Manipulation of Network Linkages.” Social Networks 14: 309–23. Levi-Strauss, Claude. 1969. The Elementary Structures of Kinship. Rev. ed. Boston: Beacon. Lind, E. A. 1986. “Substantive Contributions to Federal Court Policy and Procedures: Internal Channels for Psychological Research.” In The Impact of Social Psychology on Procedural Justice, edited by M. Kaplan, 40–55. Springfield, IL: Charles C. Thomas. Macy, Michael W., and John Skvoretz. 1998. “The Evolution of Trust and Cooperation between Strangers: A Computational Model.” American Sociological Review 63: 638– 60. Malinowski, Bronislaw. 1992. Argonauts of the Western Pacific. New York: E. P. Dutton. Markovsky, Barry, John Skvoretz, David Willer, Michael J. Lovaglia, and Jeffrey Erger. 1993. “The Seeds of Weak Power: An Extension of Network Exchange Theory.” American Sociological Review 58: 197–209. Markovsky, Barry, D. Willer, and T. Patton. 1988. “Power Relations in Exchange Networks.” American Sociological Review 53: 220–36. Marsden, Peter V. 1983. “Restricted Access in Networks and Models of Power.” American Journal of Sociology 88: 686–717. Mauss, Marcel. 1990 [1954]. The Gift: The Form and Reason for Exchange in Archaic Societies. New York: W. W. Norton. Michaels, James, John N. Edwards, and Alan C. Acock. 1984. “Satisfaction in Intimate Relationships as a Function of Inequality, Inequity and Outcomes.” Social Psychology Quarterly 47: 347–57. Molm, Linda D. 1985. “Relative Effects of Individual Dependencies: Further Tests of the Relation between Power Imbalance and Power Use.” Social Forces 63: 810–37. ———. 1990. “Structure, Action and Outcomes: The Dynamics of Power in Exchange Relations.” American Sociological Review 55: 427–47. ———. 1994a. “Is Punishment Effective? Coercive Strategies in Social Exchange.” Social Psychology Quarterly 57: 75–94. ———. 1994b.“Dependence and Risk:Transforming the Structure of Social Exchange.” Social Psychology Quarterly 57: 163–76. ———. 1997. Coercive Power in Social Exchange. Cambridge: Cambridge University Press.

Social Exchange, Power, and Inequality in Networks 217 ———. 2000.“Theories of Social Exchange and Exchange Networks.” In The Handbook of Social Theory, edited by George Rizter and Barry Smart, 260–72. London: Sage Publications. ———. 2003a. “Power, Trust and Fairness: Comparisons of Negotiated and Reciprocal Exchange.” In Advances in Group Processes: Power and Status.Vol. 20, edited by Shane R. Thy and John Skvoretz, 31–65. Oxford, UK: Elsevier Press. ———. 2003b. “Theoretical Comparisons of Forms of Exchange.” Sociological Theory 21: 1–17. ———. 2010. “The Structure of Reciprocity.” Social Psychology Quarterly 73, no. 2: 119–31. Molm, Linda D., Jessica L. Collett, and David R. Schaefer. 2006. “Conflict and Fairness in Social Exchange.” Social Forces 84: 2331–52. ———. 2007a. “Building Solidarity through Generalized Exchange: A Theory of Reciprocity.” American Journal of Sociology 113: 205–42. Molm, Linda D., and Karen S. Cook. 1995. “Social Exchange and Exchange Networks.” In Sociological Perspectives on Social Psychology, edited by Karen S. Cook, Gary A. Fine, and James S. House, 209–35. Boston: Allyn and Bacon. Molm, Linda D., Gretchen Peterson, and Nobuyuki Takahashi. 1999. “Power in Negotiated and Reciprocal Exchange.” American Sociological Review 64: 876–89. ———. 2001. “The Value of Exchange.” Social Forces 80: 159–85. Molm, Linda D., David R. Schaefer, and Jessica L. Collett. 2004. “Risk, Trust, and Solidarity: Explaining Effects of the Form of Exchange.” Annual Meetings of the American Sociological Association. San Francisco, CA. ———. 2007b. “The Value of Reciprocity.” Social Psychology Quarterly 70: 199–217. Molm, Linda D., Nobuyuki Takahashi, and Gretchen Peterson. 2000. “Risk and Trust in Social Exchange: An Experimental Test of a Classical Proposition.” American Journal of Sociology 105: 1396–427. ———. 2003. “In the Eye of the Beholder: Procedural Justice in Social Exchange.” American Sociological Review 68: 128–52. Ostrom, Elinor, and Jimmy Walker, eds. 2005. Trust and Reciprocity: Interdisciplinary Lessons from Experimental Research. New York: Russell Sage Foundation. Pfeffer, Jeffrey, and Gerald R. Salancik. 1978. The External Control of Organizations: A Resource Dependence Perspective. New York: Harper and Row. Resnick, Paul, Richard Zeckhauser, Eric Friedman, and Ko Kuwabara. 2000. “Reputation Systems.” Communications of the ACM 43, no. 12: 45–48. Rice, Eric. R. 2002. The Effect of Social Uncertainty in Networks of Social Exchange. Ph.D. dissertation. Stanford University. Ridgeway, Cecilia, Joseph Berger, and LeRoy Smith. 1985. “Nonverbal Cues and Status: An Expectation States Approach.” American Journal of Sociology 90, no. 5: 955–78. Seabright, Mark A., Daniel A. Levinthal, and Mark Fichman. 1992. “Role of Individual Attachments in the Dissolution of Interorganizational Relationships.” Academy of Management Journal 35: 122–60. Scott, W. Richard. 1992. Organizations: Rational, Natural, and Open Systems. 3rd ed. Englewood Cliffs, NJ: Prentice Hall. Shah, S. K., and Sheen S. Levine. 2003. “Towards a Theory of Large-Scale Generalized Exchange.” American Sociological Association Meetings. Theoretical Issues in Economic Sociology Session. Atlanta, Georgia. Simpson, Brent, and Michael Macy. 2001. “Collective Action and Power Inequality: Coalitions for Exchange Networks.” Social Psychology Quarterly 64: 88–100. Singelmann, Peter. 1972. “Exchange as Symbolic Interaction: Convergences between Two Theoretical Perspectives.” American Sociological Review 37: 414–24. Skvoretz, John, and David Willer. 1991. “Power in Exchange Networks: Setting and Structural Variations.” Social Psychology Quarterly 54: 224–38.

218 Karen S. Cook and Coye Cheshire ———. 1993. “Exclusion and Power: A Test of Four Theories of Power in Exchange Networks.” American Sociological Review 58: 801–18. Sprecher, Susan. 1988. “Investment Model, Equity, and Social Support Determinants of Relationship Commitment.” Social Psychology Quarterly 51: 57–69. ———. 2001. “A Comparison of Emotional Consequences of and Changes in Equity over Time Using Global and Domain-Specific Measures of Equity.” Journal of Social and Personal Relationships 18: 477–501. Stolte, John F. 1987. “Legitimacy, Justice and Productive Exchange.” In Social Exchange Theory, edited by Karen S. Cook, 190–208. Newbury Park, CA: Sage Publications. ———. 1988. “From Micro- to Macro-exchange Structure: Measuring Power Imbalance at the Exchange Network Level.” Social Psychology Quarterly 51: 357–64. Stolte, John, and Richard M. Emerson. 1977. “Structural Inequality: Position and Power in Network Structures.” In Behavioral Theory in Sociology, edited by Robert J. Hamblin and John H. Kunkel, 117–38. New Brunswick, NJ: Transaction Books. Suhonen, Emmi, Airi Lampinen, Coye Cheshire, and Judd Antin. 2010. “Everyday Favors: A Case Study of a Local Online Gift Exchange System.” Proceedings of ACM, GROUP 2010 conference, Sanibel Island, FL. Takahashi, Nobuyuki. 2000. “The Emergence of Generalized Exchange.” American Journal of Sociology 105: 1105–34. Takahashi, Nobuyuki, and Toshio Yamagishi. 1996. “Social Relational Foundations of Altruistic Behavior.” Japanese Journal of Experimental Social Psychology 36: 1–11. Thompson, James D. 1967. Organizations in Action. New York: McGraw-Hill. Thye, Shane, David Willer, and Barry Markovsky. 2006. “From Status to Power: New Models at the Intersection of Two Theories.” Social Forces 84: 1471–96. Whitmeyer, Joseph M. 1994. “Social Structure and the Actor: The Case of Power in Exchange Networks.” Social Psychology Quarterly 57: 177–89. ———. 1997a. “Applying General Equilibrium Analysis and Game Theory to Exchange Networks.” Current Research in Social Psychology 2: 13–23 . ———. 1997b. “The Power of the Middleman—A Theoretical Analysis.” Journal of Mathematical Sociology 22: 59–90. Willer, David. 1987. “The Location of Power in Exchange Networks.” Social Networks (Special Issue) 14: 187–344. ———. 2003. “Power-at-a-Distance.” Social Forces 81: 1295–334. ———, ed. 1999. Network Exchange Theory. Westport CT: Praeger. Willer, David, and Bo Anderson. 1981. Networks, Exchange and Coercion: The Elementary Theory and Its Applications. New York: Elsevier. Willer, David, and Pamela Emanuelson. 2006. “Elementary Theory.” In Contemporary Social Psychological Theories, edited by Peter Burke, 217–43. Stanford: Stanford University Press. Willer, David, and Barry Markovsky. 1993. “The Theory of Elementary Relations: Its Development and Research Program.” In Theoretical Research Programs: Studies in Theory Growth, edited by Joseph Berger and Morris Zelditch, Jr., 323–63. Stanford: Stanford University Press. Willer, David E., Barry Markovsky, and Travis Patton. 1989. “Power Structures: Derivations and Applications of Elementary Theory.” In Sociological Theories in Progress: New Formulations, edited by Joseph Berger, Morris Zeldtich, Jr., and Bo Anderson, 313–53. Newbury Park, CA: Sage Publications. Willer, David, and Travis Patton. 1987. “The Development of Network Exchange Theory.” In Advances in Group Processes. Vol. 4, edited by E. J. Lawler and B. Markovsky, 199–242. Greenwich, CT: JAI Press. Willer, David, and John Skvoretz. 1997. “Network Connection and Exchange Ratios:

Social Exchange, Power, and Inequality in Networks 219 Theory, Predictions and Experimental Tests.” In Advances in Group Processes. Vol. 14, edited by Barry Markovsky, Michael Lovaglia, and Lisa Troyer, 199–234. Greenwich, CT: JAI Press. Willer, R., F. J. Flynn, and S. Zak. 2010. “Exchange Structure and Solidarity: The Role of Group Identification in Generalized Exchange.” Working paper. Department of Sociology, University of California. Berkeley. Yamagishi,Toshio. 1995. “Social Dilemmas.” In Sociological Perspectives on Social Psychology, edited by Karen S. Cook, Gary A. Fine, and James S. House, 311–54. Needham Heights, NY: Allyn and Bacon. Yamagishi, Toshio, and Karen S. Cook. 1990. “Power Relations in Exchange Networks: A Comment on Network Exchange Theory.” American Sociological Review 55, no. 2: 297–300. ———. 1993. “Generalized Exchange and Social Dilemmas.” Social Psychology Quarterly 56, no. 4: 235–48. Yamagishi, Toshio, Karen S. Cook, and Motoki Watabe. 1998. “Uncertainty, Trust and Commitment Formation in the United States and Japan.” American Journal of Sociology 104: 165–94. Yamagishi, T., M. R. Gillmore, and K. S. Cook. 1988. “Network Connections and the Distribution of Power in Exchange Networks.” American Journal of Sociology 93: 833– 51. Yamagishi, Toshio, Masafumi Matsuda, Noriaki Yoshikai, Hiroyuki Takahashi, and Yukihiro Usui. 2009. “Solving the Lemons Problem with Reputation: An Experimental Study of Online Trading.” In eTrust: Forming Relationships in the Online World, edited by Karen S. Cook, Chris Snijders,Vincent Buskens, and Coye Cheshire, 73–108. New York: Russell Sage Foundation. Yamaguchi, K. 1996. “Power in Networks of Substitutable and Complementary Exchange Relations: A Rational-Choice Model and an Analysis of Power Centralization.” American Sociological Review 61: 308–22.

chapter

Social Capital

6

henk flap and beate völker

Social Networks and Network Studies People’s relationships show patterns, so-called social networks. These social networks affect people’s lives considerably. For example, in all Western industrialized countries between one- and two-thirds of the working population found their jobs through informal social ties (Marsden and Gorman 2001). People’s chances of illness and recovery, and even of dying, partly depend on their networks (House, Landis, and Umberson 1988). In this chapter we present a research program on social networks, that of social capital. The emergence of social networks and their consequences for individual behavior and well-being are central in this program. Before embarking on this we will describe the state of the art in network studies.

State of the Art A review of the social-scientific literature teaches that social networks studies have gained great momentum since the 1960s. There are more than ten thousand articles on social networks, and a number of handbooks (for example, Wasserman and Faust 1994), edited volumes (for example, Lin and Erickson 2008). and reviews (such as Portes 1998) have been published; network ideas are applied in many different fields in the social sciences (compare Rivera, Soderstrom, and Uzzi 2010). Thus, on the one hand social network research is flourishing. But on the other hand social network research requires general theories that answer questions on the (a) emergence of social networks and their effects, and on the (b) integration of different and diverse results. In addition, one wants to know (c) whether network effects can be generalized across situations. If this is not the case—since most regularities know exceptions—it becomes important to understand (d) why certain network effects sometimes occur and sometimes not. Answering these questions is a necessity for the progress of network research.

Social Capital 221

structuralism An important development that might integrate network studies theoretically is the elaboration of the structuralist view, an expression coined by Mullins (1973). This structuralist stance claims that structure overrides preferences. Social networks are so restrictive that, to explain people’s actions, you need to know only the structure of the social network these actors are part of: “Give me the network and I will tell you what the actors will do.” The selling point of structuralism is that it attempts beating neoclassical economics on its home turf—that is, the analysis of the economic world. All markets—for example, labor markets (Granovetter 1974), but also markets for illegal services such as abortion (Lee 1969)—are socially organized through networks. Structuralism produced valuable theoretical insights that are partly corroborated empirically, and that will be part of more general theories on social networks. In the 1970s, Harrison White developed the idea of structural equivalence and the technique of block model analysis to detect and analyze positions of actors. His major contribution was that not only the “1s,” representing the choices, but also the “0s,” indicating nonchoices, in a sociomatrix are important. The zeros do tell something about someone’s network. When one looks at Figure 6.1, a sociogram of a playgroup of children, where each arrow represents a choice for playing with a child, and Figure 6.2, a sociomatrix of the same group, one sees that children 1, 3, and 5 are similarly placed in the social network, because they do not want to play with each other, and no one picks them to play with. They have a structurally equivalent position, as they are tied to the same persons (Lorrain and White 1971). So, according to the hard-core argument of structuralism, they will also behave similarly. By rearranging the columns and the rows in the sociomatrix in a way that blocks are created with only zeros, as done in Figure 6.3, one can come to this result far more easily. One also sees that there is another group of children

1

2

3

4

figure 6.1. Sociogram of a group of five children 5

figure 6.2. Sociomatrix of the same group

1

2

3

4

5

1

0

1

0

0

0

2

0

0

0

1

0

3

0

0

0

1

0

4

0

1

0

0

1

5

0

0

0

1

0

222 Henk Flap and Beate Völker 1 3 5 2 4

1 0 0 0 0 0

3 0 0 0 0 0

5 0 0 0 0 0

2 1 0 0 0 1

4 0 1 1 1 0

figure 6.3. Block model of the same play group

figure 6.4. Image matrix of the same group

A

B

A

0

1

B

0

1

who are also similarly placed in the network: the children 2 and 4 want to play only with each other and with no one else, whereas all the others would like to play with them. (Note: There is also the notion of being role equivalent, meaning that persons are tied not to the same persons but to similar persons [see Wasserman and Faust 1994].) Aggregating structurally equivalent actors shows that there are two positions in the network—that is, that of an elite person and that of a hanger-on. So the block model can be reduced to a so-called image matrix, as in Figure 6.4, showing only two positions or roles. The basic intuition of zero-blocks must have inspired Granovetter (1973), as White’s student, to come up with his “strength-of-weak-ties” argument: although news is quickly disseminated through strong ties, strong ties will hardly ever bring news that is new, because friends of friends will be friends such that there will be inbreeding. “No strong tie will be a bridge” to other networks, a bridge through which information might flow that is nonredundant, really new, and possibly relevant to one’s goals. Only a weak tie will be a bridge to other new groups. Burt (1992), starting from the same insight that zeros are indicative of someone’s position, argues the importance of structural holes in one’s network: having ties to others, who do not have ties to each other, gives ego nonredundant information and also a control advantage. the problem: structuralist shortcomings However, there is something wrong with the structuralist argument. According to the empirical results no general assumption on effects of network structures holds. An important example is the above-mentioned strength-of-weak-ties argument. Granovetter assumed that weak ties make for quick dissemination of information on job vacancies. His study Getting a Job (1974/1995) indeed showed that weak ties bring better jobs. However, his research evidence is weak: “colleagues” in the sample are equated with weak ties, and contact frequency is used to measure tie strength. Support for the weak-ties argument has not always been found; sometimes it was even refuted (Lin, Vaughn, and Ensel 1981b; Bridges and Villemez 1986; Grieco 1987; Flap and Boxman 1999). Hence, it is not unconditionally true that informal contacts lead to better jobs. On the contrary, placement via social networks only rarely results in higher prestige jobs, and income is often even negatively affected.

Social Capital 223

Another example of a structural hypothesis refuted empirically is the “structural holes” argument. According to Burt (1992), structural holes bring advantages—for example, in occupational life: if weak ties feed back into the group—that is, form no bridge—they bring redundant information. The best assets are strong ties that bridge. They not only bring information advantages but also control advantages, since others to whom one is strongly connected are more prepared to help, and even more so if one’s strong ties remain unconnected (compare Grieco 1987). Note that Burt implicitly uses rational choice arguments. His empirical research on the occupational attainment of managers shows that structural holes work only for established men, but not for women or men who are new in an organization. According to his own explanation newcomers and women lack legitimacy, and therefore need the support of a tight personal network, or preferably the support from a higher placed male mentor (Burt 1992). Upon closer reading it turns out that structuralists often use individualistic assumptions to understand networks.Take “the strength of weak ties” argument: the mechanism producing this structural effect is based on individualistic assumptions. The effect is supposedly the outcome of a tendency toward cognitive balance as described by Heider (1958). Structuralists often supplement their core assumption—that network structure explains people’s behavior— with the idea that people infer what is in their own interest by interpreting and imitating the behavior of people who have a network position similar to their own one. More or less implicitly a social comparison or learning theory is used. For example, in his reanalysis of Coleman, Katz, and Menzel’s classic researchmonograph Medical Innovation (1966), Burt (1987) explains the influence of networks of colleagues on the decision of physicians to adopt a new drug by pointing out that physicians resolve their uncertainty about a risky choice by comparing themselves to other physicians who are located similarly within the social network at large, or who are, with a technical term, structurally equivalent. Another example can be taken from the analysis of product markets by Leifer and White (1988). They argue that producers compete for status and profit, and that they decide on prices for their products not by looking at the demand of buyers but by referring to similarly placed other producers. Burt, White, and Leifer actually use a signaling argument: in case of uncertainty, people use signals to decide about their best interests. the theory-gap in network analysis Structuralists acknowledge these points of criticism. In a—surprisingly widely neglected—small essay on “The Theory Gap in Social Network Analysis,” Granovetter (1979) gave a few examples of well-known network ideas that are refuted. He demonstrates that structural balance theories cannot be upheld, since social reality produces many counterexamples of which he explicates two. Situations of conflicting loyalties—“forbidden” by balance theory—often last for years. Furthermore, he mentions patronage networks, in which a patron has strong ties with his clients while the clients remain unconnected with each other. These open triads are “forbidden” by balance theory. Implicitly he also acknowledged that structural sociologists need individualistic theories for

224 Henk Flap and Beate Völker their explanations. In conclusion, Granovetter (ibid.: 501) gave the following diagnosis of the state of the art in network studies at that time: In my reading of the rapidly expanding literature on “social networks,” one nagging question keeps intruding: where is the theoretical underpinning for all these models and analyses? . . . [M]ost network models are constructed in a theoretical vacuum, each on its own terms, and without reference to a broader or common framework. Despite continuing progress, therefore, the point of diminishing returns is approaching, and will rapidly overtake us, unless we pay more attention to what I call the “theory gap” in network studies.

Thus what has kept social network studies together is not a full-blown theory but the orienting notion that the structure of social networks determines the actions of the network members (Watkins 1957). Why do structuralist “laws” on network effects meet with exceptions? One inroad to this question starts with a constraints-driven rational choice perspective on social networks that conceives of networks as social resources. Some social anthropologists (Kapferer 1972; Boissevain 1974) and sociologists (Fischer 1982; Fischer et al. 1977) took this perspective in the 1970s. Later, the argument has been further developed by sociologists who conceived social networks as social capital.They are rational choice sociologists who, inspired by the achievements of human capital theory, apply a utilitarian, rational choice point of view to social networks. Or they are neo-Marxist and neo-Weberian sociologists who apply an interest-driven account of human action to social networks. All emphasize the productive and investment side to social networks (Bourdieu 1973; Loury 1977; Coleman 1988; Flap 1988; Burt 1992). The remainder of this chapter sketches the research program of the theory of social capital, its core, main questions, and tentative answers. Finally, we provide research evidence showing that the program works and discuss a number of new, theoretically interesting but unresolved questions that have emerged from the program.

The Research Program of a Social Capital Theory the common view One way to solve the problem described above is to go back to the assumptions about human nature. What are the goals people strive for? Most efficient is to assume that basic preferences are the same for everybody and do not change remarkably. Human nature is the same for people at different ages and in different places, meaning that people have the same general goals. According to Adam Smith (Lindenberg 1990) these goals are to achieve physical welfare, social approval, and status. This view also holds that those with more resources will better succeed in reaching their goals. Importantly, according to Smith there are specific means in typical situations to reach these goals. For example, under the constraints of capitalism people hardly have a choice but to specialize in economic resources, since these are the means to most other goods. It is not for the love of money, why people strive for high profit and income, but because money

Social Capital 225

is instrumental for many things people want. This point is somehow lost in economics. People’s preferences or interests are socially conditioned. Given their social situation, people can only achieve the goal of having a good life by realizing the less general goals. The latter are instrumental goals, so to speak. People produce their subjective well-being by optimizing these instrumental goals (Ormel et al. 1999). Sociology contributed at least two major insights to this common view of man. First, next to economic resources there are also other means that produce general human goals. According to the neo-Weberian synthesis of sociology, there are three main types of resources to achieve a better life: economic resources (financial assets and capital goods, but also occupational knowledge and skills), political resources (voting rights, membership of political bodies such as parties), and symbolic resources (occupational prestige). Having these means enables a person to achieve a better life. Second, sociology describes several institutional conditions next to markets. In traditional societies individual action is influenced mainly by traditions and norms transmitted through the generations, whereas in modern societies individual behavior is more constrained by markets and organizations (Heilbronner 1964). A major sociological task is to specify how institutional constraints influence the returns of the above mentioned resources. the program: hard core and research questions It is plausible to interpret personal networks as yet another type of resource, analogous to the three types of resources in the classic sociological program, and treat them as social capital that is instrumental in reaching general goals. This idea ties in with neo-Weberian arguments on life chances, determined by one’s economic, political, and cultural resources. The idea is not new, though. “To have friends is to have power: for they are strengths united”; see Thomas Hobbes in his famous Leviathan (1663). This assumption, however, is of great heuristic value in social network studies. It is the hard core of a budding research program (Lakatos 1970). It serves as a guide for where to look for explanations of network phenomena. In a nutshell, it presents a research program in which social networks are treated as a specific resource important for most goals people have in life. The core of social capital theory consists of two straightforward propositions. First, the social resources hypothesis: people better equipped with social capital will be better able to attain their goals. Second, the investment hypothesis: people will invest in social capital according to its instrumental value in producing their ends.These propositions are analogous to the common view on economic, political, and symbolic resources described above. Take note: social networks are not merely seen as another constraint, but they are social capital that produces goals that cannot be attained otherwise or only at much higher cost. These thoughts lead to some important new questions, such as: 1. What are the main constituents of social capital and how do the various effects of social capital depend on its main constituents? What makes a social network productive? 2. How can social capital be measured?

226 Henk Flap and Beate Völker 3. How are social resources related to other resources? Does social capital increase the returns to human capital? 4. How do people acquire social resources? How and when do persons invest in others? 5. Why are social resources distributed unevenly? 6. Under what circumstances and in what societies are social resources most important? This series of related questions articulates the contours of a research program that is encompassed by the idea of social networks as social resources (Lakatos 1970). Such a program systematizes the results of social network research, creating a system that was lacking before. Several theoretical and empirical contributions to this budding research program can be discerned, which will be discussed next. tentative answers The first question Scattered through the social science literature there are attempts to answer the research questions of the program. Lin, Vaughn, and Ensel (1981a: 1163) pointed to an answer to the first question. Social resources consist in “the wealth, status, power, as well as the social ties, of those persons who are directly or indirectly linked to the individual.” The resources of someone’s network members are substitutes for their own resources; that is why Boissevain (1974: 158–63) called them “second order resources.” An important aspect of the latter is the diversity of the second-order assets. Greater diversity of these resources is social capital, especially if one realizes that for many goals in life only one or two other helpers are needed, and a third or more potential helpers do not bring additional benefits but are a drain on one’s time and energy (see Snijders 1999). It is usually assumed that the mere presence of another person and his social support are enough to satisfy someone’s needs, but it is often crucial what is at the other end of a tie. A mother, for example, usually wants to help her child with his studies; however, if she did not receive much education herself, this help will be of no great avail. According to Coleman (1988, 1993) American youth currently have a dim future because parents do not help their children as much as in earlier years, although American parents nowadays do have more resources than their predecessors. As a counterpoint Coleman (1990: 491) presents the example of children of Asian immigrant families who always purchased two copies of a textbook. The second copy was bought for the mother to study along with her children, in this way maximizing the help she could give to her child. Social capital consists of at least three elements: the number of others prepared or obliged to help, the extent to which they are willing to help, and “what is at the other end of the tie”—that is, their resources. One can include the structure of the network as a fourth dimension of social capital. Bourdieu (1981) and Coleman (1990) hold that there is social capital in a dense social network. It is critical, for example, to school success: a rather tightly knitted

Social Capital 227

community network is an asset as parents will continuously observe each other’s children and correct them if they go astray or notify the parents of these children. So there is social capital in the structure of networks. Dense networks not only lower the costs of information search, they also decrease the costs of norm enforcement. These arguments on the effects of dense networks are also forwarded by transaction cost economics. Coming to an agreement is easier in a closed social network. It spreads reputations for being trustworthy and thereby increases the chances of cooperation because people know that opportunistic behavior will be uncovered, and probably sanctioned by others (Granovetter 1985; Raub and Weesie 1990). Trust is higher in closed networks. People will sooner invest in particular others if they think that these people will honor trust (see below for the answer to question 4); compare Smith 2005). Burt (1992) proposes the opposite view (see, however, Coleman 1990: 313– 19)—that is, that the structural autonomy of a person within a network increases the readiness of others to provide help. If others have no alternatives for you, but you do have alternatives for them, you are autonomous.This does not bring only nonredundant information but also control benefits. You can play those who are tied to you off against each other. Being autonomous, of course, is an instance of having a favorable exchange rate in transactions with others. So being in the middle between other persons who are otherwise disconnected, having “structural holes,” can be seen as social capital. This principle was discovered earlier in studies of patronage-ridden societies: the staying power of patronage derives partly from the particular network structure—that is, an open triangle (Singelman 1975; Flap 1989). With regard to our first question it has to be said that there is too much talk about the concept of social capital and its definitions. Definitions tell nothing about the real world. Sociology is about theories, not about definitions. And Coleman’s proposal to define social capital in a functional mode—social capital is whatever aspect of a social structure that is helping actors to realize their interest (Coleman 1990: 305)—is an immunizing stratagem. The second question As empirical research on social capital is developing, it is annoying that measurements of social capital are often rather ad hoc, pragmatic, and unsystematic. How can social capital be measured, given its multidimensional nature, its goal specificity, and the institutional conditioning? In network research various methods have been used to chart networks, running from direct observation, respondents’ diaries, secondary data—for example, on memberships—to more direct questions such as: “To whom do you feel close?” or the role-relation question: “Who are your three best friends?” Usually the latter two questions are followed by questions on how well the respondent knows these persons, what activities one undertakes with them, what their occupation is, and so forth (see, for example, Laumann 1966: 171). Claude S. Fischer (1982; Fischer and McCallister 1978) improved the measurement of social capital. Instead of starting with the question about friends, and then asking about their jobs and the like, Fischer proposed to first

228 Henk Flap and Beate Völker ask about activities, for example: “With whom do you discuss your personal problems?” and thereafter ask whether this person named is a friend, family, or neighbor, as well as how strong this tie is, how frequent one sees the person, what kind of occupation the person has, and so forth. The first question is called the name generator. The latter ones are name interpreters. This way one gets information on the three major dimensions of social capital: the number of people prepared to help, the extent to which they are prepared to help, and the resources they can use to provide this help. The fourth dimension—that is, the structure of the network—can also be measured by asking whether the persons mentioned know each other. The question about discussing personal problems describes the core-network of confidants and is about the only standard question in network studies. Lin and Dumin (1986) came up with another, even simpler instrument to measure a person’s social capital, called position generator. The respondents are presented a list of job titles, all represented in the labor force and spread across the occupational prestige ladder.The next question is for all these jobs, whether the respondent knows anyone in a particular job, and if so whether this is family, a friend, or an acquaintance.This instrument again gives information on the three major dimensions of social capital—that is, on how many people are prepared to help the respondent, to what extent, and what kind of resources they have to provide this help, while assuming that family is more prepared to help than a friend, and that a friend is more prepared to help than an acquaintance (see Van der Gaag, Snijders, and Flap 2008). Inspired by Putnam’s work (see below) on collective social capital, new measurements are now used. One is a question on general trust: “Generally speaking, would you say that most people can be trusted, or that you cannot be too careful in dealing with people?” The other is a count of the number of memberships in voluntary organizations. But the link of both measures to the network theoretical idea of social capital is questionable. The third question With respect to the third question—how social resources are related to other resources—Bourdieu (1981), Coleman (1988), and Burt (1992) argue that a person’s social capital increases the returns to his other resources. The productivity of social capital is rooted in the opportunities embedded in social relationships that help to benefit from one’s other resources, especially human and financial capital. Social capital adds to their value. An interesting result comes from a study by Boxman, De Graaf, and Flap (1991) on managers of Dutch companies. They found that social capital helps to achieve a higher income at any level of human capital, but human capital makes no difference at the highest levels of social capital. Bourdieu states that the occupational prestige ladder has two dimensions at the upper end: there are jobs with a higher prestige in the financial and economic sphere, and similarly there are jobs with a higher prestige in the cultural sphere. The prestige ladder has the shape of a fork. Hansen (1996) and Flap and Völker (2008) demonstrate how resources are interconnected. Especially at the higher ranks of society children receive an extra benefit if they complete an education and get a job in the same sector as their father; they lose in income if they switch to another work sector.

Social Capital 229

The fourth question With regard to the question of how social resources are acquired, the idea of social capital implies that men spend their resources on others, not only for the efficacy of the moment but also with an eye to the future. As early as 1923, Marcel Mauss, in his famous Essai sur le don, forme archaïque de l’échange, expounded on how people acquire social capital—namely, by giving and in that way indebting others to them. Coleman (1990: 306) introduced the image of “credit slips” to indicate the amount to which a receiver was indebted to his or her helper. In their investment decisions people will always discount what they think future benefits will be and consider their value. The importance of a relation is not only determined by past investments but also by the expectation of future help (Boissevain 1974: 250). People will invest in social capital according to the present expected value of future support. There are a number of assumptions implied in the hard core of social capital theory. The importance of social capital in social life strongly relates to nonsimultaneous exchange of help (delayed or generalized reciprocity), which usually is, at the time given, considerably more valuable to the recipient than it is costly to the donor (Coleman 1990: 90–116). This means that some complementarities of the partners’ fortunes are required. If they do not expect that there might be a reversal of fate, generalized reciprocity between two partners will not occur (Litwak 1985; Cosmides and Tooby 1992). Furthermore, through time there will be a discount rate to the value of future help (Taylor 1974: 9). The faster the value of social capital has to be discounted, the smaller the expected value of support will be. Axelrod (1984: 12) catches the idea’s meaning with the image of the shadow of tomorrow.The value of social capital increases by enlarging the shadow of tomorrow. However, the future is less important than the present, for three reasons. First, players tend to value payoffs less as the time of their obtainment recedes. Second, there is always some chance that the actors will not meet again. A relationship may end when one of the other actors moves away, changes jobs, or dies. Third, there is a risk that the other behaves opportunistically and does not repay his debt. Because there is a time lag between investments and returns, one has to trust that the other person will repay the service delivered. Investing in others is similar to what game theory calls playing a trust game (Dasgupta 1988; Diekmann 2007). As to the costs of investments in others and of entertaining a particular tie, these have not been clearly envisioned in social capital research so far. They will be lower if both partners are members of a pre-existing group (Lindenberg 1998) or a community (Glaeser 2001). Sharing a context where one meets the other also decreases the costs of engaging in and maintaining a social tie. Sharing a context implies that the chance of meeting again increases drastically, so that people can trust each other more (Diekmann 2007: 51). There is a way of acquiring social capital without having to mobilize one’s resources—that is, through endowment, or more generally through ascription. The major example is being born into a family. This part of a person’s social capital develops without one’s own intervention. Further, weak ties can grow “at random,” as a by-product of actions toward other goals. The beauty of the program is that it has one key idea, explaining effects of

230 Henk Flap and Beate Völker social networks as well as their emergence and change. Such an investment theory makes it possible to explain why theories that are often used to explain personal relations, such as the exchange theory and cognitive balance theory, meet with refutations. Refutations include, for example, battered wives staying in a strained relationship with their husbands (Rusbult et al. 1991; Rusbult and Martz 1995). People invest in social networks pending their expected value of future support. Together with the direct costs and rewards of maintaining a tie, the past investments in the tie, the shadow of the future, the quality of available alternatives, and the cross-linkage between the personal networks of both partners in a relation are involved in the decision to divest or invest. The “shadow of the past” is important not only because one has learned about the trustworthiness of others but also because investments in others are nearly always relation-specific. These investments are largely gone if one switches partners. If a person’s investments in others are clustered in a dense network, it is more difficult to withdraw one’s investment because this will damage relations with others as well (Kapferer 1972). An illustration is that homeowners have a larger network in the neighborhood than those who rent their house (Glaeser 2001), not only because they have a greater interest in having ties to their neighbors but also because their shadow of the future is larger. Take note, there is not much research on investment in ties to others. Research is usually on the stability of existing ties and not on newly emerging ones. The establishment of ties is hard to observe. Any action of a person may be a repayment of an earlier investment by the other rather than a genuine act of investment in a new relationship. Experimental research might help here (see various reviews by Jackson: see, for example, Jackson 2005). The fifth question The answer to the fifth question, why social capital is unevenly distributed, is rather straight-forward. If people control more economic, symbolic, and political resources they can produce more social capital, and since the former are usually unevenly distributed, the same goes for the latter—that is, social resources. From the perspective of the others, those with more resources are an attractive target for investments. Lin,Vaughn, and Ensel (1981a) stress that the social background of one’s parents influences the social resources of the children. Combining this hypothesis (the more economic, symbolic, and political resources someone has, the more social capital he can produce) and the social resources hypothesis (the more social capital someone has, the better he can achieve his goals) leads one to expect a reproduction or even an accumulation of social inequality. Social inequality will be perpetuated intragenerationally and intergenerationally by differences in access to and use of social capital (Bourdieu 1973; Flap 1991). So there will be a kind of Matthew effect here. According to Bourdieu social closure through selective employment of social capital is a compensatory strategy used by the social elite when their position is threatened (Bourdieu and De Saint Martin 1982). Often such a closure will be an unintended effect of one group employing social ties to transmit advantages to others of their own group while other groups do not act like that. This goes for job vacancies, but it also applies, for example, to housing (Grieco 1987). If

Social Capital 231

work is organized in vacancy chains, a differential tendency to use informal channels to find a job may quickly result, unintendedly in one group taking over a whole line of work. Lin, Ao, and Song (2009), using a nationwide survey of urban residents in China, established clear reproduction and accumulation effects of social capital. Being older, well educated, and having work experience led to more social capital. While questions 1 to 5 of the research program received tentative answers, there is not yet an answer to the 6th question, on how institutional arrangements influence the returns on social capital. This question is largely equivalent to the question of why there are no pure structural effects. Below we will deal with conditions, such as level of technology and the kind of legal institutions that confound structural effects by causing differences in returns of networks—that is, in the instrumental value of social capital. Before proceeding to an answer on the 6th question, we discuss necessary preconditions for network effects: numbers and places. Numbers and Places. Without meeting there will be no mating (Verbrugge 1977). This insight sank in only recently. Basic to meeting chances are numbers and places. The circumscriptive effect of numbers on meeting chances was brought forward by Blau and Schwartz (1984) in their study on marriage patterns. Their point is summarized neatly in the one-liner: “You can’t marry an Eskimo, if no Eskimo is around.” They presented strong support for their supply-side argument (see also Blum 1995).This argument does not apply only to marriage ties, but to all kinds of relations—for example, in the United States it is far easier for a black person to have a white person for a friend than the other way round. The most obvious condition influencing contact opportunities, apart from absolute numbers, are places. Feld (1981, 1984) noticed that social interaction is often tied to certain places offering foci for interaction with other people. Obvious examples of foci that can organize social ties are public places and facilities, such as a bar, shops, schools, a disco, a restaurant, a library or public squares, but one can also think of work places, voluntary associations, or other organizations. As a result people’s networks become organized around such foci. Social ties emerging from foci are a quasi by-product of other actions, and the relational demography of such foci strongly determines which ties are actually formed (Flap, Bulder, and Völker 1998: 117–18; Lindenberg 1998; Kalmijn and Flap 2001; Mollenhorst,Völker, and Flap 2008a,b; Rivera, Soderstrom, and Uzzi 2010: 105–7). This is a supply-side argument: the composition of people’s social networks reflects the composition of the pool of people in the places that they visit—that is, the opportunity structure for selection of associates (Marsden 1990: 397). Structuralists question the assumption that ties exist because two members of a dyad want to interact with each other. In practice, many ties are involuntary in that they come as part of a network membership package. They may be ties to persons who must be dealt with at work or in the neighborhood (Wellman, Carrington, and Hall 1988). Meeting places have more effect if people are forced or required to stay in them for longer periods of time, and if it is more difficult to enter and

232 Henk Flap and Beate Völker leave a particular meeting place (Mollenhorst,Völker, and Flap 2008a). Smaller meeting places will lead to more dense networks, but the ties are usually more heterogeneous (compare Fischer 1977; 1982). One may also expect that weaker ties will be more strongly influenced through meeting places people find themselves in than stronger ties. And indeed meeting places are less decisive in finding a friend or a partner than in acquiring other types of ties. Preferences play a larger role in the choice of stronger ties (Mollenhorst,Völker, and Flap 2008a). Our idea is that through time there occurred an unbundling of contexts in the Western world, leading to individualization also of contact patterns. Coleman (1990: 579 passim) hinted that such a process may be unfolding (Pescolido and Rubin 2000; Völker, Flap, and Mollenhorst 2009). Network changes are a result of changes in the structure of meeting places. In modern society, social contexts such as work, neighborhood, family, or voluntary groups have become unbundled. There is a long-term trend of unbundling from “ascribed” (family, neighborhood) to “achieved” social contexts (work, voluntary organizations). Or as Coleman calls them, from “primordial” structures that grow without conscious design to “purposive social structures” that are consciously designed. Activities that were once enmeshed in the home, extended family, or the neighborhood, such as domestic care, child rearing, food consumption, preparation for adult tasks, spending leisure time, and work are taken over by modern state or market institutions and their agents (see Coleman 1990: 585). Organizations take over activities that were once integrated in family and neighborhood communities because they are often more efficient in conducting these activities. If social contexts are falling apart, personal networks are expected to change. The more contexts are bundled, the more an individual meets the same others and the more the networks of actors will overlap. In consequence, personal networks are expected to be denser and probably more multiplex, and will exhibit stronger ties than personal networks in separated, unbundled contexts. Once contexts become unbundled different network structures can emerge, since the supply of potential interaction partners increases. The ultimate result of context unbundling would be a radial network structure, where the focal actor is related to others through ties that are disconnected, weak, and only uniplex (Pescolido and Rubin 2000). Another cause of unbundling has been the functionalistic city planning in the last century. The spatial spread of functions hinders community development. In one place you live, in another you work, going out takes place in yet another, and so forth (see Jacobs 1961). In reaction, modern city planning, in particular the so-called new urbanism, has recently been heading toward consciously bundling contexts by locating work, residence, and recreational facilities within each other’s proximity. A basic argument is given by Lindenberg (1998), who points out that it is efficient if a relationship or a network fulfills various functions at once. In network studies, the costs of investing in new ties and maintaining existing ties are often not accounted for. Having friends in one place is less costly. Verbrugge (1979) and Mollenhorst,Völker, and Flap (2008a) found that people favor having their friends in one place, even if that results in friendships with

Social Capital 233

dissimilar others: if a person has a friend in a certain setting, the chances that a second friend will be taken from that same setting are far greater. The sixth question: technology Insecurity probably enlarges the value of social capital. Analyses of African stateless societies at a rather low level of technology (horticulture and herding) show the strong impact of technology on what can and has been done by social networks. In such a situation an extra premium is placed upon the formation and maintenance of social capital, since under subsistence conditions each family is too small to support itself (Gluckman 1965: 13–14). For example, success in hunting today is no guarantee of success tomorrow, so it is wise to share with others. Moreover, since there are “no cops and constables” in stateless societies, one had better have friends for personal protection (Flap 1988, 1997). And as food and other goods are difficult to store or trade, the main investment available to man is in personal relationships. At higher levels of technology insecurities are not banned from social life. Although some types of social capital are goal specific (for example, it takes a strong man to carry an invalid), its major advantage is that, like money or human capital, it is often a means to all ends. Litwak (1985) showed that support from informal networks is far more important in the sheer amount of services delivered and the number of people that are helped than care provided by formal organizations. Informal relations can better master events and tasks with many contingencies that are not easily subdivided, or involve problems for which the time of their occurrence is hard to tell. That is why solutions to many everyday problems cannot be routinized or standardized. Coleman (1990) provided a very apt example. Because small children growing up in a neighborhood need continuous attention, so that they, for example, do not get into traffic incidents or fall into a pool of water, a dense network of good relations among neighbors is very helpful. When solutions can be routinized and standardized through technology or universal laws and rules, the value of social capital usually shrinks. But new highly developed technologies also produce new nonstandard, nonroutine exigencies because of all kinds of unforeseen tight couplings in technological and work processes (Perrow 1986; Vaughn 1990) that can be mastered only by resourceful people and through fine tuning of social relations. The damage potential of accidents is often enormous in terms of money and lives lost. Generally, the value of social capital will increase in periods of economic contraction, even in industrial societies. If there is a labor surplus, stronger ties will be more important for getting a job (Grieco 1987: 48). In the last decades, the discovery and diffusion of new technologies has raised hopes of various people that communities might be saved or rebuilt by using the Internet. The costs of interacting have decreased drastically. Yet research shows that the Internet largely functions just like the telephone (Fischer 1992)—that is, one emails mainly with those whom one also speaks with. Internet traffic is mostly very local. So the hope that one might form new communities in an easy way while using the Internet has to be curtailed. People first have to meet in person, looking each other in the eyes, before they trust each other. In our own survey of the networks of the Dutch, we found

234 Henk Flap and Beate Völker that only 1 percent of all the contacts were first met on the Internet. The opposite is not true either; the Internet does not make people lonely (Franzen 2000). The network of those who use the Internet does not decrease in size, nor does it decrease the time spent with friends. The sixth question: institutions It is a challenge for the field to come to grips with the institutional embeddedness of social ties. Institutions, just like technology, can provide universal solutions to human problems that make particularistic solutions through mobilizing social capital more or less superfluous. The welfare state provides social rights, pre-empting much of the former value of social networks. This idea is sometimes called the “crowding out hypothesis.” But man does not live by bread alone: social prestige and an identity cannot be created through issuing social laws. Particularistic solutions are called for, and there will always be some value in social relations with others. Institutional arrangements affect the productivity of people’s resources, social capital not excluded. According to Coleman (1990: 585–87, 1993) there has been an irrevocable loss of social capital in Western industrial societies, caused by the growth of the welfare state, technological changes, and the rising number of large organizations providing services that were once produced more efficiently in the family and the neighborhood.These developments have destroyed the social capital in the family and local community. Parents, for example, will not take care of other parents’ children. They will not even invest much more in their own children, because their need for them has decreased with the availability of old age pensions. Putnam (2000) has presented quite some empirical evidence that in the United States over the last thirty years there has indeed been a decline in social capital, at least as measured by him. That is social capital measured by memberships in associations and by trust in strangers. The images of “bowling alone” and “checkbook-writing organizations” summarize the discussion (Putnam 1995). We should note, however, that such trends as the decline in membership rates of voluntary associations are to be discerned in some Western countries, but certainly not in all. Given a certain institutional arrangement, social capital may be differently productive across social groups. Van der Meer, Scheepers, and Te Grotenhuis (2008) find in their international comparative research, for example, that enforcement of rights by the state, as well as a greater national welfare, lead to more informal help given especially to and received by the poor. Institutions also influence what type of relationship might be instrumental to a good life. A major example of the last decades is provided by the political turnover in Eastern Europe from a totalitarian one-party to a democratic multiparty political system, and from a centrally controlled economy to a market economy. The institutional changes alter the returns of investments in social capital and thereby affect the (dis)investment of persons in one another, which implies that their social networks will change. See various contributions in Badescu and Uslaner (2004) as well as in Meulemann (2008) on the productiveness of social capital in the so-called transition countries. Our own research in the former German Democratic Republic (GDR)

Social Capital 235

suggests that weak ties, quite unlike the situation in Western societies, had perverse effects in communist societies, as they pose a threat (Völker and Flap, 1997, 2001). One never could be sure whether state or party organs were not spying on one’s private life, or whether third persons would not collect information that could prove to be dangerous upon disclosure. Although the regime did succeed in mixing neighborhoods socially (the professor living next door to the plumber or the pimp), people actually living next door to each other kept their dealings with each other to a minimum. The Marxist experiment that was meant to create social cohesion actually resulted in people distrusting their neighbors and having small personal social networks. Although today fear of weak ties is gone and one would expect networks to have grown years after the turnover, people’s networks were still small in the first years after the fall of the wall; in particular, the core networks did not change much. People seem to cling to what they have, probably because they do not yet know what is in their best interest, let alone how specific social relations might serve these interests (Völker 1995;Völker and Flap, 1997, 2001). Networks do not always work in concert with existing institutions. Some situations present an institutional minimum. This puts an extra premium upon the formation and maintenance of social capital. Concentration camps are such an extreme case. As Primo Levi (1979: 94–95) states, in his horrific autobiographical account of his time in Auschwitz: With the adaptable, the strong and the astute individuals even the leaders willingly keep contact, sometimes even friendly contact, because they hope to perhaps derive some benefit. But with the musselmen, the men in decay, it is not even worth speaking, because one knows already that they will complain and will speak about what they used to eat at home. Even less worthwhile is it to make friends with them, because they have no distinguished acquaintances in camp, they do not gain any extra ration, they do not work in profitable Kommandos and they know no secret method of organizing. And in any case, one knows that they are only here on a visit, that in weeks nothing will remain of them but a handful of ashes in some near-by field and a crossed-out number on a register.

Examples can also be found in modern Western societies. For example, stratification research has shown that the rise of democracy during the last century led to more equality between the higher and lower strata. Bourdieu describes how elites in modern times compensate for their loss of power by compensatory mechanisms: one is a keen marriage politics. Elites in democratic societies will close their ranks to potential marriage partners from other social classes, guard their resources, and compensate for egalitarian measures taken by social-democratic or socialist governments. It shows that social capital can be effective notwithstanding moral disapproval and legal prohibitions on the use of social connections, if relations are hidden from the public through ignorance or secrecy—for example, if rich families marry into each other (Bourdieu and De Saint Martin 1982: 42). Organizational conditions also influence the value of social capital.Whenever the quality of services and products is hard to measure or the damage potential

236 Henk Flap and Beate Völker of a job is high, social networks come in and the value of social capital goes up, because people rely more on the opinion of others they trust. For example, for jobs with a high damage potential employers and contact persons want to be certain that they do not hire or recommend the wrong person (compare Smith 2005). They accomplish this by recruitment through informal, stronger channels (Flap, Bulder, and Völker 1998; Flap and Boxman 1999, 2001; Völker and Flap 1999). In addition, strong ties also provide more leverage to ward off opportunistic acts. Networks are also affected by cultural norms. Kalmijn and Uunk (2008) studied the effects of deviance from important shared norms on people’s network. People who break traditional norms are sanctioned by others. For example, in regions where divorce is strongly disapproved of, those that nevertheless divorce from each other suffer a significant loss of social ties.

The Program Works the first round The social capital program works, as can be seen from the empirical insights on the emergence and effects of social networks. It has produced cumulative research especially in the area of occupational attainment. A boost to social networks and social capital studies was the organization of large-scale data collections on social networks and social capital. One of the first large surveys on social networks was the Detroit Area Study of 1966, originated by Laumann. His 1973 book is based on it. He used the role-relation method and collected information on respondents’ three closest friends. In his earlier book (1966) he had used data from a survey in 1963 in Cambridge and Belmont on a few hundred male respondents. Fischer et al.’s study Networks and Places (1977) is based on this data set too. Influential has been The Northern California Study conducted by Fischer (1982), among others, because it was the first time that the name generator methodology was applied. Another milestone has been the American General Social Survey (GSS) of 1984. A representative sample of all Americans were interviewed employing the by now standard question on the persons with whom the respondent discusses personal problems. These data were used in multiple articles. The International Social Survey (ISSP) of 1986 contained a network module. These data are as far as we know the first international comparative data set on social networks. In eight countries it was asked to whom respondents would go for help in case they needed a specific kind of help. Höllinger and Haller (1990) compared social networks of citizens in seven countries involved in the ISSP of 1986 and showed that people in Middle Europe—that is, in West Germany, Austria, and Hungary—have fewer friends than in other Western countries, or no friends. Using the same data set Immerfall (1997) described in detail the differences between the networks of citizens of different Western countries. Next to the size of the networks of people from different countries there are also clear differences in their composition. People from Middle and Southern European countries include more family within their networks than people from the Nordic countries, and people from Australia and the United States include even less family. Moreover, people in Middle Europe turn more often to the same people for different kinds of help.

Social Capital 237

How to account for international differences in social networks is the first issue to be discussed here. International differences in social capital might be caused by institutional differences. One hypothesis is Coleman’s idea that welfare states drive out social capital incorporated in people’s personal social networks.The contrasting hypothesis on socioeconomic security is that welfare states and economic prosperity enable individual citizens to engage in voluntary associations without time-pressure and to acquire a sense of belonging. Using Euro-barometer data from 1992 on thirteen European countries, Scheepers, Te Grotenhuis, and Gelissen (2002) corroborate that in a more developed welfare state people have less social capital (“crowding out hypothesis”). However, a more recent publication by Van der Meer, Scheepers, and Te Grotenhuis (2008) on ISSP-data on twenty Western countries from 2001 refutes this crowdingout hypothesis. Social security has no negative impact on social participation. Social capital research seems to have entered a new phase. In particular, it is new that international comparative studies allow for comparison of networks of citizens in different societies. European sociology has a special position, because sociology in Europe is catching on with the research in various areas—among others, network research. Even more important, Europe has an interesting institutional and cultural variation between various countries and regions, inviting research on the contingency of the value of social capital as a resource. The diffusion of multilevel-regression analysis is another stimulant for this kind of research (Snijders and Bosker 1999). By now there are three types of this comparative research depending on how social capital is measured.The first branch uses name generating questions, whereas the second branch uses questions on generalized trust of others, including relative strangers. And lately there has been yet another, a third branch of comparative research in the making, a literature that uses the position generator. Lin, Ao, and Song (2009) and Völker and Flap (2008) are among the first to start this latter line of comparative research. The research literature on social capital in (former) communist societies uses both the name-generator and the interpersonal trust question to measure social capital (see, for example, various contributions to Badescu and Uslaner 2004). Citizens of former communist countries of Central and Eastern Europe have less trust in their fellow citizens than those in the West. Their associational involvement is also far lower than that of their Western counterparts. Too many people have bad memories of party-controlled associations. However, their social networks today do not seem to be that different from those in Western countries. Research on the effects of social networks while applying the resources hypothesis of the social capital theory has been successful in several research areas. A good example is research on the role of networks on the labor market. This research shows quite some progress through the years. It is clear that use of informal social contacts as such does not produce better jobs. The central finding is that, not so much the number of people prepared to lend a helping hand, nor the cohesion within the network, but the resources of the persons within one’s network are critical social capital in achieving a good job. It was furthermore demonstrated (Lin, Vaughn, and Ensel 1981a; De Graaf and Flap 1988) that part of the effect on occupational success originally attributed to human capital has to be attributed in fact to the employment of social resources,

238 Henk Flap and Beate Völker human capital being partly responsible for having better social resources. See also Flap and Völker (2008). It is important to realize that this does not indicate favoritism or clientelism. If someone helped to do a favor, there are hardly any clear benefits. One does get a job, but not a good one. Usually there are all kinds of disadvantages such as too much noise, and dirty and dangerous work circumstances and the like (see Sprengers, Tazelaar, and Flap 1988). Mouw (2003) criticized existing research because reversed causality is not considered: having and achieving a better job leads to having friends who have better jobs themselves. Recently Ruiter and De Graaf (2009) demonstrated that members of voluntary associations are more likely to start a new and better job than nonmembers. Since Ruiter and De Graaf (ibid.) have the precise timepoints when respondents became a member of an association and when they accepted a new job, they can counter the criticism that the association between social capital and a good job may be the product of selection. It is curious that the second hypothesis of the hard core, the investment hypothesis, has not been tested often, although the idea is at the heart of the program. It has been tested by social psychologists for the dissolving of romantic love relationships (Rusbult et al. 1991; Rusbult and Martz 1995). And more generally for the stability of informal relationships of respondents over a period of several years, while experiencing a number of life events (Busschbach 1996; Busschbach, Flap, and Stokman 1999). Both tests support the idea. These tests show that, apart from direct rewards and costs, especially the shadows of the past and of the future determine investment and stability in social relations. Having relational alternatives is also somewhat important. Rusbult’s theory does not include expected future benefits and second-order resources, though. Busschbach showed that the embeddedness of ties contributes to their stability only in the short term. Companionship and emotional ties that are embedded will last even when a person temporarily does not invest in them. But if an embedded instrumental tie has to last over a longer period of time, people do have to make maintenance investments. Take note, existing tests are about the stability of ties, about decay and not about newly emerging or newly chosen ties. Research on the latter is scarce. As mentioned earlier, questions on tie stability are quite similar to questions on change in social networks. McDonald and Mair (2010) apply a life-course perspective to network changes. Using cross-sectional data on Americans, they find that networks tend to grow in size if persons get older, yet the closeness of the ties and the density of the networks decrease by around thirty years of age. Interestingly, resources embedded in occupational networks accumulate across the career even if the size of this network of colleagues shrinks. McPherson, Smith-Lovin, and Brashears (2006) published an article that made an impression also on the general public. As previously said, the General Social Survey of 1985 contained network questions. The same questions were posed again in the GSS of 2004. This makes possible describing the changes in American discussion networks over a period of twenty years. This has led to two main findings. First, the discussion networks were smaller in 2004 than in 1985; the average core-network decreased in size from an average of 2.9 to 2.0 members. Second, the number of people without confidants nearly tripled,

Social Capital 239

from 10 percent in 1985 to 25 percent in 2004. Other interesting findings are that the educational heterogeneity of the core-network decreased, whereas the ethnic heterogeneity went up. Furthermore, as the decrease among nonfamily ties is somewhat greater than among kin ties, people have to rely more on their kin, especially their partner to discuss their personal problems. Take note that these are usually ties that are less bridging. These empirical results of course underlined concerns about the loss of community. But see Fischer (2008), who based on various kinds of inconsistencies in the results argues that the data overestimate the number of isolates strongly because missing data and refusals have been miscoded as isolated respondents. The study by McPherson, Smith-Lovin, and Brashears was soon replicated in Germany and in the Netherlands. Wöhler and Hinz (2007) studied samples of the Family Survey of the German Youth Institute from 1988 and 2000. Comparative results are that the size of the average core networks of the Germans grew somewhat between 1988 and 2000, from 1.9 to 2.1. In addition, the percentage of respondents without a confidant decreased quite clearly, from 47 to 37 percent. The percentage of people with nonkin confidants also clearly rose. Thus there is no erosion of the core networks in Germany. Recently we also replicated the American study for the Netherlands.Völker, Mollenhorst, and Flap (2008) found that the average size of the core network of the Dutch is stable—that is, 2.4 other persons at both time points, 1999 and in 2007. Furthermore, the number of persons without a confidant decreased a little from 13 percent in 1999 to 11 percent eight years later, in 2007. Relatives and especially the partner became somewhat more important as a confidant. In our study of changes in the networks of the Dutch we used several name-generating questions. This allows us to determine in more detail what happened to the many ties that are not any longer within the core-network. Only 30 percent of the core ties at the first point of measurement remain in the core network, and 70 percent are not in the core anymore. Yet most of these former core members—that is, 80 percent of all core members—stay within the personal network and to a large part fulfill other functions. Only 17 percent of all former core ties are not mentioned anymore in whatever function. It is amazing that with so much personal change the size and composition of people’s networks are that stable. Contacts that are embedded in a tight network are most stable (ibid.). Most change is explained by relational characteristics, such as age of the tie. People who are liked, live nearby, or have a higher education also have a greater chance of staying within the core network. The program also produced new interpretations of well-known facts in established fields of study, such as stratification research. A finding that can be better understood within a social capital framework is that in larger families children profit less from their parents’ resources, probably because siblings have to compete for these resources (Downey 1995). Social capital research has opened up new research areas. Posing new questions is scientific progress too. Great impact was made by the study of Coleman and Hoffer (1987) on the differential school success of minorities in the United States, which was shown to be greater in Catholic schools. The presumed reason was the greater social capital located in the communities of parents

240 Henk Flap and Beate Völker surrounding these schools. Children achieve better educational results attending schools in which the parents of one child take care, on their own initiative, of the children of other parents. Especially children of parents who do not have many personal resources themselves profit from such schools. Migration tends to destroy this kind of capital, which is detrimental to the educational and occupational chances of children unless the father and the mother do have strong relations with their children (Hagan, MacMillan, and Wheaton 1996). Within the sociology of the family, McLanahan (1984) started research on the detrimental effects of single-parent families, especially of divorce, on the educational and occupational chances of the children. Children of oneparent families have lower educational achievements, a high drop-out rate from school, lower earnings and occupational status, and a greater chance of becoming a welfare recipient. McLanahan interprets these effects as caused by loss of social capital within the family. A divorce seems to be more incisive than the death of a parent, probably because the death of a parent does not end the support delivered by the child’s family of both parents, in contrast to the divorce, which often puts an end to relations with at least part of the extended family. Even more detrimental to a child’s chances in life seems to be a father who is imprisoned (Western and McLanahan 2000). Another new problem opened up by social capital research is the mutual influence of partners on each other’s career. Having a partner with a good education and a good job is promoting the career of the main actor. Someone’s partner is a major form of social capital to ego, always available and strongly prepared to help (see Bernasco, De Graaf, and Ultee 1997 for more on these partner-effects).When spouses support each other, education is not only human capital but also social capital. Bernasco, De Graaf, and Ultee (1997) demonstrate such “cross-effects.” There is an accumulation of advantages within a family, as partners well provided with educational and occupational resources establish coupled careers in which each partner promotes the career and income of the other. A similar argument can be made for the very employment of the spouse. Just being employed, for example, brings with it information, available to a person’s partner, that is not available to others. Research shows that having a partner with a higher education promotes a partner’s chances of being employed, of attaining a job with a higher prestige, and higher hourly wages (Ultee, Dessens, and Jansen 1988). These partner effects go against the wellknown specialization argument made by Gary Becker’s human capital theory: partners in a marriage are a kind of firm and they specialize: those, the men or the women, who are more productive in earning an income work on the labor market, while the others make a home and take care of the children. Social capital theory also has made an impact in the area of minority research. Immigrants are not isolated, as is often thought, but are frequently rather well connected (Fernandez-Kelly 1994). However, often they do not profit from their networks because there are only a few second-order resources at the other end (see, for example,Völker, Pinkster, and Flap 2008). One bone of contention is left—that is, whether networks with holes or dense networks promote performance. It is our contention that Burt’s idea only seemingly contradicts Bourdieu’s and Coleman’s idea on the positive value of integration of a group. The first idea refers to situations in which individuals

Social Capital 241

can further their ends better by competing with others, whereas the latter refers to the situation in which individuals generally can better improve their fate by cooperation. So it depends on the situation at hand whether the one or the other type of network structure is productive. Because real life is a changing mixture of both situations, it is hard to optimize one’s personal network. Another manner of reconciling both views was presented by Uzzi and Gillespie (1998). They describe how firms that entertain strong ties to a particular bank, or more precisely, an account manager at a bank, and also have ties at arm’s length to other banks, do get better deals in their transactions with the bank because they use their weak ties to evaluate the deals they make in their strong connections. In fact the bank also turns out to close a better deal, because it is far better informed about the characteristics of its client and the condition of the client’s firm (see Flap, Bulder, and Völker 1998: 134–36). the next round The first surge of studies along the lines of the research program has sharpened the theoretical formulations and produced interesting new theoretical questions. We present a number of new issues of which at least some will or should be on the research agenda for the upcoming years. The first issue promises to become large in several respects. Coleman (1990) notices that there is a collective good aspect to social capital that could lead to an underinvestment in social capital: why should people contribute to keeping their common network in good shape? It is not a pure collective good, though: individuals do not provide information for nothing. There is also an argument to be made for the opposite statement: because people do not know their future and do not know whose help they might need, they do not want to be caught on the wrong side and invest in far more people than they will ever cash in on. Glaeser (2001; Glaeser, Laibson, and Sacerdote 2002) suggests that virtuous and vicious circles will occur: people like to belong to communities that are thriving and to which the members make contributions, whereas vicious circles may also easily arise. Some members seeing other members not contributing anymore or contributing less than earlier on will also stop contributing to the general interest, setting in motion a downward spiral of disinvesting. A recent discussion, especially in the political sciences, revolves around collective social capital. Putnam (1993) examined why everything seems to go wrong economically and politically in the Italian south, and why the north is thriving. This has been a classic theme in social sciences since Carlo Levi’s novel Christ Stopped at Eboli (1945) and Banfield’s anthropological case-study The Moral Basis of a Backward Society (1958). Putnam argues that civic traditions in the north promote the growth of lateral social ties, voluntary organizations, norms, and trust. People well organized in voluntary organizations and having lateral ties to each other pressure or even force politicians and civil servants to practice good government, which strengthens local democracy and regional economic growth. He explicitly locates social capital at the collective level of regions. In the south patronage networks are a brake on any collective action. Together with his book Bowling Alone from 2000, this made such an impression on political and other social scientists and the general public that Putnam’s ideas have been applied to various areas of life rather indiscriminately.

242 Henk Flap and Beate Völker The sheer mass of publications on collective social capital is overwhelming. A quotation from Putnam (2007: 138) illustrates what is meant: My wife and I have the good fortune to live in a neighborhood of Cambridge, Massachusetts, that has a good deal of social capital: barbecues and cocktail parties and so on. I am able to be in Upsala, Sweden, confident that my home is being protected by all that social capital, even though I actually never go to the barbecues and cocktail parties. I benefit from those social networks even though I am not actually in them myself. In the language of economics, social networks often have powerful externalities.

In his lectures Putnam sometimes adds that his wife does join these barbecues and parties. A major field of application is health research. There indeed are some indications that in neighborhoods in which people are on average a member of more voluntary associations, their health is somewhat better, controlling for several individual-level characteristics (Poortinga 2006; Fagg et al. 2008). But what the mechanism could be is not directly clear. It might be the readiness to help relative strangers when they fall ill or daring to ask relative strangers for help, or it may be that citizens pressure local government to invest in regional health care facilities. Even more important, in this application of social capital ideas, the achievements of the network tradition are almost forgotten. An article critical of this development is titled “Lost in Translation” (Moore et al. 2005). The main point is whether there is something to the notion of social capital at the collective level. Is it a real context characteristic, or is it just a composition effect of individual characteristics—that is, of all kinds of individual characteristics of residents living in the region, such as education, or the personal networks of all citizens? Or is it a result of selection? There are still other issues that for reasons of space cannot be extensively dealt with. These are, among others, whether social capital is a kind of castor oil, a means for all purposes, or whether it is nearly always goal-specific. A nice example of goal-specific social capital in the educational career is given by Parcel and Menegahn (1993) and Parcel and Dufur (2001). They discovered that children at primary schools fare better at school if their parents give them emotional support, yet if their children are in higher education they thrive better if the parents help them with their homework. Another issue is whether there is such a thing as sour social capital—that is, enemies that work against someone, and how effective they are. Moerbeek and Need (2003) found, in an analysis of data on occupational careers of a representative sample of Dutch respondents, and applying event-history analysis, that somebody who gets another job through an internal hiring gains twenty points of job prestige on Blau and Duncan’s well-known occupational prestige scale (running from 0 to 100), compared with one who has to change jobs because of having troubles with colleagues. These latter persons lose prestige. The social capital idea suggests that the negative effects of having foes will be larger to the degree that these foes have more resources. Labianca and Brass (2006) argue that negative ties have a bigger effect on performance and promotion than positive ties.

Social Capital 243

Social capital evaporates if it is not used. It differs from other kinds of capital, such as cultural capital, in that it grows in use. Because social capital fades away if no action is undertaken, people often consciously try to create institutional carriers for informal networks, foci one might say, such as fraternal or other voluntary organizations with signs communicating who is “one of us.” People stabilize social production functions in which social resources are included by building institutional brickwork around it, often creating settings in which people meet each other (Müller 1986). Baron, Jennings, and Dobbin (1994) suggest that internal labor markets are created by employers to bind employees with valuable human and social capital to the company, for example those with extensive ties to customers.

Conclusions The research program of social capital structures the field of network studies through its research questions and by its hard core. Social capital theory is a research program that can help to close the theory gap in social network analysis. The beauty of the program is that one key idea, social networks being social capital, explains the emergence as well as the effects of social networks: a person’s social capital promotes his goal achievement, and he will invest in it depending on its instrumental value. The program does not only bring unity to a rather disintegrated research area, it also provides new predictions. A major example is the development of a supply side argument on numbers and places. Furthermore, the program helps to explain why there are, contrary to the structuralist claim, no pure structural effects: structural effects meet with exceptions because effects of networks often depend on the resources that are embedded in a network. Moreover, the instrumental value of social capital is contingent upon existing institutions and available technology. For example, we now understand why in totalitarian societies people are afraid of strangers. Weak ties are a liability in such a situation. Both the structuralist view as well as social capital theory take the idea of goal-directed, rational man as a point of departure; the former uses its alleged atomism to define its own position, the latter uses it to create a new research program by adding new auxiliary assumptions on networks. Although the structuralist view and social capital theory defined their position in opposition to each other, actual research conducted within the tradition of structuralism is not that far removed from rational choice sociology and social capital research. Although its popularity is rapidly growing, there is also skepticism on social capital ideas. For example, Baron and Hannan (1994) stated that its “theoretical cutting edge is lost if attention is not called to investments, rates of return, opportunity costs, the future, and the ability to appropriate the returns from the investment.” A major provision for the further development of the research program is to take seriously the analogy with human capital (see Glaeser, Laibson, and Sacerdote 2002; Esser 2008). A next step that most certainly will be made will be the organization of longitudinal or panel data on the development of networks through time. Analyses of own panel data on the networks of the Dutch in the period between 1999

244 Henk Flap and Beate Völker and 2007, the SSND, show, for example, that networks stay rather stable qua size and composition although there is quite some in- and outflow from personal networks.

Discussion Our sketch of the social capital theory did not provide definitive answers to the question on mechanisms that make personal networks productive. A review of the literature teaches that several mechanisms are important. The main mechanism is instrumental help promoting a person’s interest in some sense. Another mechanism is to cognitively frame the social situation of actors and to define what is in their interest. A third mechanism is a normative one: behavior is steered through norms or agreements enforced by dense social networks. Another critical issue is that proper tests of the investment theory are scarce. This major part of the program is actually not very well established. People choose others who can solve or help with their problems. They probably are not looking for specific others but for others who are capable of solving their problem. If they have a problem at work they look for a friend, but for a type of friend who can help with that specific problem. It is likely that this person is a colleague from the workplace. It is a potentially revolutionizing idea that people choose particular contexts instead of particular persons. If a person has a work-related problem, he wants advice from somebody from the work context (Feld 1984). This thought also provides some explanation for the fact (see above) that over time people seem to reconstruct similar networks qua size and composition although the actual network members change. In addition, it becomes less puzzling that tie characteristics have the largest share in explaining stable and new ties (Busschbach, Flap, and Stokman 1999;Völker and Flap 2008). Experimental research would probably help to depict investment processes (see Riedl and Van Winden 2005). It is hard to tell whether providing help is an act of investment or a repayment for a service rendered earlier. To represent what happens if one actor invests in the other, one can first play a dictator game and let people invest in others and next play a trust game and see whether investments pay out in this game in greater cooperation. There is also new and growing experimental research that explicitly focuses on network formation. Conventional models of exchange or dilemma games between a number of players in a pregiven constant social network are altered by assuming that the players can change their network and add or delete social ties. In economic models of dynamic networks, ties have a utility and their formation has a cost. In subsequent rounds players make choices to delete a tie, which can be done unilaterally, or to form a tie for which both partners have to agree. For more information on experimental work on social networks and social capital, see the contribution by Snijders on social network dynamics in this volume (see also Jackson 2005 for a review of dynamic models of network formation). The focus on macrolevel social capital is important and will have lasting consequences. However, the theoretical elaboration of collective social capital is not really progressing. In addition, there is a discontinuity in research problems studied, in theories and measurements between studies based on the microlevel of existing ties and networks on the one hand and the macrolevel

Social Capital 245

of the communitarian, collective dimension of social capital on the other. Understanding the interrelationship between these two dimensions of social capital is a future research problem.

References Axelrod, Robert. 1984. The Evolution of Co-operation. New York: Basic Books. Badescu, Gabriel, and Erik M. Uslaner, eds. 2004. Social Capital and the Transition to Democracy. London: Routledge. Banfield, Edgar. 1958. The Moral Basis of a Backward Society. Chicago: University of Chicago Press. Baron, James N., and Michael T. Hannan. 1994. “The Impact of Economics on Contemporary Sociology.” Journal of Economic Literature 32: 1111–46. Baron, James N., Devereaux Jennings, and Frank R. Dobbin. 1988. “Mission Control? The Development of Personnel Systems in U.S. Industry.” American Sociological Review 53: 497–514. Bernasco, Wim, Paul de Graaf, and Wout Ultee. 1997. “Coupled Careers.” European Sociological Review 14: 15–31. Blau, Peter M., and Joseph Schwartz. 1984. Cross-cutting Social Circles. New York: Academic Press. Blum, Terry. 1995. “Structural Constraints on Interpersonal Relations: A Test of Blau’s Macro-sociological Theory.” American Journal of Sociology 91: 511–21. Boissevain, Jeremy. 1974. Friends of Friends. Oxford: Blackwell. Bourdieu, Pierre. 1973. “Cultural Reproduction and Social Reproduction.” In Knowledge, Education and Cultural Change, edited by R. Brown, 71–112. London: Tavistock. ———. 1981. “Le capital social.” Actes de la recherche en sciences sociales 31: 2–3. Bourdieu, Pierre, and Monique de Saint Martin. 1982. “Le patronat.” Actes de la recherche en sciences sociales 32: 3–82. Boxman, Ed, Paul de Graaf, and Henk Flap. 1991. “The Impact of Social and Human Capital on the Income Attainment of Dutch Managers.” Social Networks 13: 51–73. Bridges, William P., and Wayne J. Villemez. 1986. “Informal Hiring and the Labor Market.” American Sociological Review 51: 574–82. Burt, Ronald S. 1987. “Social Contagion and Innovation: Cohesion versus Structural Equivalence.” American Journal of Sociology 92: 1287–335. ———. 1992. Structural Holes: The Social Structure of Competition. Cambridge: Harvard University Press. ———. 1997. “The Contingent Value of Social Capital.” Administrative Science Quarterly 42: 339–65. Coleman, James S. 1988. “Social Capital in the Creation of Human Capital.” American Journal of Sociology 94, Supplement: 95–120. ———. 1990. The Foundation of Social Theory. Boston: Belknap Press. ———. 1993. “The Rational Reconstruction of Society.” American Sociological Review 58: 1–15. Coleman, James S., and Thomas Hoffer. 1987. Public and Private Schools: The Impact of Communities. New York: Basic Books. Coleman, James S., Elihu Katz, and Herbert Menzel. 1966. Medical Innovation: A Diffusion Study. Indianapolis, IN: Bobbs-Merrill. Cosmides, Leda, and John Tooby. 1992. “Cognitive Adaptations to Social Exchange.” In The Adapted Mind, edited by J. H. Barkov, L. Cosmides, and J. Tooby, 163–228. Oxford: Oxford University Press.

246 Henk Flap and Beate Völker Dasgupta, Patha. 1988.“Trust as a Commodity.” In Trust, Making and Breaking Cooperation, edited by D. Gambetta, 49–72. Oxford: Oxford University Press. De Graaf, Nan Dirk, and Henk Flap. 1988. “With a Little Help from My Friends.” Social Forces 67: 453–72. Diekmann, Andreas. 2007. “Dimensionen des Sozialkapitals.” In Sozialkapital, Grundlagen und Anwendungen, edited by Axel Franzen and Markus Freitag, 45–65. Wiesbaden: Verlag für Sozialwissenschaften. Downey, Douglas. 1995. “When Bigger Is Not Better: Family Size, Parental Resources and Children’s Educational Performance.” American Sociological Review 60: 746–61. Esser, Hartmut. 2008. “The Two Meanings of Social Capital.” In The Handbook of Social Capital, edited by Dario Castiglione, Jan W. van Deth and Gugliemo Wolleb, 25–49. Oxford: Oxford University Press. Fagg, James, Sarah Curtis, Stephen A. Stansfeld, Vicky Cattell, Ann-Marie Tupuola, and Muna Arephin. 2008. “Area Fragmentation, Social Support for Individuals and Socio-psychological Health in Young Adults.” Social Science and Medicine 66, no. 2: 242–54. Feld, Scott. 1981. “The Focused Organization of Social Ties.” American Journal of Sociology 86: 1015–35. ———. 1984. “The Structured Use of Personal Associates.” Social Forces 62: 640–52. Fernandez-Kelly, M. 1994. “Social and Cultural Capital in the Urban Ghetto.” In The Economic Sociology of Immigration, edited by A. Portes, 213–47. New York: Free Press. Fischer, Claude S. 1982. To Dwell among Friends. Chicago: University of Chicago Press. ———. 1992. America Calling: A Social History of the Telephone to 1940. Berkeley and Los Angeles: University of California. ———. 2008. “The 2004 GSS Finding of Shrunken Social Networks: An Artefact?” American Sociological Review 74: 657–69. Fischer, Claude S., Robert Max Jackson, C. Ann Stueve, Kathryn Gerson, Lynne McAllister Jones, and Mick Baldassare. 1977. Networks and Places. New York: Free Press. Fischer, Claude S., and Lynne McCallister. 1978. “A Procedure for Surveying Personal Networks.” Sociological Methods and Research 7: 131–47. Flap, Henk. 1988. Conflict, Loyalty and Violence. Berne: Peter Lang. ———. 1989. “Patronage: An Institution in Its Own Rights.” In Social Institutions, edited by Michael Hechter, Karl-Dieter Opp, and Reinhard Wippler, 225–43. New York: Aldine de Gruyter. ———. 1991.“Social Capital in the Reproduction of Inequality: A Review.” Comparative Sociology of Family, Health and Education 20: 6179–202. ———. 1997. “The Conflicting Loyalties Theory.” L’année sociologique 47: 183–216. Flap, Henk, and Ed Boxman. 1999. “Getting a Job among Managers.” In Corporate Social Capital and Liability, edited by Roger Leenders and Shaul M. Gabbay, 197–216. Boston: Kluwer. ———. 2001. “Getting Started.” In Social Capital: Theory and Research, edited by Nan Lin, Karen Cook, and Ron Burt, 159–81. Chicago: Aldine. Flap, Henk, Bert Bulder, and Beate Völker. 1998. “Intra-organizational Networks and Performance.” Computational and Mathematical Organization Theory 4, 109–47. Flap, Henk, and Beate Völker. 2008. “Social, Cultural, and Economic Capital and Job Attainment: The Position Generator as a Measure of Cultural and Economic Resources.” In Social Capital: An International Research Program, edited by Nan Lin and Bonnie H. Erickson, 65–80. Oxford: Oxford University Press. Franzen, Axel. 2000. “Does the Internet Make Us Lonely?” European Sociological Review 16: 427–38.

Social Capital 247 Glaeser, Edward L. 2001. “The Formation of Social Capital.” ISUMA, Canadian Journal of Policy Research 2: 34–40. Glaeser, Edward L., David Laibson, and Bruce Sacerdote. 2002.“An Economic Approach to Social Capital.” Economic Journal 112: 437–58. Gluckman, Max. 1965. Politics, Law, Ritual in Tribal Society. Oxford: Blackwell. Granovetter, Mark S. 1973. “The Strength of Weak Ties.” American Journal of Sociology 78: 1360–80. ———. 1974. Getting a Job. 1995. Rev. ed. Cambridge: Harvard University Press. ———. 1979. “The Theory Gap in Social Network Analysis.” In Perspectives on Social Network Research, edited by Paul W. Holland and Samuel Leinhardt, 501–18. New York: Academic Press. ———. 1985. “Economic Action and Social Structure:The Problem of Embeddedness.” American Journal of Sociology 91: 481–510. Grieco, Margaret. 1987. Keeping It in the Family. London: Tavistock. Hagan, John, Ross MacMillan, and Blair Wheaton. 1996. “New Kid in Town: Social Capital and the Life Course Effects of Family on Children.” American Sociological Review 61: 368–85. Hansen, Marianne N. 1996. “Earnings in Elite Groups: The Impact of the Social Class Origin.” Acta Sociologica 39: 385–408. Heider, Fritz. 1958. The Psychology of Interpersonal Relations. New York: John Wiley and Sons. Heilbronner, Robert L. 1964. The Making of Economic Society. Englewood Cliffs, NJ: Prentice-Hall. Hobbes, Thomas. 1663. Leviathan. London: Penguin. Höllinger, Franz, and Max Haller. 1990. “Kinship and Social Networks in Modern Societies: A Cross-cultural Comparison among Seven Nations.” European Sociological Review 6: 103–26. House, James S., Karl R. Landis, and Debra Umberson. 1988. “Social Relationships and Health.” Science 241: 540–45. Immerfall, Stefan. 1997. “Sozial Integration in den westeuropäischen Gesellschaften: Werte, Mitgliedschaften und Netzwerke.” In Die westeuropäischen Gesellschaften im Vergleich, edited by Stefan Hradil and Stefan Immerfall, 139–73. Opladen: Leske und Budrich. Jackson, Matthew. 2005. “A Survey of Models of Network Formation: Stability and Efficiency.” In Group Formation in Economics: Networks, Clubs and Coalitions, edited by Gabrielle Demange and Myrna Wooders, 11–57. Cambridge: Cambridge University Press. Jacobs, Jane. 1961. The Death and Life of Great American Cities. New York: Vintage Books. Kalmijn, Matthijs, and Henk Flap. 2001. “Assortative Meeting and Mating: Unintended Consequences of Organized Settings for Partner Choices.” Social Forces 79: 1289– 313. Kalmijn, Matthijs, and Wilfried Uunk. 2008. “Regional Value Differences in Europe and the Social Consequences of Divorce: A Test of the Stigmatization Hypothesis.” Social Science Research 36: 447–68. Kapferer, Bruce C. 1972. Strategy and Transaction in an African Factory. Manchester: Manchester University Press. Labianca, Giuseppe, and Daniel J. Brass. 2006. “Exploring the Ledger: Negative Relationships and Negative Asymmetry in Social Networks in Organizations.” Academy of Management Journal 41: 596–614. Lakatos, Imre. 1970. “Falsification and the Methodology of Scientific Research

248 Henk Flap and Beate Völker Programmes.” In Criticism and the Growth of Science, edited by Imre Lakatos and Alan Musgrave, 91–196. London: Cambridge University Press. Laumann, Edward O. 1966. Prestige and Association in an Urban Community: An Analysis of an Urban Stratification System. New York: Bobbs-Merrill. ———. 1973. Bonds of Pluralism:The Forms and Substance of Urban Social Networks. New York: Wiley. Lee, Nancy Howell. 1969. The Search for an Abortionist. Chicago: University of Chicago Press. Leifer, Eric M., and Harrison C. White. 1988. “A Structural Approach to Markets.” In The Structural Analysis of Business, edited by Michael Schwartz and Mark Mizruchi, 85–108. New York: Cambridge University Press. Levi, Carlo. 1945. Christ Stopped at Eboli. London: Cassell. Levi, Primo. 1979 [1947]. Is This a Man? London: Penguin. First published as Se questo un uomo. Lin, Nan, Dan Ao, and Lijon Song. 2009. “Production and Returns of Social Capital: Evidence from Urban China.” In Contexts of Social Capital: Social Networks in Community, Markets and Organizations, edited by Ray-May Hsung, Nan Lin, and Ron Breiger, 163–92. London: Routledge. Lin, Nan, and Mary Dumin. 1986. “Access to Occupations through Social Ties.” Social Networks 8: 365–85. Lin, Nan, and Bonnie H. Erickson, eds. 2008. Social Capital: An International Research Program. Oxford: Oxford University Press. Lin, Nan, John C. Vaughn, and Walter M. Ensel. 1981a. “Social Resources and Occupational Attainment.” Social Forces 59: 1163–81. ———. 1981b. “Social Resources and Strength of Ties.” American Sociological Review 46: 393–405. Lindenberg, Siegwart. 1990. “Homo Socio-economicus.” Journal of Institutional and Theoretical Economics 146: 727–48. ———. 1998. “Solidarity: Its Microfoundations and Macro Dependence: A Framing Approach.” In The Problem of Solidarity:Theories and Models, edited by Patrick Doreian and Thomas Fararo, 61–111. Amsterdam: Gordon Breach. Litwak, Eugene. 1985. Helping the Elderly. New York: Guilford Press. Lorrain, Francois, and Harrison C.White. 1971.“Structural Equivalence of Individuals in Social Networks.” Journal of Mathematical Sociology 1: 49–80. Loury, Glenn. 1977. “A Dynamic Theory of Racial Income Differences.” In Women, Minorities, and Employment Discrimination, edited by P. Wallace and A. LaMond, 153– 88. MA: Lexington Books. Marsden, Peter V. 1990. “Network Diversity, Substructures and Opportunities for Contact.” In Structures of Power and Constraint, edited by Craig Calhoun, Marshall W. Meyer, and W. Richard, 397–410. Cambridge: Cambridge University Press. Marsden, Peter V., and Elizabeth H. Gorman. 2001. “Social Networks, Job Changes, and Recruitment.” In Sourcebook on Labor Markets: Evolving Structures and Processes, edited by Ivar Berg and Arne L. Kalleberg, 467–502. New York: Kluwer Academic/Plenum Publishers. Mauss, Marcel. 1923. “Essai sur le don. Forme et raison de l’échange dans les sociétés archaïques.” L’année sociologique, seconde série, 1: 30–186. McDonald, Steve, and Christine A. Mair. 2010. “Social Capital and the Life Course: Age and Gendered Patterns of Network Resources.” Sociological Forum 25: 335–59. McLanahan, Sarah. 1984. “Family Structure and the Reproduction of Poverty.” American Journal of Sociology 90: 873–901. McPherson, Miller, Lynn Smith-Lovin, and Matthew E. Brashears. 2006. “Social

Social Capital 249 Isolation in America: Changes in Core Discussion Networks over Two Decades.” American Sociological Review 71: 353–75. Meulemann, Heiner, ed. 2008. Social Capital in Europe: Similarity of Countries and Diversity of People? Leiden: Brill. Moerbeek, Hester, and Ariana Need. 2003. “Enemies at Work: Can They Hinder Your Career?” Social Networks 25: 67–83. Mollenhorst, Gerald, Beate Völker, and Henk Flap. 2008a. “Social Contexts and Personal Relationships: Different Effects of Meeting Opportunities on Similarity in Various Types of Personal Relationships.” Social Networks 30: 60–68. ———. 2008b. “Core Discussion Networks of the Dutch: The Effects of Social Contexts on Similarity in Intimate Relationships.” Social Forces 86: 937–65. Moore, S., A. Shiell, P. Hawe, and V. A. Haines. 2005. “The Privileging of Communitarian Ideas: Citation Practices and the Translation of Social Capital into Public Health Research.” American Journal of Public Health 95, no. 8: 1330–37. Mouw, Ted. 2003. “Social Capital and Finding a Job: Do Contacts Matter?” American Sociological Review 68: 868–98. Müller, Hans-Peter. 1986. Klassen, Kassifikationen und Lebensstile. Mannheim: Universität der Bundeswehr. Mullins, Nicholas C. 1973. “The Structuralists: Tracking Something.” In Theories and Theory Groups in American Sociology, edited by N. Mullins, 250–69. New York: Harper Row. Ormel, Johan, Siegwart Lindenberg, Nardi Steverink, and Lois M. Verbrugge. 1999. “Subjective Well-being and Social Production Functions.” Social Indicators Research 46: 61–99. Parcel,Toby L., and Mikaela Dufur. 2001. “Social Capital at Home and at School: Effects on Student Achievement.” Social Forces 79: 881–912. Parcel, Toby, and Elisabeth G. Menegahn. 1993. “Family Social Capital and Children’s Behaviour Problems.” Social Psychology Quarterly 56: 120–35. Perrow, Charles. 1986. Complex Organizations. New York: McGraw-Hill. Pescolido, Bernice A., and Beth A. Rubin. 2000. “The Web of Group Affiliations Revisited: Social Life, Postmodernism and Sociology.” American Sociological Review 65: 52–76. Poortinga,Wouter. 2006.“Social Relations or Social Capital? Individual and Community Health Effects of Bonding Social Capital.” Social Science and Medicine 63: 255–70. Portes, Alejandro. 1998. “Social Capital: Its Origins and Applications in Modern Sociology.” Annual Review of Sociology 24: 1–24. Putnam, Robert. 1993. Making Democracy Work: Civic Traditions in Modern Italy. Princeton: Princeton University Press. ———. 1995. “Tuning In, Tuning Out: The Strange Disappearance of Social Capital in America.” Political Science and Politics 28: 664–83. ———. 2000. Bowling Alone:The Collapse and Revival of American Community. New York: Simon and Schuster. ———. 2007. “E Pluribus Unum: Diversity and Community in the Twenty-first Century.” The 2006 Johan Skytte Prize Lecture. Scandinavian Political Studies 30, no. 2: 137–74. Raub, Werner, and Jeroen Weesie. 1990. “Reputation and Efficiency in Social Interactions: An Example of Network Effects.” American Journal of Sociology 96: 626– 54. Riedl, Arno, and Frans van Winden. 2005. “Information and the Creation and Return of Social Capital: An Experimental Study.” In Creation and Returns of Social Capital: A New Research Program, edited by Henk Flap and Beate Volker, 77–103. London: Routledge.

250 Henk Flap and Beate Völker Rivera, Mark T., Sara B. Soderstrom, and Brian Uzzi. 2010. “Dynamics of Dyads in Social Networks: Assortative, Relational, and Proximity Mechanism.” Annual Review of Sociology 36: 91–115. Ruiter, Stijn, and Nan-Dirk de Graaf. 2009. “Socio-economic Payoffs of Voluntary Involvement: A Dutch Life Course Study.” European Sociological Review 25: 425–42. Rusbult, Caryl E., and John Martz. 1995. “Remaining in an Abusive Relationship: An Investment Model Analysis of Nonvoluntary Dependence.” Personality and Social Psychology Bulletin 21: 558–71. Rusbult, Caryl E., Julie Verette, Gregory A. Whitney, Linda F. Slovik, and Isaac Lipkus. 1991. “Accommodation Processes in Close Relationships.” Journal of Personality and Social Psychology 60: 53–78. Scheepers, Peer, Manfred te Grotenhuis, and John Gelissen. 2002. “Welfare States and Dimensions of Social Capital: Cross-national Comparisons in European Countries.” European Societies 4: 185–207. Singelman, Peter. 1975. “The Closing Triangle.” Comparative Studies in Society and History 17: 389–409. Smith, Sandra S. 2005. “‘Don’t Put My Name on It’: Social Capital Activation and Job-finding Assistance among Black Urban Poor.” American Journal of Sociology 111: 1–57. Snijders, Tom A. B. 1999. “Prologue to the Measurement of Social Capital.” La Revue Tocqueville 20: 27–44. ———. 2013. “Network Dynamics.” In Handbook of Rational Choice Social Research, edited by Rafael Wittek,Tom Snijders, and Victor Nee. Stanford: Stanford University Press. Snijders, Tom A. B., and Roel J. Bosker. 1999. Multilevel Analysis: An Introduction to Basic and Advanced Multilevel Modeling. London: Sage Publishers. Sprengers, Maarten, Frits Tazelaar, and Henk Flap. 1988. “Social Resources, Situational Constraints, and Re-employment.” The Netherlands’ Journal of Social Sciences 22: 98– 116. Taylor, Michael. 1974. Anarchy and Cooperation. London: Wiley. Ultee, Wout, Jos Dessens, and Wim Jansen. 1988. “Why Does Unemployment Come in Couples?” European Sociological Review 4: 111–22. Uzzi, Brian, and James J. Gillespie. 1998. Social Embeddedness and Corporate Financing. Evanston, IL: Northwestern University. Van Busschbach, Jooske. 1996. Uit het oog, uit het hart? [Out of Sight, Out of the Heart?]. Amsterdam: Thesis. Van Busschbach, Jooske, Henk Flap, and Frans N. Stokman. 1999. “Stability and Change in Personal Networks: A Test of the Investment Theory.” Paper presented at the Sociologendagen, Utrecht, May 29. Van der Gaag, Martin, Tom Snijders, and Henk Flap. 2008. “Position Generator Measures and Their Relationship to Other Social Capital Measures.” In Social Capital: An International Program, edited by Nan Lin and Bonnie Erickson, 27–48. Oxford: Oxford University Press. Van der Meer, Tom, Peer Scheepers, and Manfred te Grotenhuis. 2008. “Does the State Affect the Informal Connections between Its Citizens?” In Social Capital in Europe: Similarity of Countries and Diversity of People?, edited by Heiner Meulemann, 42–72. Leiden: Brill. Vaughn, Diana. 1990. “Autonomy, Interdependence and Social Control: NASA and the Space Shuttle Challenger.” Administrative Science Quarterly 35: 225–57. Verbrugge, Louis M. 1977. “The Structure of Adult Friendship Choices.” Social Forces 56: 576–97. ———. 1979. “Multiplexity in Adult Friendship.” Social Forces 57: 1286–309.

Social Capital 251 Völker, Beate. 1995. Should Auld Acquaintance Be Forgot . . .? Institutions of Communism, the Transition to Capitalism and Personal Networks: The Case of East Germany. Amsterdam: Thesis. ———. 1999. “Getting Ahead in the GDR.” Acta Sociologica 43: 17–34. ———. 2001. “Weak Ties as a Liability: The Case of East Germany.” Rationality and Society 13: 397–428. ———. 2007. “Feinde am Arbeitsplatz.” In Die Analyse von Gesellschaften, Organisationen und Individuen in ihrem Zusammenhang. Theoretisch und methodische Herausforderungen, edited by H. Hummell, 133–55. Bonn: Gesis—IZ Sozialwissenschaften. ———. 2008. “Reproduction of Inequality in the Netherlands through the Creation of and Returns to Social Capital?” Paper presented at the International Social Capital Conference Academia Sinica, Taipei, Taiwán May 29–30. Völker, Beate, and Henk Flap 1997. “The Comrade’s Belief: Intended and Unintended Consequences of Communism for Neighbourhood Relations in the Former GDR.” European Sociological Review 13: 241–65. Völker, Beate, Henk Flap, and Gerald Mollenhorst. 2009. “Changing Places: The Influence of Meeting Places on Recruiting Friends.” In Contexts of Social Capital: Social Networks in Communities, Markets and Organizations, edited by Ray-May Hsung, Nan Lin, and Ronald Breiger, 28–48. Oxford: Routledge. Völker, Beate, Gerald Mollenhorst, and Henk Flap. 2008. “Core Discussion Networks of the Dutch ‘Ten Years After.’” Unpublished paper. Völker, Beate, Fenne Pinkster, and Henk Flap. 2008. “Inequality in Social Capital? Differences in Networks and Social Capital between Ethnic Minorities and the Dutch in the Netherlands.” In Special Issue of Kölner Zeitschrift für Soziologie und Sozialpsychologie, edited by F. Kalter, 325–50. Wasserman, Stanley, and Kathy Faust. 1995. Social Network Analysis: Methods and Applications. Cambridge: Cambridge University Press. Watkins, John W. N. 1957. “Between Analytical and Empirical Statements.” Philosophy 33: 112–31. Wellman, Barry, Peter Carrington, and Alan Hall. 1988. “Networks as Personal Communities.” In Social Structures: A Network Approach, edited by Barry Wellman and Steven Berkowitz, 130–84. Cambridge: Cambridge University Press. Western, Bruce, and Sara McLanahan. 2000. “Fathers behind Bars: The Impact of Incarceration on Family Formation.” In Families, Crime and Criminal Justice: Contemporary Perspectives in Family Research. Vol. 2, edited by Greer Litton Fox and Michael L. Benson, 309–24. New York: Elsevier Science. Wöhler, Thomas, and Thomas Hinz. 2007. “Egozentrierte Diskussionsnetzwerke in den USA und Deutschland.” In Sozialkapital. Grundlagen und Anwendungen, edited by Axel Franzen and Markus Freitag, 91–112. Wiesbaden:Verlag für Sozialwissenschaften.

chapter

Network Dynamics

7

tom a. b. snijders

Introduction The scientific interest in social networks arose in the early twentieth century, and grew in the second half of that century, in the first place because the networks in which individuals are embedded have important effects on their behavior, performance, and general well-being.This history is described vividly by Freeman (2004). In most of these studies, networks were taken as a given. Interest turned in the last two decades of the century toward the explanation of networks using, for example, theories of social capital or social resources.These developments examined the consequence, logical within theories of purposive actors, of the proposition that benefits and harm can accrue to social actors from their network position: social actors will attempt to obtain beneficial network positions. Studies of social capital seek to analyze and explain in detail what is beneficial in network positions (Coleman 1990; Burt 1992; Lin, Cook, and Burt 2001; Flap and Völker, this volume). Studies of network dynamics, reviewed in this chapter, seek to model and explain why networks are as they are, in particular how they change. When this explanation is based in part on the importance to actors of the consequences of their network positions, a natural next step is to endogenize not only networks but also actors’ behavior and outcomes. outline This chapter considers networks as relational structures in a given set of social actors, and provides an overview of models and some empirical results for dynamics of social networks, considered in a setting of social actors optimizing a utility function that is based, among other things, on their network embeddedness (excluding purely rule-based models). Some attention is also paid to network equilibrium, this being relevant for network dynamics as a potential final state. The chapter starts with discussing some general properties of networks, and how a rational actor perspective may be helpful to understanding them. This is a background to the rest of the chapter, focusing on models for representing

Network Dynamics 253

networks. It then turns to a discussion of game-theoretic models for network equilibrium and network dynamics.The game-theoretic approach has difficulties in coming to grips with empirical reality; the latter is the purpose of the statistical models treated next. Network dynamics are important especially in studies in which not only the network but also actor properties are endogenized.Therefore, the second part of the chapter discusses models for the joint dynamics of networks and actor characteristics, both in a game-theoretic and in a statistical framework.

Some Empirical Regularities This section considers some well-known strong empirical regularities of social networks, and sketches explanations for these regularities. number of ties The number of ties per actor is usually limited in some sense: social networks are sparse. What is sparse depends on the type of social actor and the type of relation under study. Core discussion networks of individuals, discussed as the set of close important personal relationships of an individual, are usually limited to two to six people; McPherson, Smith-Lovin, and Brashears (2006) found that from 1987 to 2004, the average size of the core discussion network of Americans decreased from 2.6 to 2.1. The core discussion network defines a highly selective relation. At the other extreme of the spectrum of tie strength, the number of acquaintances of individuals (corresponding to the question “How many people do you know?”) is estimated to be usually in the range from 150 to 1,000 (Marsden 2005). In the network defined by the whole world, this seemingly large number still leads to a sparse network. Zheng, Salganik, and Gelman (2006) estimate that the median acquaintance network size is about 610, with 90 percent of the population having between 250 and 1,710 acquaintances. Explanations of limited network sizes are typically based on the resources needed to maintain network ties, of which for sociologists time is a primary example (examples abound; for example, Hallinan 1979), but not for the evolutionary psychologists Stiller and Dunbar (2007). These authors propose that the major reason for limiting our acquaintance networks is the limitation of cognitive capacities, and find that interindividual differences in the number of close relationships can be predicted by intentionality (that is, the capacity to understand the intentions of others) and memory capacity. There can also be strategic arguments for restricting the number of ties. Gulati, Nohria, and Zaheer (2000) discuss “lock-in” in the formation of strategic alliances between firms, defined as exclusion of other partners because of fidelity to an already existing alliance; as these authors say, “Many alliances are explicitly monogamous.” reciprocity There is a tendency to dyadic reciprocation in most directed networks: the existence of the tie i → j makes it more likely that also the tie j → i exists. This was early proposed as a basic feature of social networks by Moreno (1934) and has been studied and confirmed enormously. Social exchange theory (for

254 Tom A. B. Snijders example, Emerson 1972) provides a basic explanation: actors depend on each other for valued outcomes, and benefits will be received from another actor only if they are also given in return. Axelrod (1984) offered a game-theoretic explanation by showing the good performance of the tit-for-tat strategy in the iterated prisoner’s dilemma. Gould (2002) argued that relations are liable to be terminated more rapidly when they are not reciprocated, because keeping a nonreciprocated relation implies status deference. Reciprocity need not be direct but can be indirect—that is, circulate in larger groups; see, for example, Molm, Collett, and Schaefer (2007). However, although mutual dependence and solidarity in larger groups are pervasive in societies, the number of three-cycles in empirically observed directed networks is relatively low (Davis 1970). homophily Many social networks have a tendency toward homophily—that is, actors have a larger probability of being tied to each other if they have similar characteristics. The importance of homophily was noted by Lazarsfeld and Merton (1954), and it has been found to be a strong determinant of network ties in many cases. A review is given by McPherson, Smith-Lovin, and Cook (2001). These authors mention as major explanations of homophily the spatial-geographical organization of social ties, which is closely related to opportunity arguments; the genesis of social ties within families; the role of organizational foci (compare Feld 1982) in creating and maintaining ties; and influence processes between actors who are tied. Based on theories of social capital (Lin 2001) and given the high value of having access to diverse resources, however, the reverse—heterophily—can also be expected. The importance of complementarity for creation of alliances between firms is well known; see Gulati and Gargiulo (1999). As examples, Chung, Singh, and Lee (2000) found evidence for resource complementarity as an explanation of alliance formation between investment banks; Casciaro and Lobo (2008) found that collaboration between individuals occurs more frequently when similarity on demographic variables goes together with complementary specializations. Riolo, Cohen, and Axelrod (2001) present a game-theoretic model of the iterated prisoner’s dilemma type, in which cooperation is higher when actors are similar on a trait that otherwise is arbitrary. They show that this gives a sufficient structural condition for high levels of cooperation, and reciprocity is not required to achieve this. transitivity The existence of the two ties i→j and j → h often makes it more likely that also the tie i → h exists. This is called the tendency toward transitivity or triadic closure.

figure 7.1. Transitive and Intransitive Triads

transitive (closed) triad

intransitive (open) triad

Network Dynamics 255

When social ties are based on being affiliated to the same group (as discussed by Simmel 1908), transitivity is one of the consequences. Thus transitivity follows from homophily and from arguments based on the opportunity to interact (Feld 1981). However, there are other mechanisms that show directly the advantages of, or tendencies toward, triadic closure, without being necessarily linked to common affiliations. Simmel’s teachings (1917) about the fundamental importance of triads, the constellation of relations between three actors, have been expressed in network analysis early on by the focus on triadic structures. Triads were studied systematically by Davis, Holland, and Leinhardt in the late 1960s and during the 1970s; see Davis (1979) and Holland and Leinhardt (1975). Davis (1970) found empirically, in a study based on a collection of a large number of networks of positive interpersonal affect, that transitivity is the main feature that differentiates these data from a pattern of random ties. Similar findings were obtained later in studies of interorganizational linkages—for example, by Gulati and Gargiulo (1999) in a study of strategic alliances between large production companies. Simmel (1917) brought forward the special consequences of triadic embeddedness on bargaining power of the social actors and on the possibilities of conflicts between them.The manifold consequences of triadic embeddedness, and especially the tensions between openness and closure, are basic threads in the modern literature on network analysis. Coleman (1990) stressed the importance of triadic closure for social control, where the third actor, who has information about the behavior and interrelations of the two other actors, has the potential to sanction them in case they behave opportunistically or otherwise violate norms of good behavior. This was confirmed by Gulati (1998) both theoretically and empirically in his review of studies of strategic alliances between firms, where uncertainty about potential partners and risks of opportunistic behavior can be checked by alliances that are transitively embedded. Gulati and Gargiulo (1999) argued that information from third parties, signaling capacity for cooperation, and reputational effects will favor the creation of transitive ties between firms. On the other hand, arguments for the advantages of being in a brokerage position, where an actor has information from two unrelated others, were formulated early by Granovetter (1973) in his study on the strength of weak ties. This was elaborated in studies of the strategic importance of brokerage positions (“structural holes”), and of the potential to exploit the informational and strategic advantages of the broker, by Burt (1992 and further publications). Krackhardt (1999) elaborated the consequences of triadic embeddedness by defining a Simmelian tie as a strong tie between two partners that is combined with strong dyadic ties of both partners with one or more others. He argued that Simmelian ties are more constraining, and are more difficult to break. The contrast between Burt’s and Krackhardt’s theories is most clear in the bow tie structure (see Figure 7.2). Structural hole ideas will lead to the conclusion that actor A is in an advantageous position, being the only linkage between two strongly tied groups: tertius gaudens. Simmelian tie theory, on the other hand, puts forward the idea that actor A has to satisfy the normative constraints of two distinct

256 Tom A. B. Snijders

A

figure 7.2. Bow Tie

groups: the ties that torture.The differences between situations in which network closure or openness of networks is more advantageous for social actors were elaborated, for example, by Ahuja (2000) and Burt (2000). Another argument that can differentiate between closed and open networks is uncertainty and risk aversion. If closed networks are more helpful for uncertainty reduction (Burt and Knez 1995), while open networks give the potential of new opportunities but at a higher risk, then actors who are more risk averse should have a greater tendency to form open networks and explore the new possibilities associated with ties to more distant others. This was hypothesized by Baum et al. (2005), who proposed that firms whose performance deviates from their aspirations—be it positively or negatively— will have a higher propensity to form nonembedded ties. They found this confirmed for syndication ties between investment banks. As a sideline in a paper on network dynamics, Vega-Redondo (2006) presents an interesting game-theoretic result that gives a potential explanation for transitivity. He postulates a model where actors play prisoner’s dilemma with their network partners, where ties are formed and maintained if and only if they are profitable, and where information about behavior of others is shared between actors who are tied. One of the results is that transitivity of ties enhances the conditions for stability of the network, which can be attributed to the strategic foreshadowing of the information sharing. node centrality and network centralization An important issue in social network analysis is the position of nodes, and a primary characterization of this position is the centrality of nodes in the network— a concept capturing the extent to which nodes are connected to other actors, which can be defined in various ways (Freeman 1979; Bonacich 1972, 1987). Many social networks show a fair extent of centralization—that is, differentiation between social actors with respect to their centrality. Centralization of networks can be the result of feedback processes that favor the creation of links to nodes that are already highly connected. Such a model was proposed by Price (1976) for bibliometric networks. Price called this a cumulative advantage model, in the spirit of Merton’s Matthew effect (1968): “Unto him that hath is given and from him that hath not is taken away, even that which he hath.” His model leads to power law distributions for the degrees; these are distributions with heavy tails—that is, relatively high probabilities of a few nodes with very high degrees (“hubs”). For most types of networks between human individuals this does not seem realistic because, as was discussed above, various constraints will limit the occurrence of very high degrees. This model was independently rediscovered, elaborated, and popularized by Barabási and Albert (1999), who called it a

Network Dynamics 257

preferential attachment model and gave many examples of networks—for example, the Internet, for which the distribution of degrees for large degrees is close to a power distribution. Handcock and Jones (2004) give further references about these models and their recurrent rediscoveries. Centralization of a social network reflects social organization and social opportunities. Differentiation of node centrality represents inequality between nodes; for example, centrally located actors will have more resources and greater power; they may also incur higher costs. In studies of collective action, Marwell, Oliver, and Prahl (1988) found that a strongly centralized network is conducive to the potential for collective action in organizer-centered mobilizations. The reason is that the organizers of collective action can selectively contact central actors who, in turn, will have a larger contribution to the collective action and thereby increase the probability that the collective action will be produced. Even if centralization is important for social networks, it is noteworthy that centralization is much larger in physical networks, which often are designed to have “hubs”—that is, highly central nodes—to facilitate flows through the network. Examples are electronic communication networks, air link networks, and so forth. This is indeed one of the differences noted by Newman and Park (2003) between social networks and other networks. connectedness Most social networks, even if they are large, have the property that a large portion of the nodes are connected: for most pairs of nodes it is possible via a chain of ties to traverse the network and, starting from one node, arrive at the other one. A component is defined as a maximal set of nodes that are mutually connected. If the number of nodes in the network is large and the network is sparse—that is, the largest degree is of limited size—then it is not immediately evident that most of the network would be connected; a network can also “fall apart” in a number of disconnected components. A basic property concerning connectedness can be proved mathematically for random networks—that is, networks in which ties occur independently and with the same probability between all pairs of nodes. A famous result of Erdös and Rényi (1960) is the existence of a so-called giant component. This is a set of nodes that are connected and have no further outside connections, and comprise (in the limit as n grows indefinitely) a positive fraction of the set of all nodes. Denoting the number of nodes by n and the average degree by dn, they proved that if n → ∞ and dn tends to a finite limit greater than 1, there will be a giant component, and no more than one, with probability tending to 1. This can be regarded as a mathematical micro-macro result, because on the basis of a trivial model for tie formation—random links—it proves a nontrivial result at the level of the whole network. The existence of a giant component holds also for many networks that do not have the pure randomness of the Erdös-Rényi construction, both in mathematical models and in the empirical world. A beautiful example is the structure of romantic relations illustrated by Figure 7.2 of Bearman, Moody, and Stovel (2004). This is not a random network at all but it does consist of a giant component together with a large number of very small components and isolates. Kogut, Urso, and Walker (2007: fig. 3) demonstrate empirically the genesis of a giant component in the network of venture capital ties between

258 Tom A. B. Snijders firms, where the links are defined as common investments in the same target company. shortness of path length s Another basic empirical observation for social networks has been that geodesic distances, defined as the shortest path lengths in the graph connecting a given node to another node, tend to be relatively small. This was first studied in the experiment of Milgram (1967) and Travers and Milgram (1969), discussed extensively in the literature collected in Kochen (1989), and it has led to the insight that through acquaintance links almost all people in the world are linked by paths of a length of at most 6. Watts and Strogatz (1998) noted that high transitivity by itself will tend to increase geodesic distances, so that—given the tendencies toward transitivity— it is remarkable that many social networks indeed have short geodesic distances. They defined a small world as a network with many nodes, having degrees that are not too high, without dominating nodes—in other words, with a maximum degree much smaller than the number of nodes, having a high tendency toward transitivity (or clustering, as it is also called), and short average path lengths. They wondered about the possibility of proposing a mathematical model that would combine these properties, and proposed such a model in Watts and Strogatz (ibid.) and Watts (1999). The model is defined by placing the n nodes on a circle, first linking each node to the k nearest neighbors, and then reconsidering each link i ↔ j and with some small probability p replacing it by a link from i to a randomly chosen node. They showed that this model indeed produces a small-world network, if the parameters n, k, and p are chosen appropriately. This in itself is not a plausible social network model (compare Robins, Pattison, and Woolcock 2005), but it does give an existence proof of a probability model for networks that has the small-world properties. Many subsequent publications have investigated properties of this model and of slight modifications intended to make it better mathematically tractable. Robins, Pattison, and Woolcock (ibid.) noted that the exponential random graph model, or p* model (Wasserman and Pattison 1996; Robins et al. 2007), a flexible and quite general distribution on the space of graphs and digraphs, can also be specified so that it yields stochastic network models that have the smallworld properties; other parameter choices can yield well-interpretable but quite different types of networks. This opens up the possibility of combining insights into small-world structures with empirical data analysis based on the exponential random graph model. In situations in which networks represent opportunities for communication, flow of resources, transport, and the like, short path lengths have a direct benefit. Indeed one of the most studied game-theoretic specifications of network formation is a model where direct ties are costs, and short distances are benefits; see the utility function (1) discussed below.

Network Games Networks can define opportunities for interaction that are fruitfully modeled as games. Examples are Raub and Weesie (1990) and the theory of network

Network Dynamics 259

exchange (see below). Reviews are given by Goyal (2007) and Jackson (2004, 2008). In the context of this chapter, especially interesting are situations in which the network evolves endogenously because the network determines the payoffs, and the actors (nodes in the network) respond to this. stability and efficiency When discussing network dynamics, it is important to pay attention also to network statics. Stable networks can be reference points in studies of network dynamics, and in idealized situations they can be the structures obtained as limits of a dynamic process. An essential element in optimization of networks is the dyadic nature of ties: two actors are involved. For nondirected networks, a natural assumption often is that the two actors concerned have to agree for a tie to be created (Myerson 1991; Jackson and Wolinsky 1996).This can be formalized as follows. A strategy for actor i is the vector Yip* = (Yip1,..., Yinp ) of proposals (constraining Yiip to be 0, regarding it as a meaningless variable); the realized network Y is the disjunction of the reciprocally proposed ties, defined by Yij = min {Yijp ,Yjip }. The two actors involved have to enter in some kind of negotiation or cooperation process about the creation of the tie. The general concept of a Nash equilibrium is not quite satisfactory for nondirected networks because it does not account for this type of cooperation. An example is provided by the case where ties are formed if and only if both participating actors decide to establish a tie, while utility functions are non-negative, and zero only for isolated actors. Then all actors prefer to be part of a nonempty network. However, the situation in which all actors propose no ties at all is a Nash equilibrium.This defect of the Nash concept for network games has led to the proposal of various stability concepts. equilibrium concepts for network games Jackson and Wolinsky (1996) defined pairwise stability for nondirected networks. Their basic idea is that both actors involved in a tie have to agree for the tie to exist. A network is stable if, for every existing tie, the utility of the network would become smaller or remain equal for both actors when the tie is deleted; and for every nonexisting tie the utility would either remain constant, or become smaller for at least one actor, by creating the tie. The definition of pairwise stability is distinct from the definition of a Nash equilibrium because it involves simultaneous choices by two actors, which are not considered in Nash equilibrium; and the latter also considers changes by a given actor in more than one tie, which is not considered in pairwise stability. The combination is made by Goyal and Joshi (2006) and Gilles et al. (2006). A network is defined to be pairwise Nash (the concept was mentioned by Jackson and Wolinsky 1996, section 5, and Gilles et al. call it strongly pairwise stable) if it is a network corresponding to a Nash equilibrium and also pairwise stable as defined above. Thus, no pair of actors would both prefer to have an additional tie between them, and no single actor would prefer to dissolve a subset of his ties. Buskens and van de Rijt (2008) propose a further strengthening of this equilibrium concept.They define a network to be unilaterally stable if no actor i would strictly gain by changing his ties in such a way that this would entail no utility loss to any of the other actors with whom i establishes a new tie. Bloch

260 Tom A. B. Snijders and Jackson (2006) discussed diverse definitions of equilibrium for networks, focusing on equilibrium concepts where transfers between the players are allowed, and on the relations between these with the preceding concepts. some theoretical results The literature on network formation games is rapidly growing, and here we can give only an outline of a few results. Jackson and Wolinsky (1996) considered utility functions where each tie i ↔ j has a cost cij while the benefit of the existence of a path from i to j is positive but decreases exponentially with the geodesic distance.Thus the utility function for actor i is ui (Y ) = ∑ wij b j ≠i

(

d ij (Y )

− cijYij

)

(1)

where dij(Y) is the geodesic distance (length of the shortest path) between i and j, the wij and cij are positive constants, and b is a constant between 0 and 1. They proved that for such utility functions, the only possible pairwise stable networks are the empty network, the stars (where one actor is connected to all the others, but the others are mutually disconnected), and the complete network; which of these is stable depends on the parameters of the utility function. Since the star is also the most centralized network (see above), this can be related to the empirical occurrence of highly centralized networks. Jackson and Wolinsky went on to study the relation between stability and efficiency for general utility functions, and found essential tensions between these two network properties. A network is efficient if the total payoff to all actors is maximal. For intermediate tie costs, the uniquely efficient structure is the star, where the central actor links all other actors. However, this network is not pairwise stable because the central actor contributes disproportionately to the collective good represented by the efficiency of this network structure. The only pairwise stable network here is the empty network, which, however, is Pareto dominated by other networks. This tension between stability and efficiency, which must be expected to have consequences for network dynamics, as actors may be assumed to be attracted to stability but also to efficiency, was further investigated by, among others, Jackson (2003) and Jackson and van de Nouweland (2005). To give a flavor of their results, let us mention one theorem from the latter paper. The main assumptions are that the payoff function is such that within connected components, every actor gets the same (“componentwise egalitarian allocation”), and the payoff of actors in a connected component is not affected by the ties within other components (“component balance”).The first is a very specific condition (ruling out differential payoffs caused by power differences, for example); the second is reasonable in many situations. Jackson and van de Nouweland proved that under these conditions, the set of efficient networks is the same as the set of pairwise stable networks if and only if the efficient networks also maximize the payoff per individual actor. Note that the latter is not the case if a subset of actors can exclude some of the others and thereby gain a payoff per actor that is, within this subset, higher than the payoff per actor in the efficient network. A loose interpretation of these results is that

Network Dynamics 261

if conflict of interests is high (as in social dilemma situations), efficiency and stability in networks can be contradictory; in situations with less possibility for opportunistic action and with egalitarian payoffs, the combination of efficiency and stability is less problematic. Bala and Goyal (2000) consider directed networks. They study utility functions for actor i that depend only on the number of actors reachable from i (via directed paths in the “one-way flow” case, and via paths where tie direction does not count in the “two-way flow” case), and on the number of outgoing ties of i (the out-degree), in which the number of reachable actors is a benefit and the out-degree is a cost. They find that for the one-way flow, Nash equilibria must be the empty network or a wheel (all nodes circularly connected in one direction without any other ties). For the two-way flow, Nash equilibria are either the empty network or a star. For utilities also depending on geodesic distances the Nash equilibria can also be more complicated networks, called flowers, defined by a central actor who is part of a number of further disconnected wheels. Buskens and van de Rijt (2008) studied nondirected networks with the utility function being defined as the negative of Burt’s constraint measure (1992). This reflects the postulated wish of all the actors in the network to be involved in structural holes. They found that pairwise stable networks must be connected and have all geodesic distances smaller than or equal to 2. A number of further stability results were obtained for multipartite networks (where the actors are divided into several nonempty subsets and there are no ties within subsets) and bipartite networks (where there are two such subsets). empirical results for network formation games There have been a variety of experimental studies in which the utility function is determined by the researcher and supplied to the actors, and the question is investigated to what extent the players empirically converge on equilibria of some kind. Kosfeld (2004) gives a survey. It turns out that for many network games the players do succeed rather well in converging to equilibria, as long as the equilibria do not contradict fairness conditions. Details of the experimental design can have important consequences. Berninghaus, Ehrhart, and Keser (1999) propose to use a continuous-time design instead of the more usual design where games are played in discrete rounds. In the continuous-time implementation, players have opportunities to change their ties at any moment while time is flowing on. This is more realistic and adds fluidity for reaching equilibrium states. In their game, a modification of the game of Bala and Goyal (2000), a periphery-sponsored star is the unique strict Nash equilibrium. Berninghaus, Ehrhart, and Ott (2006), as well as Berninghaus, Ehrhart, Ott, and Vogt (2007), find that in such a game, where the central position is most advantageous, which contradicts an even distribution of payoffs, many groups tried to combine fairness with the Nash equilibrium—for example, by occasionally rotating the central position between the players. Another way to deal with the problems resulting from fairness considerations is to introduce heterogeneity between the actors. Goeree, Riedl, and Ule (2009) conducted experiments in which again the game was a variation of Bala and Goyal (2000), now with actors who differed in their benefit to others or in their costs of tie

262 Tom A. B. Snijders formation. It turned out that introducing one actor with high benefit to others led to star networks, with this actor in the center; that was not the case when introducing one actor with lower tie costs. By estimating actor-dependent parameters for envy and guilt according to the utility function proposed by Fehr and Schmidt (1999), representing inequality aversion (not part of the payoff function supplied to the actors, but implicit in their decisions), Goeree, Riedl, and Ule explained this result from the fact that the envy parameters tended to be sufficiently high that the actors operated (“subjectively”) with a utility function in which the star network was a strict Nash equilibrium. Burger and Buskens (2009) studied network formation in small groups while differentiating between a competitive and a cooperative context. They found that in the cooperative context networks tended to be created with a high extent of transitivity, whereas in the competitive context, where it is advantageous to be in the center, the players did not permit each other to occupy the central position.

Statistical Models To empirically test theories of network dynamics it is important to have tractable models for statistical inference for network data, but the interdependence of network ties has long been an obstacle to their development. This section presents such models that have been constructed explicitly on the basis of rational action models. An interesting statistical model for network dynamics was used by Gulati and Gargiulo (1999) in their study of strategic alliances in three production industries over a twenty-year period. Their data was collected on a yearly basis, and for the creation of an alliance Yijt between organizations i and j in year t they used a random effects probit model, P{Yijt = 1} = Φ(α′ xij + β′ zij(t – 1) + uij)

(2)

where xij is a vector of time-constant characteristics of organizations i and j, zij(t – 1) is a vector of changing characteristics of these organizations as measured for the preceding year, uij is a random effect capturing unexplained nonchanging characteristics of the pair (i, j), and Φ is the cumulative distribution function of the standard normal distribution. The changing characteristics of the pair (i, j) included measures of network embeddedness, such as past alliance history, centrality, and transitivity-related measures. This model is quite flexible. Limitations of this approach, as discussed by Gulati and Gargiulo (ibid.: 1483), are the lack of representing the dependence between decisions by any given actor (except by the included covariates), and the lag of one year in taking account of the changing covariates zij. actor-based statistical models An approach to network dynamics that is based on modeling the choices made by the actors in the network was proposed by Snijders (1996, 2001). It was noted in Snijders (1996) that the usual approach to theory testing in the social sciences is to deduce implications from the theory with the boundary conditions, and then test these empirically using statistical models that have no

Network Dynamics 263

particular connection to the theory being tested; and that it can be preferable to use a statistical model that is itself a direct formulation of the theoretical model. This aligns well with studying revealed preferences, because such models allow estimating parameters in preference or utility functions. A traditional example of such an approach is McFadden’s model (1973) of choice among a set of discrete alternatives. Integrating theoretical and statistical models is fruitful especially for studying network dynamics, because traditional statistical models have difficulties representing theoretically plausible dependencies by suitable stochastic dependence conditions.This leads to actor-oriented statistical models for network dynamics, in which the ties are controlled in some way by the nodes representing the social actors. This model (Snijders 1996, 2001) is composed of a stochastic process model for the evolution of a network on a given set of nodes, where ties can be created and terminated, and defined with a continuous time parameter. It can be used for the case that the network data consists of a sequence (“panel”) of two or more successively observed networks, but also if a continuous record of changes in network ties are available. The following ingredients are used to specify the model. • The control of actors over part of the network. The control is specified in line with the game-theoretic models discussed above. For directed networks, a natural assumption is that actors control their outgoing ties. For nondirected networks, a natural assumption is that ties are formed only if both corresponding actors agree with the existence of the tie, and some kind of negotiation process has to take place to decide upon creation of ties. • The actions that the actors can choose, and the time schedule for doing so. In the case of sequential network experiments, the actions permitted and their timing are defined by the experimental design. When the game is played in rounds where each player can select new a collection of ties at each round (for example, Callander and Plott 2005), it is natural to use a discrete time parameter t and allow the actors in the mathematical model to make the same changes as the actors in the experiment. When the game is played in quasi-continuous time (for example, Berninghaus, Ehrhart, and Keser 1999), it is natural to use a model with a continuous time parameter and to allow only one change in a tie at any given (“infinitesimal”) moment. Holland and Leinhardt (1977) proposed using a continuous time parameter also for networks observed repeatedly outside the laboratory according to a panel design.The arguments are the theoretical plausibility of allowing network changes in continuous time (compare Berninghaus, Ehrhart, and Keser 1999; and Berninghaus, Ehrhart, and Ott 2006) and the greater simplicity that is possible when using continuous-time models (already argued by Coleman 1964). Such a continuous-time model for observations in a panel design implies that there will be a sequence of unobserved changes between each pair of consecutive observations. In a model with a continuous time parameter, and in models with fixed rounds where only one or some pairs of actors meet in any round, it has to be

264 Tom A. B. Snijders specified which actors, or pairs of actors, change their ties at given moments. For example, in Watts (2001), at each round one pair of actors is randomly chosen to redefine their tie. When the salience of the network is different for different actors, it may be plausible that actors for whom the network has a higher salience will also have, or take, more frequent opportunities for changing their ties. Next to differential salience, there may be differential meeting opportunities; compare Verbrugge (1977); Feld (1981); and Pattison and Robins (2002). Holland and Leinhardt (1977) further proposed to allow no more than one tie change at any time point. This decomposes the change model into its smallest constituents and excludes coordination between actors. • The utility or preference functions of actors, ui(Y). To have a flexible specification of models, allowing to estimate the weights of several mechanisms or influences simultaneously, a linear combination of several components can be employed, ui(Y) = ∑k βk sik(Y)

(3)

The utility function (1) mentioned above combines a cost (negative β1) for si1 defined as the number of links of actor i and a benefit (positive β2) from si2 defined as a function of short geodesic distances to other nodes. More generally, any functions sik(Y) representing the network position and network neighborhood of actor i can be chosen to express the theories investigated, along with alternative mechanisms and known dependencies between ties (reciprocity, homophily, transitivity, and so forth), while their weights βk can be treated as statistical parameters whose values are estimated from the data. • The information available to each actor. A simple first-order approximation is to assume that each actor has full information about the entire network. In large networks this may not be a reasonable assumption. Actors will not know about all the ties, and they may even have a limited knowledge about the composition of the network. The perception of network ties is known to be more precise for the part of the network that is close to the focal actor (Kumbasar, Romney, and Batchelder 1994). Here the concept of a social setting (Pattison and Robins 2002) can be useful to circumscribe the knowledge available to actors, depending on their social, institutional, and geographical environment. • The type of optimization performed by the actors. Full strategic rationality does not seem to be a very realistic option, as it may be too hard even for the modeler. Myopic optimization is an alternative, where actors, when they have the opportunity to change a tie, use a myopic best response strategy: they choose a best change given the rest of the network existing at the current moment. Such models were proposed for network dynamics, for example, by Snijders (1996); Bala and Goyal (2000); and Watts (2001). • The role of probability, or randomness.

Network Dynamics 265

As in other statistical models, randomness has to be included to account for deviations between what is actually observed and the best choices that actors would have had, according to the model. A tradition in game theory is to model deviations by “trembles” (for example, Jackson and Watts 2002; Goyal and VegaRedondo 2005), where actors choose either a best action or, with some small probability, a random other action. The econometric modeling tradition is to model these deviations by random utility components (McFadden 1973), where actors optimize their utility function to which independent random components are added. This yields models where the probability of choices is an increasing function of their utility, which seems more reasonable for statistical modeling than the use of trembles. • Lack of antisymmetry between creation and termination of ties. For nondirected networks, Jackson and Wolinsky (1996) already made clear the difference that creation of a tie will require consent of both involved actors, whereas for breaking the tie it will be sufficient for one of the actors to discontinue the tie. Another deviation from antisymmetry is that breaking an existing tie may yield a difference in the utility function that is not equal to the negative of the difference obtained from creating that tie, even if the remainder of the network is exactly the same. As an example, the cost of terminating a reciprocal friendship tie might be greater than the gain in creating such a tie— one could say that the existence of a reciprocated tie gives a reward without cost; this is argued by van de Bunt, van Duijn, and Snijders (1999). This is similar to the endowment effect in microeconomics (Thaler 1980), defined as the difference between selling prices and buying prices, and related to loss aversion and framing as discussed, for example, by Kahneman, Knetsch, and Thaler (1991) and Lindenberg (1993). Using these ingredients, actor-based models for network dynamics were proposed by Snijders (1996) for networks defined by rankings and Snijders (2001) for networks defined by digraphs. These methods were applied, for example, in studies of friendship formation (van de Bunt, van Duijn, and Snijders 1999; van Duijn et al. 2003; Schaefer, Kornienko, and Fox 2011), dynamics of work-related trust (van de Bunt, Wittek, and de Klepper 2005), and job mobility of managers in venture capital firms (Checkley and Steglich 2007). An overview is in Snijders, van de Bunt, and Steglich (2010).

Network Structures and Utility Arguments After this review of utility-based models for networks, let us briefly consider the type of utilities that have been postulated. First we consider theoretical work, subsequently some empirical work. Much work in game theory models for network formation is based on the connections model (Jackson and Wolinsky 1996), which assumes that there is a cost to ties and a benefit to indirect connections, leading to a weighted sum of the number of indirect connections (as the benefit) and the degree (as the cost) as the utility function for an actor, as in (1). This utility function can be compounded by spatial considerations (for example, Johnson and Gilles 2000; Jackson 2008). Experimental studies of these games have mostly used the same

266 Tom A. B. Snijders utility functions as the theoretically proposed ones. An interesting exception is Goeree, Riedl, and Ule (2009), who in their empirical analysis added to the utility function components for envy and guilt as proposed by Fehr and Schmidt (1999). In the coauthorship model of Jackson and Wolinsky (1996), ties are nondirected, and each tie is interpreted as a collaboration project between the two actors. Actors have a given amount of time that they distribute over their projects, so that actor i devotes 1/yi+ to each collaborative project, where yi+ is the degree of this actor. The utility function for actor i is 1 + ∑ yij j

( y1

j+

+ y1y i+ j+

)

figure 7.3. Utility Function

where the interpretation is that the term 1 is the result of the efforts of the actor on her own projects, and each collaborating actor j provides the fruit of her activities 1/yj+ to which is added a synergy reward 1/(yi+yj+). Thus, if i and j collaborate, then the collaboration of j with third parties is a negative externality for i as it diminishes the time that j spends on the collaboration with i. Jackson and Wolinsky examined stability and efficiency for this game. The implication of the value of structural holes (Burt 1992) for network formation was elaborated by Goyal and Vega-Redondo (2007), Kleinberg et al. (2008), and Buskens and van de Rijt (2008). The first of these papers used a utility function for actor i that depends on the number of times that i is necessary for linking any pair of two other actors, and the number of others with whom i shares this strategic position. The second does something similar, considering only shortest paths of length two. The third uses Burt’s constraint measure (1992) as a utility function, which, however, leads to a study of brokering without considering the parties who are to be brokered. Belleflamme and Bloch (2004) and Goyal and Joshi (2006) consider a model for free trade agreements in which the utility for actor i is a function that increases in its own degree and decreases in the degrees of those to whom i has a tie. The models of Price (1976) and Barabási and Albert (1999), although only implicitly utility-based, come close to using utility functions that depend increasingly on the degrees of those to whom the focal actor is tied. Such models are also analyzed in Jackson and Rogers (2007, section V). Now let us turn to empirical work. For studies of interfirm alliances, rationality arguments are quite natural (Gulati, Nohria, and Zaheer 2000). Evidently, for commercial firms the final utility function is profit, or expected future profits. In practice, cognitive limitations imply that network formation is driven by more proximate goals. Gulati, Nohria, and Zaheer mention endogenous constraints such as the costs implied by ties, which limit the number of ties that actors can sustain; and alliance loyalty, which leads to the exclusion of ties to competitors of one’s allies.They also argue that competition can move from the firm level to the level of multiparty alliances (clusters) of firms, while internally these multiparty alliances are characterized by a combination of aligned and opposed interests. Stuart and Sorensen (2007) discuss the fact that the utility functions used

Network Dynamics 267

in the economic game-theoretic literature do not correspond well to the mechanisms in the entrepreneurship literature. They suggest mechanisms leading as components in utility function to resource complimentarity; access to diverse information; homophily (explainable, for example, as compatibility of procedures and culture, both conducive to collaboration); and transitive embeddedness (conducive to trust, sanctioning potential, and reputation). For friendship networks, utility-based approaches were proposed, for example, by Zeggelink (1995), Stokman and Zeggelink (1996), van de Bunt, van Duijn, and Snijders (1999), van Duijn et al. (2003), and Jackson and Rogers (2007).Van de Bunt, van Duijn, and Snijders proposed an explicit list of utility arguments, including the need to have friends; the desire to affiliate with popular others, where popularity is measured by indegree and regarded as a reflection of status (formally similar to the consideration of the Matthew effect by Price 1976); the preference for proximate and similar others; balance (friends who make the same friendship choices as oneself; Heider 1958; this is quite close to transitivity); reciprocity; and aversion to loss of the investments accumulated in a reciprocal friendship.Van Duijn et al. (2003) differentiated between similarity on visible and invisible characteristics, and hypothesized that the former would be more important in early stages of the friendship formation process, and the latter in later stages. Stokman and Zeggelink (1996) distinguished between attributes leading to homophily (where a similar value is preferred), aspiration (where an ideal value is preferred), and complementarity (heterophily). Mayer and Puller (2008), elaborating a model proposed by Jackson and Rogers (2007), use a very restricted utility function determined by homophily only.

Coevolution of Networks and Behavior If individual outcomes are influenced by individual behavior as well as by network position, actors will attempt to choose their behavior as well as their ties so as to improve their outcomes. This leads to dynamic models in which individual behavior as well as networks are endogenous. If we regard the individual behavior as a representation of the microlevel, and the network as the social context defining the macrolevel, then such models are integrated micro-macro models (Wippler and Lindenberg 1987)—although one might also call them micro-meso. empirical research Questions about the interdependent dynamics of networks and individual outcomes are studied in a variety of fields. In the broad domain of adolescent development, for example, Ennett and Bauman (1994) studied influence of friends on smoking behavior, Haynie (2001) effects of friends on delinquency, and Dijkstra et al. (2010) influence of friends on weapon carrying. The general observation here is that friends tend to be similar; and that this might be due to mutual influence between friends, to preferential selection of friends who are similar on salient characteristics, or to friends being subject to similar contextual influences. A central question then is which of these three types of processes is the major determinant of the observed similarity between friends. It may be noted that the term “homophily” sometimes is used for the

268 Tom A. B. Snijders descriptive phenomenon of similarity between tied actors, and sometimes more specifically for the preferential choice of similar network partners. When discussing simultaneous evolution of networks and individual outcomes, usually the more restrictive definition is used, so that homophilous selection is contrasted with influence between network partners. The descriptive phenomenon is referred to as network autocorrelation (Doreian 1989)—that is, the correlation of outcomes between those who are tied in the network. In alliances between firms, the explicit purpose of creating ties is to improve the firm’s competitiveness, and therefore in this domain the dynamic interdependence between network ties and firm success is a natural point of view, and argued for by Gulati, Nohria and Zaheer (2000). They mention the example (Gomes-Casseres 1997) of alliances between minicomputer industries in the 1980s, where each cluster of alliances developed its own standards, and where the competitive dynamics became a competition at the level of alliances and between the standards. Alliance formation here was intimately linked to the commercial success of the standard, and thereby, of the firms. Stuart and Sorensen (2008) argue that the expected effects of network positions on firm success, and the strategic behavior of entrepreneurs, imply that network formation is endogenous, and discuss the resulting difficulties for the empirical identification of network effects. In the next three sections, a variety of models will be discussed that have been put forward for the coevolution of a network and an actor-level variable. This literature is rapidly expanding, and we can only indicate the issues being studied without attempting to be even close to complete.

Homophily Coevolution Models Several coevolution models have recently been proposed where actors prefer to be tied to similar others, and can change their ties as well as their behavior. Macy, Kitts, Flache, and Benard (2003) consider a model with actors situated in a nondirected valued network and having a binary behavior variable. The tie values as well as the behavior are updated in a way favoring homophily: behavior is influenced toward the weighted average of the behavior of the network neighbors, and tie values tend to become stronger between similar actors and weaker (or stronger negative) between dissimilar actors. This can be regarded as a balance model (Heider 1958). Simulations show that the dynamics are most likely to lead to polarization between two antagonistic camps, but under some parameter combinations it is possible to converge to pluralistic alignments. Holme and Newman (2006) study a model of a nondirected network where behavior of actors is a categorical attribute. At each discrete time step a random nonisolated actor is selected, and with some probability an edge of this actor is moved to connect to another actor with the same behavior, and else (with the complementary probability) the actor’s behavior is changed to the behavior of a randomly selected tied actor. This model leads to a situation of isolated groups of actors having homogenous behavior, and Holme and Newman study by simulation how this limiting situation is approached depending on various parameters.

Network Dynamics 269

Ehrhardt, Marsili, and Vega-Redondo (2007) study the coevolution of a nondirected network and similarity between actors in continuous time. The similarity between actors i and j is denoted by dij. For nontied pairs of actors at a distance dij ≤ d0, new ties are randomly formed at a rate of 1 per unit of time; for actors at a larger distance, new ties are randomly formed at a rate % much lower than 1; existing ties are dissolved randomly at a rate λ. Three different specifications of the distance are considered. In the first, it is the geodesic distance—thus the model represents a tendency to cohesion and transitive closure. In the second specification the actors have one categorical attribute, and they form ties only to those having the same value for the attribute. In the third model the actors have a real-valued attribute that changes in a diffusion model with a preferred direction toward their network partners. Each actor changes his attribute at a rate ν where values held by a larger number of neighbors have higher probabilities. It is assumed that ν is very large, such that the conditional distribution of attributes, given the network, may be assumed to be in stochastic equilibrium. These three types of distance-dependent coevolution all represent that ties are formed preferentially with similar actors (as defined by the distance), and also that distance tends to be, or to become, small between tied actors. In all three models there is a stable state consisting of networks with cohesive subgroups of actors with high similarity that can be regarded as networks with coordination. In addition, contrasting with the model of Holme and Newman (2006), there is a stable state consisting of a sparse network with low similarity between network partners.

Game-Theoretic Coevolution Models An interesting class of coevolution models is those in which behavior includes the strategy played in games between the actors. Most published models are about models of nondirected networks where, in each round, actors may play social dilemma (cooperation—defection) games with all their network neighbors, and where in any round of the game each actor follows the same strategy in all games played; or where the actors play coordination games, following the same setup. cooperation Skyrms and Pemantle (2000: 9340) give the following sketch to describe the simultaneous evolution of strategy and social network they are studying. “Individual agents begin to interact at random. The interactions are modeled as games. The game payoffs determine which interactions are reinforced, and the social network structure emerges as a consequence of the dynamics of the agents’ learning behavior.” Specifically, they study schemes for network dynamics where there is a propensity matrix with elements wij, and repeated realizations of stochastic networks, regarded as encounters. Time progresses in discrete rounds and at every round the probability of an encounter i→j depends on wij compared with wik for k ≠ j. When an encounter takes place the two partners play a game that yields a payoff to each of them, and the propensities wij are updated as a result of the payoffs. Individuals follow the same strategies

270 Tom A. B. Snijders in all their encounters in a given round; they may change strategies between rounds, subject to inertia. The individual variables are the strategies followed and the accumulated payoffs; the network variables are the propensity matrix and the realized encounters. Updating rules for the propensity matrix either depend only on who encounters whom (that is, with trivially defined payoffs), or are defined by the result of a coordination game played in each encounter; with varying rules of reciprocity, noise, and discounting of the past. Bonacich and Liggett (2003) study a similar model for the case of updating based on encounters only. Skyrms and Pemantle (2000) show that the propensity matrix tends to a structure where interaction is concentrated in small groups following a coordinated strategy. In the case of the coordination game, the probability to reach the payoff dominant (that is, socially efficient) strategy depends on the degree of inertia in updating strategies—less inertia leads to more easily achieving the efficient strategy. Eguìluz, Zimmerman, Cela-Conde, and San Miguel (2005) study the coevolution of ties and individual strategies in a network of actors playing cooperation games. Time proceeds in discrete steps, and at each step each actor plays prisoner’s dilemma games with all his network partners; as in Skyrms and Pemantle (2000), each actor follows one behavior (cooperation, C, or defection, D) toward all partners, but this behavior may change freely from step to step. The strategies are assumed to be chosen according to “replicator dynamics,” where the actor adopts the behavior of that actor in his personal network (including himself) who obtained the highest payoff in the preceding step. If the actor adopts the behavior of a defecting partner, with some probability she replaces this partner by a randomly chosen other actor. This process determining interaction partners is simpler than that of Skyrms and Pemantle (2000) in the sense of assuming less information and less calculation by the actors. Stable states of the network turn out to be states in which actors can be distinguished in three classes: leaders, who are cooperators and have the highest payoff in their personal network; conformists, who are cooperators not having the highest payoff in their personal network but who imitate another cooperator; and exploiters, who are defectors who have a higher payoff than the cooperators in their personal network. The stable networks are necessarily trees. It is clear that a model in which strategies are prescribed yields less insight than one in which strategies emerge from the analysis. Vega-Redondo (2006) studies a population of actors connected by a network, where connected actors play prisoner’s dilemma games.The payoff structures in the PD games differ across dyads, and change randomly with some frequency (“payoff volatility”). The network specifies who plays with whom and also determines transmission of information about behavior. Specifically, at each step in the sequential process, all linked pairs of actors share the information they have acquired concerning the behavior of any other; evidently, they have direct information about those who played with them in a dyad. This information sharing has an impact on the set of pairwise stable networks, enlarging this set compared with the situation where there would be no sharing of information. The network changes by a random choice of one player i who may change one tie. The player either carries out a search for the best change within her own component; or is confronted with a randomly chosen other player from the

Network Dynamics 271

whole population. Whether the tie change is made depends on the collection of dyadic payoff structures, in which player i chooses the partner with the most advantageous payoffs. Links that do not give a profit to both partners are removed. The main result is that when payoff volatility increases, the network becomes less dense (links become nonprofitable more frequently) but also more “cohesive”; one could say that the transitivity becomes stronger, although the measure used for cohesiveness is not precisely a transitivity measure. Also the size of the giant component decreases. Santos, Pacheco, and Lenaerts (2006) study a model in which actors can play social dilemma games with their network neighbors, being a cooperator or a defector in all their games played in a given round. Actors try to be connected to cooperators rather than to defectors. A random actor i is selected and a random network neighbor j of this actor. They play the social dilemma game with all their neighbors and compare their total payoffs from the games in the current round. The round can lead to a network change with probability p and to a behavior change with probability 1 – p. In case of a network change, if actor j is a defector, then actor i will attempt to switch the tie i ↔ j to a tie linking i with one of the other neighbors of j. In case of a behavior change, j will attempt to influence i to adopt the same behavior as j. In both cases, the probability that the attempt is successful depends on the difference between i’s and j’s total payoff in the games they played in this round. They find that if the probability p is high enough, cooperation will eventually dominate. Takács, Janky, and Flache (2008) study the joint equilibrium states of partner choices in a network and contributions to a collective good. They obtained results where, because of the network structure and the assumption of behavioral confirmation by network partners who have the same contribution behavior, it is possible to obtain stable situations in which a fraction between 0 and 1 of the population contributes, and where contributors and noncontributors are linked to each other. coordination Goyal and Vega-Redondo (2005) study a model where actors can unilaterally establish ties to others; ties are costly; and actors play 2×2 coordination games with their network partners, where at any moment their behavior is the same toward all their interaction partners. The game is symmetric and is defined such that one option is efficient and the other is risk-dominant. All payoffs are positive, so that there is no reason to refuse any offered ties. At each time step actors have a certain probability of being allowed to revise their ties and their behavior; when this happens, they choose myopic best-response strategies. In addition, there is a small probability that the actor chooses a random strategy; stable strategies are studied that are obtained as this probability tends to 0. The Nash equilibria are either the empty network where nobody plays; or the complete network (that is, all pairs of actors play with each other) where everybody has the same behavior; or two disconnected internally complete subnetworks, each of which has a homogeneous behavior different from that of the other subnetwork. It turns out that if costs of ties are high, actors will eventually coordinate on the efficient behavior option; if costs of links are low, they will coordinate on the risk-dominant behavior.

272 Tom A. B. Snijders Buskens, Corten, and Weesie (2008) study the conditions under which a group of actors, playing coordination games with those to whom they are tied, and having the possibility of changing their ties, polarize into opposite “camps.” They also study what influences the extent to which the socially efficient solution is reached. These conditions are described in terms of the initial network structure. They find that initially denser networks lead to more homogeneous behavior, while initially less dense and more segregated networks lead to more heterogeneous behavior. The final proportion of choices for the efficient solution depends on the initial proportion; this lock-in effect is much stronger in initially dense than in initially sparse networks.

Statistical Coevolution Models Statistical actor-oriented models for the coevolution of networks and behavior were proposed by Snijders, Steglich, and Schweinberger (2007) and Steglich, Snijders, and Pearson (2010), extending the actor-oriented models for network dynamics discussed above. The actors control their outgoing ties as well as their behavior, and in a continuous-time model network and behavior change in mutual dependence.To decompose the model in the smallest possible steps, it is assumed that at any moment where a change occurs, this can be a change by only one actor in either one outgoing tie or in one behavior variable. As in the stochastic actor-oriented models for network dynamics only, the actions result from myopic optimization of a stochastically perturbed utility function.What is called a utility function here must be regarded as a resultant of costs, benefits, and constraints, and these will depend on a mixture of distant and more nearby goals as well as practical considerations. Since statistical modeling requires the flexibility to adapt the model to fit well to the data, this leads to the desirability of employing potentially different utility functions for network choices and behavior choices. One way of arguing this is by regarding network choice and behavior choice as being determined by different decision frames (Lindenberg 2001, this volume).The fact that network change is codetermined by behavior is called behavior-dependent selection (or selection briefly), represented by letting the utility function for the network depend also on the behavior of the focal actor and the interaction partners. Similarly, the change probabilities for behavior, defined by the utility function for the behavior, can depend also on the current state of the network as well as the behavior—for example, the behavior of the interaction partners—and this will be called influence. This terminology was also used, for example, by Ennett and Bauman (1994). This yields models in which the preference for being tied to similar others can be expressed in the utility function for the network as well as the utility function for behavior. Given longitudinal data, the model can be used for testing whether observed similarity between network neighbors can be attributed to influence (the behavior dynamics), selection (the network dynamics), or both. Some examples are given in Steglich, Snijders, and Pearson (2010) and Dijkstra et al. (2010) to friendship networks between adolescents and behaviors such as smoking, drinking, and delinquency; by Agneessens and Wittek (2008) to interemployee relations and well-being; by Berardo and Scholz (2010) to

Network Dynamics 273

collective action dilemmas in policy networks; and by Lewis, Gonzalez, and Kaufman (2012) to cultural tastes and online friendships.

Outlook Social networks have long been studied because of their effects on behavior, well-being, and performance of social actors. Research on the endogenous formation of social networks has come up more recently, with a number of important papers in the 1990s and a flurry of activity since the turn of the millenium. In addition to models of optimizing actors as treated in this chapter, there have been closely related and very interesting developments in agentbased models and models inspired by statistical mechanics, of which reviews may be found in Gross and Blasius (2008), Jackson (2008), Newman (2003), and Vriend (2006). In all these models one can distinguish between agent-based or computational models, game-theoretic models, and statistical models. The boundary lines are blurry, as the myopic best-response models in game theory as well as the actor-based statistical models are fundamentally of the same type as agent-based models. The differences often are in the details, and in the tradition and focus of research. Game-theoretic models mostly have the aim of studying stability and efficiency.To keep the analysis tractable and make possible obtaining explicit analytical results, such models are restricted to including only a small number of cost and benefit terms in the utility function. Agent-based models usually have the purpose of studying how global properties emerge from local rules of behavior, often with a special interest in phase transitions— that is, discontinuities in global properties as a function of the parameters of the model. Statistical models have the purpose of being used in data analysis, which requires flexibility in incorporating a large number of utility components, together with principled methods of parameter estimation and hypothesis testing. These diverse strands in the literature are not well integrated and are mostly oriented toward different disciplinary or subdisciplinary audiences. Much still is to be gained from more linkages between these literatures. The empirical study of network dynamics, and of the dynamic interdependence of networks and individual behavior, is only just beginning, as is argued, for example, by Stuart and Sorensen (2008). Especially for the more detailed investigation of the mutual dependence between networks and individual outcomes, theoretical arguments as well as data analysis methods have to be further elaborated. The collection of longitudinal network data is a painstaking process, and for generalizing across networks it is necessary to collect such data for multiple networks (Snijders and Baerveldt 2003), posing even stronger requirements.The insights into micro-macro processes that will be generated by such developments are ample reasons for making these efforts.

References Agneessens, F., and R. P. M. Wittek. 2008. “Social Capital and Employee Well-being: Disentangling Intrapersonal and Interpersonal Selection and Influence Mechanisms.” Revue Française de Sociologie 49: 613–37. Ahuja, G. 2000. “Collaboration Networks, Structural Holes, and Innovation: A Longitudinal Study.” Administrative Science Quarterly 45: 425–55.

274 Tom A. B. Snijders Axelrod, Robert. 1984. The Evolution of Cooperation. New York: Basic Books. Bala, Venkatesh, and Sanjeev Goyal. 2000. “A Noncooperative Model of Network Formation.” Econometrica 68: 1181–229. Barabási, A. L., and R. Albert. 1999. “Emergence of Scaling in Random Networks.” Science 286: 509–12. Baum, J. A. C., T. J. Rowley, A. V. Shipilov, and Y.-T. Chuang. 2005. “Dancing with Strangers: Aspiration Performance and the Search for Underwriting Syndicate Partners.” Administrative Science Quarterly 50: 536–75. Bearman, P. S., J. Moody, and K. Stovel. 2004. “Chains of Affection.” American Journal of Sociology 110: 44–91. Belleflamme, P., and F. Bloch. 2004. “Market Sharing Agreements and Collusive Networks.” International Economic Review 45: 387–411. Berardo, R., and J. T. Scholz. 2010. “Self-Organizing Policy Networks: Risk, Partner Selection, and Cooperation in Estuaries.” American Journal of Political Science 54: 632– 49. Berninghaus, S. K., K.-M. Ehrhart, and C. Keser. 1999. “Continuous-time Strategy Selection in Linear Population Games.” Experimental Economics 2: 41–57. Berninghaus, Siegfried K., Karl-Martin Ehrhart, and Marion Ott. 2006. “A Network Experiment in Continuous Time: The Influence of Link Costs.” Experimental Economics 9: 237–51. Berninghaus, Siegfried K., Karl-Martin Ehrhart, Marion Ott, and Bodo Vogt. 2007. “Evolution of Networks—An Experimental Analysis.” Journal of Evolutionary Economics 17: 317–47. Bloch, Francis, and Matthew O. Jackson. 2006. “Definitions of Equilibrium in Network Formation Games.” International Journal of Game Theory 34: 305–18. Bonacich, P. 1972. “Factoring and Weighting Approaches to Status Scores and Clique Identification.” Journal of Mathematical Sociology 2: 113–20. ———. 1987. “Power and Centrality: A Family of Measures.” American Journal of Sociology 92: 1170–82. Bonacich, Phillip, and Thomas M. Liggett. 2003. “Asymptotics of a Matrix Valued Markov Chain Arising in Sociology.” Stochastic Processes and Their Applications 104: 155–71. Burger, M. J., and V. Buskens. 2009. “Social Context and Network Formation: An Experimental Study.” Social Networks 31: 63–75. Burt, R. S. 1992. Structural Holes. Cambridge: Harvard University Press. ———. 2000. “Structural Holes versus Network Closure as Social Capital.” In Social Capital: Theory and Research, edited by N. Lin, K. S. Cook, and R. S. Burt, 31–56. New York: Aldine de Gruyter. Burt, R. S., and M. Knez. 1995. “Kinds of Third-party Effects on Trust.” Rationality and Society 7: 255–92. Buskens, Vincent, Rense Corten, and Jeroen Weesie. 2008. “Consent or Conflict: Coevolution of Coordination and Networks.” Journal of Peace Research 45: 205–22. Buskens, Vincent, and Arnout van de Rijt. 2008. “Dynamics of Networks If Everyone Strives for Structural Holes.” American Journal of Sociology 114: 371–407. Callander, S., and C. Plott. 2005. “Principles of Network Development and Evolution: An Experimental Study.” Journal of Public Economics 89: 1469–95. Casciaro, T., and M. S. Lobo. 2008. “When Competence Is Irrelevant: The Role of Interpersonal Affect in Task-related Ties.” Administrative Science Quarterly 53: 655– 84. Checkley, Matthew, and Christian E. G. Steglich. 2007. “Partners in Power: Job Mobility and Dynamic Deal-Making.” European Management Review 4: 161–71. Chung, S. A., H. Singh, and K. Lee. 2000. “Complementarity, Status Similarity and

Network Dynamics 275 Social Capital as Drivers of Alliance Formation.” Strategic Management Journal 21: 1–22. Coleman, J. S. 1964. Introduction to Mathematical Sociology. New York: Free Press of Glencoe. ———.1990. Foundations of Social Theory. Cambridge: Belknap Press. Davis, James A. 1970. “Clustering and Hierarchy in Interpersonal Relations: Testing Two Graph Theoretical Models on 742 Sociomatrices.” American Sociological Review 35: 843–52. ———.1979. “The Davis/Holland/Leinhardt Studies: An Overview.” In Perspectives on Social Network Research, edited by P. W. Holland and S. Leinhardt, 51–62. New York: Academic Press. Dijkstra, J. K., S. Lindenberg, R. Veenstra, C. Steglich, J. Isaacs, N. A. Card, and E. V. E. Hodges. 2010. “Influence and Selection Processes in Weapon Carrying during Adolescence: The Roles of Status, Aggression, and Vulnerability.” Criminology 48: 187–220. Doreian, P. 1989. “Network Autocorrelation Models: Problems and Prospects.” In Spatial Statistics: Past, Present, Future, edited by D. A. Griffith, 369–89. Ann Arbor: Michigan Document Services. Dutta, Bhaskar, and Matthew O. Jackson, eds. 2003. Networks and Groups: Models of Strategic Formation. Berlin: Springer. Eguìluz, Vìctor M., Martìn G. Zimmerman, Camilo J. Cela-Conde, and Maxi San Miguel 2005. “Cooperation and the Emergence of Role Differentiation in the Dynamics of Social Networks.” American Journal of Sociology 110: 977–1008. Ehrhardt, George C. M. A., Matteo Marsili, and Fernando Vega-Redondo. 2007. “Emergence and Resilience of Social Networks: A General Theoretical Framework.” Annales d’Economie et de Statistique 86: 1–13. Emerson, Richard M. 1972. “Exchange Theory, Part II: Exchange Relations and Networks.” In Sociological Theories in Progress. Vol. 2, edited by Joseph Berger, Morris Zelditch, Jr., and Bo Anderson, 58–87. Boston: Houghton-Mifflin. Ennett, Susan T., and Karl E. Bauman. 1994. “The Contribution of Influence and Selection to Adolescent Peer Group Homogeneity:The Case of Adolescent Cigarette Smoking.” Journal of Personality and Social Psychology 67: 653–63. Erdös, P., and A. Rényi. 1960. “On the Evolution of Random Graphs.” Matematikai Kutató Intézet Közleményei 5: 17–61. Fehr, E., and K. M. Schmidt. 1999. “A Theory of Fairness, Competition, and Cooperation.” Quarterly Journal of Economics 114: 817-–68. Feld, S. 1981. “The Focused Organization of Social Ties.” American Journal of Sociology 86: 1015–35. ———. 1982. “Structural Determinants of Similarity among Associates.” American Sociological Review 47: 797–801. Freeman, Linton C. 1979. “Centrality in Social Networks: Conceptual Clarification.” Social Networks 1: 215–39. ———. 2004. The Development of Social Network Analysis: A Study in the Sociology of Science.Vancouver, BC: Empirical Press. Gilles, R. P., S. Chakrabarti, S. Sarangi, and N. Badasyan. 2006. “Critical Agents in Networks.” Mathematical Social Sciences 52: 302–10. Goeree, J. K., A. Riedl, and A. Ule. 2009. “In Search of Stars: Network Formation among Heterogeneous Agents.” Games and Economic Behavior 67: 445–66. Gomes-Casseres, B. 1997. “Alliance Strategies of Small Firms.” Small Business Economics 9: 33–44. Gould, R.V. 2002 “The Origins of Status Hierarchies: A Formal Theory and Empirical Test.” American Journal of Sociology 107: 1143–78.

276 Tom A. B. Snijders Goyal, Sanjeev. 2007. Connections: An Introduction to the Economics of Networks. Princeton: Princeton University Press. Goyal, Sanjeev, and Sumit Joshi. 2006. “Unequal Connections.” International Journal of Game Theory 34: 319–49. Goyal, Sanjeev, and Fernando Vega-Redondo. 2005. “Network Formation and Social Coordination.” Games and Economic Behavior 50: 178–207. ———. 2007. “Structural Holes in Social Networks.” Journal of Economic Theory 137: 460–92. Granovetter, M. S. 1973. “The Strength of Weak Ties.” American Journal of Sociology 78: 1360–80. Gross, Thilo, and Bernd Blasius. 2008. “Adaptive Co-evolutionary Networks—A Review.” Journal of the Royal Society Interface 6: 259–71. Gulati, R. 1998. “Alliances and Networks.” Strategic Management Journal 19: 293–317. Gulati, R., and M. Gargiulo. 1999. “Where Do Interorganizational Networks Come From?” American Journal of Sociology 104: 1439–93. Gulati, R., N. Nohria, and A. Zaheer. 2000. “Strategic Networks.” Strategic Management Journal 21: 203–15. Hallinan, Maureen T. 1979. “The Process of Friendship Formation.” Social Networks 1: 193–210. Handcock, M. S., and J. Jones. 2004. “Likelihood-based Inference for Stochastic Models of Sexual Network Formation.” Theoretical Population Biology 65: 413–22. Haynie, Dana L. 2001. “Delinquent Peers Revisited: Does Network Structure Matter?” American Journal of Sociology 106: 1013–57. Heider, F. 1946. “Attitudes and Cognitive Organization.” Journal of Psychology 21: 107– 12. ———. 1958. The Psychology of Interpersonal Relations. New York: Wiley. Holland, P. W., and S. Leinhardt. 1975. “Local Structure in Social Networks.” Sociological Methodology—1976: 1–45. ———. 1977. “A Dynamic Model for Social Networks.” Journal of Mathematical Sociology 5: 5–20. Holme, P., and M. E. J. Newman. 2006. “Nonequilibrium Phase Transition in the Coevolution of Networks and Opinions.” Physical Review E 74, 056108–5. Jackson, M. O. 2003. “The Stability and Efficiency of Economic and Social Networks.” In Networks and Groups: Models of Strategic Formation, edited by B. Dutta and M. O. Jackson, 99–141. Heidelberg: Springer Verlag. ———. 2004. “A Survey of Models of Network Formation: Stability and Efficiency.” Ch. 1 in Group Formation in Economics: Networks, Clubs and Coalitions, edited by Gabrielle Demange and Myrna Wooders. Cambridge: Cambridge University Press. ———. 2008. Social and Economic Networks. Princeton: Princeton University Press. Jackson, M. O., and B. W. Rogers. 2007. “Meeting Strangers and Friends of Friends: How Random Are Social Networks?” American Economic Review 97: 890–15. Jackson, M., and A. van de Nouweland. 2005. “Strongly Stable Networks.” Games and Economic Behavior 51: 420–44. Jackson, M. O., and A. Watts. 2002. “On the Formation of Interaction Networks in Social Coordination Games.” Games and Economic Behavior 41: 265–91. Jackson, Matthew O., and Asher Wolinsky. 1996. “A Strategic Model of Social and Economic Networks.” Journal of Economic Theory 71: 44–74. Johnson, C., and R. P. Gilles. 2000. “Spatial Social Networks.” Review of Economic Design 5: 273–99. Kahneman, D., J. Knetsch, and R. Thaler. 1991. “Anomalies: The Endowment Effect, Loss Aversion and Status Quo Bias.” Journal of Economic Perspectives 5: 193–206. Kleinberg, J., S. Suri, E.Tardos, and T.Wexler. 2008. “Strategic Network Formation with Structural Holes.” Proceedings 9th ACM Conference on Electronic Commerce, 284–93.

Network Dynamics 277 Kochen, M., ed. 1989. The Small World. Norwood, NJ: Ablex Press. Kogut, B., P. Urso, and G. Walker. 2007. “Emergent Properties of a New Financial Market: American Venture Capital Syndication, 1960–2005.” Management Science 53: 1181–98. Kosfeld, M. 2004. “Economic Networks in the Laboratory: A Survey.” Review of Network Economics 3: 20–41. Kossinets, G., and D. J.Watts. 2006. “Empirical Analysis of an Evolving Social Network.” Science 311: 88–90. Krackhardt, D. 1999. “The Ties That Torture: Simmelian Tie Analysis in Organizations.” Research in the Sociology of Organizations 16: 183–210. Kumbasar, Ece, A. Kimball Romney, and William H. Batchelder, 1994. “Systematic Biases in Social Perception.” American Journal of Sociology 100: 477–505. Lazarsfeld, P. F., and R. K. Merton. 1954. “Friendship as a Social Process.” In Freedom and Control in Modern Society, edited by M. Berger, T. Abel, and C. Page, 18–66. New York:Van Nostrand. Lewis, K., M. Gonzalez, and J. Kaufman. 2012. “Social Selection and Peer Influence in an Online Social Network.” PNAS 109: 68–72. Lin, Nan. 2001. Social Capital: A Theory of Social Structure and Action. Cambridge University Press. Lin, N., K. Cook, and R. S. Burt, eds. 2001. Social Capital: Theory and Research. New York: Aldine de Gruyter. Lindenberg, S. 1993. “Framing, Empirical Evidence and Applications.” In Jahrbuch für Neue Politische Ökonomie, edited by P. Herder-Dorneich, K.-E. Schenk, and D. Schmidtchen, 11–38. Tübingen: Mohr (Siebeck). ———. 2001. “Social Rationality versus Rational Egoism.” In Handbook of Sociological Theory, edited by J. Turner, 635–68. New York: Kluwer/Plenum. Macy, M.W., J. Kitts, A. Flache, and S. Benard. 2003. “Polarization in Dynamic Networks: A Hopfield Model of Emergent Structure.” In Dynamic Social Network Modelling and Analysis: Workshop Summary and Papers, edited by R. Breiger, K. Carley, and P. Pattison, 162–73. Washington, DC: National Academies Press. Maddala, G. S. 1983. Limited-dependent and Qualitative Variables in Econometrics. 3rd ed. Cambridge: Cambridge University Press. Marsden, Peter V. 1987. “Core Discussion Networks of Americans.” American Sociological Review 52: 122–31. ———. 2005. “Recent Developments in Network Measurement.” In Models and Methods in Social Network Analysis, edited by Peter J. Carrington, John Scott, and Stanley Wasserman, 8–30. New York: Cambridge University Press. Marwell, G., and P. Oliver. 1993. The Critical Mass in Collective Action: A Micro-social Theory. Cambridge: Cambridge University Press. Marwell, G., P. E. Oliver, and R. Prahl. 1988. “Social Networks and Collective Action: A Theory of the Critical Mass. III.” American Journal of Sociology 94: 502–34. Mayer, A., and S. L. Puller. 2008. “The Old Boy (and Girl) Network: Social Network Formation on University Campuses.” Journal of Public Economics 92: 329–47. McFadden, D. 1973. “Conditional Logit Analysis of Qualitative Choice Behavior.” In Frontiers in Econometrics, edited by P. Zarembka, 105–42. New York: Academic Press. McPherson, M., L. Smith-Lovin, and M. E. Brashears. 2006. “Social Isolation in America: Changes in Core Discussion Networks over Two Decades.” American Sociological Review 71: 353–75. McPherson, M., L. Smith-Lovin, and J. M. Cook. 2001. “Birds of a Feather: Homophily in Social Networks.” Annual Review of Sociology 27: 415–44. Merton, R. K. 1968. “The Matthew Effect in Science.” Science 159: 56–63. Milgram, S. 1967. “The Small World Problem.” Psychology Today 2: 61–67. Molm, L. D., J. L. Collett, and. D. R. Schaefer. 2007. “Building Solidarity through

278 Tom A. B. Snijders Generalized Exchange: A Theory of Reciprocity.” American Journal of Sociology 113: 205–42. Moreno, J. L. 1934. Who Shall Survive? A New Approach to the Problem of Human InterRelations. New York: Beacon House. Myerson, R. 1991. Game Theory: Analysis of Conflict. Cambridge: Harvard University Press. Newman, Mark E. J. 2003. “The Structure and Function of Complex Networks.” SIAM Review 45: 167–256. Newman, Mark E. J., and Juyong Park. 2003. “Why Social Networks Are Different from Other Types of Networks.” Physical Review E 68, 036122. Pattison, Philippa E., and Garry L. Robins. 2002. “Neighbourhood Based Models for Social Networks.” In Sociological Methodology. Vol. 22, edited by Ross M. Stolzenberg, 301–37. Boston: Blackwell Publishing. Price, Derek de Solla. 1976. “A General Theory of Bibliometric and Other Advantage Processes.” Journal of the American Society for Information Science 27: 292–306. Raub, Werner, and J. Weesie. 1990. “Reputation and Efficiency in Social Interactions: An Example of Network Effects.” American Journal of Sociology 96: 626–54. Riolo, Rick L., Michael D. Cohen, and Robert Axelrod. 2001. “Evolution of Cooperation without Reciprocity.” Nature 414: 441–43. Robins, G., T. A. B. Snijders, P. Wang, M. Handcock, and P. Pattison. 2007. “Recent Developments in Exponential Random Graph (p*) Models for Social Networks.” Social Networks 29: 192–215. Robins, G. L., P. Pattison, and J. Woolcock,. 2005. “Small and Other Worlds: Global Network Structures from Local Processes.” American Journal of Sociology 110: 894– 936. Santos, Francisco C., Jorge M. Pacheco, and Tom Lenaerts. 2006. “Cooperation Prevails When Individuals Adjust Their Social Ties.” PLoS Computational Biology 2: 1284– 90. Schaefer, David R., Olga Kornienko, and Andrew M. Fox. 2011. “Misery Does Not Love Company: Network Selection Mechanisms and Depression Homophily.” American Sociological Review 76: 764–85. Simmel, G. 1955 [1908]. Conflict and the Web of Group-Affiliations. Translated by Kurt Wolff. New York: Free Press. ———. 1950 [1917]. Individual and Society. In The Sociology of Georg Simmel, edited by K. H. Wolff. New York: Free Press. Skyrms, B., and R. Pemantle. 2000. “A Dynamic Model of Social Network Formation.” PNAS 97: 9340–46. Snijders, T. A. B. 1996. “Stochastic Actor-oriented Dynamic Network Analysis.” Journal of Mathematical Sociology 21: 149–72. ———. 2001. “The Statistical Evaluation of Social Network Dynamics.” In Sociological Methodology, edited by M. Sobel and M. Becker, 361–95. Boston and London: Basil Blackwell. Snijders, T. A. B., and C. Baerveldt. 2003. “A Multilevel Network Study of the Effects of Delinquent Behavior on Friendship Evolution.” Journal of Mathematical Sociology 27: 123–51. Snijders,T. A. B., P. Pattison, G. L. Robins, and M. Handcock. 2006. “New Specifications for Exponential Random Graph Models.” Sociological Methodology 36: 99–153. Snijders, Tom A. B., Christian E. G. Steglich, and Michael Schweinberger. 2007. “Modeling the Co-evolution of Networks and Behavior.” In Longitudinal Models in the Behavioral and Related Sciences, edited by Kees van Montfort, Han Oud, and Albert Satorra, 41–71. Mahwah, NJ: Lawrence Erlbaum. Snijders,T. A. B., G. G. van de Bunt, and C. E. G. Steglich. 2010. “Introduction to Actorbased Models for Network Dynamics.” Social Networks 32: 44–60.

Network Dynamics 279 Steglich, C. E. G., T. A. B. Snijders, and M. Pearson. 2010. “Dynamic Networks and Behavior: Separating Selection from Influence.” Sociological Methodology 40: 329–93. Stiller, J., and R. I. M. Dunbar. 2007. “Perspective Taking and Social Network Size in Humans.” Social Networks 29: 93–104. Stokman, F. N., and E. P. H. Zeggelink. 1996. “‘Self-organizing’ Friendship Networks.” In Frontiers in Social Dilemmas Research, edited by W. B. G. Liebrand and D. M. Messick, 385–418. Berlin: Springer. Stuart, Toby E., and Olav Sorensen. 2007. “Strategic Networks and Entrepreneurial Ventures.” Strategic Entrepreneurship Journal 1: 211–27. Takács, K., B. Janky, and A. Flache. 2008. “Collective Action and Network Change.” Social Networks 30: 177–89. Thaler, R. 1980. “Toward a Positive Theory of Consumer Choice.” Journal of Economic Behavior and Organization 1: 39–60. Travers, J., and S. Milgram. 1969. “An Experimental Study of the Small World Problem.” Sociometry 32: 425–43. Van de Bunt, G. G., M. A. J. van Duijn, and T. A. B. Snijders. 1999. “Friendship Networks through Time: An Actor-oriented Statistical Network Model.” Computational and Mathematical Organization Theory 5: 167–92. Van de Bunt, G. G., R. P. M. Wittek, and M. C. de Klepper. 2005. “The Evolution of Intra-Organizational Trust Networks: The Case of a German Paper Factory: An Empirical Test of Six Trust Mechanisms.” International Sociology 20: 339–69. Van Duijn, M. A. J., E. P. H. Zeggelink, M. Huisman, F. N. Stokman, and F. W. Wasseur. 2003. “Evolution of Sociology Freshmen into a Friendship Network.” Journal of Mathematical Sociology 27: 153–91. Vega-Redondo, F. 2006. “Building up Social Capital in a Changing World.” Journal of Economic Dynamics and Control 30: 2305–38. Verbrugge, L. M. 1977. “The Structure of Adult Friendship Choices.” Social Forces 56: 576–97. Vriend, N. 2006. “ACE Models of Endogenous Interactions.” In Handbook of Computational Economics. Vol. 2: Agent-Based Computational Economics, edited by L. Tesfatsion and K. L. Judd, 1047–79. Amsterdam: Elsevier. Wasserman, Stanley, and Philippa E. Pattison. 1996. “Logit Models and Logistic Regression for Social Networks: I. An Introduction to Markov Graphs and p*.” Psychometrika 61: 401–25. Watts, Alison. 2001. “A Dynamic Model of Network Formation.” Games and Economic Behavior 34: 331–41. Watts, Duncan J. 1999. “Networks, Dynamics, and the Small-World Phenomenon.” American Journal of Sociology 105: 493–527. Watts, Duncan J., and Steven H. Strogatz. 1998. “Collective Dynamics of ‘Small-world’ Networks.” Nature 393: 440–42. Wippler, R., and S. Lindenberg. 1987. “Collective Phenomena and Rational Choice.” In The Micro Macro Link, edited by J. C. Alexander, B. Giesen, R. Münch, and N. J. Smelser, 135–52. Berkeley: University of California Press. Zeggelink, E. P. H. 1995. “Evolving Friendship Networks: An Individual-oriented Approach Implementing Similarity.” Social Networks 17: 83–110. Zeggelink, Evelien P. H., Frans N. Stokman, and Gerhard G. van der Bunt. 1996. “The Emergence of Groups in the Evolution of Friendship Networks.” Journal of Mathematical Sociology 21: 29–55. Zheng, T., M. J. Salganik, and A. Gelman. 2006. “How Many People Do You Know in Prison?: Using Overdispersion in Count Data to Estimate Social Structure in Networks.” Journal of the American Statistical Association 101: 409–23.

chapter

Rational Choice Research in Criminology: A Multi-Level Framework

8

ross l. matsueda

Introduction A challenging puzzle for rational choice theory concerns the causes and control of criminal behavior. Crime is a difficult case for rational choice. Compared with market behavior, financial decisions, and corporate crime, in which institutionalized norms frame decision-making in the terms of rationality, street crimes are often characterized as irrational and suboptimal. Street criminals are commonly portrayed by the media and a few social scientists as impulsive, unthinking, and uneducated, and their behaviors as beyond the reach of formal sanctions (for example, Gottfredson and Hirschi 1990). Consequently, support of rational choice principles for criminal behavior would provide strong evidence for the perspective (Matsueda, Kreager, and Huizinga 2006). Crime is an important arena for investigating rational choice for another reason: utilitarian principles, and their accompanying psychological assumptions, undergird our legal institution (for example, Maestro 1973). This connection is rooted in writings of members of the classical school, particularly Jeremy Bentham and Caesare Beccaria. Bentham (1948 [1789]) argued that happiness is a composite of maximum pleasure and minimum pain, and that the utilitarian principle—the greatest happiness for the greatest number—underlies morals and legislation. Punishment by the state constitutes one of four sanctions—political, moral, physical, and religious—that shape pleasures and pains. Influenced by the moral philosophers of the Enlightenment, Beccaria (1963 [1764]) assumed that criminal laws reflect the terms of a social contract between members of society and the state. Individuals receive protection of their rights to personal welfare and private property in exchange for relinquishing the freedom to violate the rights of others. The rights of individuals are protected by the state through deterrence, threatening potential transgressors with just enough punishment to outweigh the pleasures of crime. With his writings, Beccaria attempted to reform the unjust and brutal legal system of eighteenth-century Europe by developing a rational system in which laws are specified clearly and a priori (so individuals have full information about the consequences of their acts), judicial

284 Ross L. Matsueda discretion is eliminated (so all citizens are equal in the eyes of the law), and punishments are made certain, swift, and no more severe than needed to deter the public from crime (Matsueda, Kreager, and Huizinga 2006). Because of the obvious implications for public policy, theory and research on rational choice and crime have focused primarily on the question of deterrence: Does the threat of punishment by the state deter citizens from crime (see Zimring and Hawkins 1973)? Recent research concludes that the threat of formal sanction does deter, but that the effects are modest in size and perhaps conditioned by social context (for example, ibid.; Nagin 1998). Less research has moved beyond deterrence to examine incentives outside the scope of formal punishment, such as psychic rewards and costs, within a rational choice theory of crime. This modest but growing literature has underscored the importance of rational choice theory for understanding and explaining criminal behavior (for example, Clarke and Cornish 1985; Cornish and Clarke 1986). At this time, rational choice remains an important but still minority position in criminology. This is partly because of the historical dominance of sociologists in criminology, many of whom continue to take a jaundiced view of rational choice theory. Such views are holdovers of old sociological debates that persist today, such as free will versus determinism, macro- versus microexplanations, and liberal political views versus conservative individualist ideologies. Skepticism over rational choice theories of crime has diminished recently as neoclassically trained economists and rational choice sociologists have increasingly turned their attention to the problem of crime. But, with a few notable exceptions, particularly in the policy realm, economic research has not been well integrated into the mainstream of criminological thought. At the same time, during the last decade, criminologists have made substantial theoretical and empirical advances in uncovering important causes of crime. Most of this research is rooted in sociological perspectives. For example, research has underscored the importance of life course transitions—such as developing a committed marriage, serving in the military, becoming a mother, and successfully entering the labor force—in altering trajectories of criminal offending (for example, Sampson and Laub 1993; Giordano, Cernkovich, and Rudolph 2002). Research has found that incarceration of residents undermines the strength of local communities, and that re-entry of felons into communities may also have negative consequences for both the former offender and the community (for example, Western 2007; Pager 2007; Clear 2007). Sociologists have identified important dimensions of community social capital upon which residents can draw to solve local neighborhood problems, such as crime and disorder, and which help to explain the effects of urban structure on community rates of crime (see Sampson, Morenoff, and Gannon-Rowley 2002). Research has also provided detailed ethnographic descriptions of innercity gangs (for example,Venkatesh 2000), street violence (Anderson 1999), and organized crime (Gambetta 1993).With a few notable exceptions (for example, Gambetta 1993), most of this research is not explicitly rooted in rational choice perspectives. This chapter uses a multilevel framework to discuss advances in rational choice research on crime. Rather than providing an exhaustive review of

Rational Choice Research in Criminology

285

pertinent research, I instead organize the discussion around one important theoretical issue, the integration of micro- and macrolevels of explanation.Thus the underlying assumption that gives structure to the chapter is that rational choice principles offer a parsimonious microfoundation for macrosociological concepts and causal mechanisms. The task, then, is to identify how macrolevel social contexts condition microlevel processes (individual decisions), and how microprocesses, in turn, produce macrolevel outcomes (social organization) (for example, Coleman 1990). I begin by discussing an individual-level model of rational choice, deterrence, and criminal behavior. A rich and voluminous literature has developed around the question of general deterrence—do threats of formal sanction by the legal system deter the general public from crime? I review the models and different research designs used in empirical studies, and then discuss the individual-level rational addiction model of drug use (Becker and Murphy 1988; Becker 1996). To link individual-level models to macrosociological models, I review the micro-macro problem in sociology, and the potential utility of using a rational choice model as a microfoundation for macrolevel causal relationships. Here, I summarize Coleman’s position (1990), which emphasizes the crucial task of identifying micro-to-macro transitions. I then use this multilevel framework to analyze two productive lines of research in criminology: (1) social capital, collective efficacy, and neighborhood controls (Sampson, Raudenbush, and Earls 1997); and (2) the protection racket of organized crime (Gambetta 1993). Theoretically, I treat these processes as examples of what Edwin Sutherland (1947) termed “organization against crime” and “organization in favor of crime” as the defining features of his theory of differential social organization (see Matsueda 2006). In each instance, I stress the utility of rational choice at the individual level, the broader context that conditions individual purposive action, and the micro-to-macro transitions that lead to social organization either against or in favor of crime. The extent to which these lines of research capitalize on a rational choice microfoundation varies considerably. For example, collective efficacy theory has been treated as a purely macrolevel process linking social disorganization, social capital, and informal social control into a macrostructural theory of crime. Therefore, I show how rational choice can provide a microfoundation for social capital and collective efficacy that opens new theoretical puzzles and empirical research questions. In contrast, Gambetta’s analysis (1993) of the Sicilian Mafia’s protection racket draws explicitly on a rational choice perspective to explain the origins and functioning of privatized protections. Therefore, I explicate the individual-level rational choice argument and show how it links to a macrolevel system of illicit action. In the final section, I discuss avenues for future research within a multilevel framework.

Individual-Level Model of Criminal Behavior rational choice, deterrence, and criminal acts Rational choice theories of crime are rooted in the seminal writings of Gary Becker (1968), who argues that the same principles explaining decisions

286 Ross L. Matsueda by firms and members of households should also explain criminal behavior. Drawing on the expected utility theory of risky decisions under uncertainty by von Neumann and Morgenstern (1944), Becker (1968: 177) specifies a simple utility function for committing crimes: E(UC) = (1 – pc) U(R) + pc U(R – C)

(1)

where E(UC) is the expected utility of crime, pc is the probability of getting arrested and punished, (1 – pc) is the probability of getting away with crime, R is the return (both monetary and psychic) from crime, and C is the cost of punishment (for example, a fine or prison sentence), and U is a utility function translating punishments and rewards to a common metric. The expected utility model assumes that individuals have complete and transitive preference orderings for all possible decision outcomes. As von Neumann and Morgenstern (1944) famously pointed out, expected utilities can differ from expected values. For example, the expected income from crime will not differ when an increase in the probability of punishment p is compensated by an equal percentage decrease in severity of punishment, C (Becker 1968). E(R) = (1 – pc) (R) + pc (R – C) = R – pc C

(2)

Such a change in pc and C, however, can change expected utility because it will alter risk. The change in expected utility depends on the individual’s attitude (or taste) toward risk. If a person has a preference for risk, the utility function is convex, and an increase in pc will reduce expected utility more than an equal increase in C (ibid.). Conversely, if a person is risk-averse, the utility function is concave, and C will have a greater effect than pc. Finally, if a person is riskneutral, the utility function is linear, and pc and C will have identical effects. If we ignore the role of legitimate opportunities, that is, assume the expected utility from noncrime is zero, E(UN) = 0, we can specify that a crime will occur when E(UC) > E(UN) = 0, so that from equation (1), a crime will occur when the following holds: U(R) > pc U(C)

(3)

That is, when the returns to crime exceed the punishment, weighted by the probability of detection, an individual will commit a crime. The policy implication here is that by increasing the certainty and severity of punishment, the probability of crime will be reduced. Crime can also be reduced by lowering the rewards to crime—by defending public spaces through increasing surveillance, employing security guards, and using technological advances in metal detection, alarms, locks, fences, and the like. Historically, following Becker’s work (ibid.), most microeconomic research on crime has focused on the policy implications of increasing the certainty and severity of punishment. Of course, legitimate opportunities are important for criminal decisions, as most members of society obtain some utility from noncriminal activities.1 Bueno de Mesquita and Cohen (1995) present a simple model that considers legitimate opportunities by specifying a utility function for noncriminal activity:

Rational Choice Research in Criminology

E(UN) = pi U(I ) + (1 – pi) U(W )

287

(4)

where I is income (returns to conventional activity), pi is the probability of obtaining I (through having high social status, resources, or talent), and W is welfare or the social safety net for those who cannot obtain I (that is, pi = 0). Then, the utility function for criminal behavior becomes: E(UC) = (1 – pc) [U(R) + pi U(I ) + (1 – pi) U(W )] (5) + pc U(R +W – C ) In other words, the utility from crime is a function of the returns to crime plus income from conventional activity (each weighted by the probability of getting away with crime), plus the returns to crime and conventional activity minus the punishment for crime (each weighted by the probability of getting caught and punished). This assumes that the criminal’s booty from crime is not confiscated upon arrest (Becker 1968). Note that when the probability of getting caught is zero (pc = 0), the utility from crime is equal to the returns to crime plus the returns to noncrime. When the probability of getting caught is 1.0 (pc = 0), the utility from crime is the returns to crime, plus welfare, minus the penalty. A crime will be committed when E(UC) > E(UN); therefore, from (4) and (5), a crime will occur when the following holds: (1 – pc) [U(R) + pi U(I ) + (1 – pi) U(W )] + pc U(R +W – C ) (6) > pi U(I ) + (1 – pi) U(W ) Or, equivalently, stated in terms of the risk of punishment, crime will occur when pc < U(R) / U(C ) + pi U(I –W )

(7)

That is, crime occurs when the probability of detection is less than the ratio of the reward to the sum of the punishment plus the returns to noncriminal activity weighted by the probability of realizing those returns. From a policy point of view, the probability of crime can be altered not only through criminal justice policies that increase the certainty and severity of punishments or that change defensible space (and thereby reduce opportunities for crime), but also through policies that increase conventional alternatives to crime. For example, job training, higher education, and other programs to enhance human and social capital may reduce the attractiveness of crime by increasing pi, the probability of obtaining a desired income from legitimate activities. Returns to conventional activity include not only income but also social status and prestige, self-esteem, and happiness; policies that increase these quantities by inculcating strong commitments to conventional institutions may help to reduce crime. empirical research on rational choice and deterrence Early empirical tests of Becker’s model used statistical models of aggregate crime rates, focusing on the deterrent effects of objective risk of punishment, using, for example, risk of imprisonment (measured by imprisonment per capita) or risk of arrest (measured by arrests per crimes reported to police). Ehrlich (1973) found deterrent effects of risk of imprisonment, but scholars criticized his simultaneous equation models for using implausible solutions

288 Ross L. Matsueda to the identification problem—the problem of finding good instrumental variables to identify reciprocal effects between rates of imprisonment and rates of crime—such as assuming population age, socioeconomic status, and region have zero direct effects on crime (Nagin 1978). Recent work using aggregate data includes more plausible instrumental variables to address the problem of reverse causality, and found deterrent effects. Sampson and Cohen (1988) follow the work of Wilson and Boland (1978) and use aggressive policing as an instrument for risk of arrest, finding a deterrent effect. Levitt (1997) employs the timing of mayoral elections as an instrument of number of police per capita—such elections should have a direct effect on investment in the police force (as newly elected mayors seek to crack down on crime), but only an indirect effect on crime (but see McCrary 2002). For a review of aggregate deterrence research, see Nagin (1998) and Durlauf and Nagin (2011). These tests of the deterrence hypothesis assume that actors know the objective certainty of arrest and imprisonment (Nagin 1998). By contrast, subjective expected utility models relax this assumption, replacing the single known objective probability with a distribution of subjective probabilities. Subjective utility models are still rational models because the statistical mean of the subjective probability distribution is assumed to fall on the value of the objective probability (ibid.). Empirical research from a subjective expected utility framework uses survey methods to measure perceived risk of punishment directly from respondents, rather than inferring it from behavior through the method of revealed preferences (for example, Kahneman, Wakker, and Sarin 1997). Early empirical research by sociologists used cross-sectional data and found small deterrent effects for certainty of punishment but not for severity (for example, Williams and Hawkins 1986). Respondents who perceive a high probability of arrest for minor offenses (such as marijuana use and petty theft) report fewer acts of delinquency. Such research has been criticized for using cross-sectional data in which past delinquency is regressed on present perceived risk, resulting in the causal ordering of the variables contradicting their temporal order of measurement. To address this criticism, sociologists have turned to two-wave panel models and found, for minor offenses, little evidence for deterrence (perceived risk had little effect on future crime) and strong evidence for an experiential effect (prior delinquency reduced future perceived risk) (see ibid.; Paternoster 1987). Piliavin et al. (1986) specify a full rational choice model of crime, including rewards to crime as well as risks, and find, for serious offenders, that rewards exert strong effects on crime, but perceived risks do not. Recent longitudinal survey research has used more sophisticated measures of risk, better-specified models, and better statistical methods. Matsueda, Kreager, and Huizinga (2006) specify two models based on rational choice. First is a Bayesian learning model of perceived risk, in which individuals begin with a baseline estimate of risk, then update the estimate based on new information, such as personal experiences with crime and punishment or experiences of friends. Second is a rational choice model of crime in which crime is determined by prior risk of arrest, perceived opportunity, and perceived rewards to crime, such as excitement, kicks, and being seen as cool by peers (see also McCarthy 1995; Hagan and McCarthy 1998). Using longitudinal data from

Rational Choice Research in Criminology

289

the Denver Youth Survey, Matsueda, Kreager, and Huizinga (2006) find support for both hypotheses: perceived risk conforms to a Bayesian updating process (see also Pogarsky, Piquero, and Paternoster 2004; Anwar and Loughran 2011), and delinquency is determined by perceived risk of arrest, rewards to crime, perceived opportunities, and opportunity costs (see also Pogarsky and Piquero 2003). Similarly, Lochner (2007) uses two national longitudinal datasets and finds support for an updating model of “beliefs about the criminal justice system” and a deterrent effect of perceived risk. Sherman (1990) has observed that the deterrent effect of interventions, such as police crackdowns or passage of more punitive legislation, often has an initial deterrent effect that diminishes with time. A simple explanation of this decay in deterrent effect is that criminals initially overestimate the effect of the policy change on certainty of getting caught, and consequently through Bayesian updating, adjust their risk perceptions downward (Nagin 1998). A second explanation of initial decay in deterrence derives from decision theorists’ concept of “ambiguity aversion.” In contrast to risk aversion, which refers to an event in which a probability can be assigned to every outcome, ambiguity aversion refers to an event in which the probabilities of outcome are unknown (Epstein 1999). A new intervention may increase the uncertainty of the risk perceptions of potential offenders, which will create a deterrent effect if offenders find uncertainty or ambiguity aversive. Over time, this ambiguity over risk may diminish, as offenders adapt to the new policy and sharpen their estimates of true risk. The important point here is that, even if the policy did not change the true certainty of punishment or the mean values of offenders’ subjective perceptions of risk, it may change the variance of risk perceptions, which will deter crime if offenders are risk averse (Nagin 1998). Sherman suggested that a policy of varying police crackdowns over time and space may increase ambiguity in risk perceptions, and thereby more effectively deter crime. Loughran et al. (2011) found support for the deterrent effect of ambiguity aversion for crimes that did not involve contact between victims and offenders: at low levels of certainty of sanction, ambiguity reduced offending, whereas at high levels of offending, ambiguity increased offending. Another way of addressing the causal order problem is with scenario or vignette methods. Here a specific crime scene is depicted in a written scenario and the respondent is asked to assess the probability of getting caught or getting rewards from the crime depicted. Then the respondent is queried for his intentions to engage in the crime. The method has the strength of embedding reported risk perceptions in the situation in which they should apply (Nagin 1998). Moreover, intentions data may be reasonable proxies for actual behavior (see Manski 1990; Dominitz and Manski 1997). The vignette method has the additional strength of random assignment of scenario characteristics—such as presence of witnesses, time of day, potential monetary returns—to vignettes in a factorial design, creating orthogonal regressors that allow one to obtain precise estimates of characteristics on outcomes (for example, Rossi and Nock 1982). A weakness is the potential for a response effect: respondents who report high risk of arrest may be unlikely to admit to an intention to commit the crime because of social desirability effects.Vignette studies of deterrence and rational choice generally find robust effects of deterrence: certainty has a substantial

290 Ross L. Matsueda effect on criminal intentions, while severity has modest effects. This holds for tax evasion (Klepper and Nagin 1989), drunk driving (Nagin and Paternoster 1994), sexual assault (Bachman, Paternoster, and Ward 1992), and corporate crime (Paternoster and Simpson 1996). In sum, empirical research on an individual model of rational choice, deterrence, and crime finds consistent support for the model. As deterrence theory suggests, certainty of sanctions exerts a consistent deterrent effect on crime, although the severity of punishment exerts a small and inconsistent effect. Consistent with rational choice, returns to crime—particularly psychic returns, such as excitement and high status among peers—and opportunity costs are both important predictors of future criminality. Note that models of deterrence and crime are essentially depicting a two-person game between the criminal and the criminal justice system. Most research on deterrence, however, treats individual criminal behavior as endogenous with respect to the actions of the criminal justice system, which are assumed exogenous (that is, the endogeneity of legal actors is treated as a nuisance to be overcome). Nagin (1998) and Swaray, Bowles, and Pradiptyo (2005) review economic research on the effects of interventions on the criminal justice system—in which the intervention is truly exogenous. A more complete treatment would model the legal system and the criminal as interdependent actors, using game theory—the use of mathematical models to tease out interdependent decision-making. McCarthy (2002) reviews applications of game theory, particularly two-person games, to the relationship between criminals and the legal system (see also Bueno De Mesquita and Cohen 1995). McAdams (2009) reviews the relevance of game theory beyond the prisoner’s dilemma for law and legal analysis. By extending the equations used earlier, I can give an illustrative example, based on research by Bueno de Mesquita and Cohen (1995), of the utility of game theory in theorizing about criminal behavior, and drawing links between macrostructures and social interactions. Bueno de Mesquita and Cohen (1995) show how an unjust social structure— containing selective barriers to human and social capital that undermine job attainment and wages—can change the incentive structure for criminal decisions. For individuals, there is uncertainty about fairness or justice in the social system.Therefore, we can define pj as a measure of individual perceptions of the probability of justice or fairness in social institutions, and (1 – pj) as a measure of perceived probability that society is unfair. The likelihood that an individual will be treated fairly by social institutions will affect the probability of returns to conventional activity. A fair society will allow individuals to gain income from conventional sources (I ) based on pi, the probability of getting a good job, which is based on ability, human capital, and social capital. An unfair society will prevent some qualified individuals from getting good jobs, which implies that those individuals will receive zero income from conventional jobs (I = 0), making total benefits equal to welfare, W. Therefore, if we incorporate fairness into our earlier equation (4), the utility from noncrime becomes: E(UN) = pj [ pi U(I ) + (1 – pi) U(W )] + (1 – pj) U(W )

(8)

In a completely fair society, in which all members perceive fairness, pj = 1, utility from noncrime is pi U(I ) + (1 – pi) U(W ), as above. But in a completely

Rational Choice Research in Criminology

291

unfair society, in which all members perceive unfairness, pj = 0, utility from noncrime is reduced to welfare, U(W ). Then, modifying equation (5), the utility from crime, allowing fairness to vary, is: E(UC) = (1 – pc) {U(R) + pj [ pi U(I ) + (1 – pi) U(W )] (9) + (1 – pj) U(W )} + pc U(R +W – C ) A crime will be committed when E(UC) > E(UN); therefore, from (8) and (9), a crime will occur when the following holds: (1 – pc) {U(R) + pj [ pi U(I ) + (1 – pi) U(W )] + (1 – pj) U(W )} + pc U(R +W – C) > pj [ p i U(I ) + (1 – pi) U(W )] + (1 – pj) U(W )

(10)

Stated in terms of perceived probability of injustice, crime will occur when pj < U(R) – pc U(C) / pc pi U(I –W )

(11)

and it follows that pj pc pi U(I –W ) < U(R) – pc U(C)

(12)

This shows that as perceived justice increases, crime becomes less likely because the returns to conventional activity increase. Using these equations, Bueno de Mesquita and Cohen (1995) provide a game-theoretic analysis of changes in society’s fairness, certainty and severity of punishment, probability that an individual will gain conventional income, and welfare policies. Their simulations reveal three important patterns. First, by increasing social justice, crime is reduced substantially. Second, the effect that poverty reduction policies have on crime depends on the policy: reducing poverty by welfare programs increases crime in the short run; conversely, reducing poverty by increasing the human capital skills of individuals reduces crime sharply. Third, crime is reduced substantially when policies of increasing human capital skills are combined with policies of increasing the probability of punishment. theory of rational addiction to illicit drugs The rational choice model can also be applied to the consumption of illicit drugs. In their path-breaking article, Becker and Murphy (1988) note that addiction or habit formation is pervasive throughout society. People often become addicted not only to drugs, alcohol, and cigarettes but also to work, eating, music, and many other activities. Therefore, Becker and Murphy (ibid.) suggest that the explanatory power of rational choice theories would be seriously compromised if addictions required separate theories. They show how addiction, including drug addiction, can be explained within a rational choice framework in which individuals maximize expected utility subject to constraints and incorporate both past and future behavior in decision-making. In this way, addictive behavior is consistent with the usual assumption of optimization with stable preferences. This explanation consists of two parts. First is a backward-looking model, or “learning by doing,” in which increases in past drug use (consumption) increase current drug use by raising the marginal utility of current drug use. Second is a forward looking model, in which current consumption is a function of anticipated future utility: an individual

292 Ross L. Matsueda expecting to consume drugs in the next period will consider the utility from that future drug use when maximizing utility of current drug consumption. Individuals recognize that consumption of beneficial goods (for example, sex) increases future utility, whereas consumption of harmful goods (for example, illicit drugs) reduces future utility. Thus, in making current decisions, rational actors trade off the present utility of drug consumption with the future utility of drug addiction. The model implies strong intertemporal complementarity for drug consumption: consuming drugs at time one will be highly correlated with drug consumption at time two. A myopic (or backward looking) model is a special case in which individuals fail to consider utility of future behavior on current choices. Empirical research on rational addiction models of drug use models the relationship between drug prices and drug use over time (for example, Becker, Grossman, and Murphy 1994; Grossman and Chaloupka 1998). Drug use at time t is specified as a function of price at time t, drug use at time t – 1 (backward-looking), and drug use at time t + 1 (forward-looking). Ct = θCt – 1 + βθCt – 1 + θ1Pt + θ2 εt + θ3 εt + 1 where Ct is present consumption, Ct – 1 is past consumption, Ct + 1 is future consumption, θ is a parameter reflecting addiction, β is a time discount factor (1/[1 + r]) assumed to be less than one, θ1 is a coefficient for price Pt, and Ct – 1 = θCt – 2 + βθCt + θ1Pt – 1 + θ2 εt – 1 + θ3 εt Ct + 1 = θCt + βθCt + 2 + θ1Pt + 1 + θ2 εt + θ3 εt + 2 To address the obvious endogeneity problem, price at time t – 1 is used as an instrument for drug use at time t – 1, price at time t is used as an instrument for drug use at time t, and price at time t + 1 is used as an instrument for drug use at t + 1. Identification is achieved by the perhaps plausible assumption that price at t – 1 and price at t + 1 have no effects on drug use at time t, net of price and time t. Such models have the weakness of assuming perfect foresight, although partial foresight models are tractable here. Using data from the national Monitoring the Future dataset as well as data on marijuana prices (from Drug Enforcement agents’ attempts to purchase marijuana in nineteen cities for 1982–1992), Pacula et al. (2000) estimate price elasticity of demand, estimating that a 1 percent increase in price reduces demand by about 30 percent.They find, however, that peer effects and attitudes are the strongest predictors of marijuana use. Using the same data, Chaloupka et al. (1999) find that youth living in decriminalized states were more likely to use marijuana than in other states, and that youths’ consumption patterns were responsive to median fines for possession of marijuana. In contrast, Farrelly et al. (1999), using fixed-effects models on the National Household Survey on Drug Abuse, find no relationship between fines and marijuana use. This line of research assumes that youth are aware of the objective costs of marijuana use, and use those costs in their decision-making. It has been criticized for assuming that youth are able to anticipate future prices of marijuana accurately. On this point, with respect to cigarettes, Gruber and Köszegi (2001) argue that a more reasonable assumption is that individuals are able to anticipate future changes in excise taxes because they tend to be publicized, whereas increases in cigarette

Rational Choice Research in Criminology

293

prices are rarely announced in advance. Using data on excise taxes, Gruber and Köszegi (ibid.) find support for a forward-looking model of rational addiction for cigarette smoking. The theory of rational addiction is an audacious attempt to explain addictive behavior—an act that is almost always deemed irrational—within a conventional rational choice framework. It has received remarkably substantial empirical support on a wide variety of addictive behaviors. With respect to illicit drug use, such as marijuana and cocaine, future research is needed to explore whether youth are able to anticipate future prices accurately, how they acquire that information, and whether effects of future prices persist when controlling for other time-varying covariates, such as changes in the certainty of arrest, peer effects, and local supply of the drug.2 Nevertheless, these results allow us to apply forward-looking rational choice principles for addiction, crime, and conventional behaviors as a microfoundation for macrosociological theories.

The Micro-Macro Problem in Sociology Sociologists have long attempted to overcome the bifurcation of the discipline into separate subdisciplines of social psychology and social organization by identifying specific linkages between micro- and macrolevels of explanation (for example, Hechter 1983; Alexander, Giesen, Münch, and Smelser 1987; Huber 1991). Such linkages would presumably help overcome criticisms lodged at myopic theorizing and research operating at single levels. For example, structural theories—and the macrolevel research they stimulate— typically explain system outcomes based on causal mechanisms operating at the macrolevel, thus ignoring the role of individual actors. Such theories have been criticized for being crudely functionalist (a system outcome is explained by a system characteristic defined by its function), obviously teleological (a system outcome is explained by a system-level purpose), and unlikely to identify effective interventions to bring about positive social change (for example, Coleman 1990). Individual-level theories of purposive action—and the microlevel research they stimulate—explain individual outcomes based on causal mechanisms operating at the individual level, with macro outcomes assumed to be mere aggregations of such processes. These theories have been criticized for trivializing the role of social organization and oversimplifying the micro-macro problem.3 Macro-Level Context

Macro-Level Outcome

4

1

3 Micro-Level Predictor

2

Micro-Level Outcome

figure 8.1. Links between micro- and macro-level mechanisms. Source: Coleman 1990.

294 Ross L. Matsueda Among the many proffered solutions to the micro-macro problem (for example, Sawyer 2001), perhaps the most distinctive approach, outlined in a series of papers and chapters by Coleman (1983, 1986, 1990), specifies that macrolevel relationships are brought about by microlevel processes, and vice versa, through a series of micro-macro transitions. Figure 8.1 illustrates these relationships. Macrosocial theories focus on link 4 between a macrolevel context (for example, social structure) and a macrolevel outcome (for example, rates of crime). Microindividual theories focus on link 2 between a microlevel predictor (for example, human capital investment) and a microlevel outcome (for example, earnings). These two levels are connected by two cross-level linkages. Link 1, commonly investigated in sociological studies of individual behavior, shows how macrocontext (for example, social class) conditions individual attributes (such as human capital investments), which in turn produce microlevel outcomes (for example, earnings) through a microlevel theory (such as microeconomic theory). The other cross-level relationship, link 3, is less studied and more complicated. Here, individual outcomes combine to produce macrolevel outcomes (for example, social organization). Stated differently, the question becomes, “How are interdependencies formed among individual actors to organize action?” Here, Coleman uses the concept of emergence to show how “collective phenomena are collaboratively created by individuals yet are not reducible to individual action” (Sawyer 2001). For Coleman (1990: 5), emergence is tied to purpose in interaction: “The interaction among individuals is seen to result in emergent phenomena at the system level, that is, phenomena that were neither intended nor predicted by the individuals.” This allows for more complexity than the simple assumption, made by reductionists and some economists, that collective phenomena are merely the aggregations of individual actions. The ways in which individual purposive actions combine to create macrolevel outcomes vary by the complexity of the social organization being constituted and reconstituted. In the simplest case of bilateral exchange between two actors, an agreement or contract governing the exchange is the macrolevel outcome. In this case, the macro outcome is intended by the individuals. Bilateral exchange between two parties can also result in externalities, which are costs or benefits to third-party stakeholders—usually in the form of a public good—for which compensation is neither collected nor paid. Thus parties to the exchange do not necessarily reap all the costs or benefits of the transaction. This can be seen as an application of Merton’s concept (1936) of unanticipated consequences of action to exchange relationships.4 Externalities, which can be positive or negative, constitute the most elementary form of moving from individual action to system-level properties. Nevertheless, externalities may be the most prevalent micro-macro link in any society, and exemplify the notion of emergence. I will use this elementary form of building social organization to link individual rational choice to neighborhood social capital. Bilateral exchange can be generalized to multilateral exchange, such as a market, in which the system-level outcome is a set of prices. This is perhaps the prototypical micro-to-macro transition, because it demonstrates that certain outcomes (such as the exchange price of goods) cannot be reduced to an aggregation of individual behaviors, but rather entail a broader social

Rational Choice Research in Criminology

295

organization—in this case the organization of the market. Prices are an emergent explained by equilibrium theory, in which individual capital and preferences combine to produce equilibrium prices through competitive exchange. Another micro-macro link concerns authority and control. Individuals that trust others may give up rights of control of certain actions to those others. Such vesting of authority in others provides the basis for the emergence of social norms, an emergent property of social systems based on common interests of the individuals. Authority relations and norms governing those relations, of course, are key elements of hierarchical organizations, authority structures, and formal organizations. I will illustrate the micro-to-macro transition with examples drawn from recent criminological research. To frame these examples theoretically, I use Sutherland’s classical criminological concept of differential social organization.

Differential Social Organization and Crime Edwin Sutherland, perhaps the most important criminologist of the twentieth century, is best known for coining the term “white collar crime” and developing his individual-level learning theory of crime, differential association. Sutherland (1947) also developed the concept of differential social organization—a macrolevel counterpart to his individual-level theory—to explain the distribution of aggregate rates of crime: the crime rate of a group or society is determined by the extent to which that group or society is organized against crime versus organized in favor of crime. Sutherland, however, failed to expound on the macro portion of the theory, leaving the conception of organization empty of content, except by illustration. For example, organization against crime includes strong conventional institutions that inculcate conventional commitments in individuals, such as having a job, investing in education, owning a home; organization in favor of crime includes nefarious organizations such as delinquent gangs, professional theft rings, and criminal organizations like the Mafia. Clearly, the theory would be more powerful if the concrete content and causal mechanisms of such organization were specified explicitly.5 In the following sections I will attempt to specify such concrete causal mechanisms, drawing on rational choice theory as a microfoundation, and showing how social organization is built up by identifying micro-to-macro transitions.The next section specifies mechanisms of organization against crime using recent research on social capital and collective efficacy.This is followed by a discussion of organization in favor of crime using the protection racket of the Sicilian Mafia.

Social Capital, Collective Efficacy, and Organization against Crime social capital theory One of the most important recent theoretical innovations in the social sciences has been the development of the concept of social capital. The concept has been popularized by Putnam (1995, 2001), who defines social capital as elements of social organization, such as “networks, norms, and trust, that facilitate coordination and cooperation for mutual benefit,” and laments

296 Ross L. Matsueda the decline of civic participation and social capital in contemporary society. Similarly, Bourdieu (1986) defines social capital as the “aggregate of the actual or potential resources that are linked to possession of a durable network of more or less institutionalized relationships of mutual acquaintance and recognition— or in other words, to membership in a group—which provides each of its members with the backing of the collectively-owned capital, a ‘credential’ that entitles them to credit in the various senses of the word,” and shows how unequal access to social capital helps to reproduce social inequality. Perhaps the most rigorous and developed conceptualization of social capital is due to James Coleman (1988, 1990). A distinctive feature of Coleman’s version, which separates it from others, such as Putnam (2001) and Bourdieu (1986), is its explicit value in making the micro-to-macro transition. Indeed, in his early writings about the micro-macro problem, Coleman (1986, 1988) discusses the role of exchange relationships, authority relations, social norms, and information flows—all of which he later captures under the umbrella of social capital—as examples of the micro-to-macro transition. Coleman’s version of social capital builds on Granovetter’s argument (1985: 487) that purposive action of individuals is “embedded in concrete, ongoing, systems of social relations,” which generate interpersonal trust. Social capital is defined by two characteristics: it inheres in the structure of social relationships, and not within an individual, and it facilitates certain forms of purposive action (Coleman 1990: 302). From the standpoint of individuals, social capital is a resource that can be used by members of social systems to realize their interests. In this way, it is a capital asset, as is physical capital and human capital, although one that is much less tangible and not “owned” by individuals. From the vantage point of the social system, social capital is the stuff that binds individuals, the fundamental elements of social organization, the medium through which social structure facilitates purposive actions of individuals, and, as important, the medium through which those actions constitute and reconstitute that structure. In this way, social capital accounts for interdependencies among otherwise atomized individuals. More specifically, for Coleman (1990), social capital consists of four dimensions: (1) obligations and expectations, (2) informational potential, (3) norms and effective sanctions, and (4) authority relations. Obligations and expectations, or reciprocated exchange, constitute the most elemental form of social relationships. Actors seek to realize their interests by engaging in bilateral exchange with others—doing favors for each other, which is made possible by the norm of reciprocity and the existence of trust in the social system. When one actor does a favor for a second, the second actor is now indebted to the first who can call in the favor at a future date when it is needed to attain an important objective. Favors are unpaid obligations to be fulfilled at the time of one’s choosing. Social systems with dense social networks of outstanding obligations are said to be rich in social capital (ibid.). A second form of social capital is the information potential that inheres in social relationships and is principally transmitted interpersonally. Information can be used by individuals to facilitate purposive action. As Granovetter (1973) has argued, the form and utility of information may vary by the strength of social

Rational Choice Research in Criminology

297

relationships. Strong ties within a homogenous group lead to the circulation and recirculation of similar ideas and information. Weak ties between members of heterogeneous groups—so-called bridging ties—may expose members of each group to novel information and new ideas because the information is coming from dissimilar individuals occupying disparate roles. Such information can have more utility for certain purposive actions, such as finding a job. Other information derives from the media, a social institution. A third form of social capital consists of norms, which specify proper or improper conduct, and are enforced by sanctions (for an excellent overview, see Hechter and Opp 2001). Norms are needed when an externality affects a collection of individuals (third parties) similarly, and can be resolved by neither bilateral exchange between the perpetrator and the third parties, nor a market solution in which third parties purchase the right of control from the perpetrator (Coleman 1990). Like other forms of social capital, norms are properties of social structure and are more effective when structures are closed because enforcers of the norm can then coordinate their monitoring and sanctioning. A norm entails a transfer of control over behavior from an individual to a collective—thus, it is a form of multilateral control. Norms facilitate purposive action by coordinating otherwise atomized individual actions. The final form of social capital is authority relations, in which individuals transfer control of certain behavior to another individual, the authority, who now exercises power over the others. Weber’s notion (1978 [1921/22]: 241) of charismatic authority is a special case of an authority relation in which a single leader, endowed with “exceptional powers or qualities” is given control over the behaviors of many. Authority relations are the most elaborate form of social capital, and appear in large hierarchical structures and other complex forms of social organization. A number of empirical studies have examined the relationship between social capital and criminal and deviant behavior. For example, Rosenfeld, Messner, and Baumer (2001) use individual-level data from the General Social Survey to measure social capital with attitudes about trust, fairness, and being helpful, as well as voting behavior and membership in Elks clubs. They then aggregate the responses to the 99 GSS primary sampling units and, using a simultaneous equation model, find social capital to predict homicide rates. Using a different survey dataset of individuals within forty geographic areas, Messner, Baumer, and Rosenfeld (2004) find that aggregate measures of trust are negatively associated with homicide rates. Using survey data on Berlin youth, Hagan, Merkens, and Boehnke (1995) find that family and school social capital are negatively associated with right-wing extremism and school delinquency. Finally, Lederman, Loayza, and Menéndez (2002) use data from the World Values Survey (2002) and find (using instrumental variables to address simultaneity) that trust within the community is consistently negatively associated with rates of homicide across thirty-nine countries. These studies suggest that social capital may be important for the etiology of homicide and delinquent behavior.

298 Ross L. Matsueda neighborhood social capital, collective efficacy, and informal social control Perhaps the best application of social capital to crime has been carried out by Sampson and colleagues (for example, Sampson, Raudenbush, and Earls 1997; Sampson, Morenoff, and Earls 1999; Sampson and Raudenbush 1999). They merge the dimensions of social capital of Coleman (1990) with Bandura’s notion (1986, 1997) of “collective efficacy.” Bandura (1986: 391) is well known for his concept of self-efficacy, which he defines as “people’s judgments of their capabilities to organize and execute courses of action required to attain designated types of performances.” For Bandura, if the level of skill and opportunity are held constant, individuals who perceive a high degree of personal efficacy will outperform those with little self-efficacy because they can act with persistence, overcome obstacles, and capitalize on narrow opportunities. Self-efficacy is learned through self-observations of performance, vicarious observations of others, making social comparisons, and the like. The perceived efficacy of a group, a shared belief in acting collectively to achieve an objective, is not the mere sum of the individual personal efficacies of members. Instead, collective efficacy—members’ perceptions of the efficacy of the collectivity— will “influence what people do as a group, how much effort they put into it, and their staying power when group efforts fail to produce results” (ibid.: 449). Again, for Bandura, individuals’ perceptions of the group’s ability to “solve their problems and improve their lives through concerted effort” are more important than the objective ability of the group (ibid.). The insight made by Sampson and colleagues is to apply the concept of collective efficacy to neighborhood action, tie it to Coleman’s concept (1990) of social capital, and obtain operational indicators of it taken from previous neighborhood surveys (for example, Taylor 1996). Sampson, Raudenbush, and Earls (1997: 918) treat collective efficacy as a task-specific property of neighborhoods—namely, “the capacity of residents to control group level processes and visible signs of disorder,” which helps reduce “opportunities for interpersonal crime in a neighborhood.” This definition echoes the flip side of Shaw and McKay’s concept (1969 [1942]) of social disorganization, defined succinctly as “the ability of local communities to realize the common values of their residents or solve commonly experienced problems” (Bursik 1988: 521; Kornhauser 1978: 63). For Shaw and McKay, social disorganization is tied directly to the absence of local community institutions, organizations, and social ties. For Sampson et al., collective efficacy is tied directly to the presence of neighborhood social capital: “[It] is the linkage of mutual trust and the willingness to intervene for the common good that defines the neighborhood context of collective efficacy” (Sampson, Raudenbush, and Earls 1997: 919). Thus, collective efficacy translates the resource potential of neighborhood social networks—that is, social capital—into “active support and control of children” and thereby reduces the rate of youth crime (Sampson, Morenoff, and Earls 1999: 635). This formal definition of collective efficacy emphasizes the objective capacity of a neighborhood to intervene for the common good, rather than members’ perceptions of that capacity, as emphasized by Bandura (1986). Ironically, in operationalizing collective efficacy, Sampson et al. (1997; 1999) use measures of residents’ perceptions of collective efficacy.

Rational Choice Research in Criminology

299

In their empirical research, Sampson, Raudenbush, and Earls (1997: 919) treat collective efficacy as an objective characteristic of neighborhoods, emphasizing that “it is the linkage of mutual trust and the willingness to intervene for the common good that defines the neighborhood context of collective efficacy,” which results in informal social control. They identify two neighborhood-level concepts, “social cohesion and trust” and “informal social control,” that constitute collective efficacy, and collect measures of each using residents of 343 neighborhood clusters in Chicago—from the community survey of the Project on Human Development in Chicago Neighborhoods (PHDCN). For each construct, they use respondents as informants on the neighborhood characteristic, asking them, for example, “[A]re people in the neighborhood willing to help neighbors?” and “[Do] neighbors trust each other?” (cohesion and trust), and “[W]ould you agree that your neighbors could be counted on to intervene if children were skipping school and hanging out on street corners?” (informal social control). After combining the two constructs into a single collective efficacy variable, Sampson, Raudenbush, and Earls (1997) find that disadvantage, immigration, and residential mobility are associated with collective efficacy in the expected negative direction. They also find that collective efficacy is negatively associated with homicide and violent victimization, and to some extent mediates the effects on violence of neighborhood structural covariates. Sampson, Morenoff, and Earls (1999) use the PHDCN, but operationalize collective efficacy slightly differently, and examine spatial processes—spillover effects from one locale to another—across neighborhoods. They retain the concept of informal control, renaming it child-centered control, drop social cohesion and trust, and add two new neighborhood-level constructs: intergenerational closure (relationships among parents and children in the neighborhood) and reciprocated exchange (exchange of favors between neighbors). After combining the new constructs into a single index of adultchild exchange, they find the index to be positively associated with concentrated affluence and residential stability, and negatively associated with population density. Sampson, Morenoff, and Earls (ibid.) find child-centered social control to be positively associated with affluence and negatively associated with disadvantage, immigration, and population density. Finally, they find positive spatial effects: net of other covariates, a given neighborhood’s collective efficacy is positively associated with that of contiguous neighborhoods. Moreover, this effect is racially patterned: white neighborhoods disproportionately enjoy the advantage of spillover effects from surrounding high efficacy neighborhoods, while black neighborhoods are doubly disadvantaged, suffering from low average efficacy and the absence of surrounding efficacious neighborhoods. Using the PHDCN data, research has also found that collective efficacy is related to rates of violence. Morenoff, Sampson, and Raudenbush (2001) find that spatial proximity to neighborhoods with high homicide rates is strongly related to increased homicide rates. Concentrated disadvantage and low collective efficacy are also positively associated with homicide. Finally, Sampson and Raudenbush (1999) use the PHDCN dataset and find that collective efficacy is strongly related to homicide, burglary, and robbery. Moreover, they test the “broken windows” hypothesis of Kelling and Coles (1997), which argues

300 Ross L. Matsueda that physical disorder, or incivilities, such as graffiti, broken windows, and litter, directly induces crime by signaling to criminals that residents are indifferent to crime. Using a simultaneous equation model to control for reciprocal effects of disorder on collective efficacy, they find that the correlation between disorder and crime is spurious because of the confounding variable, collective efficacy. Therefore, they conclude that collective efficacy theory is supported over the broken windows hypothesis. Research on neighborhood collective efficacy is one of the best applications of social capital theory to a specific social problem. As a theory of neighborhood social organization, however, it operates exclusively at the macrosociological level, implicitly treating the neighborhood as a corporate actor, and ignoring— or at least remaining agnostic about—microlevel processes and potential links between individual actors and neighborhoods. An important theoretical task would specify a microlevel model of purposive action compatible with the macrolevel concept of collective efficacy. an individual-level model of investment in social capital Empirical studies of collective efficacy specify macrolevel neighborhood models that estimate macrorelationships, labeled link (4) in Figure 8.1 above (see Figure 8.2). This specification is appropriate and consistent with the conceptualization by Sampson, Raudenbush, and Earls (1997) of collective efficacy as a macrolevel (neighborhood) concept produced by macrostructures (community structural characteristics)—a position, of course, known as methodological holism, in which an internal analysis of social systems is eschewed in favor of identifying causal mechanisms at the system level. Adopting a position of methodological individualism, however, may provide a window for examining the collective action dynamics by which social capital is translated into collective efficacy. Moreover, as Coleman (1990) argues, there are distinct advantages to adopting a position of methodological individualism, in which macrolevel processes are linked to an internal analysis of the social system. From a theoretical standpoint, specifying an individual-level model of purposive action helps address the teleological problem in macrolevel theories, in which outcomes are explained by future states or purposes. In our case, collective efficacy theory may be vulnerable to the accusation that it is a functionalist explanation: the theory assumes consensus among residents in a desire for a safe neighborhood and argues that neighborhood collective efficacy functions to ensure a safe neighborhood. By treating consensus not as an assumption but as a goal that must be achieved by residents, by specifying purpose at the individual level, by allowing for unintended consequences of purposive action, and by explaining outcomes in terms of efficiency rather than final states, we Neighborhood Structure

Collective Efficacy 4

• • •

Child-Centered Social Control Intergenerational Closure Reciprocated Exchange

figure 8.2. Links between micro- and macro-level mechanisms. Source: Coleman 1990.

Rational Choice Research in Criminology Neighborhood Structure

301

Collective Efficacy

4

1

3 2

Individual Resources

Reciprocal Obligations

figure 8.3. Links between micro- and macro-level mechanisms. Source: Coleman 1990.

can move away from teleological explanation and vulnerability to functionalist critique. An individual-level model of purposive action eventuating in collective efficacy begins with a utility maximization model of neighborhood social interaction (see Figure 8.3). A number of economists, focusing on memberships in civic associations, have found that investments in social capital follow a standard economic investment model: individuals invest in social capital when there are private incentives to do so—such as home ownership, close spatial proximity, fewer opportunity costs for time, and complementarities (peers with more social capital). They theorize that aggregation is complex because of externalities, which can be positive (networks) or negative (status) (for example, Glaeser, Laibson, and Sacerdote 2002). Durlauf and Fafchamps (2005), in particular, have reviewed the economics of social capital literature and identified conditions under which social capital will increase Pareto optimality. I focus on the most elementary and fundamental form of social capital, social exchange—the practice of exchanging obligations and favors—and draw from the classic writings of Peter Blau (1964). Blau (ibid.: 91) defines social exchange as “voluntary actions of individuals that are motivated by the returns they are expected to bring and typically do in fact bring from others.” In contrast to economic exchange, in which a formal contract stipulates the precise nature and quantity to be exchanged, social exchange entails only a general expectation of future reciprocation, whose nature and quantity is left unspecified and open-ended.Whereas economic exchange is depersonalized by institutional rules and expectations, social exchange is personal, and “engenders feelings of personal obligation, gratitude, and trust” (ibid.: 94). Most favors contain an implicit promise to be repaid at some future date. Of course, as in all promises, there is extreme asymmetry of information over the promissory property of favors: the party receiving the favor knows much more about the likelihood of honoring it than the party giving it. We assume that individuals seek to maximize utility under constraints in asking for favors and doing favors for neighbors. Thus residents ask neighbors to borrow tools to facilitate achieving a goal of fixing a car, repairing a home, or shoveling snow. But other potential benefits may accrue, such as deriving pleasure from an enjoyable interaction, gaining social approval or a degree of respect, and

302 Ross L. Matsueda building solidarity with the neighborhood. Residents may ask their neighbors to watch their home or monitor their children when they are away. Here, residents are seeking assistance in protecting their property, a necessary goal in a context in which trust of others—particularly new acquaintances—is imperfect. Failure to reciprocate will produce distrust and eventually end the relationship; repeated reciprocation builds trust, commitment, and strengthens relationships.6 Repeated reciprocation within organized groups often produces norms of reciprocity, which include sanctions for failure to reciprocate.7 Finally, repeated reciprocal exchanges are subject to diminishing marginal utility: continual social exchange between the same pair of actors reduces the benefits each receives (Blau 1964). Why do rational actors do favors for neighbors when there is a risk of nonreciprocation, and even if reciprocated, the return favor will likely have the same or less utility as the initial favor? One rational reason pertains to the timing of the return favor. The initial favor creates an unspecified debt whose settlement is postponed. A rational actor can specify when the debt should be paid—for example, at a time when the actor is in dire straits, and the utility of the return favor is amplified (Coleman 1990). Thus, reciprocated exchange can be explained using a simple utility maximization model. But how does social exchange translate into social capital and collective efficacy? from reciprocated exchange to collective efficacy There are two intersecting neighborhood social systems relevant to the generation of social capital and collective efficacy. The first is a system that generates reciprocated exchange among neighbors, creating social capital; the second is a system translating social capital into collective efficacy, the capacity to solve local problems collectively. Each system entails links between micro- and macroprocesses, and illustrates how communication and consensus building can produce more efficient (collective) forms of purposive action. By describing the two systems, and their interrelations, I provide a picture of the obstacles facing neighborhoods in developing high levels of social capital, collective efficacy, and ultimately, personal safety. Creation of Social Ties. Let us begin with the creation and maintenance of social ties among residents. This is the most elementary form of social capital, described by Coleman (ibid.) as obligations and expectations among individuals, and Sampson, Raudenbush, and Earls (1997) as neighborhood reciprocated exchange. Start with a set of residents who engage in reciprocated exchange with their neighbors for their private instrumental purposes— borrowing tools to fix the plumbing, helping to pull out a tree, lending a hand to fix a car. Coleman (1990) points out that the resulting social ties among neighbors can have a positive externality for the neighborhood as a whole—the creation of neighborhood social capital. Once created, social capital becomes a resource available for individuals to facilitate purposive action, such as maintaining a safe neighborhood through informal social control. Neighborhoods rich in social capital (in the form of dense social ties) will have a large capacity to solve local problems—in other words, they will have collective efficacy (Sampson, Raudenbush, and Earls (1997). But how are those social ties increased and maintained over time? How are they

Rational Choice Research in Criminology

303

translated into the neighborhood’s capacity to accomplish shared goals, such as maintaining a safe neighborhood? Neighborhoods with residents who, individually, have a high propensity for interacting and doing favors for each other will enjoy a high degree of social capital in the form of reciprocal obligations and expectations. These social ties translate into social capital as a positive externality, which generally facilitates residents’ purposive actions, including individual instrumental behavior and collective behavior in behalf of the neighborhood as a whole. Some neighborhoods—and specifically within those neighborhoods, some key residents—may become aware of the relationship between dense social ties and the ability of neighborhoods to solve shared problems collectively.8 They may recognize that some residents are relatively isolated, and realize that if they were more involved, the neighborhood would be better off. Consequently, they have an incentive to encourage those isolated residents to become involved, and urge their neighborhood friends to encourage involvement as well. Over time, they may convert some neighbors with persuasion and rewards in the form of informal approval such as smiles, pats on the back, and kudos, while at the same time questioning, gossiping about, or even demeaning neighbors who remain isolated. In this way, social capital is increased in the neighborhood over and above the sum of effects of high individual propensities to interact. But in a neighborhood in which residents become aware of the link between neighborhood ties and neighborhood solidarity, some residents will realize that they can enjoy the fruits of neighborhood social capital—because it has a public goods aspect—and not contribute to such ties. In the parlance of rational choice theory, they have an incentive to free ride on the actions of others. To reduce the number of free riders, residents might provide selective incentives, such as informal approval or disdain, and even coordinate sanctioning in pairs, which would be facilitated by social ties between pair members (Olson 1971). A more efficient way of eliciting compliance than the use of selective incentives by relatively unorganized individuals would be to create a norm—a general rule backed by collective sanctions—prescribing being “neighborly.” Such a norm necessitates building a working consensus over the value of being neighborly, the transfer of control from individual residents to the neighborhood as a whole, and the appropriate sanctions for violators.9 This consensus, in turn, requires communication and social ties.Thus, neighborhoods in which a critical mass of residents have developed social ties for personal instrumental reasons would have the social capital necessary to facilitate creation of more social capital through creating norms of being neighborly. This may follow a threshold model, in which a critical mass of social ties is needed to communicate and create consensus over a norm of neighborhood participation. Social capital, then, builds upon itself: social ties created for one purpose provide positive externalities facilitating the creation of new forms of social capital, which create more social ties. The norm requires group members to enforce the norm by sanctioning, which can entail a cost, particularly if the sanction is negative. Here, rational actors will again have an incentive to free ride, relying on neighbors to sanction norm-violators, without contributing to sanctioning themselves. To overcome this problem—the second-order public

304 Ross L. Matsueda goods problem—residents might use only positive, relatively costless sanctions, such as informal approval.10 Creation of Collective Efficacy. The existence of neighborhood social ties is a prerequisite for residents to act collectively to combat youth crime and incivilities. Youth crime, which violates essential conjoint norms, may be rational from the standpoint of youth, but also provides negative externalities for local residents.11 For example, vandalism may upset the victimized home owner, but also reduce the attractiveness and consequently property values of the neighborhood as a whole. Residents can respond by attempting to intervene in isolation, confronting the youth, scaring the youth off, or threatening to call the police. Such monitoring and sanctioning entails a cost—youth could fight back, retaliate, or threaten, and even calls to the police take time and energy. Isolated acts of intervention require the individual to shoulder the entire cost of intervening, including investing time and energy, absorbing opportunity costs, and facing potential retaliation or unpleasant interactions with the offender. Regardless of the self-efficacy of the individual, the probability of intervention is probably low because of its high cost. When the negative externalities affect multiple residents similarly—for example, costs such as creating an unsafe environment for children, reducing property values, or inducing fear and anxiety—the potential for a collective response e