Aggressive Political Participation 9781400870455

Table of contents :
Cover
Contents
Acknowledgments
1. What is to be Explained and How
2. Explanations of Aggressive Political Participation
3. Measurement of Aggressive Political Participation
4. The Expectancy-Value- Norms Theory
5. Frustration-Aggression Theory
6. Left-Out Variables
7. Cross-Validity of the Expectancy-Value-Norms Model
8. Uses and Limitations of the Expectancy-Value- Norms Model
9. Macro-Micro Linkages
Appendices
Index

Citation preview

Aggressive Political Participation

EDWARD Ν. MULLER

Aggressive Political Participation

PRINCETON UNIVERSITY PRESS PRINCETON. NEW JERSEY

Copyright © 1979 by Princeton University Press Published by Princeton University Press, Princeton, NewJersey In the United Kingdom: Princeton University Press, Guildford, Surrey All Rights Reserved Library of Congress Cataloging in Publication Data will be found on the last printed page of this book Clothbound editions of Princeton University Press books are printed on acid-free paper, and binding materials are chosen for strength and durability. Printed in the United States of America by Princeton University Press, Princeton, NewJersey

To MADELON

CONTENTS

ACKNOWLEDGMENTS

ONE Two

THREE

FOUR

FIVE

What Is To Be Explained and How

3

1.1.

7

Research Design

Explanations of Aggressive Political Participation

11

2.1. 2.2. 2.3. 2.4. 2.5.

15 18 20 23 31

Deprivation and Frustration Change in Group Control of Power Resources Utilitarian Incentive for Aggression An Expectancy-Value-Norms Theory Testing the Explanations

Measurement of Aggressive Political Participation

37

3.1. 3.2.

Zones of Aggressive Political Participation Summary

54 66

The Expectancy-Value-Norms Theory

69

4.1.. 4.2. 4.3. 4.4. 4.5. 4.6.

70 79 95 100 110 116

The Expectancy-Value Concept Personal Normative Beliefs Social Norms Availability for Collective Action Testing the Expectancy-Value-Norms Model Summary

Frustration-Aggression Theory

121

5.1. 5.2.

138

5.3. 5.4. 5.5.

Six

IX

Measurement of Frustration Frustration and Aggressive Political Participation The Role of Frustration in the ExpectancyValue-Norms Model Rank Disequilibrium and Socioeconomic Disadvantage Summary

Left-Out Variables 6.1. 6.2.

Specific Political Performance Dissatisfaction Internal-External Control of Reinforcement

150 164 167 180

183 184 207

6.3. 6.4. 6.5.

SEVEN

EIGHT

NINE

RejectionofIndividualism Personal and Political Resources Summary and Discussion

213 219 226

Cross-Validity of the Expectancy-ValueNorms Model

236

7.1.

240

DirectionofCausality

Uses and Limitations of the ExpectancyValue-Norms Model

244

8.1. 8.2.

246 257

Projections from the Model The Model as a Diagnostic Tool

Macro-Micro Linkages

264

9.1.

277

Coda: On the Effectiveness of Repression

APPENDICES A. Comparison of Political Trust-Distrust and Political Support-Alienation as Predictors of Aggressive Political Participation B. Summary Characteristics of the Variables C. GermanTextoftheVariables

280 286 289

INDEX

303

ACKNOWLEDGMENTS

I am particularly grateful to Rudolf Wildenmann for his sage advice and continual encouragement since our initial discus­ sions about this research while we were colleagues at the State University of New York at Stony Brook in 1972. The project, entitled "Gesellschaftliche und Politische Indikatoren fur U nterstutzung / Opposition, Zufriedenheit / Unzufriedenheit und Beherrschung/Machtlosigkeit," was funded by the Deut­ sche Forschungsgemeinschaft; I am naturally grateful to the DFG, for awarding the sizable grant that made the work pos­ sible, and to the project sponsors who administered the grant from the University of Mannheim, Professors Wildenmann and Wolfgang Hirsch-Weber. In addition to the DFG, two other institutions that provided financial support are deserving of thanks: the University of Mannheim, which invited me to take up a visiting appointment in the summer of 1972, at which time the research proposed was written; and the Ford Founda­ tion, from which I received a Faculty Research Fellowship for the academic year 1972-1973 that enabled me to remain at the University of Mannheim and begin wprk on the pilot study phase of the project. Quite a number of individuals contributed to this study in important ways. Jonathan Pool of the University of Washington (formerly at the State University of New York at Stony Brook) participated as a co-investigator in the pilot study, carried out in 1973-1974, when many versions of the interview scheduled were proposed, tested, and revised. Without his unusual lin­ guistic talents, the final version of the interview schedule would have been much the poorer. Sampling, interviewing, and preparation of the data tapes was carried out under the very able direction of Yola Laupheimer and Dorothea Reppart of Infratest, Miinchen. The quality control exercised by them over the "guts" of the study was first-rate. Helpful research assistance during the pilot study was provided by Walter Wehrli of the

Zentrum fur Umfragen, Methoden and Analysen (ZUMA). Hans-Jiirgen Hippler, Inge Kostka, and Silke Wollweber assisted in the research at ZUMA during the main phase of the project. The director of ZUMA, Max Kaase, is deserving of special thanks for making the excellent resources of that insti­ tution available to me during the many summers that I spent in Mannheim working on the analysis of the data. Last but not least, I am indebted to TomJukam of the State University of New York at Stony Brook for the give and take of many lengthy discussions that benefited this book; and to those who read all or parts of it in one version or another, including Ray DuVall of the University of Minnesota, Vladimir Konecni of the University of California, San Diego, Bill Linehan and Joe Tanenhaus of the State University of New York at Stony Brook, Erich Weede and Uli Widmaier of the University of Mannheim, and John Wahlke of the University of Iowa.

Aggressive Political Participation

ONE

What Is To Be Explained and How

This book reports an effort to formulate and test a general multivariate theory of individual participation in acts of politi­ cal aggression. Due to the infrequency of its occurrence, as well as difficulties of measurement, this kind of political behavior has rarely been subjected to rigorous scientific investigation at the micro level of analysis. Yet, despite the fact that aggres­ sive political participation is an unusual event under normal circumstances in most countries, on occasion it can have dramatic consequences, contributing to major change in, or to the downfall of, established systems of government. Thus, it is worthwhile to try to understand the causes of aggressive political participation. It is sometimes thought that the causes of aggressive political participation can be uncovered simply by studying charac­ teristics of persons who actually have engaged in such action. Of course, this is a fallacy. To understand why some people engage in aggressive political action, one also must know why other persons do not take part. Causal analysis therefore re­ quires a research design that captures variation in the phe­ nomenon to be explained as well as in the putative causal agents. The empirical base for this study consists of data gathered from structured interviews with representative samples of per­ sons residing in selected rural, urban, and university milieus in the Federal Republic of Germany. The intent of the research design was to elicit variation in individual attitudes and be­ havior sufficient for reliable multivariate analysis, without at

4 I What Is to be Explained and How

the same time unduly sacrificing representativeness according to basic social and economic variables such as education, in­ come, age, sex, marital status, and place of residence. Because of the focus on general theory, this book does not offer the richness of detail and colorful narrative that often characterize historical and journalistic accounts of political protest, violence, and revolution. But my emphasis here on formal models, mathematical reasoning and statistical evidence has the compensating feature that it affords, at least potentially, an understanding of aggressive political participation that is not limited to particular, historically unique cases. The inten­ tion is to develop an explanation of aggressive political par­ ticipation per se, not just aggressive political participation in West Germany. Since the data come from a large and heter­ ogenous sample of persons residing in various communities of West Germany, the findings should be generalizable to other nations having similar macro social, economic, and political characteristics: the subset of nations called "advanced indus­ trialized democracies." Of course, it would be desirable to carry out replications in one or more of the other such democra­ cies in order actually to determine this. In the meantime, West Germany serves as the first test case. 1 The dependent variable of this study, aggressive political participation, is a subset of collective political participation, where collective political participation may be defined as behavior intended to influence authorities, engaged in by groups of persons who do not themselves occupy positions of authority in a political system. Collective political participa­ tion occurs in a profusion of concrete manifestations. In societies with regularly scheduled, competitive elections, collective political participation is focused on the electoral process: groups of people proselytize for parties and candidates, attend meet­ ings, rallies, and demonstrations, perform a variety of tasks to 1 Although I assume that West Germany is as suitable as any of the other industrialized democracies (excepting, perhaps, Italy) for testing the theory, its selection as the first test case is not due to any special scientific value of the site but rather to idiosyncratic reasons—mainly the interest shown in the re­ search by Professors Wildenmann and Hirsch-Weber of the University of Mannheim.

What Is to be Explained and How / 5

help a party or candidate during a campaign, become fullfledged members of political parties or clubs. This is politics as usual, "ordinary" political participation. 2 But the domain of collective political participation is much broader than this. It includes many forms of extraordinary or unconventional behavior: participating in illegal strikes, seizing public build­ ings, battling with police or with other demonstrators, becom­ ing involved with a group that wants to dislodge the govern­ ment by violent means. To be sure, unless a rebellion is brewing or a society actually is undergoing a revolution, only a rather small minority of the population ever engages in unconven­ tional political behavior. Yet, as recent history attests, uncon­ ventional action can ebb and flow dramatically even in polities such as polyarchies that supposedly are quite stable. 3 To under­ stand collective political participation in general, one must necessarily consider the unconventional as well as the conven­ tional, extraordinary as well as ordinary forms of political participation. There is another compelling reason for coming to grips with unconventional action. It is "strong" as opposed to "weak" political behavior in the sense that John Wahlke and Bertrand deJouvenel have used these labels. 4 "Strong" political behavior is action that, at least potentially, can have dangerous conse­ quences for a regime, threatening its stability or persistence. The resort to unconventional action signifies that, to some extent, institutionalized channels of representation and con­ flict resolution may be inadequate or defective. Of course, in 2 The most comprehensive conceptualization and measurement of this sub­ set of political behavior appears in Sidney Verba and Norman H. Nie, Par­ ticipation in America (New York: Harper and Row, 1972). 3 1 prefer to use the neutral term, "polyarchy," coined by Robert Dahl as a replacement for "democracy," as the latter term has too many diverse ideo­ logical connotations to be useful in social scientific discourse. A polyarchy is a political system characterized by both free electoral competition to contest the government and widespread voting rights. See Robert A. Dahl 3 Polyarchy (New Haven: Yale University Press, 1971). 4 John C. Wahlke, "Policy Demands and System Support: The Role of the Represented," British Journal of Political Science, 1 (July 1971), p. 282; Bertrand de Jouvenel, "On the Nature of Political Science," American Political Science Review, 55 (September 1961), p. 777.

6 / What Is to be Explained and How

and of itself, "strong" political behavior is just a warning signal. But in the longer run the probability that a regime will stand, change, or fall certainly is linked to the incidence of unconventional action. Hence to get at the question of why political systems stand, change, or fall, one will want to have an understanding of what causes variation in the incidence of "strong" political behavior. It should be emphasized that the labels "strong" and "weak" are not to be taken in any pejorative sense. "Strong" political behavior is neither better nor worse than "weak" behavior— indeed, from the perspective of those committed to a given regime, "weak" political behavior is always the preferred mode of interest representation and conflict resolution. Furthermore, what constitutes "strong" or "weak" behavior will vary with regime type. In inclusive hegemonies, to follow Robert Dahl's terminology, 5 where the right to vote is widespread but elec­ toral competition to contest the government is not allowed, many campaign activities that are ordinary in polyarchies be­ come "strong" political behavior; by contrast, such campaign activities are "weak" behavior in polyarchies. Let us turn now to a more precise definition of the kind of "strong" political behavior under scrutiny in this study. Draw­ ing on Douglas Hibbs' definition of mass political violence, 6 aggressive political participation will be defined as behavior that possesses these properties: (1) it must be anti-regime in the sense of deviating from legal or formal regime norms re­ garding political participation, that is, it must be political action that is illegal; (2) it must have political significance, that is, it must be an attempt to influence the government that inconveniences it or disrupts its normal functioning; (3) it must involve group activity on the part of non-elites. Aggres­ sive political participation, by this definition, may or may not involve violence. If it does not involve violence, it will be called civil disobedience, as distinguished from political violence. Excluded 5

Dahl 5 Polyarchy, pp. 7-8. A. Hibbs, Jr., Mass Political Violence (New York: Wiley, 1973), p. 7. 6 DougIas

What Is to be Explained and How / 7

by this definition are—in addition to conventional electoral politics—legal protest actions such as boycotts; ordinary labor strikes without political objectives; and individualistic actions such as refusal of military service and assassinations. Coups d'etat by dissident factions of the military are excluded be­ cause they involve intra-elite conflict. Also excluded is violence initiated by agents of the government to repress dissident groups, since, although collective in nature, this normally (though not always) is legally sanctioned aggressive action, and by definition is behavior engaged in by elites.7

1.1

RESEARCH DESIGN

What kinds of objective measurement procedures are feasible for the study of aggressive political participation at the micro level? Experimentation under controlled laboratory conditions raises ethical problems.8 This leaves non-experimental field research in natural settings. There are two ways objectively to

7 In Eckstein's terminology, my focus here is on insurgents instead of in­ cumbents—see Harry Eckstein, "On the Etiology of Internal Wars," in Anger, Violence, and Politics, ed. Ivo K. Feierabend, Rosalind L. Feierabend, and Ted Robert Gurr (Englewood Cliffs, N.J.: Prentice-Hall, 1972), pp. 17-18; ori­ ginally published in History and Theory, 4 (no. 2, 1965), 133-162. Certainly, as he argues, to predict the likelihood of internal war it is necessary to consider characteristics of incumbents as well as insurgents. Characteristics of incum­ bents will be taken up in the last chapter. 8 Psychologists have sidestepped the ethical constraints on aggression re­ search by developing a "shock" paradigm for the experimental study of inter­ personal aggression in the laboratory. This involves use of an "aggression machine" that, so the subject of the experiment is told, will deliver electric shocks of varying intensity to a "victim" (really a confederate of the experi­ menter) seated in an adjacent room. The experiments usually are conducted under the guise of studying the effect of punishment on learning, wherein the subject, or "teacher," is instructed to shock the confederate, or "learner," every time a wrong response is made in a learning task. Those who administer especially painful shocks or press the shock button for especially long periods of time are considered to be acting aggressively. Of course, unknown to the subject, the shock electrodes attached to the victim are inoperative. For a dis­ cussion of this paradigm, see Albert Bandura, Aggression: A Social Learning Analysis (Englewood ClifFs, N.J.: Prentice-Hall, 1973), pp. 120-139. The shock paradigm would appear to have limited applicability to the study of collective political aggression.

8 / What Is to be Explained and How

measure participation in aggressive political action by means of field research, but only one would appear generally practical. One method maximizes objectivity by having a trained corps of research personnel closely monitor the behavior of individuals over a period of time. This method has been used, for example, to study aggressive behavior among adolescents attending a summer camp.9 However, to study aggressive political behavior among adults, this direct observation method would not only usually be impractical but, more importantly, would constitute an unconscionable invasion of privacy. The alternative is to rely on indirect evidence gathered from individuals' self-reports of their behavior, given through self-completion questionnaires or personal interviews. The data used here come from personal interviews carried out with 2,662 adults in the Federal Republic of Germany during the fall of 1974 by Infratest, 49 percent of whom were reinterviewed in the fall of 1976.10 The interview protocol, averag­ ing slightly over 60 minutes to complete, represents the culmina­ tion of a research program begun in the United States in 1968. The attitude and behavior measures are instruments that have emerged from a process of trial-and-error testing. In the course of three separate studies, two done in the United States, one in the Federal Republic, various ways of operationalizing the attitudes and behavior of concern here were explored.11 On the basis of this research, and attention to the results from 9 See Muzafer Sherif and Carolyn W. Sherif, Groups in Harmony and Tension (New York: Harper and Row, 1953). 10 The interview schedule was prepared by the author and Jonathan Pool Helpful advice was received from many colleagues at the University of Mann­ heim, especially Rudolf Wildenmann and Uwe Schleth, and from Yola Laupheimer, director of the fieldwork, and Dorothea Reppart, assistant di­ rector, both of the Economics Research Bureau, Infratest GmbH & Co., Munich. 11 Reports of findings from these studies include Edward N. Muller, "Corre­ lates and Consequences of Beliefs in the Legitimacy of Regime Structures," Midwest Journal of Political Science, 14 (August 1970), 392-412; Muller, "The Representation of Citizens by Political Authorities: Consequences for Regime Support," American Political Science Review, 64 (December 1970), 1149-1166; Muller, "A Test of a Partial Theory of Potential for Political Violence," American Political Science Review, 66 (September 1972), 928—959; Bernard N.

What Is to be Explained and How / 9

parallel studies conducted by others,12 those instruments which appeared to be the most promising were selected for inclusion in the interview protocol. There were twelve sampling sites in all, four rural, two urban, and six university communities.13 Each was selected because, in the aggregate, opposition to the regime had been manifested there during the preceding five years at higher than average levels. In the rural and urban sites, opposition had taken the form of voting support for extreme left and extreme right political parties; in the universities it had taken the form of civil disobedience and political violence. Two major considerations of the research design were (1) to elicit variation in individual attitudes and behavior sufficient for reliable multivariate analysis and (2) to investigate the Grofman and Edward N. MulIer j "The Strange Case of Relative Gratification and Potential for Political Violence: The V-Curve Hypothesis," American Political Science Review, 67 (June 1973), 514—539; Muller, "Behavioral Corre­ lates of Political Support," American Political Science Review, 71 (June 1977), 454-467. 12 I have benefited especially from interaction with colleagues at several conferences and meetings of the Research Group on Political Alienation and Support, coordinated by Jack Dennis of the University of Wisconsin, Madison. 13 The rural sites were four small villages: Friederichskoog and Neuenkirchen in northern Germany, Erpolzheim and Mauchenheim in southern Germany. From these sites a total of 569 persons was interviewed, of which 479 were drawn randomly from lists of eligible voters, and 90 were drawn from lists of community influentials obtained from discussions with the mayor and other community leaders by the chief of the team of interviewers for each site. The urban sites were working-class sections of Bremen in northern Germany and Niirnberg in southern Germany. From these sites a total of 990 persons were interviewed, of which 928 were drawn randomly from lists of eligible voters, and 62 were drawn from lists of community influentials compiled from nominations submitted by persons in the eligible voters sample who were active in local organizations. The university sites were six of the major West German universities: Berlin, Bochum, Frankfurt, Heidelberg, Koln, and Miinchen. A total of 1,104 students and faculty from the arts and sciences at these universites was interviewed, of which 956 were drawn by quota sampling, and 148 were drawn from lists of influential persons in various university organizations com­ piled from nominations submitted by persons in the quota sample. Quota sampling was used in the universities because, in previous studies of students, it had been difficult to acquire a proper sampling frame. A detailed report of the sampling procedure (in German) is available from Dorothea Reppart, Infratest, Wirtschaftsforschung, 8 Munchen 21, Landsberger Strasse 338, Federal Republic of Germany.

10 I What Is to be explained and How

effect of community context on relationships between attitudinal variables and behavior. An additional consideration was to avoid completely sacrificing representativeness at the altar of enhanced variation. While the communities chosen are by no means representative of West Germany as a whole, they do cap­ ture basic regional and community-size differences.

TWO

Explanations of Aggressive Political Participation

The study of aggressive domestic political conflict has a vener­ able tradition in political science and sociology, and has yielded an abundance of explanatory propositions. But these propositions have as yet borne little fruit in the form of reliable knowledge about what it is that motivates men to take part in aggressive political action. Fundamentally, this is a problem of research methodology. As Eckstein pointed out some years ago in a seminal article surveying problems and prospects of re­ search in the area of political violence and rebellion, the methodological problem is that "most propositions about the causes of internal war have been developed in historical studies of particular cases (or very limited numbers of cases) rather than in broadly comparative, let alone genuinely socialscientific studies."1 And since the single case or handful of cases can prove nothing about behavior in general, these propositions have the status of untested hypotheses. In the ensuing decade, a number of scholars sought to remedy the methodological weakness inherent in the historical, casestudy approach. They collected data on various societal con­ ditions and on rates of collective protest and violence for as many nations as possible, then carried out scientifically rigor­ ous, quantitative analyses of the etiology of domestic civil 1 Eckstein, "On the Etiology of Internal Wars," pp. 11-12 (full citation in footnote 7 of Chapter One).

12 I Explanations of APP

strife.2 However, by the nature of their unit of analysis (the nation-state as a single entity), such macro level, cross-national studies have been limited to the testing of what Eckstein terms "structural" hypotheses: "A structural hypothesis singles out, so to speak, 'objective' social conditions as crucial for the oc­ currence of internal war: aspects of a society's 'setting,' such as economic conditions, social stratification and mobility, or geo­ graphic and demographic factors."3 And this unavoidable focus on structural variables entails a cost, namely, exclusion of indi­ vidual motivational variables from playing a direct role in the explanation of aggressive political participation. On what grounds would it be tenable to exclude individual motivational variables? Alternatively, why might this be re­ garded as a cost? To answer these questions, it is useful to con­ sider the anatomy of a structural hypothesis, for there are really two parts to any such hypothesis invoked to account for aggres­ sive political behavior among men. The first part states the objective social and economic conditions that function as ante­ cedents of aggressive political participation. The second part identifies a rationale or causal mechanism for the objective conditions by assuming that the relevant objective conditions are connected closely to certain attitudes of individuals provid­ ing explicit motivation for the behavior. If there is an almost 2 See, among others, Ivo K. Feierabend, Rosalind L. Feierabend, and Betty A. Nesvold, "Social Change and Political Violence: Cross-National Patterns," in Violence in America: Historical and Comparative Perspectives, ed. Hugh Davis Graham and Ted Robert Gurr (New York: Signet Books, 1969), 606-670; William H. Flanigan and Edwin Fogelman, "Patterns of Political Violence in Comparative Historical Perspective," Comparative Politics, 3 (October 1970), 1-20; Ted Robert Gurr 5 "A Causal Model of Civil Strife: A Comparative Analysis Using New Indices," American Political Science Review, 62 (December 1968), 1104-1124; Ted Robert Gurr and Raymond Duvall, "Civil Conflict in the 1960s: A Reciprocal Theoretical System with Parameter Estimates," Com­ parative Political Studies, 6 (July 1973), 135-169; Hibbs, Mass Political Violence (New York: Wiley, 1973); Michael C. Hudson, "Conditions of Political Vio­ lence and Instability: A Preliminary Test of Three Hypotheses," Sage Pro­ fessional Papers in Comparative Politics, Series Number 01-005, 1971; Bruce M. Russet, "Inequality and Instability: The Relation of Land Tenure to Politics," World Politics, 16 (April 1964), 442-454. 3 Eckstein, "On the Etiology of Internal Wars," pp. 18-19.

Explanations of APP / 13

perfect correlation between structural variables and attitudes, of course there is no need to pay special attention to the attitudinal part of the schema, since knowledge about attitudinal variables would not markedly increase one's ability to predict the occurrence of aggressive political behavior. Under this assumption, it is tenable to exclude attitudes. Objective condi­ tions serve as the bones and muscle of the explanation; attitudes merely flesh it out. Exclusion of subjective variables from direct consideration becomes a cost as soon as one questions the assumption of a very high correlation between objective conditions and sub­ jective reaction to such conditions. If objective conditions may often be uncorrelated and at best only imperfectly correlated with attitudes that motivate individual participation in political protest and violence, then it is likely to be rather difficult to establish general laws for prediction of the occurrence and magnitude of aggressive political participation on the basis of knowledge about objective social and economic conditions alone. As an example, take the Rise and Drop or J-Curve hypothesis formulated by James Davies, widely regarded as one of the more intuitively plausible hypotheses about the causes of revo­ lution in particular and of aggressive political participation in general. This hypothesis states that violent civil disturbances such as rebellion and revolution "are most likely to occur when a prolonged period of objective economic and social develop­ ment is followed by a short period of sharp reversal."4 The causal mechanism of the Rise and Drop hypothesis is the assumption that an improvement trend produces a rise in sub­ jective expectation of need satisfaction. When a sharp deteriora­ tion trend follows hard upon the heels of an improvement trend, expectation of need satisfaction will remain high while perception of actual need satisfaction will decline apace with deteriorating objective conditions. The end result is a sizable 4 James C. Davies, "Toward a Theory of Revolution," American Sociological Review, 27 (February 1962), p. 6.

14 I Explanations of APP

gap between subjective expectation of need satisfaction and subjective experience of need satisfaction, a gap postulated to be sufficiently intolerable or frustrating to motivate men to take aggressive political action. Davies cites impressionistic historical evidence to support the Rise and Drop hypothesis—Dorr's rebellion in nineteenthcentury America, the Russian revolution of 1917, the Egyptian revolution of 1952.5 But the relative calm of the industrialized nations in the mid-19 70's appears to fly in the face of the hypoth­ esis. After World War II, these nations experienced steady improvement in economic and social conditions until the oil embargo instituted by OPEC in late 1973; then a sharp reversal began. Yet the highest rates of violent civil conflict coincided with the zenith of the improvement trend, reached during the late 1960's and early 1970's, while the deterioration trend has coincided with decreasing rates of violent civil conflict—exactly the opposite of what the hypothesis predicts. It could well be that an intolerable gap between subjective expectation and achievement is a motivational variable that generally disposes men to take aggressive political action, regardless of the par­ ticular time and place. But if this motivational variable were neither always nor closely correlated with objective improvement-followed-by-deterioration trends, then, as seems to be the case, the structural conditions referred to in the hypothesis would not be systematically related to aggressive political par­ ticipation. Hence, one could not reliably predict the occurrence of aggressive political behavior on the basis of a Rise and Drop trend. The primary emphasis in this book is on the individual as the unit of analysis. This does entail a cost. With a shift of atten­ tion from whole nations to individual men, cross-national generalizability immediately becomes an open question. But research at the micro level of analysis, carried out on heterog­ eneous samples of persons from even a small number of coun­ tries can, in principle, yield universal or at worst semi-universal 5

Ibid., passim.

Explanations of APP / 15

laws of behavior.6 Of course, demonstration of the universality of putative micro-level laws of behavior requires that the rela­ tionships found to hold across individuals sampled from any one population also hold up, at the minimum, in at least one other population. The findings reported here should be re­ garded as preliminary evidence, the result of the testing of hypotheses in one instance, but in need of further validity generalization studies before qualifying as confirmed scientific knowledge. Focusing on the micro-level does not mean that objective social and economic conditions are ignored. They are regarded as predetermined variables, that is, factors that are not caused by any of the variables postulated to determine individual aggressive participation, though they may themselves affect some of these variables. We now review some of the major approaches to the explana­ tion of individual differences in level of aggressive political participation, and attempt to specify a general micro-theory of the causes of such behavior. Is it possible to identify certain attributes of individuals that, when present, make the occur­ rence of aggressive political behavior highly probable, and when absent, highly improbable? Or is such behavior largely a re­ sponse either to fortuitious circumstances or to macro-level influences beyond the scope of this research design? 2.1

DEPRIVATION AND FRUSTRATION

What might be the likely preconditions (if there are any) of aggressive political participation? Adverse social circumstances, by themselves, are not a promising candidate. Sometimes the poor revolt, but most of the time they do not. Hardship and economic exploitation do not necessarily lead to an aggressive response; unemployment, inflation, and a host of other objec6 For a good discussion of this and related topics in the area of comparative hypothesis-testing, see Adam Przeworski and Henry Tenune, The Logic of Comparative Social Inquiry (New York: Wiley, 1970).

16 I Explanations oMPP tive privations are as likely (perhaps more likely) to be asso­ ciated with political inactivity as with activity. Empitical tests of hypotheses asserting a relationship between objective de­ privation and aggressive political participation generally have found that such behavior occurs as often under conditions of objective well-being as under conditions of objective hard­ ship. 7 But if the preconditions of aggressive political participa­ tion are not to be found in hardship, then where is one to look? One could look to relative hardship, to a person's perception of hardship in relation to his expectations. Perception of rela­ tive hardship need not correspond to objective conditions. Many people in a state of objective deprivation adjust their expectations to their situation: they come to feel that what they are getting is what they are entitled to, that low socio­ economic rewards are simply their just deserts. On the other hand, persons who are objectively well-off are not necessarily satisfied with what they are getting. Some may develop expec­ tations that are much higher than their actual achievements; they may come to feel that even their relatively great socio­ economic rewards are not at the level they deserve. Relative hardship could be an important precondition of aggressive political participation even if objective hardship is not, so long as the two are not closely correlated. Ted Gurr has developed the most comprehensive theory linking relative hardship to aggressive political participation.8 7 See the review of micro-level research on participation in civil disorders given in Clark McPhail 3 "Civil Disorder Participation: A Critical Examina­ tion of Recent Research," American Sociological Review 36 (December 1971), 1058-1073; the review of macro-level research on political protest and violence given in Erich Weede5 iiUnzufriedenheit, Protest und Gewalt: Kritik an einem makropolitischen Forschungsprogram " Politisehe Vierteljahressehrift i 16 (Heft 3, 1975), 409-428; the analysis of determinants of leftist radicalism in Alejandro Portes, "On the Logic of Post-Factum Explanations: The Hypothesis of Lower-Class Frustration as the Cause of Leftist Radicalism." Social Forces, 50 (September 1971), 26—44; and the historical analysis of protest and violence in France, Italy, and Germany in Charles Tilly, Louise Tilly, and Richard Tilly, The Rebellious Century (Cambridge: Harvard University Press, 1975). 8 Ted Robert Gurr, Why Men Rebel (Princeton: Princeton University Press 5 1970).

Explanations of APP / 11

His basic premise is that the necessary precondition of aggres­ sive political participation is relative deprivation, where rela­ tive deprivation is defined as actors' perception of discrepancy between their value expectations and their value capabilities. Value expec­ tations are the goods and conditions of life to which people believe they are rightfully entitled. Value cap­ abilities are the goods and conditions they think they are capable of getting and keeping. . . . The emphasis . . . is on the perception of deprivation; people may be sub­ jectively deprived with reference to their expectations even though an objective observer might not judge them to be in want.9 The motivational rationale (or causal mechanism) that con­ nects relative deprivation to aggressive political participation is frustration. Relative deprivation is postulated to be a frus­ trating psychological condition. According to the general theory of aggression formulated by John Dollard and his colleagues and subsequently clarified by Neal Miller, man is motivated to engage in aggressive action by a frustration-induced "drive."10 Frustration does not necessarily result in aggression; but the occurrence of aggression presupposes the existence of frustra­ tion. In Gurr's application of frustration-aggression theory to specifically political aggression, the occurrence of political aggression presupposes relative deprivation; but the effect of relative deprivation on political aggression is inhibited or am­ plified depending on the level of normative justification for aggression, utilitarian justification for aggression, and the bal­ ance of power between dissident forces and the agents of social

9

Ibid., p. 24.

10 John

Dollard, Leonard W. Doob, Neal E. Miller, O. H. Mowrer, and Robert R. Sears, Frustration and Aggression (New Haven: Yale University Press, 1939). Also see Neal E. Miller, "The Frustration-Aggession Hypothesis," Psychological Review, 48 (1941), 337-342,

18

I

Explanations of

APP

control. The theory is expressed by the following equation: (2.1)

MPV = RD + {RD* JUST* BALANCE) + ε,

where the magnitude of political violence is denoted by MPV, relative deprivation by RD, the sum of normative and utilitar­ ian justifications for violence by JUST, and the balance of organizational and coercive capacities between dissidents and the government by BALANCE. (The ε symbol is an error term representing variation in MPV that cannot be explained by the variables in the model.) This model postulates (1) that relative deprivation has a causal effect on magnitude of political vio­ lence regardless of normative and utilitarian justification and regardless of the balance of power between dissidents and agents of social control; (2) that justifications and balance of power have no effect on magnitude of political violence independent of relative deprivation; but that, (3) when combined multiplicatively with relative deprivation, justification and balance do affect magnitude of political violence.11 2.2

CHANGE IN GROUP CONTROL OF POWER RESOURCES

Gurr's Relative Deprivation theory is basically a psychological explanation; the single most important cause, and two of the three causes, entail subjective motivation for participation in aggressive political action. Subjective motivation in general, and relative deprivation in particular, are rejected by scholars who have advocated an approach that focuses on group control of power resources.12 According to this view, aggressive politi­ cal participation does not stem primarily from psychological motivation, but rather from situations in which a dissident or 11 This is Gurr's most compact statement of his theory. See Gurr and Duvall, "Civil Conflictin the 1960's," p. 137. 12 Especially Anthony Oberschall, Social Conflict and Social Movements (Englewood Cliffs, N.J.: Prentice-Hall, 1973); David Snyder and Charles Tilly, "Hardship and Collective Violence in France, 1830 to 1960," American Socio­ logical Review, 37 (October 1972), 520-532; Charles Tilly, "Does Moderniza­ tion Breed Revolution?" Comparative Politics, 5 (April 1973), 425-447; Tilly, Tilly, and Tilly, The Rebellious Century.

Explanations of APP / 19

challenging group is able to mobilize sufficient power resources to make claims on values, privileges, and resources controlled by the government, and these claims are then resisted. The laying of claims by groups and their rejection by the govern­ ment are a normal part of the political process, and aggressive political participation is seen to be simply a byproduct of this process. As formulated by David Snyder and Charles Tilly, the major hypothesis of the Control of Power Resources ap­ proach is that "collective violence results from changes in the relations between groups of men and the major concentrations of coercive power in their environments."13 More specifically, Tilly has proposed a curvilinear relationship between aggres­ sive political participation and the degree to which groups laying claim to values controlled by the government are losing or acquiring power resources.14 The relationship can be dia­ grammed as: Change in Group Control of Power Resources Aggressive Political Participation

Negative

High

n%

Low

(100-n)

Minimal Change 0% 100

Positive n%

(100 - n )

Tilley postulates that groups that are either gaining or losing power resources "have a special propensity to articulate strongly moral definitions of their situations."15 This means that, in the normal bargaining process of politics, such groups should be especially likely to resist compromising their claims—with the result that they will be especially prone to aggressive conflict with agents of the government. Note, however, that the hypothesis about change in control of power resources predicts either high or low aggressive political 13

.

Snyder and Tilly, ibid., p. 520. Tilly, "Does Modernization Breed Revolution?" esp. pp. 437-439. 15 Ibid., p. 438. 14

20 / Explanations of APP

participation as a result of gain or loss of power resources. Such positive or negative change is postulated to be a necessery pre­ condition of aggressive political participation, but it is not also a sufficient condition. Groups gaining and losing power re­ sources are predicted to engage in some form of political action; but whether this action takes an aggressive turn is "a contingent outcome of interactions among contenders and governments, in which the agents of government commonly have the greater discretion and do most of the injury and damage." 16 The Control of Power Resources approach is thus a theory about the conditions that make aggressive political participation extremely unlikely to occur; it does not identify conditions that make the occurrence of such behavior extremely likely.17 The Control of Power Resources explanation is cast exclu­ sively at the macro level of analysis. Implied in this approach are the predictions that, at the micro level of analysis, if any attributes of individuals are at all systematically associated with participation in aggressive political action, first, no indi­ vidual attributes should be anywhere close to necessary and sufficient conditions of such participation; second, the most likely candidates for necessary but not sufficient conditions should be objective attributes such as an individual's avail­ ability for collective action of any kind; and third, general availability for collective action should override the effects of any subjective psychological variables, especially relative de­ privation or other indicators of frustration.

2.3

UTILITARIAN INCENTIVE FOR AGGRESSION

The Control of Power Resources approach regards motiva­ tional concepts such as relative deprivation as either theoreti16 17

Ibid., p. 439.

Tilly does propose a set of conditions that lead to political revolution, but it is unclear whether these conditions represent variables that can actually be measured and used to predict the likelihood of revolution or whether they are intended more just to summarize temporally distinct sequences or stages in the development of revolution. See Charles Tilly, "Revolutions and Collective Violence," in Handbook of Political Science, vol. 3, ed. Fred I. Greenstein and Nelson W. Polsby (Reading, Mass.: Addison-Wesley, 1975).

Explanations of APP / 21

cally unnecessary or empirically unsupported. But the question of empirical support is still open, since no decisive evidence has yet been gathered bearing on the hypothesis that participation in aggressive political action will vary strongly with relative deprivation in the sense specifically of perceived discrepancy between entitlement and achievement. A persuasive case for the theoretical necessity of integrating relative deprivation into the Control of Power Resources approach has been made by Walter Korpi on the grounds that "any theory of conflict that focuses on mobilization of power resources will at least implicitly have to come to grips with motivational concepts like relative deprivation, since motivational factors are generally accorded a central place in theories of mobilization."18 Korpi's concern is the development of a model to explain why people become mobilized for conflict entailing political aggression. Following the general theory of motivation formu­ lated by John W. Atkinson, he proposes that the probability of an actor A being mobilized for conflict with another actor B is a multiplicative function of actor A's expectancy of success and the utility of his reaching the goal.19 The concept of utility, or value of the goal to the actor, is considered difficult to measure, but is assumed to vary inversely with actor A's expected costs of reaching the goal and to vary directly with the degree of relative deprivation that actor A experiences with respect to the goal. Korpi regards change in the control of power resources as an indirect cause of mobilization for conflict. Change in control of power resources is seen as linked to conflict via its effect on the three intervening motivational variables: expectancy of success, expected costs, and relative deprivation. Korpi hy­ pothesizes that a positive change or gain in power resources by an actor is (1) positively related to his expectancy of success

18 Walter Korpi, "Conflict, Power and Relative Deprivation," American Political Science Review, 68 (December 1974), p. 1570. 19 Atkinson's theory is given in John W. Atkinson, An Introduction to Motiva­ tion (Princeton, N.J.: Van Nostrand 5 1964). Korpi's theory also draws upon Hubert M. Blalock, Toward a Theory of Minority Group Relations (New York: Wiley, 1967).

22 I Explanations of APP

in a potential conflict; (2) negatively related to his expected costs in a potential conflict; and (3) positively related to his sense of relative deprivation (positive change is assumed to increase an actor's level of aspiration, that is, what he believes he is justifiably entitled to). Korpi's model thus differs from the Control of Power Re­ sources approach in two key respects. First, and most impor­ tantly, subjective attributes of individuals—psychological motivation—are crucial intervening variables that relate changes in control of power resources to the probability of conflict between groups in general, and between dissidents and agents of the government in particular. The psychological motivation variables are direct preconditions or causes of con­ flict because they provide the impetus or "drive" for an actor to become mobilized for conflict. Second, according to Korpi's model, only positive change or gain in power resources results (via the mediation of motivational variables) in an increased probability of conflict; negative change or a loss of power re­ sources is predicted to result in a decreasing probability of con­ flict because it decreases the motivational impetus stemming from expectancy of success, expected costs, and relative depriva­ tion. In simplified form, Korpi's model looks like this:20

Degree of Gain in Power Resources of A (Weaker Person or Group)

A's Expectancy of Success Expected Costs to A A's Relative Deprivation

Utility to A

Probability of Conflict between A and B

The direct antecedents of aggressive political participation in Korpi's model are variables that justify aggressive action on 20 Arrows denote causal effects; the arrow stemming from expectancy of success and utility denotes a multiplicative interaction.

Explanations of APP / 23

utilitarian grounds. His UtilitarianJustification approach syn­ thesizes the Relative Deprivation and Control of Power Re­ sources approaches, casting them as determinants of the variables that afford utilitarian incentive for aggressive political participation.

2.4

AN EXPECTANCY-VALUE-NORMS THEORY

The essential issue raised by these approaches is the question of whether aggressive political participation results primarily from group influences operating on the individual regardless of his attitudes, or primarily from individual attitudes the deter­ mination of which may be subject to group influences. The controversy boils down fundamentally to the question of atti­ tudes and behavior: is there a causal link between attitudes of individuals and their participation in acts of political aggres­ sion? And if so, how strong is this causal link? Strong enough to be of real predictive value? Orjust statistically but not sub­ stantively significant ? Empirical research on the relationship between attitudes of individuals and their overt behavior has generally turned up statistically significant but weak correlations. A comprehensive review of such studies by A. Wicker concluded that "only rarely can as much as 10 percent of the variance in overt behavioral measures be accounted for by attitudinal data."21 The studies covered by Wicker dealt with a variety of attitude objects, ranging from concrete things such as jobs and ethnic groups to concrete actions such as civil rights activities, cheating, student political activity, and breast feeding. The vast majority of these studies, however, have followed a simple bivariate model: they have attempted to relate a person's evaluative feelings (pro or con, favorable or unfavorable) about an object to his behavior toward that object. 21 A. W. Wicker, "Attitudes versus Actions: The Relationship of Verbal and Overt Behavioral Responses to Attitude Objects," in Attitudes and Be­ haviour, ed. Kerry Thomas (Harmondsworth, England: Penguin Books, 1971), p. 161; originally published in Journal of Social Issues, 25 (1969), 41-78.

24 I Explanations of APP

Failure to find strong confirmation of the bivariate attitudebehavior model has led many investigators to question the measure of attitude used. One of the more extensive attempts to come to grips with the measurement question is a study carried out by C. R. Tittle and R. J. Hill.22 To a sample of 301 upper-division undergraduate students they administered five different measures of attitude toward participation in student political activities: a Likert scale, a Guttman scale, a self-rating scale, a semantic-differential scale, and a Thurstone scale. Also, they obtained five different measures of overt participation in various types of student political activity. The results confirmed the expectation that how an attitude was measured would make a difference in the degree of correlation between attitude and behavior. The attitude measure with the greatest reliability, the Likert scale, showed an average corre­ lation of 0.54 with the behavior measures. By contrast, the Thurstone scale, which showed the least reliability, also was the worst predictor of behavior, showing an average correla­ tion with the behavior measures of only 0.26. This "measurement improvement" approach has been challenged by Martin Fishbein on the grounds that, within the confines of the bivariate model, it is an insufficient answer to the problem of low attitude-behavior correlation.23 Fishbein is one of a number of scholars who have stressed a multivariate approach and, as noted by Wicker, "although a number of fac­ tors in addition to attitudes have been suggested as influences upon overt behaviors, Fishbein is the only writer who has at­ tempted to combine several factors into a systematic formula­ tion."24 Fishbein begins by rejecting the assumption that a person's attitude toward an object is the major determinant of his behavior toward that object. Instead, he argues that atti2 2 C. R. Tittle and R. J. Hill, "Attitude Measurement and Prediction of Behavior: An Evaluation of Conditions and Measurement Techniques," Sociometry, 30 (1967), 199-213. 2 3 Martin Fishbein, "Attitude and the Prediction of Behavior," in Readings in Attitude Theory and Measurement, ed. Martin Fishbein (New York: Wiley, 1967), 477-492. 24

Wicker, "Attitudes versus Actions," p. 172.

Explanations of APP / 25

tude about the behavior itself is the more important, though not the only, determinant: Rather than viewing attitude toward a stimulus object as a major determinant of behavior with respect to that ob­ ject, the theory identifies three kinds of variables that function as the basic determinants of behavior: (1) atti­ tudes toward the behavior; (2) normative beliefs (both personal and social); and (3) motivation to comply with with the norms. Furthermore, although the theory sug­ gests that other variables can also influence behavior, it indicates that these other variables operate indirectly, that is, by influencing one or more of the basic determinants.25 Attitudes about behavior are defined as the individual's beliefs about the consequences of his behavior multiplied by their subjective value or utility to him. Normative beliefs refer to an individual's own belief in the justifiability of his behavior as well as to his perception of significant others' (parents, peers) expectations about it. Motivation to comply with the norms reflects such factors as an individual's personality and his per­ ception of the reasonableness of expectations of others. Fishbein's initial formulation of the theory was expressed in two equations: (2.2)

A act = Σ B i U i I= 1

(2.3)

BEHA VIOR = [A a c t ]w 0 + [(JiB p ) (MC p )]w\ + [(NB s ) (MC s )]w r

The first equation (2.2) is Fishbein's expectancy-value model of attitudes.26 It defines ^act, or attitude about an action, as the

Fishbein, "Attitude and the Prediction of Behavior," p. 490. A useful discussion of this model and how it relates to other conceptualiza­ tions of attitude appears in Bobby J. Calder and Michael Ross, Attitudes and Behavior (Morristown, N.J.: General Learning Press, 1973), esp. pp. 20-25. 25

2 6

26 I ExpIanationso - MPP

sum of an individual's beliefs about the consequences of per­ forming an action (Bi) multiplied by his evaluation of the consequences (a{). The second equation (2.3) is his theory of behavior.27 Here, Aact is as defined in equation 2.2; NBp de­ notes personal normative beliefs; MCp denotes motivation to comply with personal normative beliefs; NBs denotes social normative beliefs, that is, perceived expectations of others; MCs denotes motivation to comply with social normative beliefs; and w 0 , W 1 , and W 2 are empirically determined weights that express the degree to which a variable on the right side of the equation affects the behavior variable. According to this theory, the generally weak attitude-behavior correlations found in social-psychological research are due not simply to measure­ ment inadequacy but, more importantly, to the fact that such research has been misguided, has concentrated on an indirect cause of behavior—attitude about the object—instead of focus­ ing on the direct determinants—Aact, (NBp)(MCp), and (NBs) (MCs). Only when attitude about the action and norma­ tive beliefs are taken into account can one expect to obtain accurate prediction of behavior. Fishbein and his colleague, I. Ajzen, have carried out some interesting tests of the model. Ajzen found encouragingly high multiple correlations between behavior (cooperative versus competitive) in the Prisoner's Dilemma game and measures of attitude about the behavior and social normative beliefs.28 Ajzen and Fishbein also found consistently high multiple corre­ lations between intention to perform a wide variety of behaviors and attitude about behavior, personal normative beliefs, and social normative beliefs.29 In general, normative beliefs carry 27 Fishbein differentiates between behavior and behavioral intention, but argues that under appropriate conditions they are approximately the same, i.e., B ~ Bi. See Fishbein, "Attitude and the Prediction of Behavior," pp. 489-490. 28 I. Ajzen, " Attitudinal versus Normative Messages: An Investigation of the Differential Effects of Persuasive Communications on Behavior," Sociometry , 34 (1971), 262-280. 29 I. Ajzen and M. Fishbein, "The Prediction of Behavioral intentions in a Choice Situation," Journal of Experimental Social Psychology, 5 (1969), 400—416.

Explanations of

APP

/ 27

greater predictive weight than attitude about behavior, al­ though both usually make important independent contributions to predictive accuracy. Because of difficulties in operationally differentiating between normative beliefs and motivation to comply with normative beliefs, motivation to comply usually has not been found to be important. In some instances, multi­ plying normative beliefs by motivation to comply does not yield significant improvements in the correlation with the behavior variable; in other instances it even reduces the correlation. Since one can assume that an individual normally is motivated to comply with his normative beliefs, Ajzen and Fishbein have eliminated the MCp component in their revised formulation of the model:30 (2.4)

BEHAVIOR =[A 0 C t ]w ^[JVB p Iw

Fishbein's general theory of behavior, as expressed by equation 2.4, will serve as the basis for the theory of aggressive political participation proposed here. Fishbein's general behavior theory subsumes some of the variables in Gurr's Relative Deprivation theory and Korpi's Utilitarian justification theory. Take the general Aact term, attitude about the action. The B component of Aact (see equa­ tion 2.2), denoting beliefs about the consequences of performing an act, is similar to the expectancy of success variable in Korpi's theory, since expectancy of success refers to subjective expec­ tancies concerning the probability of successful outcomes. And, as Ajzen and Fishbein note, the a component of Aact (see equa­ tion 2.2), the evaluative aspect, is meant to correspond directly to the concept of utility or subjective value of an action.31 Since Aact is defined as the product of B times a, it follows that Aact is quite similar to the multiplicative interaction of expec­ tancy of success and utility in Korpi's model. Turning to Gurr's theory, recall that the justification for violence variable con­ sists of two components, utilitarian and normative. Although 30 31

Ibid., pp: 400-402. Ibid., p. 402.

28 I Explanations of

APP

Gurr has not defined the concept of utilitarian justification in precise mathematical terms, it would appear from the context of his discussion of micro-level referents of utilitarian justifica­ tion that its intended meaning is similar to the meaning of A 32 - ClCt · The personal normative beliefs term ( N B p ) in Fishbein's theory is similar to the concept of normative justification for violence, the other component of the justification variable in Gurr's theory.33 Korpi's theory does not include a normative beliefs term. This is because it is an application of Atkinson's motivation theory, which is quite similar to the SEU (Subjec­ tively Expected Utility) decision-theory model elaborated by W. Edwards, neither of which makes use of the concept of normative beliefs as a general determinant of behavior.34 Fishbein's theory is distinct from the utility models precisely in its emphasis on normative beliefs as an important determinant of behavior; and the claim of superiority for Fishbein's theory rests on the assumption that normative incentive to perform behavior will account for much of the variance in actual be­ havior independent of utilitarian incentive.3 5 A unique feature of Fishbein's theory, not taken into account in either the Relative Deprivation or Utilitarian Justification theories, is the social normative beliefs term (MBs) as multi­ plied by motivation to comply (MCs). This variable introduces social context explicitly into the explanation of individual behavior. Because individuals do not perform in a social vacuum, it would seem quite important that any general be­ havior theory be sensitive to the context of social normative beliefs to which the individual is exposed. Indeed, it seems eminently plausible to expect that, for most people, exposure rl

32 33

Gurr l Why Men Rebel, pp. 210-223. Ibid., pp. 160-192.

34 See W. Edwards, "The Prediction of Decisions Among Bets," Journal of Experimental Psychology y 50 (1955), 201-214; W. Edwards, "The Theory of Decision-Making," Psychological Bulletin, 51 (1954), 380-417. A useful review of variants of decision theory appears in Ν. T. Feather, "Subjective Probability and Decision Under Uncertainty," Psychological Review, 66 (1959), 150-164. 35 Ajzen and Fishbein, "The Prediction of Behavioral Intentions," p. 415.

Explanations of APP / 29

to social norms favoring or disfavoring a given behavior will serve as an impetus to perform or not perform the behavior, above and beyond motivational incentive stemming from purely personal normative and utilitarian beliefs. Fishbein's theory, of course, is quite abstract. There are many ways one might give concrete meaning to the terms in equation 2.4. The adaptation of this equation to the phe­ nomenon of aggressive political participation is shown in a model for what I shall call the Expectancy-Value-Norms theory of aggressive political participation: (2.5)

APP = a + b 0 (E* V) +- b ,(MBA p ) + b 2 {FSN*MCJ + β,

where APP = degree of participation in aggressive political action by an individual; E = the degree to which an individual expects that aggressive political action will have beneficial consequences; V = the value or utility of aggressive political action

to an individual; JVBA p = the degree to which an individual regards ag­ gressive political action as justifiable on the basis of personal normative beliefs; FSJV = the degree to which social normative beliefs to which the individual is exposed are facilitative (favor) of aggressive political action; MC s n = the degree to which an individual is motivated to comply with social norms; b Q , b v b 2 = empirically determined weights that express the degree to which ^i3Pis expected to change, given a unit change in the describing variable;36 36 Note, these are unstandarized weights, in contrast to the standardized weights specified by Fishbein in his models. Standardized weights are useful because they are based on the same unit of measurement in the response and describing variables and hence always range between — 1 and +1. Therefore 5 they provide a means of comparing the size of the effects on the response variable of describing variables expressed in different original units of measure-

30 I Explanations of APP

a = the expected APP score when all of the predictor variables are equal to zero; ε = a stochastic disturbance or "error" term repre­ senting the assumption that APP will not be exactly determined by the variables in the model. The task of converting the terms in the Expectancy-ValueNorms model into measured variables for the purpose of esti­ mating the parameters of equation 2.5 will be taken up in the next two chapters. Prior to that, it is useful to consider alterna­ tive specifications of the Expectancy-Value-Norms model. Equation 2.5 says that participation in aggressive political action is determined (1) in part by an individual's utilitarian motivation to perform such behavior (Ε* V), regardless of his personal normative motivation (JVBAp) and regardless of his exposure to, and motivation to comply with, facilitative social norms; (2) in part by an individual's personal normative motivation, regardless of the other two variables; and (3) in part by the influence of facilitative social norms, regardless of the other two variables. The model entails the assumption that no predictor variable is highly collinear with any other predictor variable, as well as the assumption of additivity of effects—that the effect due to each of the predictors is uniform or constant, regardless of the particular value of the other predictors. An equally plausible alternative specification assigns a more central role to facilitative social norms and questions the

ment. Standardized weights are given by the formula w = b ( s x j s y ) , where w is the standardized weight, b is the unstandardized weight, s x is the standard deviation of the describing variable, and s y is the standard deviation of the response variable. An undesirable feature of standardized weights arises from the fact that they are afunction of the variability in the .*· and/ variables. If the ratio of the variability in * to the variability in y were to change substantially from one sample to another, the size of the standardized weight would be altered, even though the unstandardized weight remained the same. Thus, if one is interested in stating an estimate of a general law expected to hold across different samples (and populations), it is preferable to use unstandardized weights.

Explanations of APP / 31

additivity of effects assumption. This version of the ExpectancyValue-Norms theory can be expressed as: (2.6)

APP=a + c 0 [{E*V)*{FSN*MCJ) + c l [MBA p *{FSM*MCJ] + ε.

According to equation 2.6, the variables providing personal motivational incentive on utilitarian (E*V) and normative (NBAp) grounds will not affect participation in aggressive political action independent of facilitative social norms. Such norms will serve to inhibit or amplify the effect on APP due to E*V and NBAp. When social norms disfavor participation in aggressive political action, the individual will be unlikely to take such action even if E* V and NBAp are highly favorable. Likewise, when Ε* Vand NBAp are not favorable, the individual will be unlikely to take part in aggressive action, even if social norms are favorable. But if social norms favor participation in aggressive political action, and if E*V and NBAp also are favorable, then the individual will be especially likely to engage in such action. The model given in equation 2.6 still presents utilitarian in­ centive and personal normative incentive as having separate, additive effects on participation in aggressive political action, albeit in interaction with facilitative social norms. One could eliminate the additivity of effects assumption entirely by speci­ fying a model of the form: (2.7)

APP = a + c 0 [{E*V)*NBA p *(FSN*MCJ] + e.

However, this complete non-additivity model begins to move well away from Fishbein's theory, the distinctiveness of which derives partly from the assumption that normative beliefs are an important determinant of behavior independent of utilitarian incentive. 2.5

TESTING THE EXPLANATIONS

The Expectancy-Value-Norms, Utilitarian Justification, and Relative Deprivation theories postulate psychological attributes

32 / Explanations of APP

o f individuals as the direct antecedents of aggressive political

participation. Thus they can be subjected to empirical tests with the present research design. The Control of Power Re­ sources theory cannot be tested here because it regards attri­ butes of groups, not individuals, as central to the determination of mass political aggression. However, if support is found for the Expectancy-Value-Norms, Utilitarian Justification, or Relative Deprivation theories, then the assumption of the Con­ trol of Power Resources theory, that attributes of individuals are irrelevant to the explanation of aggressive participation, at least can be called seriously into question. In this case, there would be justification for provisionally relegating change in group control of power resources to the status of a predeter­ mined macro variable that possibly plays an indirect role in the determination of aggressive participation through effects on individual psychological antecedents—precisely the role given to it by Korpi. In testing the Expectancy-Value-Norms, Utilitarian Justifi­ cation, and Relative Deprivation theories, it is useful to classify the terms in the models for these theories into three categories: explanatory variables, irrelevant variables, and left-out vari­ ables. An explanatory variable is one that occurs in an explana­ tory model for a given response variable and is estimated to have a direct effect on that response variable. An irrelevant variable is one that occurs in an explanatory model for a given response variable and is estimated to have no direct effect on that response variable. A left-out variable is estimated to have a direct effect on a given response variable that is unspecified in the explanatory model for that response variable. Models for the Expectancy-Value-Norms theory were speci­ fied by equations 2.5 and 2.6. All terms on the right-hand side of those equations are explanatory variables. The Relative Deprivation theory, based on frustration-aggression theory, is both an expansion and a modification of the Expectancy-ValueNorms theory. It says that the Expectancy-Value-Norms theory must be (1) supplemented by incorporation of the con­ cept of relative deprivation as an additional independent

Explanations of APP / 33

determinant of aggressive political participation and (2) modified to reflect the paramount status of relative deprivation (or frustration) as the principal cause of aggressive participa­ tion. The modification entails an interaction between relative deprivation and utilitarian as well as normative justification such that the effect of E*V and NBAp on aggressive political participation is contingent upon the presence of high levels of relative deprivation. (It should be noted that balance of power between dissidents and regime, the BALANCE term in equa­ tion 2.1, is a variable intended to be observed at the group or societal level and to help account for the overall magnitude of political violence in a society; hence it is eliminated from this micro-level specification of the Relative Deprivation model.) The Relative Deprivation model can be expressed as: (2.8)

APP = a + b 0 ( R D ) + c 0 [ ( R D ) (E*V)] + c j (RD s f NBA p ) + ε,

where RD = amount of relative deprivation experienced by an individual and the other terms are as defined in equation 2.5. According to the model for the Relative Deprivation theory, all the terms on the right-hand side of equation 2.8 are left-out variables in the Expectancy-Value-Norms model. In addition, the FSN*MCm term from the Expectancy-Value-Norms model is an irrelevant variable. The Utilitarian Justification theory is a contraction of the Expectancy-Value-Norms theory. It singles out the expectancyvalue concept (Ε* V), utilitarian incentive for aggression, as the dominant direct determinant of aggressive political participa­ tion and can be expressed as: (2.9)

APP = a + b 0 (E*V) + ε.

According to the model for the UtilitarianJustification theory, all the terms involving normative beliefs (NBA p and FSN* MC s n ) from the Expectancy-Value-Norms model are irrelevant vari­ ables. We now have a set of explicit models for prediction of indi­ vidual differences in level of participation in aggressive political

34 I Explanations of APP

action. They express the hypotheses central to three alternative theories about general micro-level preconditions of aggressive political participation. The Relative Deprivation theory em­ phasizes a certain kind of frustration (experience of discrepancy between value expectations and value capabilities) as the most important direct determinant of aggressive participation; the Utilitarian Justification theory singles out utilitarian incentive stemming from expectancy that political aggression will have beneficial consequences weighted by the value of politically aggressive action; the Expectancy-Value-Norms theory stresses the role of normative beliefs in the justifiability of politically aggressive action. The models for the three micro-level motivational theories will be tested in the fourth and fifth chapters.37 First, of course, careful attention must be given to the development of a satis­ factory quantitative indicator of aggressive political participa­ tion. This will be the topic of Chapter Three. In Chapter Four, operational indicators of the terms in the Expectancy-Value-Norms model will be defined so as to test its predictions. For the model to be supported, the indicator of 37 Proper testing of these models requires quantitative (interval scale) variables. Although the data from the personal interviews are, strictly speaking, categorically (ordinal scale), numerical values are assigned to all variables in this analysis. The distinction between ordinal and interval measurement often is difficult to make in practice, and should not be followed to the point where information in the data is lost that otherwise could be uncovered by the ju­ dicious assignment of numbers to ordered categories and the use of statistical techniques that effectively exploit the properties of numbers. (An intelligent discussion of this point appears in Edward R. Tufte 5 "Improving Data Analysis in Political Science," World Politics , 21 [July 1969], 641-654.) Because of the advantages of quantitative measurement, social scientists dealing with attitudes and behavior do generally assign numbers to the categories of their variables so that these can be treated as if they were interval scales. (The psychometric rationale for this procedure, particularly as related to the summative model for scaling people with regard to psychological traits, is presented ably in Jum C. Nunnally, Psychometric Theory [New York: McGraw-Hill, 1967]). This practice can be defended on the grounds that, when the number of categories is relatively large, the difference between ordinal and interval measurement often becomes miniscule. The assignment of numbers to ordered categories helps to uncover additional information in the data, since substantially more powerful statistical techniques can be used to analyze it.

Explanations of APP / 35

each term on the right-hand side of equation 2.5 or 2.6 must be estimated to have a significant effect on the APP variable, that is, these terms must turn out to be explanatory variables. By contrast, for the UtilitarianJustification model to be supported, only the Ε* V term from equation 2.5 should be an explanatory variable. Chapter Five will focus on the model for the Relative Depri­ vation theory. Since the notion of relative deprivation is open to quite a variety of different operational interpretations, special emphasis will be given to the analysis of alternative indicators of that concept. Then a test will be carried out to determine if the terms from equation 2.8 do indeed represent left-out var­ iables from the perspective of the Expectancy-Value-Norms and UtilitarianJustification models. In addition, this chapter will consider the role played by objective indicators of depriva­ tion and frustration in the determination of aggressive political participation. Chapter Six will take up a number of psychological variables not included in the three general motivational theories: indi­ cators of dissatisfaction with specific political performance (evaluation of the policy performance of the incumbent ad­ ministration, evaluation of treatment received from public officials), internal-external control of reinforcement, and rejec­ tion of belief in individual self-sufficiency. These will be investi­ gated in order to determine if any of them function as left-out variables, calling for revision of the motivational model found to hold on the basis of considering only the terms in the Expec­ tancy-Value-Norms, Utilitarian Justification, and Relative Deprivation theories. Also to be investigated in Chapter Six are the roles of certain social background characteristics— personal and political resources such as level of education, organizational involvement, and general interest in politics— which might turn out to be left-out variables that Function as "facilitators" of aggressive political participation through in­ teraction with belief in the normative justifiability of political aggression. The theory-testing phase of this analysis will be completed in

36 I Explanations of APP

Chapter Seven, where the question of the validity of the moti­ vational model will be considered from the important perspec­ tive of whether the particular explanatory model found to hold at one point in time is replicable in subsequent testing.

THREE

Measurement of Aggressive Political Participation

Our first priority is to develop a measure of aggressive participa­ tion. Aggressive political behavior has been defined as action that is illegal, has political significance by virtue of disrupting the normal functioning of government, and involves group activity on the part of non-elites. This is a sensitive topic and its measurement by means of personal interviews poses a challenge. In this study, it was made clear to the interviewee that the research was being carried out under academic auspices for purely scientific purposes. Over and above the usual assurance of anonymity, Infratest guaranteed respondents from university communities that their university affiliations would be held in strict confidence. In addition to these safeguards, the questions themselves were presented in what was hoped would be a nonthreatening manner. Respondents were given a deck of ten cards. A specific political action was listed on each card. Five were aggressive, five were not. The interviewer explained that the items listed on the cards should be considered as a set of possible actions for exerting influence on the government. Respondents were asked: (1) whether or not they approved of each behavior; (2) in their view, how large a percentage of citizens in the Federal Republic would approve of each behavior; (3) whether or not they per­ sonally would engage in each behavior; and (4) whether or not they had done each behavior. The cards were shuffled before each question. For the last one, dealing with actual participa-

TABLE 3.1. Response to the Various

Political Behavior Items

Response Refused to

Have Not

Answer

Done

Have Done

Behavior Items

(JV = )

%

(JV = )

%

(JV = )

%

Participation in a petition-signing campaign

(149)

5.6

(1247)

46.9

(1266)

47.5

Sacrifice of time in order to work for a political party or a candidate in an elec­ tion campaign

(173)

6.4

(1986)

74.6

(505)

19.0

Refusal of military service

(251)

9.4

(2250)

84.5

(161)

6.1

* Seizure of factories, offices, and other buildings

(206)

7.7

(2364)

88.8

(92)

3.5

* Refusal to pay rent, taxes

(199)

7.5

(2299)

86.4

(164)

6.1

* Participation in fights (battle with police, battle with other demonstrators)

(209)

7.8

(2366)

88.9

(87)

3.3

Participation in a legally permitted political demon­ stration

(172)

6.4

(1590)

59.7

(900)

33.8

* Participation in a group that wants to dislodge the govern­ ment by violent means

(208)

7.8

(2420)

90.9

(34)

1.3

Measurement of APP / 39

TABLE 3.1. ( continued)

Refused to Answer

Have Mot Done

Have Done

Behavior Items

(JV =)

%

(JV =)

%

(JV =)

* Participation in a wildcat strike

(197)

7.4

(2312)

86.9

(153)

5.8

Attempting to win converts to one's own political views

(164)

6.1

(1220)

45.9

(1278)

48.0

%

* Aggressive political behavior, i.e., as a means of influencing the government the behavior is illegal, disruptive of the normal functioning of government, and involves group activity.

tion, respondents did not give their answers directly to the interviewer. Instead, they were asked to sort the cards into four boxes on a sheet of paper, where the boxes were labeled "I have tried it and it paid off," "I have tried it but it is difficult to say if it paid off or not," "I have tried it but it did not pay off," "I have not tried it." The aggressive actions were thus embedded in a set covering a wide range of behaviors instead of being specially singled out. The question about actual behavior came as part of a series of questions rather than in isolation. Responses regarding actual behavior were given according to a mechanical sorting pro­ cedure instead of by means of direct verbal communication to the interviewer. And the response options pertaining to actual behavior were not confined to a dichotomous (and perhaps ominous-sounding) "have you done it or not" choice of alternatives. Did the guarantee of anonymity and the attempt to present the questions in a non-threatening way help to reassure re­ spondents that the survey did not have an ulterior motive behind it? There is no way to answer this question definitely,

40 / Measurement of APP

but a comparison of rates of non-response to the aggressive and non-aggressive items suggests an affirmative answer. Response rates are given in Table 3.1. The mean nonresponse rate to the five aggressive actions is 7.6 percent. The mean non-response rate to the four clearly non-aggressive items (excluding Refusal of Military Service, a behavior that is not part of the aggressive set only because it normally does not involve collective activity) is 6.1 percent. As one would expect, the refusal rate is greater for the aggressive than for the nonaggressive actions. But the size of the difference between the non-response rates is encouragingly small. Even though the rate of refusal to respond to the aggressive items is not markedly out of line with the refusal rate for the four clearly non-aggressive items, the percentage of persons responding positively to the aggressive items is quite small. This is not surprising. Yet it is also likely that at least some people who say they have not engaged in a particular aggressive behavior may, in fact, have done it, and are reluctant to say so to a strange interviewer. There is no way unambiguously to identify such persons. It is possible that some of them might be found among those who responded positively to the question on behavioral intention, though negatively to the question on actual behavior. This suggests that it might be useful to try to combine behavioral intention and actual behavior into an overall participation measure. An even better reason for doing this is that not all of those who say they have done a behavior also express a positive intention to repeat it. Consider Figure 3.1, where behavioral intention is crossclassified by participation in behavior. Those who have engaged in a given action do not necessarily say that they would perform it again. This is especially so in the case of aggressive actions that involve violence. To treat the "Have Done" category as all of a piece clearly would be misleading. Those who say they have performed an action and would do so again should be differentiated from those who say only that they might do it again; and those who say that they would not perform the action again also should be differentiated from those with con­ ditionally or unconditionally positive intention. Turning to

FIGURE 3.1 Behavioral Intention by Participation in Aggressive Behavior

42 I Measurement of APP

the "Have Not Done" category, one sees that the vast majority in this group consistently express a negative intention toward future performance, while only a miniscule proportion state an unconditionally positive future intention. Still, the minority that does express a positive intention should not be treated as if its members were exactly the same as those who express a negative intention—especially if some few of the tiny "Would Do" group might be persons who have actually performed the behavior, but are unwilling to make that known in the context of a personal interview. The problem is how to build a participation measure sensi­ tive to these distinctions. One solution is to assign a score of 1 to those who express a negative intention, a score of 2 to those who express a conditionally positive intention, and a score of 3 to those who express an unconditionally positive intention, and then multiply the intention score by actual behavior, where lack of participation is given a score of 1 and participation is given a score of 2. Chart 3.1 shows the participation scores re­ sulting from this procedure. CHART 3.1. Scoring Procedure for Aggressive Political Behavior Items Participation in Behavior (- PB )

Aggressive Participation Score

Mot Done

Have

Have Done

(Bi* PB)

(1)

(2)

Behavior Intention

Would Do

(3)

3

6

Might Do

(2)

2

4

(Bi)

Would Not Do

(1)

1

2

People who have participated in a given behavior and say unconditionally that they will do it again stand apart from the

Measurement of APP / 43

others with a score of 6. The next highest score is a 4, received by those who say they have engaged in the behavior, but express only a conditionally positive intention toward repetition. Those who have engaged in the behavior, but definitely will not repeat it, receive a score of 2, the next to lowest in the series. The lowest score, naturally, is received by people who have not engaged in the behavior and will not do it in the future. People who have not done it, but express other than negative intention toward future performance, get somewhat higher scores. Comparing across the categories of intention, one sees that actual participation in the behavior is treated as being twice as great as lack of participation: 2 as compared to 1 for negative intention, 6 as compared to 3 for unconditionally positive intention. The overall participation scores thus give an appropriately heavier weight to actual participation, while at the same time incorporating additional information about participation that can be gleaned from the way in which inten­ tion toward future performance qualifies past performance. Yet these participation scores are still deficient. This is be­ cause they weight each action equally. A person who has par­ ticipated in an illegal strike and will do so again receives a score of 6—but so does a person who has participated in fights with police or other demonstrators and will do this again. Surely strikes and fights are not on the same footing. The former is an act of civil disobedience, the latter entails political violence. How can this information be built into the participation scores? Fights should get more weight than strikes—but how much more? One alternative is to weight each particular be­ havior by its cost in terms of social disapprobation as indexed by the degree to which it is perceived to be disapproved in the society at large. Recall that respondents were asked to estimate what proportion of Germans would approve of each behavior. Subtracting mean perceived social approval from 100 gives a mean perceived social disapproval score for each behavior. Figure 3.2 shows the results. Respondents perceive (fairly accurately, one, presumes) that non-violent protest such as illegal strikes and refusal to pay rent or taxes is disapproved by over 80 percent of the population, while violent protest is perceived to be frowned upon by more than 90 percent.

B,

Bg

B1: B2: B3: B1:

B(j

B7

Bs

Bn

B0: Fight police, other demonstrators B 7 : Participate in legal demonstrations B„: Use violence against government B T ,: Participate in illegal strike B m : Proselytize for own political views

BEHAVIORS

Participate in petition-signing drive Sacrifice time to work in an election campaign Refusal of military service Seize buildings B5: Refuse to pay rent, taxes

Key

B1

•§

FIGURE 3.2 Mean Perceived Social Disapproval: Various Political Behaviors

B1

Measurement of APP / 45

CHART 3.2. WeightingofAggressive Political Participation Items Weighted Aggressive Participation Score

Mean Perceived Social Disapproval

Raw Aggression Participation Score

IS w

=

.82

* (•BI IS *PB IS )

Refuse Pay Rent, Taxes

RP w

=

.85

* (BI rp * PB rp )

Seize Buildings

SB w

=

.86

* BISB* PB sb)

Fight Police, Demonstrators

FP w

=

.91

* (BIfp*PBfp)

Use Violence against Government

VG w

=

.93

* (•BI vg * PB va )

Illegal Strike

(

To take into account the fact that the set of aggressive be­ haviors is not all of a piece, each participation score is weighted —specifically, multiplied—by the mean perceived social dis­ approval rating for the particular behavior, as shown in Chart 3.2. This procedure yields a set of weighted aggressive partici­ pation scores with the following ranges: IS W \ 0.82-4.92; RP w '0.85-5.10; SBw: 0.86-5.16; FPw: 0.91-5.46; VGw: 0.93-5.58. The resulting variables thus afford a fair degree of discrimina­ tion in magnitude of aggressive response, and they certainly capture more information than the conventional way of scoring, which is to simply assign a 1 to participation and a 0 to lack of participation.1 1 It is important to note that these participation variables are constructed from components of the question set dealing with behavior that do not entail normative judgments about the behavior. Specifically excluded is the question of whether or not the respondent approves of the behavior. This question elicits a normative assessment that, were it included in the participation measure, would artificially bias the dependent variable in favor of the Expectancy-ValueNorms theory.

46 I Measurement of APP

The next question one must ask of these behaviors concerns their "dimensionality." Must they be conceived as five dis­ tinct phenomena, each requiring separate analysis, or can they be treated more parsimoniously as interrelated components of one (or more) general syndrome ? Research with macro-level variables has shown that events involving mass political conflict appear to form two empirically separate clusters. Using the most comprehensive cross-national aggregate data file now available (produced by the World Data Analysis program at Yale University), Hibbs replicated the results of many other studies in finding that, when subjected to factor analysis, events such as riots, anti-government protest demonstrations, and political strikes formed one distinct factor, while events such as armed attacks, deaths from political vio­ lence, and assassinations formed another.2 The first factor typically has been referred to as a "turmoil," "anomic vio­ lence," or "collective protest" dimension of mass political aggression, while the second factor has been defined as an "in­ ternal war" or "revolutionary" dimension. By contrast, in my own work at the micro level, I have argued that mass political aggression should be conceived not as a bidimensional but as a unidimensional syndrome of behavior.3 This is because it seems highly illogical to expect that, among individuals, intention to participate and actual participation in riots, illegal strikes, or demonstrations will be unrelated to intention to participate and actual participation in violent or armed conflict. To be sure, individuals who have participated or express positive intention to participate in "collective pro­ test" actions would not necessarily be expected to act or feel 2 Hibbs, Mass Political Violence (New York: Wiley, 1973), pp. 7-17. Results from a variety of factor analytic studies are compared in Raymond Tanter, "Dimensions of Conflict Behavior Within Nations, 1955-1960: Turmoil and Internal War," Peace Research Society Papers, III, Chicago Conference, 1965, 159-183; see also Tanter, "Dimensions of Conflict Behavior Within and Between Nations, 1958—60," Journal of Conflict Resolution, 10 (March 1966), 41-64. A third factor, involving elite political conflict and defined by variables such as coups d'etat, is also often found. 3 Muller, "A Test of a Partial Theory of Potential for Political Violence," pp. 933—936 (full citation in footnote 11 of Chapter One).

Measurement of APP / 47

similarly regarding "internal war" actions. But it is highly likely that persons participating or intending to participate in "internal war" actions also will have participated or intended to participate in "collective protest" actions. Therefore, sig­ nificant positive relationships should obtain between aggres­ sive actions subsumed under the "collective protest" dimension and aggressive actions subsumed under the "internal war" dimension. This unidimensional conceptualization of aggressive political participation is based on what seems to be a quite plausible assumption, namely, that "collective protest" and "internal war" represent different levels of aggressive response instead of qualitatively distinct dimensions. "Internal war" is a higher or more extreme level of aggressive response than "collective protest." Persons participating at the higher "internal war" level are likely to have escalated in their aggressive response from the lower "collective protest" level. The notion that aggressive political participation is a uni­ dimensional behavioral syndrome characterized by component actions reflecting different levels of aggressive response was tested and supported in a study carried out in the United States in 1970. 4 These results were replicated in a more elaborate and comprehensive test conducted in the United Kingdom in 1972. 5 Since then, studies done in Austria, the Netherlands, the United Kingdom, the United States, and West Germany have pro­ vided additional support. 6 Why, then, this divergence between micro- and macro-level studies, the former supporting a unidimensional, the latter a bidimensional conception of aggressive political participation? The answer, it would appear, has been uncovered by William Linehan in a recent analysis of macro-level political conflict

4 5

Ibid.

Alan Marsh 5 "Explorations in Unorthodox Political Behavior: A Scale to Measure 'Protest' Potential/" European Journal of Political Research, 2 (1974), 107-129. 6 Results will appear in a book dealing with value change and political be­ havior to be edited by Samuel Barnes and Max Kaase.

48 / Measurement of APP

indicators. 7 He argues that the factor analyses that have shown two dimensions of mass political aggression should be regarded very cautiously because they are based on raw events data that are not theoretically meaningful (raw events being joint prod­ ucts of both a polity's level of instability and its population size). To achieve a valid macro-level indicator of mass political aggression, differences in population size must be controlled. (For example, if Monaco and India each have ten riots, one would not want to consider them equally unstable.) Therefore, Linehan concludes, to be theoretically meaningful, macro indicators of aggressive political action should be percapitized. And when this is done, it turns out that indicators of "internal war" as well as indicators of "collective protest" load on a single factor. Linehan performed his analysis on a data set highly similar to that used by Hibbs. 8 His results strongly suggest that acts of mass political aggression, when properly measured at the macro level are, indeed, unidimensional, in keeping with the microlevel results and with the most plausible theoretical expectation. Of course there are definite analytical differences between some of the actions in the set of aggressive behaviors under con­ sideration here. The Illegal Strike and Refuse Pay Rent, Taxes variables are instances of non-violent civil disobedience; the Fight Police, Demonstrators and Use Violence against Govern­ ment variables are instances of political violence; the Seize Buildings variable could fall either way and, as such, is some­ thing of a bridge between civil disobedience and political vio­ lence. But if the civil disobedience actions simply represent a different level of aggressive response from political violence, instead of an entirely separate dimension, then the set of aggres­ sive behaviors should show a pattern of relatively high positive intercorrelation. Moreover, in the event that the civil dis­ obedience variables do constitute a lower level of aggressive response than the political violence variables, then, when one ' William Linehan, "Models for the Measurement of Political Instability," Political Methodology, 3 (Fall 1976), 441-486. 8 One variable, man-days of participation in civil conflict, is from data provided by Ted Gurr.

Measurement of APP / 49

builds a composite aggressive political participation scale by summing scores across the five component actions, the majority of those with high scores on the civil disobedience variables should fall in the medium range on the composite scale, whereas almost all of those with high scores on the political violence variables should fall in the high range on the composite scale, with very few located in the medium range. This would reflect a set of underlying relationships according to which people with high scores on the political violence variable would show a strong tendency also to have high scores on the civil disobe­ dience variables, but people with high scores on the civil dis­ obedience variables would show only a moderate tendency to score high on the political violence variables. 9 The correlation matrix shown in Table 3.2 indicates that TABLE 3.2. GorrelationMatrix:

Aggressive Political Participation Items

Illegal Strike: IS Refuse Pay Rent, Taxes: RT Seize Buildings: SB

IS

RT

1.00

.48 1.00

ED

VG

.59

.45

.40

.55

.34

.33

1.00

.60

.52

1.00

.58

SB

Fight Police, Demonstrators: ED Use Violence against Government: VG

1.00

Kuder-Richardson Reliability Coefficient (KR-20), r kk = .824 Estimated correlation of test with true scores, s fr^ k = .908

9 Because the component items are not dichotomies, the Guttman scaling technique is inappropriate for testing this expectation. Also, the Guttman scale method would require positing ordinal distinctions of dubious general validity between variables within the civil disobedience and political violence subsets. I rely on plots and cross-tabulations of the composite summated scale with components in order to determine if the expectation holds.

50 / Measurement of APP

the five aggressive participation variables are components of a unidirectional behavioral trait. This is corroborated by the convergence of results across two factor analyses, given in Table 3.3, one using unities in the diagonal of the correlation matrix, the other using squared multiple correlations ( SMC'S) in the diagonal. In both instances, only one dimension with an eigenvalue greater than 1.0 emerges.

TABLE 3.3. Factor Analyses of Aggressive

Political Participation Items Principal Components Analysisa Factor

Variance

Loading

Reproduced

Illegal Strike

.736

.542

Refuse Pay Rent, Taxes

.779

.606

Seize Buildings

.823

.744

Fight Police, Demonstrators

.762

.580

Use Violence against Government

.690

.476

Variables

Principal Factor Analysisb Factor

Variance

Loading

Reproduced

Illegal Strike

.646

.418

Refuse Pay Rent, Taxes

.708

.501

Seize Buildings

Variables

.860

.739

Fight Police, Demonstrators

.682

.465

Use Violence against Government

.589

Ml

3

Unities in the diagonal.

B

SMC'S in the diagonal.

Measurement of APP / 51

Every aggressive participation item correlates at least moder­ ately with each of the others. No item correlates so strongly with any other as to suggest that they are overlapping indicators of the same aspect of aggressive political participation. Like any relatively complex behavioral trait, aggressive political par­ ticipation has several features. Because the intercorrelations are neither especially high nor especially low, this set of behaviors appears to do a pretty good job of encompassing all the different facets of the phenomenon, while not obscuring the single general property, call it "political aggressiveness," they have in com­ mon. 10 Scores on the summated Aggressive Political Participation (APP) scale range from a low of 4.37 to a high of 26.22. The mean is 5.72 and the standard deviation is 2.52. To facilitate interpretation, scores are converted to a standardized scale with a mean of 10 and a standard deviation of 10. The lowest score on the standardized scale is 4.66. People receiving this score are slightly more than one-half of a standard deviation below the mean. Of all the people who received scores on the APP scale, 10.9 percent have a score of 4.66. This represents total inactivity. By contrast, the highest standardized score is 91.43. The two people who received this score (0.1 percent of the total) are slightly greater than eight standard deviations above the mean. This represents total activity. The distribution of the APP score is extremely abnormal. In a normal distribution of scores, only 0.13 percent of the total should be greater than three standard deviations above the mean. A score of 40 on the APP scale is three standard deviations above the mean. In 2.3 percent of the cases the score is higher than this, as can be seen from Figure 3.3, where the distribution 10 This latter point is borne out by the fact that the Reliability CoefBcient yields an estimated correlation of test scores with true scores that is slightly higher than the desired level of .9, meaning that a summated scale built from these five items can be considered a reliable measure of the concept of aggressive political behavior. The standard Kuder-Richardson reliability formula was used to compute scale reliability. See Allen L. Edwards, The Measurement of Personality Traits by Scales and Inventories (New York: Holt, Rinehart and Win­ ston, 1970), pp. 20-21, and Nunnally, Psychometric Theory (New York: McGrawHill, 1967), pp. 191-197.

FIGURE 3.3 Distribution of Aggressive Political Participation Scores 72-

71-

0-10

10-15

15-20

20-25

25-30

30-35

35-40

40-95

SCORES ON THE AGGRESSIVE POLITICAL PARTICIPATION SCALE

Measurement of APP / 53

on the APP scale is shown according to the following intervals: (1) +3 SD. 11 From Figure 3.3 it is apparent that the distribution of scores on the APP scale is not only markedly skewed to the right, but also tails up abnormally at the right-hand extreme. What can be done to make the distribution less skewed, more symmetric? The APP scale is a member of a class often referred to as "rare event" variables. Rare event variables are characterized by distributions that peak sharply at the very low end of the continuum, where most of the cases are clustered. The number of defective goods in a lot of manufactured articles or the number of accidents sustained by persons in a particular line of work are examples of rare event variables. Reliable statistical analysis of relationships between a rare event variable and a set of predictor variables is enhanced if the distribution can be made more symmetric so as better to satisfy certain statistical assumptions, especially the assumption of homoscedasticity or stability of variance. 12 The usual procedure is to convert the original scores of the rare events variable into logarithms. 13 This helps to reduce the skewness of the distribution and to stabilize the variance by spreading out the tightly clustered scores at the low end of the scale and pulling in the very large outlying scores at the high end toward the middle of the dis­ tribution. In addition, the logarithmic transformation is a useful theoretical tool. When a rare event variable is expressed as a natural logarithm (base e), the parameter that describes how it changes as a function of an increase in an explanatory or predictor variable is approximately equal to the percent in­ crease in the rate event variable, given a unit increase in the explanatory variable. 14 11

The mean is denoted by X, the standard deviation by SD.

12

See the discussion in Edward R. T ufte, Data Analysis for Politics and Policy (Englewood Cliffs, N.J.: Prentice-Hall, 1974), pp. 108-113. 13 As was done, for example, with aggregate measures of aggressive political behavior in Hibbs, Mass Political Violence, p. 14. li

Proof and an illustration is given in Tufte, Data Analysis, pp. 124-128.

54 I Measurement of APP

When the standardized APP scores are logged to the base e, the resulting APP ln variable has a mean of 2.01 and a standard deviation of 0.68. The range is 1.54 to 4.52. This substantially reduces the abnormality at the upper end of the scale. There are no scores greater than four standard deviations above the mean, as should be the case in a normal distribution. Also, only 0.6 percent of the cases have scores greater than three standard deviations above the mean, in contrast to the unlogged APP variable, where almost four times as many people score in this extreme range (two-fifths of whom score at greater than four standard deviations above the mean). Thus, the extreme scores are pulled in toward the middle. The low scores also are pushed up toward the middle, al­ though, with the large concentration of cases at the very lowest score, it is only possible to make a modest dent in the overall peakedness of the distribution. Figure 3,4 shows the distribu­ tion on the logged APP variable, divided according to the same intervals used to display the unlogged distribution. The four middle intervals (-f -JySn to + 2-JSD) encompass con­ siderably more cases in the logged as compared with the un­ logged version of the APP scale. In comparison to APP, two and one-half as many cases are located in the third interval (-F- TPSD to +ISD) of 1LogeAPP, about the same number of cases fall in the fourth interval (+ ISD to + 1-^SD), almost three times as many cases occur in the fifth interval (+ 1 ^SD to + 2SD) , and almost twice as many cases populate the sixth interval ( + 2SD to I 2^SD;. Despite the constraint imposed by the heavy con­ centration of cases at the lowest position, the logarithmic trans­ formation still produces a clearly noticeable shift of cases away from the extremes and into the middle of the scale, thus reducing (though by no means eliminating) the abnormality of the distribution. 3.1.

ZONES OF AGGRESSIVE POLITICAL PARTICIPATION

To this point it has been shown that the concept of aggressive political participation appears to be a unidimensional trait when viewed at the micro level. Aggressive behavior mani-

FIGURE 3.4 Distribution of Logged Aggressive Political Participation Scores

SCORES ON THE AGGRESSIVE POLITICAL PARTICIPATION SCALE LOGGED TO THE BASE e

56 I Measurement of APP

fested in the form of civil disobedience does not appear to be qualitatively different from explicitly violent aggressive action. Because of this unidimensionality it is justifiable to construct a single summary measure of aggressive political participation by adding together the amounts of participation a respondent registers on the five aggressive actions. Given the overall unidimensionality of the scale, are there certain scores along the continuum that serve to demarcate different levels or zones of aggressive participation? For example, on a Fahrenheit scale of temperature, 32° and 212° are points that mark important changes with respect to water. Are there similar points on the aggressive participation con­ tinuum at which substantively important or noteworthy changes occur? What kinds of discontinuity might be of special interest here? One major change would entail a transition from complete inactivity to a potential for activity. A second such change would entail a transition from potential for activity to actual participation. Within an actual participation zone, a third major change would entail a transition from non-violent to violent action. This construct implies four zones of aggressive participation: (1) a zone defined by absence of participation and low intention to participate; (2) a zone defined by absence of participation but increasingly high intention to participate; (3) a zone defined by participation in acts of non-violent civil disobedience and increasingly high intention to participate in acts of political violence; (4) a zone defined by participation in acts of political violence as well as civil disobedience. To deter­ mine whether it is possible to identify such zones, scores on the composite measure of aggressive participation must be plotted against actual participation scores as well as against scores for intention to participate. Figure 3.5 shows a plot of the relationship between the APP scale and an index of actual participation in aggressive be­ havior. Recall that each measure of actual participation was scored 1 for absence of participation, 2 for participation. Thus, when these are added up, lack of participation in any of the five aggressive actions is indicated by a score of 5, participation

Large circles denote 9 or more cases.

INDEX OF AMOUNT OF PARTICIPATION

FIGURE 3.5 Plot of the Relationship between the Aggressive Political Participation Scale and Amount of Participation in Actual Behavior

58 I Measurement of APP

in only one out of the five is registered by a score of 6, and so on up to a score of 10 for participation in all of the behaviors. The product-moment correlation coefficient, r, for this relationship is 0.79, meaning that the two variables show a very strong tendency to vary as a linear function. As arrount of actual participation increases, so do scores on the composite scale; this comes as no surprise. Noteworthy, however, is the fact that once a person reaches a score of 8 on the participation index, meaning that he has participated in three or more aggres­ sive behaviors, he is very likely to receive a score on the APP scale that is three or more standard deviations above the mean (a score of 40 or greater). If illegal seizure of buildings is in­ cluded within the domain of violent political action, then three of the five behaviors represent examples of political violence. By this classification, it follows that a person who has partici­ pated in any three of the five actions must have participated in at least one violent action. But does it follow that a score of 40 approximates a transition point? Almost all people who have participated in only two actions score between one and three standard deviations above the mean on the APP scale (20-40), while a majority of those who have participated in only one aggressive action score in a range from slightly less than one standard deviation to three standard deviations above the mean. In what kind of aggressive action have most of these people been involved ? The answer is: civil disobedience but not violence. Of those participating in only one or two aggressive actions, 46 percent were involved in rent or tax strikes, 39 percent in a wildcat strike, 17 percent in the seizure of a factory, office, or other type of building, 16 percent in fights with police or other demonstrators, and 4 percent with a revolutionary group. Thus, scores from slightly less than 20 up to 40 on the APP scale appear to demarcate a zone of par­ ticipation primarily in acts of civil disobedience but not vio­ lence. After a score of 40 is reached, violent political action is added to the individual's repertoire of aggressive political action. It has been possible to identify two distinct zones on the aggressive participation continuum: one with scores ranging

Measurement of APP / 59

from slightly less than 20 to 40, and one with scores of 40 or more. What of the range from slightly less than 20 down to 0? Is there a point somewhere within this range that marks a transition from inactivity to potential for aggressive action? To answer this question one must look to behavioral intention. Since behavioral intention was scored 1 for negative inten­ tion, 2 for conditionally positive intention, and 3 for uncon­ ditionally positive intention, an index constructed by summing the scores on the five behavioral intention variables ranges from 5 to 15. The plot of APP against the behavioral intention index is given in Figure 3.6. It appears that a score of slightly less than 10 on the APP scale marks a transition from a combination of inactivity and consistently negative intention to a combination of inactivity and some positive intention. Of the 1,460 people with con­ sistently negative intention, only 6 score at the mean or higher on APP. Of those who score lower than 10 on APP, the vast majority are located at the lowest position on the index of behavioral intention. Between the mean APP score and slightly less than one standard deviation above the mean, almost all the cases fall in the range of 6 to 9 on the behavioral intention index. This range represents positive intention primarily toward civil disobedience behaviors; to achieve a score of 10 or more on the behavioral intention index a person has to have expressed at least a conditionally positive intention toward performance of one violent action. The people who score in the 6 to 9 range of behavioral intention tend to be inactive: 74 percent are totally inactive, 21 percent have participated in one aggressive action, 4 percent have participated in two aggressive actions, and less than 1 percent (0.8) have participated in more than two. Thus, the zone between slightly less than 10 and slightly less than 20 on the APP scale is characterized by scores that primarily reflect only willingness to engage in some aggressive action—a poten­ tial for participation mainly in acts of civil disobedience. Lower than this, even potential for civil disobedience fades away. The interval from 0 to slightly less than 10 on the APP scale is represented by scores from 0 to 2.35 on the APPlri scale; 2.35 on APPln is one-half a standard deviation above the mean.

Large circles denote 9 or more cases.

INDEX OF AMOUNT OF POSITIVE BEHAVIORAL INTENTION

FIGURE 3.6 Plot of the Relationship between the Aggressive Political Participation Scale and Amount of Positive Behavioral Intention

Measurement of APP / 61

Thus, one standard deviation below the mean to one-half a standard deviation above the mean on APPlti (1.33-2.35) can be called an Inactivity zone. Slightly less than 10 to slightly less than 20 on APP corre­ sponds to scores from 2.35 to 3.03 on APPla. Thus, one-half to one and one-half standard deviations above the mean on APPln can be called a potential for civil disobedience or Low Civil Disobedience zone. Scores from slightly less than 20 up to 40 on APP correspond to scores from 3.03 to 3.71 on APPln. It has been shown that these scores reflect participation in acts of civil disobedience. The plot of APP against the behavioral intention index also shows that many of the people in the 20-40 range have be­ havioral intention scores of 10 or greater, meaning that they might or would participate in one or more violent actions. Thus, with respect to APPln, it can be said that scores from one and one-half to two and one-half standard deviations above the mean fall in a civil disobedience and potential violence or High Civil Disobedience zone. Finally, scores of 40 or greater on APP correspond to scores of 3.71 or greater on APPln. Thus, people who score higher than two and one-half standard deviations above the mean on APPln fall in a Political Violence zone. Figures 3.7 and 3.8 show plots of APP1n with the indices of actual participation and behavioral intention. Slashed grid lines demarcate the four zones of APPln. These grid lines help visually to apprehend the distinctive nature of each zone. To facilitate interpretation, they will be included on all graphs of relationships between APPln and explanatory variables. The distinctive nature of each zone also can be grasped from Table 3.4, where relationships are given between each aggres­ sive participation variable, divided into low and high categories, and the composite APPln scale, divided into the four zones. The high category of each participation variable includes only people who have performed the behavior and might or would do so again. Of course, no person who scores high on any participation variable is located in the Inactivity zone of APPln. Virtually

Large circles denote 9 or more cases.

INDEX OF AMOUNT OF PARTICIPATION

FIGURE 3.7 Plot of the Relationship between the Logged Aggressive Political Participation Scale and Amount of Participation in Actual Behavior

Large circles denote 9 or more cases.

INDEX OF AMOUNT OF POSITIVE BEHAVIORAL I N T E N T I O N

FIGURE 3.8 Plot of the Relationship between Logged Aggressive Political Participation Scale and Amount of Positive Behavioral Intention

0

100%

7.3

16.7

74.7

100%

High Civil Disobedience

Low Civil Disobedience

Inactivity

16.5

63.0

100%

73.9

16.5

8.7

0.8

1.2

PoliticalViolence

20.5

Low

(1-3)

High

(4-6)

Low

(1-3)

100%

0

21.0

46.7

32.4

(4-6)

High

Illegal Strike

^ones of the Logged Aggressive Political Participation Scale

Refuse Pay Rent, Taxes

100%

72.8

17.2

9.4

0.6

(1-3)

Low

100%

0

1.4

43.6

54.9

(4-6)

High

Seize Buildings

100%

72.3

17.1

10.0

0.8

(1-3)

Low

100%

0

1.8

34.5

63.6

(4-6)

High

Fight Police, Demonstrators

Aggressive Political Behaviors

Aggressive Political Participation Scale (Collapsed into Four Zones)

100%

70.8

16.7

10.5

1.9

(1-3)

Low

100°/O

0

0

16.7

83.3

(4-6)

High

Use Violence against Government

TABLE 3.4. Relationships between Participation in Each Particular Aggressive Behavior and Logged

Measurement of APP / 65

nobody who scores high on the three violence variables, and a relatively small proportion of those who score high on the two civil disobedience variables, are located in the Low Civil Dis­ obedience zone. Also, only a small proportion of those who score high on the civil disobedience variables is located in the Political Violence zone. In one instance a large majority, in the other instance slightly less than a majority, of those who score high on the civil disobedience variables are located in the High Civil Disobedience zone. People who score high on the three violence variables are most likely (by an absolute major­ ity) to fall in the Political Violence zone; and they are less and less likely to fall in the High Civil Disobedience zone, more and more likely to fall in the Political Violence zone, as the violence implied in the behavior is more explicit and premeditated. Thus, the relationships between the composite scale and each component variable provide additional backing for the charac­ terization of these zones. It is instructive to compare this scale of aggressive political participation with a measure of potential for political violence I developed some years ago. 15 The present measure is an im­ provement over the former for the following reasons: (1) It builds reports of actual behavior into the measure instead of focusing solely upon dispositions (approval, intention) regarding participation; (2) the component items afford a much closer approximation to interval-scale variables than the gross "yes/no" dicho­ tomies used previously; (3) the composite measure is not based on a scaling model (the cumulative or Guttman scale model) that requires that com­ ponent items be dichotomized, resulting in a measure sensitive only to crude levels of aggressive response; rather, this measure is sensitive to finely discriminated, continuous zones of aggressive response, as befitting a quantitative in­ stead of an ordinal variable.

15 Muller, "A Test of a Partial Theory of Potential for Political Violence," pp. 933-936.

66 I Measurement of APP

3.2. SUMMARY The conceptualization of aggressive political participation proposed in this chapter takes both intention to participate and actual participation into account. Participation in a given aggressive action is measured by the product of intention to participate times actual participation. This operational defi­ nition is preferable to definitions based only on intention or only on actual participation. Measures solely of behavior do not discriminate between persons who have performed an action and would repeat it and persons who have performed an action but would not do it again. The operational definition proposed here is sensitive to such distinctions, scoring positive intention and performance higher than either positive intention without performance or performance without positiveintention. An additional refinement is the weighting of participation scores for a given action by a constant that expresses the cost of the action in terms of its social disapprobation. Aggressive par­ ticipation scores are thus rendered sensitive to differences in the extremity of given actions. Participation scores for actions that entail non-violent civil disobedience are deflated relative to scores for actions that entail the use of violence. It was expected that the actions in the aggressive behavior set would turn out to be unidimensional. This was confirmed. Since the various actions have an underlying property of political aggressiveness in common, it was also expected that the variables entailing non-violent disobedience would repre­ sent a different level of aggressive response from the variables entailing political violence. Specifically, it was expected that the political violence variables would represent the highest level of aggressive response, a level to which people normally progress after having engaged in the next highest level of response, acts of civil disobedience. This implies that those who have participated in political violence will tend also to have participated in civil disobedience, whereas those who have par­ ticipated in civil disobedience will not necessarily have esca­ lated to political violence. This, too, was confirmed. And since

Measurement of APP / 67

the component items have interval-scale characteristics, these "levels" are not simply crude categories denoted by a single score, but rather approximations to continuous "zones" of ag­ gressive response. The summated scale of aggressive political participation shows good reliability and can be interpreted in a plausible and substantively meaningful way. The overall aggressive par­ ticipation continuum can be demarcated roughly into four zones of increasingly aggressive response. Moving from low to high on the scale, one moves from (1) inactivity and generally negative intention toward par­ ticipation in aggressive action, to (2) predominant inactivity but positive intention toward per­ formance of non-violent aggressive action, to (3) participation predominantly in non-violent aggressive ac­ tion coupled with positive intention toward performance of violent action, to (4) participation in acts of political violence as well as non­ violent civil disobedience. The distribution of aggressive political participation is, as one would expect, markedly skewed, with the vast majority of cases scoring in the lowest zone. But a logarithmic transforma­ tion helps to reduce the extreme skewness, thereby enabling the variable better to satisfy requirements for reliable statistical analysis of quantitative variables. One of the more important features of the Aggressive Political Participation scale is simply the degree to which aggressive response is finely discriminated, giving a good approximation to a quantitative continuum. This makes it suitable for regres­ sion analysis, a statistical technique permitting more precise and comprehensive statements about relationships between variables than are possible with techniques for statistical analy­ sis of non-quantitative variables. Previous studies of mass political aggression have suffered because of inadequacies in the dependent variable: inadequa­ cies due to exclusion of actual behavior from the measuring

68 / Measurement of APP

instrument and to the definition of the measuring instrument in terms of a limited number of crude categories or levels of aggressive response. The present measure of aggressive political participation is intended to take a large step in the direction of overcoming such deficiencies.

FOUR

The Expectancy-ValueNorms Theory

We now have a satisfactory quantitative indicator of participa­ tion in aggressive political behavior. Without this, the theories and models of political protest and violence elaborated in Chapter Two are mere speculation. But armed with such a measure, we can begin to tackle, scientifically, the engineering question that has long engaged political philosophers and men of action: what does it take to mobilize men for acts of collec­ tive political aggression? This analysis is guided by the general social-psychological theory of behavior proposed by Fishbein. The real challenge is to adapt Fishbein's abstract concepts to specifically aggres­ sive political behavior. This requires the formulation of an auxiliary theory to connect the concepts in the ExpectancyValue-Norms theory to observable indicators. 1 Since all definitions are arbitrary, any abstract concept can be defined operationally in a potentially infinite number of ways. The task of an auxiliary theory is to provide a rationale for specifying in advance which indicators of an abstract con­ cept are likely to be most useful. Of course, the ultimate test of the utility of such an indicator is its predictive accuracy in relation to those variables with which it has been hypothetically 1 On the topic of an auxiliary theory designed for testing purposes see the brief discussion in Hubert M. Blalock, Theory Construction (Englewood Cliffs, N.J.: Prentice-Hall, 1969), pp. 151-154; and the more lengthy treatment in Blalock, "The Measurement Problem: A Gap Between the Languages of Theory and Research," in Methodology in Social Research, ed. Hubert M. Blalock and Ann B. Blalock (New York: McGraw-Hill, 1968), 5-27.

70 / Expectancy-VaIue-NormsTheory

linked. But prior to the determination of predictive accuracy there must be grounds for believing that the indicator is (1) a unidimensional measure of the abstract concept and (2) cap­ tures the distinctive meaning or causal mechanism inherent in the abstract concept. 2 Thus the purpose of an auxiliary theory is the formulation of operational definitions such that a given operational definition of an abstract concept measures only that concept and no others, and corresponds well to the unique meaning of that concept.

4.1

THE EXPECTANCY-VALUE CONCEPT

The causal mechanism underlying the expectancy-value con­ cept (Ε* V) is the assumption that strong motivational incentive for performing a behavior is provided by expectation that the action will have beneficial or rewarding consequences, weighted by the value of performing that action as compared with al­ ternative actions. With regard to the definition of the Ε* V term from equation 2.5, utilitarian motivational incentive to par­ ticipate in collective political aggression will be said to derive from the following condition: collective political aggression is seen as having beneficial consequences and as being of value to a person because he feels that his ability to influence govern­ ment through normal channels is insufficient. Conditions postulated to produce low and high utilitarian justification for political aggression can be diagrammed as follows:

Utilitarian Justification for Political Aggression Political Influence Capability

Insufficient Sufficient/ Unnecessary

Consequences of Collective Political Aggression Harmful Beneficial Low

High

Low

Low

2 Statistical criteria can be used to evaluate empirically the former require­ ment; the latter is a matter of judgment concerning the correspondence be­ tween the face content of an instrument and the verbal exposition of the abstract concept which the instrument is supposed to measure.

Expectancy-Value-Norms Theory / 71

According to this formulation, belief that the consequences of collective political aggression will be beneficial is not expected to produce high utilitarian justification if a person either (1) already has sufficient political influence or (2) feels that political influence is unnecessary. Only those who both perceive bene­ ficial consequences from collective political aggression and have reason to value such action because their own political influence capability is insufficient are assumed to attach high utilitarian justification to collective political aggression. But how shall we operationally define the concept of belief in the efficacy of collective political aggression? The social learning theory of aggression set forth by Albert Bandura sug­ gests an answer.3 Bandura's general social learning theory is an application of reinforcement learning theory to specifically human behavior.4 According to the school of thought asso­ ciated especially with the work of B. F. Skinner, directly experi­ enced reinforcement contingencies are the major determinant of behavior; but with respect to human social behavior, Bandura regards the direct experience approach as incomplete, for human social behavior also is significantly under the control of observational learning or vicarious reinforcement, as well as internalized, self-regulative influences or self-reinforcement.5 Self-reinforcement for participation in collective aggression derives essentially from normative beliefs about the justifi­ ability of such action. Measurement of this will be taken up shortly. Vicarious reinforcement for participation in collective aggression derives from perception of the degree to which such action has had beneficial or rewarding consequences for the dissident groups involved. Direct reinforcement for par­ ticipation in collective aggression derives from an individual's own experience of rewarding or punishing consequences con­ tingent upon his performance of such action. 3 Albert Bandura, Aggression: A Social Learning Analysis (Englewood Cliffs, N.J.: Prentice-Hall, 1973). 4 A compact statement is given in Albert Bandura, Social Learning Theory (Morristown, N.J.: General Learning Press, 1971). 5 Bandura, ibid.; see also B. F. Skinner, Science and Human Behavior (New York: Macmillan, 1953) and B. F. Skinner, Cumulative Record (New York: Appleton-Century-Crofts, 1961).

72 I Expectancy-VaIue-NormsTheory

Bandura argues that self-reinforcement and vicarious rein­ forcement for participation in collective aggression are likely to override directly experienced response consequences: At the individual level, the reinforcements for participat­ ing in aggressive activities are varied and complex, and they can change even for the same person in the course of continuing activism. Some of the protesters, who adhere to high social and ethical principles, become distressed by institutional practices that exploit and oppress disad­ vantaged people. In such cases, coercive actions are sus­ tained, even in the face of punishing consequences, by self-approval for upholding one's convictions and by expectations that continued pressure may eventually pro­ duce humane social reforms.6 Self-approval (deriving from personal normative belief in the justifiability of aggression) as well as expectation of beneficial consequences (deriving from belief that collective aggression has led to goal-attainment for dissident groups in the past) are thus seen as variables that affect the likelihood of participation in collective aggression regardless of consequences directly experienced by an individual—consequences that may, more­ over, often be too complex and varied to function as a systematic general determinant of participation in collective aggression. Vicarious reinforcement serves to instigate and sustain par­ ticipation in collective aggression because of its informative func­ tion and because it has incentive motivational effects. Response consequences that accrue to others give information about the kinds of actions which, for groups in general, are likely to have rewarding or punishing outcomes. And, sis Bandura points out, "given knowledge about probable response consequences, people will generally do the things they have seen well-received and avoid those that they have seen punished"; also, "seeing others positively reinforced can serve as a motivator by arousing in observers expectations that they will be similarly rewarded for

6

Bandura, Aggression, p. 235.

Expectancy-VaIue-NormsTheory / 73

analogous performances."7 With respect to participation in collective political action, longer-range, generalized percep­ tion of positive reinforcement outweighs short-range, immediate experiences of success or failure: "Observational incentives play an especially important role in social activism, for here the chances of quick success are poor, but protest behavior is partly sustained by the long-range attainments of groups that have persevered in their efforts."8 Following Bandura's emphasis on the importance of observa­ tional or vicarious reinforcement as a motivational incentive for participation in collective aggression, it is assumed that belief in the efficacy of collective political aggression can be operationalized most usefully by assessing an individual's per­ ception of whether a variety of aggressive political actions have been beneficial or harmful to the goals of the groups involved. Respondents were asked to report whether they believed that the following collective political actions had helped, hurt, or neither helped nor hurt the achievement of group goals: (1) student groups which have disrupted lectures and meetings and damaged or destroyed rooms and furniture in order to bring about changes in the universities; (2) groups which have seized public buildings and fought with police in order to protest against actions of the government; (3) secret organizations—such as the Baader-Meinhof group— which have taken to guerrilla tactics in order to protest against the present social and political structure in the Federal Republic. A composite Efficacy of Collective Aggression (ECA) measure is constructed by assigning a score of O to a "hurt" response, a 1 to a "neither/nor" response, 2 to a "helped" response, then summing scores across items.9 7

Ibid., pp. 205, 206. Ibid., p. 206. 9 The unidimensionality of these three items is strongly suggested by their face content alone. In addition, they show positive empirical correlation of moderate strength. Item (1) correlates (r) with items (2) and (3) at .44 and .30 respectively; item (2) correlates with item (3) at .47. (Tau b correlations are virtually identical.) 8

74 / Expectancy-Value-NormsTheory

The value or utility of collective political aggression to the individual can also be defined operationally. It is plausible to assume that, of the alternative modes of political participation open to an individual, the value of behavior that deviates from legally permitted action will be highest when he believes that his capacity for influencing the government is insufficient, lowest when he believes that involvement in politics is unnec­ essary or undesirable. The value of legally deviant political behavior will fall somewhere between high and low for people who see themselves as having sufficient capability to influence the government. To define conditions of insufficient, sufficient, and unnecessary political influence capability, the following index is proposed:

Political Influence Capability Desirability of Political Involvement

High Medium Low

Amount of Personal Political Influence

Low

Medium

High

I

I S U

S S U

/ U

where / denotes the insufficient condition, S denotes the suffi­ cient condition, and U denotes the unnecessary condition. According to this index, influence capability is insufficient when the desirability of political involvement is greater than the amount of personal political influence. Influence capability is unnecessary whenever the desirability of political involve­ ment is low. Influence capability is sufficient when desirability of political involvement is not low and is equal to or less than the amount of personal political influence. Personal political influence was measured by a question ask­ ing respondents to give their opinion as to whether they had a large influence, a fair amount of influence, a little influence, or no influence at all on how their country is governed. In a country with a large population and a complex political sys­ tem, an individual who believes that he wields even a fair amount of influence over its government presumably has a

Expectancy-Value-Norms Theory / 75

relatively high sense of political efficacy: thus the High category on the Amount of Personal Political Influence variable is de­ fined by combining the "large" and "a fair amount" responses; the Medium category is defined by the "a little" response; and the Low category is defined by the "no influence at all" response. Desirability of political involvement was measured by a question asking respondents to give their opinion as to whether, for people like themselves, involvement in politics was very desirable, desirable, not very desirable, or completely un­ desirable. The High category on the Desirability of Political Involvement variable is defined by the "very desirable" re­ sponse; the Medium category is defined by the "desirable" response; and the Low category is defined by combining the "not very desirable" and "completely undesirable" responses. The Political Influence Capability (PIC) variable is con­ structed by cross-classifying Amount of Personal Political In­ fluence by Desirability of Political Involvement. Persons in the Low category on Desirability of Political Involvement are assigned to the Unnecessary condition of PIC. Otherwise, those whose amount of influence equals or exceeds their feeling about the desirability of involvement are assigned to the Sufficient condition of PIC, while those whose strength of feeling about the desirability of involvement exceeds their amount of influ­ ence are assigned to the Insufficient condition of PIC. The un­ necessary condition of PIC is scored as 0, the Sufficient condition as 1, and the Insufficient condition as 2. A quantitative indicator of degree of utilitarian incentive for collective political aggression is defined by multiplying the ECA variable times the PIC variable. Chart 4.1 shows the scores on Utilitarian Justification for Aggression (UJA) that result from this procedure. People who feel that it is undesirable to get involved in politics are automatically given a score of 0 on UJA because, to them, differences in degree of belief in the efficacy of collec­ tive political aggression are presumed to be irrelevant. Such a weighting is consistent with the assumption that aggressive

76 I Expectancy-VaIue-NormsTheory

CHART 4.1. Scoring Procedure for the Utilitarian Justification for Aggression Variable Efficacy of Collective Aggression {ECA)* BeneUtilitarian Justification Score {PIC* EC A) Political Influence Capability

Insufficient (2) Sufficient (1) Unnecessary (0)

Harmful ficial 1 2 3 4 2 1 0

4 2 0

6 3 0

8 4 0

5

6

7

10 5 0

12 6 0

14 7 0

* An increment of 1 is added to ECA so as to avoid giving everyone with the lowest score on ECA an automatic 0 on UJA, regardless of their PIC score.

action—or any kind of political action—has no value to indi­ viduals who consider the exercise of political influence to be unnecessary. Differences in degree of belief in the efficacy of collective political aggression are given twice as much weight among people who feel that they have an insufficient amount of in­ fluence as compared with people who feel that their influence is sufficient. This weighting is consistent with the assumption that collective political aggression has potentially greater value for individuals whose influence capability is insufficient than for those who already have sufficient influence capability. If we divide the UJA scores in the middle and call scores in the 0-7 range "low" and scores in the 8-14 range "high," it can be seen that this weighting scheme reflects the assumption that a person's utilitarian justification for participating in col­ lective political aggression will be high only when the conse­ quences of collective aggression are perceived to be beneficial and such action has value because existing influence capability is insufficient. All persons in the Sufficient condition of PIC are scored in the low range of UJA, regardless of their ECA score; whereas only those in the Insufficient condition of PIC

Expectancy-VaIue-NormsTheory / 77

and at the midpoint (4) or greater of ECA receive scores in the high range of UJA. The operational definition of the E*V term proposed here thus corresponds well to the meaning of the expectancy-value concept. As measured by the UtilitarianJustification for Ag­ gression variable, utilitarian incentive for political aggression becomes increasingly positive to the extent that expectancy of success is positive and the action itself has potential value as an alternative. In this sample, Utilitarian Justification for Aggression is quite low, as the mean UJA score is 2.3 (the standard devia­ tion is 2.6). Only 6 percent of the cases score at 8 or higher on the UJA variable. Figure 4.1 shows the plot of the relationship between Utili­ tarian Justification for Aggression scores and Aggressive Politi­ cal Participation scores. The thick line denotes the slope, summarizing the increase in APP ln units when UJA changes by one unit. The dashed lines demarcate the different zones of APP ln . The grid lines demarcate Very Low, Low, Medium, and High zones of UJA. From the angle of the slope one can see that magnitude of aggressive political participation is predicted to increase fairly rapidly as degree of utilitarian incentive for such action in­ creases. It is estimated that people with scores of 9 or more on UJA will, on the average, score in the High Civil Disobedience zone of APP ln . Scores of 9 or more on UJA can be received only by persons who (1) feel that their influence capability is insufficient and (2) perceive that collective political aggres­ sion has, in the past, led to goal attainment for dissident groups. This combination of insufficient influence and beneficial per­ ceived consequences appears to provide substantial motiva­ tional incentive for aggressive action. Yet the overall fit to the estimated slope, as indexed by the r 2 of .234, is only moderate. This is due especially to the many extreme deviations from predicted APPln scores that occur in the lower range of UJA. For example, it is estimated that people with scores in the range 0-4 on UJA will, on the average, score in the Inactivity zone of APP ln . Yet sizable numbers of

Summary

Statistics

Large circles denote 9 or more cases.

U T I L I T A R I A N JUSTIFICATION FOR AGGRESSION

(UJA)

FIGURE 4.1 Aggressive Political Participation as a Linear Function of Utilitarian Justification for Aggression

Expectancy-VaIue-NormsTheory / 79

people with these low UJA scores actually fall in the High Civil Disobedience zone of APPln. This clearly indicates that utilitarian justification is by no means the only—or even, probably, the major—determinant of participation in aggres­ sive political behavior. There are just too many people with quite high aggressive behavior scores who see little utilitarian justification for such action. But what might drive them to participate, despite the feeling of low utilitarian justification? A consideration of personal normative beliefs may provide an answer.

4.2

PERSONAL NORMATIVE BELIEFS

What kinds of beliefs might lead a person to feel that collective political aggression is or is not proper or ethically justifiable behavior? In other words, how shall the NBAp term from equa­ tion 2.5 be defined operationally? Feelings about the legitimacy of political authority are surely among the most important sources of normative justification for or against aggressive political action. People who believe that the structure of political authority or system of government is worthy of support can be expected to feel that aggressive behavior is unwarranted. Although not all of those who be­ come alienated from the system will feel that aggressive be­ havior is proper, they certainly are more likely to see it as normatively justifiable than are those who support the system. Attempts to define political support and alienation have given rise to what Gurr aptly terms "a dense thicket of scholarly concepts and distinctions."10 One of the most carefully worked out and comprehensive conceptualizations is that formulated by David Easton.11 He distinguishes between specific and diffuse support. Specific support refers essentially to satisfac­ tion-dissatisfaction with policy outputs from the government. Diffuse support refers to generalized positive-negative affect 10 Gurr, Why Men Rebel, (Princeton: Princeton University Press, 1970), p. 183. 11 David Easton, A Systems Analysis of Political Life (New York: Wiley, 1965), pp. 247-340.

80 I Expectancy-Value-Norms Theory

for the political community, for the values, norms, and institu­ tions of the regime, and for the authorities, including not only national officials and leaders of government but also the whole panoply of bureaucrats, judges, and police who administer the power of the state. Legitimacy sentiment is seen as a funda­ mental component of diffuse support. Defined in its most general sense as belief that the authorities and regime "in some vague or explicit way . . . conform to a person's own sense of what is right and proper in the political sphere," the inculcation of legitimacy sentiment is regarded as "the single most effective device for regulating the flow of diffuse support in favor both of the authorities and of the regime." 12 More specifically, legitimacy beliefs encompass evaluation of how well political institutions conform to a person's sense of what is right, of how well the system of government upholds basic political values and norms in which he believes, and of how well authori­ ties conform to his sense of what is right and proper behavior. To measure belief in the legitimacy of the system of govern­ ment, respondents were given a deck of cards in which each card contained a statement designed to capture an aspect of legitimacy sentiment. They were asked to say how strongly they agreed or disagreed with each statement, using a 7-point scale which ranged from +3, labeled "agree completely," to — 3, labeled "disagree completely." The deck of cards con­ tained eleven statements. Eight of these were selected for inclu­ sion in a measure of support for the political system, as follows: 13 S 1 : It makes me concerned when I think about the dif­ ference between what people like me value in life and what actually happens in our political system. •V 2 : I have great respect and affection for the political in­ stitutions in the Federal Republic. 12

Ibid., p. 278. Items .S11 and are inspired by analogous items used by David Schwartz in a measure called Political Alienation from the National Polity, and items and S 5 are inspired by analogous items used by him in a measure called Conformity-Acceptance of the American Political System—see David C. Schwartz, Political Alienation and Political Behavior (Chicago: Aldine 5 1973), pp. 265, 266, 273. 13

Expectancy-VaIue-NormsTheory / 81

S,: My friends and I feel that we are quite well-represented in our political system. S 4 : I find it very alarming that the basic rights of citizens are so little respected in our political system. S5:

At present, I feel very critical of our political system.

S6:

The courts in the Federal Republic guarantee every­ one a fair trial regardless of whether they are rich or poor, educated or uneducated.

S 7 : Looking back, the leading politicians in the Federal Republic have always had good intentions. Sg:

Considering everything, the police in the Federal Republic deserve great respect.

Items S 1 , S 3 , and S 4 are intended to measure evaluation of how well the political system upholds basic values. Items S 2 and S 5 are intended to measure evaluation of how well political institutions and the political system in general conform to a person's sense of what is right and proper. Items S g , S 7 , and S g are intended to measure evaluation of broad classes of politi­ cal authorities: leading national political officials, the courts, and the police. The courts and the police are included because, as Easton and Dennis have argued in the context of the United States, such objects symbolize the law; and in a constitutional, legal society (the Federal Republic is an even more legalistic society than the United States), feelings about the legitimacy of the courts and police are probably bound up fairly closely with belief in the legitimacy of political autority in general. 14 Support for the political system is a multifaceted concept for which single-item indicators are quite inappropriate. On the other hand, parsimony also is an important criterion in mea­ surement. The ultimate goal is a measure suitable for inclusion in national cross-sectional surveys. Since these are very costly to administer, the fewer the items the better. One wants a measure that is at once sensitive to the various analytically dis14 See David Easton and Jack Dennis Children in the Political System: Origins 5 of Political Legitimacy (New York: McGraw-Hill, 1969), p. 293.

82 / Expectancy-VaIue-NormsTheory

tinct facets of legitimacy belief, is unidimensional, and yet does not contain highly overlapping or unnecessarily redundant items. The eight items selected for the measure of support for the political system are diverse in respect to face content and are unidimensional in the sense both· of showing high reliability and of loading highly on the first component when subjected to a principal components analysis.15 Of the three items not included in the measure of political support, one was eliminated because it showed low correlations with the other eight, one was eliminated because it was redundant in meaning and showed a very high correlation with S8, and one was eliminated because, on the grounds of face content, it belongs with items that have been used to measure trust in government.16 15 The Reliability Coefficient (KR = 20) is .822, yielding an estimated correlation of test scores with true scores of .907. The inter-item correlation matrix is:

s, 52 53 54 55 56 57

58

S1

S2

S3

S4

S5

S6

S7

S8

1.00

-.26 1.00

-.37 .58 1.00

.41 -.34 -.36 1.00

.39 -.31 -.37 .36 1.00

-.25 .40 .41 -.26 -.21 1.00

-.34 .49 .50 -.30 -.32 .40 1.00

-.25 .55 .43 -.22 -.22 .47 .47 1.00

Before performing the principal components analysis, the scoring of items S 2 , S 3 , S 6 , S 7 , and S b was reversed so that all items would run in the direction, from low to high, of most supportive response to most alienated response. The loadings on the first component of the principal components solution are: S 1 = .588; S 2 = .758; S 3 = .770; S 4 = .582; S 5 = .570; S 6 = .637; S 7 = .731; S8 = .692. 16 The item that showed low correlations with the others is: "The true political power in the Federal Republic is held by only a small elite—even citizens who have a great desire to take part in politics have little chance to influence political decisions." This item correlates with S 1 at .28, S 2 at —.24, S 3 at —.24, S 4 at .44, S 5 at .25, S 6 at —.25, S 7 at —.21, S 8 at —.17. Its loading on the first component of the principal components solution is .438, a consider­ ably smaller value than that observed for the other variables. The item re­ dundant in meaning is: "In general, the police in the Federal Republic treat everyone equally, regardless of whether they are rich or poor, educated or uneducated." This item correlates very highly (r = .63) with item S s , also dealing with the police. The item assigned to the trust set is: "In general, one can rely on the federal government to do the right thing." There are theoretical

Expectancy-VaIue-NormsTheory / 83

The items used in the index of political support were assigned scores from O to 6, coded so as to range from positive to negative evaluation.17 When scores are summed across the eight items the resulting Political Support-Alienation {PSA) variable ranges from O at the most positive or supportive position to 48 at the most .negative or alienated position. The Political Support-Alienation variable tends to be nor­ mally distributed in this sample. Mean PSA is 24.9 (the standard deviation is 9.4) and only 4 percent of the cases register very low PSA (scores less than 10) while only 8 percent register very high PSA (scores of 40 or more). The relationship between PSA and APP ln is plotted in Figure 4.2. From this plot one sees that when PSA is relatively low rather few persons score at other than the Inactivity zone of APPltl. Since these people are supportive of the structure of political authority, they have little normative incentive to participate in political action that deviates from what is legally permissible. In fact, it is not until very high PSA is reached that Inactivity disappears. Because of this, the slope is not a good summary of the relationship. At low PSA, differences in degree of support-alienation among these basically supportive people are of little consequence for aggressive political response. Only as one moves away from the low levels of PSA do differences in degree of support-

reasons—see the discussion in Appendix A and also see David Easton 3 "A Re-Assessment of the Concept of Political Support," British Journal of Political Science, 5 (October 1975), pp. 447-450—for dealing with trust separately, despite correlations with the support items sufficiently high to warrant, on purely atheoretical dimensional criteria, inclusion of the trust item in the support scale. The trust item correlates with S 1 at —.30; S 2 at .48; S 3 at .52; S i at —.26; S 5 at -.36; S 6 at .35; S 7 at .53; S 8 at .39. 17 The original response options were scores of +3, +2, +1, 0, —1, —2, — 3, ranging from strong agreement ( + 3) to strong disagreement ( — 3). First, items S 2 , S 3 , S 6 , S 7 , and S g were reflected so as to range from very posi­ tive sentiment ( — 3) to very negative sentiment ( + 3). Then, in order to eliminate negative signs (which are mathematically undesirable when com­ puting terms for the purpose of testing interaction hypotheses—such as that between support-alienation and ideology discussed subsequently in the text), the scores from —3 to +3 were recoded to 0 to 6.

Statistics

N = 2003 r- — .251 APP,„ = 1.130 + -m(PSA)

Summary

Large circles denote 9 or more cases.

POLITICAL SUPPORT-ALIENATION (PSA)

FIGURE 4.2 Aggressive Political Participation as a Linear Function of Political Support-Alienation

Expectancy-VaIue-NormsTheory / 85

alienation begin to be associated with differences in magnitude of aggressive participation until, at the very high level of PSA, a clear majority of cases fall in the High Civil Disobedience and Political Violence zones of APPln. The relationship between PSA and APPln does not tend to become linear until after a certain threshold is reached (somewhere in the range of 10 to 20 on PSA). To describe this relationship adequately, differ­ ences in degree of low PSA should be compressed relative to differences in degree of high PSA, since the former are of much less consequence to increases in aggressive response than the latter. Squaring the scores on the Political Support-Alienation variable is a transformation that accomplishes this. The meaning of this transformation can be grasped from Figure 4.3, where PSA is plotted against PSA2. (PSA scores were first divided by 10 to avoid having to deal with extremely large PSA 2 quantities). Note that scores from 0 to 20 on PSA (the low to very low range) are now compressed into the range 0-4 on PSA2. By contrast, the very high range of PSA, scores of 40 or more, now is spread out in the interval from 16 to 24 of PSA2. Persons who score exactly at the midpoint (12) of PSA 2 are those with a score of 35 on PSA. The PSA 2 variable accounts for 28 percent of the variation in APPlrl. This is an increase—though hardly of dramatic pro­ portion—in predictive accuracy relative to that afforded by PSA. The relationship between PSA2 and APPln is shown in Figure 4.4. A more important improvement is in the slope. Cases that deviate markedly from the predicted scores are distributed more evenly instead of being bunched together in the upper right-hand quadrant of the plot. This means that the slope now gives a more reliable summary estimate of the way in which magnitude of aggressive political participation is linearly de­ pendent on degree of support-alienation—more reliable in the sense that marked departures from the estimated linear de­ pendency do not occur disproportionately within a particular subset of slope values. In addition to greater reliability, the slope has more substantive interest, for it now tells us that when PSA2 reaches the very high range (19-24), scores in the High

Large circles denote q or more cases.

POLITICAL SUPPORT-ALIENATION SQUARED (PSA2)

FIGURE 4.3 Relationship between Untransferred and Transferred Political Support Scales

Summary

Statistics

Large circles denote 9 or more cases.

POLITICAL SUPPORT-ALIENATION SQUARED (PSA 2 )

FIGURE 4.4 Aggressive Political Participation as a Linear Function of Political Support-Alienation Squared

88 / Expectancy-VaIue-NormsTheory

Civil Disobedience zone of APPln are the likely result. By con­ trast, very high PSA was predicted to lead only to APPln scores in the Low Civil Disobedience zone. Squaring the PSA variable is a transformation that enables it to conform better to the theoretically expected relationship between political support and aggressive political participation. Stated in terms of crude dichotomous variables, the prediction is that low political support will be a necessary but not suffi­ cient condition of aggressive action.18 With quantitative vari­ ables that give fine discrimination between different magni­ tudes of aggressive political participation and different degrees of support and alienation, it is possible to formulate a stronger version of the necessary-but-not-sufficient hypothesis: namely, that once political support has declined to a certain critical point, further decline (or increase in alienation) will be associated with steadily increasing magnitude of aggressive political par­ ticipation. This is a stronger hypothesis, in the causal sense, because it says that in the range of low political support, the level of aggressive participation will be greater, the lower the level of political support. The most unsatisfactory feature of the relationship plotted in Figure 4.4 is the sizable number of persons in the intermediate range (9.6-14.4) of PSA2 who score at the very lowest position on APPla. Predicted aggressive response scores for these people are considerably higher than their actual scores. For some rea­ son they (as well as those few high on PSA1 whose predicted APPln scores deviate substantially from their actual scores) manifest Inactivity despite motivational incentive on norma­ tive grounds to respond more aggressively. This suggests the possibility that our ideas about the incentive for aggressive action furnished by alienation from the political system may have to be sharpened by taking account of other sources of 18 See Ada W. Finifter, "Dimensions of Political Alienation," American Political Science Review, 64 (June 1970), esp. Figure 1 at p. 407; Jeffery M. Paige 5 "Political Orientation and Riot Participation," American Sociological Review, 36 (October 1971), esp. Chart 1 at p. 812; David C. Schwartz, Political Aliena­ tion and Political Behavior, esp. p. 156; Muller, "Behavioral Correlates of Political Support," American Political Science Review, 71 (June 1977).

Expectancy-VaIue-NormsTheory / 89

normative justification before a really strong linear relation­ ship will be observed. An independent source of normative justification not yet considered is ideology. As Gurr has put it: "Men's ideational systems, including their political ideologies, usually incorporate norms about the desirability of violence. They may prohibit the use of violence as an instrument of political competition, or prescribe violence as an historically justified response to political oppression."19 The argument applies to aggressive behavior in general, not just to violence. In West Germany today—though obviously not in the past— aggressive political action is more acceptable to the left than to the right. Broadly speaking, the left embraces a social change ideology that condones collective political aggression when it is necessary in the interests of social reform. By contrast, the right tends to be more concerned with social control and to feel that rules should be obeyed rather than challenged. Addition­ ally, specific leftist ideologies that provide a coherent rationale for collective political aggression—Marxism-Leninism, Mao­ ism, the welter of "New Left" strains of thought that arose in the 1960's—probably are more widely respected on the left than are similar specific ideologies—National Socialism, Fas­ cism—on the right. Thus, for this sample, at this time, it is most plausible to expect that a person's belief in the normative justifiability of aggressive political behavior will be higher, the more he manifests leftist ideological commitment. This also is probably true in the United States and in many other Western nations. It certainly is not a general principle, however. In a country such as contemporary Italy, normative justification for aggression would be expected to vary curvilinearly with degree of left-right ideological commitment, highest among both the extreme left and the extreme right, lowest among the center. To the extent that alienation from the structure of political authority is reinforced by ideological approbation of political aggression, normative incentive for aggressive political par19

Gurr, Why Men Rebel, p. 194.

90 I Expectancy-VaIue-NormsTheory

ticipation should be at its strongest. If alienation from the structure of political authority is not reinforced by ideological approbation of political aggression, normative incentive for participation in aggression should be relatively weak. Also, if a person is not alienated from the structure of political authority, then his normative incentive for participation in aggression should be relatively weak, even if his ideological commitment furnishes justification for such action. Finally, of course, when alienation is low and ideological commitment does not furnish justification, then normative incentive should be very weak. This line of reasoning implies that NBAp should be operationally defined as the product of degree of supportalienation and degree of commitment to an ideological posi­ tion that condones political aggression. A person's ideological commitment is measured by the LeftRight scale, previously shown to be a meaningful yardstick in the German context by Hans Klingemann.20 This instru­ ment is a horizontal scale containing ten boxes, arrayed from left to right:

LEFT

RIGHT

Respondents are told: Many people use the concepts "left" and "right" when it comes to characterizing different political viewpoints. We have here a measure which runs from left to right. When 20 Hans D. Klingemann, "Testing the Left-Right Continuum on a Sample of German Voters," Comparative Political Studies, 5 (April 1972), 93—106. It also appears to be the only ideological yardstick that has substantive meaning for mass publics in a variety of European countries—see Hans D. Klingemann and Ronald Inglehart, "Party Identification, Ideological Preference, and the Left-Right Dimension Among Western Publics," in Party Identification and Beyond, ed. Ian Budge, Ivor Crewe, and Dennis Farlie (New York: Wiley, 1976), 243-273.

Expectancy-Value-Norms Theory / 91

you think about your own political views, where would you place them on this scale ? Respondents then check off the box corresponding to their own position on the left-right ideological continuum. The variable, Left Ideological Commitment (LIC), is scored by assigning an integer from 1-10 to each box, beginning with 1 at the extreme right-hand box and ending with 10 at the extreme left. The NBA p term may now be defined operationally by the variable, Normative Justification for Aggression (NJA), set equal to the product of PSA 2 and LIC. Since the range of PSA 2 is 0-23.04 and the range of LIC is 1-10, NJA ranges from 0-230.4. Persons who score low on NJA do so either because they have low PSA 1 scores or because they have low LIC scores; those who score high on NJA must have high PSA2 and high LIC scores. Mean NJA is 47.25 (SD = 45.88), reflecting a heavy concen­ tration of cases in the very low range of NJA (0-48), which contains two-thirds of all respondents. Only 42 of 2,147 re­ spondents with valid scores on NJA score in the very high range (193-240). Does the NJA variable show a stronger linear relationship with APP|n than either PSA2 or LIC alone? The results, given in standardized regression weights, are: (4.1)

APP l n = .531(P&42), where r 2 = .282 and N= 2,003.

(4.2)

APP l n = .451(1/(7), where r 2 = .203 and N= 1,952.

(4.3)

APP xn = .628(NJA), where r 2 = .395 and N= 1,838.

The multiplicative measure of degree of belief in the normative justifiability of aggression clearly has a stronger relationship with aggressive political participation than either of its com­ ponents considered separately. Do PSA 2 and LIC have any linear effect on APP l a over and above that attributable to their multiplicative interaction as defined by the NJA variable? It is not possible to estimate reliably the impact of PSA 2 on ΛΡΡ, η , independent of NJA,

92 I Expectancy-Value-Norms Theory

because NJA is so very highly correlated with PSA 2 (r = .920) that there is hardly any independent variation between them. However, NJA and LIC do not show such extremely close asso­ ciation (r = .624). With LIC entered into the prediction equa­ tion, the result is: (4.4)

APP in = .620{NJA) + .097 (LIC),

where R 2 = .404 and N= 1,838. Since the weight attached to LICis so small (less than 0.1), it may be regarded as tantamount to zero estimated causal impact. Also, the accuracy of predic­ tion changes little when LIC is in the equation. Thus the NJA variable alone appears to capture the impact that PSA2 and LIC jointly have on APP la . The plot of the relationship between NJA and APP 1 n is given in Figure 4.5. The quite steep angle of the slope shows that NJA is estimated to have substantial linear impact on APP ln . Also, a rather sizable proportion of the variation in APP1n is accounted for by NJA. To illustrate, take Mr. X, a person with a score of 50 on the scale of normative justification for political aggression. Since NJA is very highly correlated with PSA2, Mr. X most likely supports the system of government. His predicted APP1n score falls in the Inactivity zone. He will not have taken part in any kind of aggressive political action and will have no intention of doing so. But now suppose that something happens to Mr. X that causes his support for the system to decline and/or his sympathy for a leftist, social change ideology to increase such that his NJA score goes up by 50 units. Now he registers a score of 100 on the normative justification scale. Since 1.622 plus .010 times 100 equals 2.622, a predicted APPln score solidly in the Low Civil Disobedience zone, Mr. X would be expected to express a positive intention toward participation in acts of non-violent civil disobedience. Were Mr. X's NJA score to increase by another 50 units, he would be expected actually to engage in some sort of non-violent but illegal protest action. His predicted APPln score now would fall in the High Civil Disobedience zone (1.622 plus .010 times 150 equals a predicted APPln score of 3.122). Finally, were Mr. X's NJA score to

Summary

Statistics Large circles denote 9 or more cases.

NORMATIVE JUSTIFICATION FOR AGGRESSION

(NJA)

FIGURE 4.5 Aggressive Political Participation as a Linear Function of Normative Justification for Aggression

94 / Expectancy-Value-Norms Theory

increase by 60 units more, reaching a score of 210 on the scale of normative justification for political aggression, he would be expected actually to engage in some sort of violent protest action, since 1.622 plus .010 times 210 equals a predicted APP xn score of 3.722, which falls in the Political Violence zone. Although the fit to a linear function is markedly improved when APPlri is predicted from NJA in place of PSA2, a fairly sizable proportion of the variation in APP\n still remains to be accounted for. Many of the deviations from predicted APPln scores represent people who manifest relatively low belief in the justifiability of aggression on normative grounds, yet score fairly high on the scale of participation in aggressive action. Perhaps some of these individuals score high on participation in aggressive action because they have high utilitarian incentive for such behavior, despite low normative incentive. Recall also that a fair number of people who scored low on the scale of utilitarian justification for aggression also registered rather high aggressive participation scores. Perhaps utilitarian and normative justification have separate effects on participa­ tion in aggressive action such that if one is high, aggressive participation scores will tend to be fairly high, even if the other is low. But for the purpose of estimating their separate effects on aggressive participation, they must not be too closely correlated. The correlation between UJA and NJA is .438. This is only a moderate relationship. It attests to the fact that two empiri­ cally distinct sources of incentive for participation in collective political aggression have been successfully measured. But does each make an independent contribution to prediction of par­ ticipation in aggressive action such that, when both are taken into account, predictive accuracy is clearly improved? The estimated standardized weights with both variables in the pre­ diction equation are: (4.5)

APP lri = .257 (UJA) + .516( N J A ) ,

where R 2 = .448 and N - 1,838. Each kind of motivational incentive is estimated to have a separate influence in deter­ mining magnitude of aggressive participation, but the influence

Expectancy-Value-Norms Theory / 95

of normative incentive is estimated to be twice as great as the influence of utilitarian incentive.21 Accuracy of prediction is improved, as the two types of incentive together account for 45 percent of the variation in magnitude of aggressive partici­ pation. 4.3

SOCIAL NORMS

The final task prior to testing the Expectancy-Value-Norms theory is to formulate an operational definition of the FSN* MCsn term from equation 2.5, facilitative social norms regarding participation in political aggression weighted by motivation to comply with such norms. The motivation to comply com­ ponent is difficult to define operationally and appears to be of questionable explanatory value. In their study of a variety of behaviors, Ajzen and Fishbein used an unweighted measure of social normative beliefs because weighting it by motivation to comply actually reduced the correlations with behavior.22 Since the number of questions that could be included in the interview schedule for this study was limited, no attempt was made to measure the motivation to comply component. It is excluded from this analysis on the assumption that it will not increase the size of the observed correlation between aggressive political participation and the indicator of FSN. Fishbein defines social normative beliefs as "the individual's belief about what 'society' (i.e. most other people, his 'signifi-

21 The possibility that the two types of incentive for aggressive action may interact multiplicatively also should be tested. The bivariate relationship is estimated to be: APP in = .554 (UJA*NJA), where r 2 = .307 and JV = 1,838. Predictive accuracy for the multiplicative hypothesis is much lower than for the additive hypothesis represented by equation 4.5. In fact, multiplying NJA by UJA yields considerably lower predictive accuracy than is afforded by the NJA variable alone. Also, when the multiplicative term is introduced into the equation with UJA and NJA as describing variables, it adds nothing to pre­ dictive accuracy and has a rather small weight with a theoretically implausible sign: APP ln = .329(UJA) + .580(jVJ^) - A29{UJA*NJA), where R 2 = .452 and JV = 1,838. 22 Ajzen and Fishbein, "The Prediction of Behavioral Intentions in a Choice Situation," Journal of Experimental Social Psychology, 5 (1969).

96 I Expectancy-VaIue-NormsTheory

cant others,' etc.) 'says' he should do (i.e. a social or group norm."23 But there are certain disadvantages in adopting this definition. For it focuses on the individual's subjective perception of social norms, a perception that may be heavily influenced by the individual's own normative beliefs. And if the indi­ vidual's perception of social normative beliefs coincides closely with his personal normative beliefs, then it will be difficult to disentangle any possible separate effects that these variables might have on behavior. Also, of course, the individual's per­ ception of social norms may not necessarily closely reflect objec­ tive reality, in which case it will not be a valid indicator of the prevailing social norms. What one really wants to determine is the degree to which individual behavior, independent of the individual's own atti­ tudes and beliefs, is affected by prevailing social norms about what is or is not proper behavior—by peer group influences, by the norms operative in the friendship groups or social organ­ izations to which the individual belongs, by the norms generally held in the community in which the individual lives. If social normative beliefs are defined as the average level of belief in the normative justifiability of a given action held by a person's "significant others" or by most other people in his community, then a direct method of measuring social normative beliefs would be to interview these "significant others," defined as people with whom the person has face-to-face contact, as well as a sample of people from the person's community. Measurement of the normative beliefs about political aggres­ sion held by a person's "significant others" would require inter­ views with close friends, work colleagues, and members of social and political organizations to which the person belongs. But interviewing the associates of each of the 2,662 people in this study would have been a prohibitively expensive undertaking. However, since samples of residents of the communities in which the respondents lived and/or worked were interviewed, it is possible to assess the impact of community norms on individual behavior. 23 Fishbein, "Attitude and the Prediction of Behavior," in Readings in Attitude Theory and Measurement, ed. Martin Fishbein (New York: Wiley, 1967), p. 489.

Expectancy-VaIue-NormsTheory / 97

Belief in the normative justifiability of collective political aggression differs markedly by type of community. Table 4,1 shows how Normative Justification for Aggression scores, collapsed into five intervals ranging from very low to very high, are distributed in the rural, urban, and university communities. In both the rural and urban groups, practically nine-tenths of the population falls in the very lowest interval on the norma­ tive justification scale. By contrast, less than one-half of those in the university sample fall in the very lowest interval. Medium to very high levels of normative justification (scores of 97 and over on NJ A) are manifested by less than one in fifty people in the rural communities and by less than one in thirty in the urban communities, whereas slightly more than one in four in the university communities manifest medium to very high levels. With regard to aggressive political participation, the rural and urban communities can be said to represent an inhibitory condition of social normative belief in the justifiability of ag­ gression. By comparison, the university communities can be said to represent a facilitative condition. The interpretation can be tested by means of OLS regression using the dummy variable technique.24 Dummy variables—a TABLE 4.1. DistributionofNormativeJustification

for Aggression by Type of Community Rural (JV=)

CO

193-240 145-192 97-144 49- 96 0 1

Mormative Justifica­ tion for Aggression

Total

(O) (O) (?)

(50) (377)

Urban %

(JV = )

%

0.0 0.0 1.6 11.5 86.9

(3) (5) (13) (56) (618)

0.4 0.7 1.9 8.0 88.9

(434) 100%

(695) 100%

University (JV = )

(39) (71) (156) (294) (458)

%

3.9 6.9 15.3 28.9 44.9

(1018) 100%

2 4 A useful discussion of dummy variables is given in Potluri Rao and Roger LeRoy Miller, Applied Econometrics (Belmont, Calif.: Wadsworth, 1971), pp. 88-105.

98 I

Expectancy-Value-Norms Theory

variable with two values (0 to 1) corresponding to the absence or presence of a qualitative condition—can be used to estimate the mean effect of a qualitative condition on a quantitative variable. In this case there are three qualitative conditions: rural milieu, urban milieu, university milieu. Because the data from Table 4.1 have shown that rural and urban communities have a much lower level of social normative belief in the justifiability of aggression than is manifested in university communities, one expects that the mean level of aggressive participation for indi­ viduals residing in rural and urban communities will be much lower than that for individuals residing in university communi­ ties. The mean level of aggressive participation for individuals in each milieu can be estimated by the following equation: (4.6)

APP ln = a + b 0 {URB) + 4,( UNV) + ε,

where APP l n = natural logarithms of the Aggressive Political Participation scale; URB = a dummy variable coded as 1 if the individual lives in an urban community, 0 otherwise; UMV = a dummy variable coded as 1 if the individual lives in a university community, 0 otherwise; a = a parameter that gives the estimated mean APPin score when both URB and UNV are equal to zero (by definition this is equal to the condition of residence in a rural community); b Q , b , = parameters that describe the estimated impact of residence in an urban community and resi­ dence in a university community, respectively, on APP ln ε = the error term expressing the assumption that APPln will not be determined exactly by the variables in the model. If residence in a rural or urban community has an inhibitory effect on individual participation in aggressive behavior be­ cause the social norms in these communities disfavor such action, then two conditions should hold: (1) a and bQ should

Expectancy-VaIue-NormsTheory / 99

be small, indicating that the average level of aggressive par­ ticipation in these communities is low; (2) bQ should be nearly zero because, just as there is little difference in the distribution of scores on the Normativejustification for Aggression variable between these communities, so should there be little difference between them in magnitude of aggressive participation. If residence in a university community has a facilitative effect on individual participation in aggressive behavior because the social norms in the universities favor such action (at least rela­ tive to the norms prevailing in the rural and urban communi­ ties) , then b should be large. The parameter estimates for the model expressed by equa­ tion 4.6 are: (4.7)

APPln = 1.629 + .050(ί/ΛΒ) + .876( £7JVF), (.031)

(.031)

where R2 = .374 and jV = 2,199. The standard errors of the regression coefficients are given in parentheses. To be con­ sidered reliably different from zero, a regression coefficient should be greater than at least two and preferably two and one-half times its standard error.2 5 The mean level of aggressive participation for persons resid­ ing in rural areas is estimated to be 1.629, a very low level. To determine mean aggressive participation for residents of urban communities, one takes a -\- bQ. This turns out to be 1.679, again a very low level. The URB regression coefficient is less than twice its standard error, indicative of essentially zero impact on aggressive participation. Mean aggressive participation is thus of similar magnitude in the rural and urban milieus. Mean aggressive participation in the university context, estimated by a -)- b v is 2.505, a much higher level than that estimated for the rural and urban settings. This finding is con25 For large samples, the null hypothesis may be rejected with 95 percent confidence if the /-ratio exceeds a value of 1.960, and with 99 percent con­ fidence if the ί-ratio exceeds a value of 2.576; hence the rule of thumb that the estimated parameter should be two or two and one-half times its standard error. I prefer the more conservative "two and one-half' rule because, with a very large sample such as this, very small parameter estimates, indicative of a trivial effect, can often turn out to exceed a t value of 1.960.

100 / Expectancy-VaIue-NormsTheory

sistent with the observation that social norms are facilitative in the universities relative to the norms prevailing in the non-uni­ versity communities. Since there is effectively no difference in mean aggressive participation between rural and urban communities, the URB dummy variable is unnecessary. The social normative beliefs term, FSJV, can be represented by the university versus nonuniversity dichotomy, the UJVV dummy variable. However, before we can take this as a satisfactory indicator of differing social norms, we must remove the embedded effect of a variable that has nothing to do with social norms but is sure to be highly correlated with the university versus non-university dichotomy: availability for collective action. 4.4

AVAILABILITY FOR COLLECTIVE ACTION

Certain objective socioeconomic and demographic attributes render certain people especially available for participation in collective action of any kind. Single people, students, the un­ employed or only part-time employed have larger amounts of unscheduled or uncommitted time at their disposal than married people and persons employed full-time. Age is a par­ ticularly important summary indicator of availability. The young, because they are more likely to be unmarried, to be students, and to be unemployed or only part-time employed, have more free time than the old (excepting retired persons). In addition, the young, on the average, are more likely to be available for collective action than the old by virtue of better health, greater energy, and perhaps higher general enthusiasm for social activism. All of these "background" characteristics that promote greater availability are more likely to be found among individuals at the universities than among those in the rural and urban communities. The distribution of marital status, employment status, and age in the rural, urban, and university communities is given in Table 4.2. In the rural and urban communities, 78.3 and 73.2 percent, respectively, are married. In the universities only 28.3 percent are married. The rural and urban communities

100%

(1104) 100%

(990) 100%

(569)

Total

73.5 16.6 6.2 2.2 1.6

(811) (183) (68) (24) (18)

4.7 15.2 15.7 17.8 46.7 (47) (150) (155) (176) (462)

15.5 21.4 19.3 18.1 25.7

(88) (122) (110) (103) (146)

Under 30* Thirties* Forties Fifties 60 and up

100% (1080) 100%

68.8 0.1 3.0 27.8 0.3 0.1

(743) (1) (32) (300) (3) (1)

(978)

(569)

Total

100%

(1083)

100%

0.9 1.6 4.2 36.9 18.5 38.0

Student* (5) Unemployed* (9) Part-Time Employed* (24) (210) Full-Time Employed Retired (105) Housewife (216)

100%

(978)

(762) (12) (3) (306)

/0

0Z

70.4 1.1 0.3 28.3

University

(* = )

1.0 0.8 6.1 36.4 32.8 22.8

100%

(566)

Total

5.1 4.1 17.6 73.2

Zo

0Z

(50) (40) (172) (716)

Urban (JV = )

(10) (8) (60) (356) (321) (223)

7.4 1.2 13.1 78.3

(42) (7) (74) (443)

/0

0Z

Single* Divorced* Widowed* Married

* Condition presumed to foster availability.

Age

Employment Status

Marital Status

Rural (JV = )

Type of Community

TABLE 4.2. Distribution of Availability Characteristics by Type of Community

102 I Expectancy-VaIue-NormsTheory

contain many housewives and retired persons, but these condi­ tions cannot be regarded as particularly conducive to collective action since the free time attendant upon absence of employ­ ment in these instances is counterbalanced by lack of avail­ ability due to marriage and old age. The university samples, of course, contain many students, and their free time attendant upon absence of employment is not counterbalanced by marital status or age. Ofspecial significance is the vast difference among the communities in the proportion of persons under thirty years of age. Since age reflects or stands for a variety of characteristics that render persons more or less available for participation in collective action, let us take age as the basic indicator of avail­ ability. What kind of general relationship might we expect be­ tween age and aggressive political participation? It may safely be assumed that the direction of the relationship will be nega­ tive. But what about its form? A quite plausible hypothesis is to predict that, instead of being linear, the relationship will be curvilinear, such that aggressive participation scores decline rapidly to a very low level in the range of 18 to 40 or so years of age, and this very low level remains constant beyond ap­ proximately age 40. In addition, it also is plausible to expect that the accuracy of the curvilinear slope will be much better among the old than among the young. That is, the relationship probably will be to some extent non-functional, in the sense that youth is only a necessary, not a necessary and sufficient, condition of non-low aggressive participation scores. And any slope, curved or straight, that describes a functional relation­ ship will therefore be rather inaccurate among the young. A plot of the relationship between AGE and APPin is given in Figure 4.6. This plot affords an instructive example of how mis­ leading a correlation sometimes can be. Taking the linear r2 value alone, we would say that a fairly strong negative rela­ tionship obtains between a person's age and his magnitude of aggressive political participation. But when we look at the linear slope, it is apparent that age is not a very good predictor of magnitude of aggressive participation. The r2, corresponding to a correlation of —0.51, is relatively high only because so

Summary

Statistics: Linear

Large circles denote 9 or more cases.

AGE

Statistics:

Curvilinear N = 2199 r" = .295 APP1„ = 1.123 = 31.1(1/AGE)

Summary

FIGURE 4.6 Aggressive Political Participation as a Linear and Curvilinear Function of Age

104 / Expectancy-VaIue-NormsTheory

few people over the age of about 40 register anything other than inactivity scores on APPln. Among the young, where the vast majority of non-low aggressive participation occurs, there is great variability about the linear slope. Virtually all people registering in the High Civil Disobedience and Political Vio­ lence zones of APPln are concentrated in the 18-32 years of age group. Yet according to the linear slope; most of these people are predicted to manifest low aggressive political participation. The young thus are predicted to manifest low aggressive politi­ cal participation despite the fact that the preponderance of non-low aggressive participation is confined to them. A curvilinear function that might well describe the rela­ tionship between age and aggressive participation is a negative power function of the form (4.8)

APP ln = a + b 0 (AGE~ ! ),

where AGE~ 1 may be expressed as 1 /AGE, i.e., the reciprocal of age. In words, equation 4.8 says this: magnitude of aggressive response will decline rapidly with each yearly increase in age among people at the low end of the age scale, but as the middle range of the age scale is approached, the rate of decline in aggressive response will slow dramatically, leveling off at very low values among middle-aged, old, and very old respondents. The reciprocal of age accounts for 30 percent of the variation in magnitude of aggressive response, indicating that the curvi­ linear model expressed by equation 4.8 is preferable to the linear formulation. More sensitive than the linear function to the concentration of non-low aggressive participation in the 18-32 age bracket, the curvilinear function also gives a more accurate representation of the fact that, once a person reaches middleage, he is quite unlikely to score at anything other than the very lowest position on the scale of aggressive political par­ ticipation.26 26 Of course, one would not want to extrapolate about ΑΡΡ λ η scores, on the basis of the reciprocal of age, beyond the age range actually observed. For persons under age eighteen, such an extrapolation leads to amusing results, since the curve described by the reciprocal transformation ascends very sharply and asymptotically to they (APP ln ) axis as the χ (AGE) variable approaches

Expectancy-VaIue-NormsTheory / 105

Lack of employment and being unmarried also are poten­ tially facilitative conditions. A general index of availability for collective action should take such conditions into account, in addition to age. Marital status is scored as 1 if it conforms to one of the condi­ tions denoted by an asterisk in Table 4.2; otherwise it is scored as 0.2 7 Employment status is given a score of 2 if it conforms to one of the conditions denoted by an asterisk in Table 4.2; otherwise it is scored as 0. Thus, conditions of marital status and employment status presumed to foster availability are differentiated from conditions presumed to limit availability such that they can be combined into a composite quantitative index of availability, defined as:28 (4.9)

A = AGEr + MS+ ES,

where A is the overall availability score (range: 1.1-8.6), AGEr is the reciprocal of a person's age in years,29 MS denotes

zero. To illustrate, extrapolation from the equation given in Figure 4.6 gives expected APP ln scores of 3.715 for a child of twelve, 5.011 for a child of eight, 8.898 for a child of four, 16.673 for a child of two and 32.223 for a one-year-old infant! 27 The facilitative conditions of marital status include only persons with no dependent children. 28 In building the composite index, greater weight is given to facilitative conditions of employment status than to facilitative conditions of marital status because the former is estimated to have greater impact on aggressive political participation than the latter. The regression equation (with parameters expressed as standardized weights) for prediction of a person's aggressive participation behavior score on the basis of the reciprocal of age, marital status, and employment status is APP ln = .288(,46^) + A09(MS) + .257(ES). Em­ ployment status has a little more than twice the impact of marital status. The R 2 for this equation is .343, virtually identical to the r 2 of .338 (see Figure 4.7) obtained for the composite index of availability with employment status given twice the weight of marital status. 29 The reciprocal of a person's age is multiplied by 100 in order to avoid having to deal with very small numbers. The effect of the reciprocal transforma­ tion is to reverse the direction of age, radically compress the middle to very old ages into a very small range at the low end of the transformed scale, and spread out the young over the remainder of the range. This can be seen from the following conversion table: AGE: AGE r :

18, 5.6,

25, 4.0,

32, 3.1,

39, 2.6,

46, 2.2,

53, 1.9,

60, 1.7,

67, 1.5,

74, 1.4,

81, 1.2,

88 1.1

106 I Expectancy-Value-Norms Theory

marital status, and E S denotes employment status. The effect of equation 4.9 is to boost an individual's availability score derived from AGEr alone by one unit if his marital status affords additional availability, by two units if his employment status affords additional availability, and by three units if both marital and employment status afford additional availability. The relationship between A and APPln is shown in Figure 4.7. Accuracy of prediction is improved, relative to AGEr alone, as the composite index of availability accounts for 34 percent of the variation in aggressive political participation. This is primarily because young people who are fully employed and/or married receive lower A scores than their unmarried and nonemployed counterparts. And since the employed and/or married young are less likely to register non-low aggressive participation scores than the non-employed and unmarried young, purging them from the high end of the index of avail­ ability means that fewer sizably deviant APPln scores occur at high availability. The correlation between the index of availability and the indicator of inhibitory-facilitative social norms is .740. As anticipated, the inhibitory-facilitative social norms variable is heavily contaminated by association with availability for collective action, reflecting to a substantial degree differences in this latter variable. One may test for the effect of inhibitory-facilitative social norms on aggressive political participation, independent of availability, by estimating the parameters of the following model: (4.10)

APPlu = a +b0(UJW) +b^A) +ε,

where the variables are as previously defined; the a parameter gives the estimated mean APPln score for persons residing in

Thus, all people over roughly the age of fifty years are located in the interval 1.0 to 2.0 of AGE r ; people from about fifty down to their early thirties are located in the interval 2.0 to 3.0 of AGE r ; people from their early thirties down to age twenty-five are located in the interval 3.0 to 4.0 of AGE r ; and the young twenties and late 'teens group is located in the interval 4.0 to 5.6 of AGE r .

Summary Statistics N = 2004 r2 = .338 APP in= 1.341 + .17 s(A)

Large circles denote 9 or more cases.

AVAILABILITY INDEX (A)

FIGURE 4.7 Aggressive Political Participation as a Linear Function of the Index of Availability

108 I Expectancy-VaIue-NormsTheory

rural and urban communities, independent of their availabil­ ity; a + bQ gives the estimated mean APPln score for persons in the university milieu, independent of their availability; and the b j parameter describes the estimated mean effect of each unit increase in availability on APPln, independent of com­ munity type. The parameter estimates are: (4.11)

APP ln = 1.444 + .552 (UNV) + .083 {A) , (.035)

(.007)

where R 2 = .410 and N = 2,004. Removing the effect of dif­ ferences in availability from persons in the university milieu results in an estimated mean aggressive participation score for them of 1.996. The initial estimate of the effect on aggressive response of exposure to facilitative social norms therefore must be revised significantly downward, for it was contaminated by the fact that persons exposed to facilitative social norms also are likely to have attributes that render them especially avail­ able for participation in acts of collective political aggression. Given the same level of availability, persons exposed to facilita­ tive social norms still manifest higher mean aggressive response than those exposed to inhibitory norms, but the difference is hardly dramatic. Figure 4.8 depicts the effect of differences in degree of avail­ ability on aggressive participation among persons in the rural and urban communities and among those in the university milieu. Even when the effect due to availability is added to the mean rural and urban effect, predicted aggressive response does not rise out of the Inactivity zone. But when the avail­ ability effect is added to the mean university effect, predicted aggressive response rises into the Low Civil Disobedience zone at about the midpoint of the index of availability. Recall that the Low Civil Disobedience zone is characterized by positive intention to engage in such action, but absence of actual par­ ticipation in acts of civil disobedience. This, then, is the level of aggressive response predicted by the model for persons who are exposed to facilitative social norms and whose age, marital status, and employment status render them especially available for collective action.

AVAILABILITY INDEX (A)

FIGURE 4.8 Estimated Effect of Availability on Aggressive Participation by Type of Community

110 I Expectancy-VaIue-NormsTheory

4.5

TESTING THE EXPECTANGY-VALUE-NORMS MODEL

The auxiliary theory specifying operational indicators (mea­ sured variables) of the concepts in the Expectancy-ValueNorms theory may be expressed by the following model: (4.12)

APP l a = a + b 0 ( U J A ) + b 1 [NJA) + b 2 {UNV] + b , { A ) + c0(UNV* UJA) + c j (UNV*NJA) + ε,

where APP l a = Aggressive Political Participation logged to the base e, range 1.54-4.52; U J A = Utilitarian Justification for Aggression, range 0-14 (indicator of attitude about the action); N J A = Normative Justification for Aggression, range 0-230.4 (indicator of personal normative be­ liefs) ; U N V = Inhibitory-Facilitative Social Norms, scored as 0 for rural and urban communities, 1 for univeristy communities (indicator of social norma­ tive beliefs); A = Availability for Collective Action, range 1.1-8.6 (not part of the Expectancy-Value-Norms theory, but necessary to the auxiliary theory in order to achieve an indicator of social norms that is not confounded with availability); UNV* U J A = an indicator of interaction between social norms and attitude about the action; UNV* N J A = an indicator of interaction between social norms and personal normative beliefs. Two different versions of the Expectancy-Value-Norms theory have been proposed. An "additivity of effects" version, expressed by equation 2.5, postulates that the Ci parameters in equation 4.12 should equal approximately zero. Under the "additivity of effects" model, it is assumed that the effect of utilitarian incentive and normative incentive on aggressive political participation is the same, whether a person comes from a milieu where social norms are inhibitory or from one where they are facilitative.

Expectancy-VaIue-NormsTheory /111

An "interaction" version, expressed by equation 2.6, postulates that the bt parameters (excepting b 3) in equation 4.12 should equal approximately zero. According to the "inter­ action" model, a person's belief in the utilitarian justifiability of aggressive action, as well as his belief in the normative justi­ fiability of such behavior, have no effect on his degree of aggres­ sive political participation unless the social norms to which he is exposed condone such behavior. A "complete interaction" model, expressed by equation 2.7, would have the bi parameters (excepting b and the Ci pa­ rameters equal to zero. Under such a model the only signifi­ cant effect would be due to a triple interaction of the form (UNVr* UJA*NJA). A term representing this interaction is not included in equation 4.12 because it has already been found that UJA and NJA do not interact multiplicatively in the determination of APP1n (see footnote 21). There is a crucial difference between the "additivity of effects" and "interaction" versions of the Expectancy-ValueNorms theory. The former states that utilitarian and normative incentives for political aggression are general preconditions of participation in such action. The latter attaches a condition, namely, that social norms first must be facilitative. The "inter­ action" version is a model for a restricted Expectancy-ValueNorms theory. Should the bt parameters (excepting b in equation 4.12 turn out to be zero, this means that the Expec­ tancy·Value-Norms theory holds only in communities where social norms condone political aggression, that is, it is restricted to that condition. The "additivity of effects" version is a model for a general Expectancy-Value-Norms theory. But note that in this case it is not necessary for the Ci parameters to equal zero if a general Expectancy-Value-Norms theory is to hold. So long as the bi parameters are non-zero, the generality of the theory is upheld. If the Ci parameters turn out to be significantly different from zero, this simply means that the effects of utilitarian justification and normative justification for aggres­ sion are amplified under the condition of facilitative social norms. Equation 4.12 thus states a model that is congruent with the general Expectancy-Value-Norms theory, but in this

112 / Expectancy-VaIue-NormsTheory

case, if all the parameters are significantly different from zero, both additive and interactive effects occur. The model expressed by this equation will be referred to as an "additivity-amplification" model. Now let us estimate the parameters for the model of the Expectancy-Value-Norms theory. The results are: (4.13)

APP la = 1.396 + .037 (UJA) + .003 {NJA) + .221( UNV) (.008)

(.0006)

(.042)

+ .047 (4) + .010 ( UNV* UJA) (.007) (.010) + .003 (UNV* NJA), (.0007) where R 2 = .569 and JV = 1,838. Since the parameter estimate for UNV* UJA is equal to its standard error, one may infer that this parameter is not reliably different from zero. Setting cQ equal to zero (which amounts to exclusion of the UNV* UJA term from the equation) gives the following parameter esti­ mates : (4.14)

APP Xn =\.Z9Q + m^{UJA) + .m{NJA) + .2?,&{UNV) (.005)

(.0006)

(.039)

+ .046(4) + .0M{UNV*NJA), (.007) (.0006) where R 2 = .569 and JV = 1,838. Now all the parameter esti­ mates are greater than two and one-half times their standard error, indicating that, unless one or more important exogenous variables, highly correlated with the response and describing variables, have been left out of the model, the impact on ag­ gressive participation of each describing variable is reliably greater than zero. Since all of the b parameters are reliably greater than zero, the restricted Expectancy-Value-Norms theory, expressed by the "interaction" model, may be rejected. The "additivity of effects" version of the general Expectancy-Value-Norms theory may also be rejected, since the C1 coefficient is non-zero. The equation these data fit is the "additivity-amplification" version of the general Expectancy-Value-Norms theory.

Expectancy-VaIue-NormsTheory /113

The amplification effect due to the interaction of personal and social normative beliefs can be grasped visually from Figure 4.9. In the rural and urban communities, where social norms are inhibitory, expected aggressive participation is 1.390, when all other describing variables are held constant. But even under this condition of inhibitory social norms, and inde­ pendent of the other describing variables, every increase of 10 units of NJA is expected to add 0.03 units to a person's degree of aggressive political participation. The simple fact of being in a university community, where social norms regarding aggressive political behavior are comparatively facilitative, is expected to add .236 units to a person's degree of aggressive political participation, independent of all other variables in the model. Now comes the amplification effect. Where social norms are facilitative, every increase of 10 units of NJA is expected to add 0.07 units to a person's degree of aggressive political participation. Thus the effect of NJA on APPln is amplified—more than doubled—by the presence of facilitative social norms. The effect of utilitarian incentive on aggressive political participation is not amplified by facilitative social norms. Therefore, the slope of the relationship between APPln and UJA in the university communities is parallel to the slope that obtains in the rural and urban communities. Figure 4.10 de­ picts this. Motivation for political aggression afforded by the expecta­ tion of utilitarian gain appears to operate at a relatively con­ stant level, impervious to differences in prevailing social norms about the justifiability of aggressive action. Attitude about the action, in the sense of utilitarian incentive, appears to be a very general motivating force, as assumed in various SEU (subjec­ tively expected utility) decision-theory models of behavior.30 Independent of the other describing variables, availability for collective action has only a relatively small impact on ag­ gressive political participation, as shown by the flatness of the 50 A succinct review of various SEU models of behavior is given in Ν. T. Feather, "Subjective Probability and Decision Under Uncertainty," Psychological Review i 66 (1959), 150—164.

NORMATIVE JUSTIFICATION FOR AGGRESSION

(NJA)

FIGURE 4.9 Estimated Effect of Normative Justification for Aggression on Aggressive Participation by Type of Community Holding All Other Describing Variables (Xt) Constant

UTILITARIAN JUSTIFICATION FOR AGGRESSION (V]A)

FIGURE 4.10 Estimated Effect of Utilitarian Justification for Aggression on Aggressive Participation by Type of Community Holding All Other Describing Variables Constant

116 / Expectancy-VaIue-NormsTheory

slopes depicted in Figure 4.11. For example, in the absence of utilitarian and normative incentive, the expected aggressive participation of a person who is not fully employed, unmarried, and in his early twenties (an Availability score of about 8) is less than 2.00 even when social norms are facilitative (uni­ versity milieu). And the expected aggressive participation of his counterpart in a rural or urban community, in the absence of utilitarian and normative incentive, is less than 1.66. In­ activity thus is the expected level of aggressive participation in both instances of persons with very high availability for col­ lective action. This calls into question the hypothesis put for­ ward by Clark McPhail, who has argued that the major pre­ condition of civil disorders, riot participation in particular, is high availability.31 Of course, although availability does not appear to play the dominant explanatory role in the deter­ mination of aggressive political participation, it is estimated to have some direct, independent impact, thereby attesting to the correctness of McPhaiPs insight in calling attention to this factor. The Expectancy-Value-Norms theory must be expanded to take into account the role of pure availability for collective action, in addition to utilitarian belief, personal normative belief, and social norms.

4.6

SUMMARY

As derived from Fishbein's general behavior theory, three abstract concepts, E*V, NBAp, and FSN were postulated to have independent causal influence on aggressive political par­ ticipation. The E*V term represents utilitarian incentive for such behavior; the NBAp term represents normative incentive for such behavior," and the FSN term represents exposure to favorable social normative beliefs about such behavior that might facilitate individual participation in politically aggres­ sive action.

31 See McPhail, "Civil Disorder Participation: A Critical Examination of Recent Research," American Sociological Review, 36 (December 1971), 1058— 1073.

AVAILABILITY INDEX (A)

FIGURE 4.11 Estimated Effect of Availability on Aggressive Participation by Type of Community Holding All Other Describing Variables {Xt) Constant

118 / Expectancy-Value-Norms Theory

It was then necessary to formulate an auxiliary theory stating (1) how indicators of these abstract concepts should be defined operationally and (2) specifying the form of the relationship between the indicator variables and the aggressive participation variable. An indicator of the Ε* V term, UtilitarianJustification for Aggression, was constructed by weighting a person's degree of belief that collective political aggression has led to goal attainment by his political influence capability, such that scores on the scale of utilitarian justification are highest when collec­ tive political aggression is perceived as having been beneficial and the person feels that he has insufficient influence capability. An indicator of the JVBAp term, Normative Justification for Aggression, was constructed by weighting a person's degree of alienation from the political system (using the square of his alienation score) by the degree to which he registers leftist ideological commitment, such that scores on the scale of norma­ tive justification are highest for persons who are both alienated from the system and manifest leftist commitment. Normative and utilitarian incentive are distinct conceptually and they do not show especially close empirical correlation. It was predicted that the utilitarian and normative justification variables each would have independent linear impact on an individual's magnitude of aggressive political participation. This turned out to be the case, although a standard unit increase in normative incentive was estimated to produce twice as much increase in aggressive participation as a standard unit increase in utilitarian incentive, suggesting that any motivational theory of aggressive political action that is limited to utilitarian incen­ tive will be insufficient. Variables representing normative in­ centive for such behavior are a quite important part of the explanation. On the basis of these results, the Utilitarian Justification model, as expressed by equation 2.9, was rejected on the grounds of being radically incomplete. The next step was to develop an indicator of exposure to favorable social normative beliefs about political aggression. The effect of facilitative social norms on participation in acts of political aggression was indexed by a variable scored as 1

Expectancy-VaIue-NormsTheory /119

for communities where prevailing social norms are compara­ tively favorable to collective political aggression, and O for communities where prevailing social norms heavily disfavor it. Since this indicator strongly reflected differences in pure availability for collective action of any kind (communities where norms are comparatively favorable also containing much larger concentrations of persons with high availablity than communities where norms are unfavorable), a separate com­ posite measure of availability was incorporated into the model for the Expectancy-Value-Norms theory, in order to be able to assess more directly the effect of facilitative social norms on behavior, independent and therefore uncontaminated by any effect due to differences in availability. Having defined indicators of the concepts in the ExpectancyValue-Norms theory, the next question faced was the appro­ priate specification of the model for the theory. Would the data fit a general Expectancy-Value-Norms model, expressed either as an "additivity of effects" version or as an "additivityamplification" version? Or would the data fit only a restricted Expectancy-Value-Norms model, according to which motiva­ tional incentive for aggressive action afforded by utilitarian and normative justification would affect aggressive political participation only under the condition that social norms were facilitative? The results supported an "additivity-amplification" version of the general Expectancy-Value-Norms model. Both the Utilitarian Justification for Aggression and Normative Justi­ fication for Aggression variables were found to affect a person's magnitude of aggressive participation under conditions of inhibitory as well as facilitative social norms. And the effect of the Normative Justification for Aggression variable was con­ siderably amplified where prevailing social norms were com­ paratively favorable to political aggression. However, the effect of the Utilitarian Justification for Aggression variable on Aggressive Political Participation was not amplified by the presence of facilitative social norms. This suggests that utili­ tarian incentive may be a rather general determinant of ag-

120 I Expectancy-VaIue-NormsTheory

gressive participation, having impact that is relatively imperv­ ious to differences in contextual variables such as prevailing social norms. The "additivity-amplification" model of the general Expec­ tancy-Value-Norms theory accounted for nearly three-fifths of the variation in Aggressive Political Participation. Given that the variables were measured at the micro level, and that the model was tested across a very large and heterogeneous sample of persons, this is a satisfyingly high level of predictive accuracy. It suggests that the Expectancy-Value-Norms theory may have real value as a general explanation of individual differences in degree of aggressive participation. But before we can even begin to accept the model estimated by equation 4.14 as essentially correct, we must provide answers to the questions of how complete it is and how robust its parameter estimates are. Have important explanatory variables been ignored? If so, when they are taken into account, what will happen to the parameter estimates obtained in equation 4.14? Will some be reduced or even cut down to approximately zero? Will there be amplification effects due to variables not yet considered that partially or drastically alter some of the parameters? These are the questions to be taken up in the next two chapters. First, the model for a comprehensive alternative explanation, the Relative Deprivation theory, will be subjected to an extensive investigation. Following this, consideration will be given to a number of additional variables that, it is plausible to hypothe­ size, might have been omitted incorrectly from the models elaborated in Chapter Two.

FIVE

FrustrationAggression Theory

For many years the frustration-aggression hypothesis, put forth by Dollard and his associates, was the standard explanation of aggressive behavior in general.1 The strong version of this hy­ pothesis said that frustration, defined as interference with goaldirected activity, always produces aggression. A slightly weaker version, stated by Miller, said that frustration does not always lead to aggression (other reactions to frustration may occur), but that the occurrence of aggression always presupposes the existence of frustration.2 However, Miller argued that if frustration persisted long enough, alternative non-aggressive reactions would be weakened, and aggression eventually would occur as an inevitable response to continuing frustration. Leonard Berkowitz has proposed a modification of the frustration-aggression hypothesis.3 He rejects the monocausal frustration-aggression argument as being at once too simplistic and too sweeping: too simplistic because it does not recognize other motivational factors—such as utilitarian incentive—that can lead to aggression even in the absence of frustration; too

1 Dollard et al., Frustration and Aggression (New Haven: Yale University Press, 1939). 2 Neal E. Miller, "The Frustration-Aggression Hypothesis," Psychological Review, 48 (1941). 3 See Leonard Berkowitz, "The Frustration-Aggression Hypothesis Re­ visited," in Roots of Aggression, ed. Leonard Berkowitz (New York: Atherton Press, 1969), 1-28; Berkowitz 5 "Frustration, Comparisons, and Other Sources of Emotional Arousal as Contributors to Social Unrest," Journal of Social Issues, 28 (1972), 77-91.

122 I Frustration-Aggression Theory

sweeping because it postulates a deterministic (aggression is certain) instead of a probabilistic (aggression is in some degree likely) relationship between the occurrence of aggressive be­ havior and the existence of frustration. Essentially, Berkowitz's revised frustration-aggression hypothesis proposed that (1) the occurrence of frustration merely increases the likelihood of an aggressive response—it does not inevitably lead to such; (2) persisting frustration does not increase the likelihood of an aggressive response because it may often lead to apathy; (3) the occurrence of aggression does not presuppose frustration because it may have been caused by other kinds of motivational incentive. In fact, Berkowitz's reformulation of the frustrationaggression hypothesis, while not an out-and-out rejection of it, constitutes a considerably watered-down version. As Berkowitz says: Where Dollard and his colleagues at Yale, together with many others, had posited a very close connection between frustration and aggression, my own thinking contends that the linkage is fairly tenuous. Frustration enhances the probability of aggression somewhat, but other factors also help determine whether any violence will actually occur. These conditions are very important and may even override the influence of thwarting [interference with goal-directed behavior] per se. Frustration is by no means the inevitable precursor of aggression.4 Similarly, after a thorough review of laboratory studies in which conditions regarded as frustrative were varied and the effect on aggressive response then observed, Bandura concluded that "aversive events commonly subsumed under the term frustra­ tion—be they physical assaults, insults, reinforcement with­ drawal, or thwarting of goal-directed activities—are, at best, facilitative rather than necessary or sufficient conditions for aggression. 'Frustration' is most likely to provoke aggression

4

Berkowitz, "Frustrations, Comparisons, and Other Sources," p. 80.

Frustration-AggressionTheory / 123

in individuals who have learned to behave aggressively and for whom aggression has functional value."5 Berkowitz and Bandura both hypothesize that frustration is a Jacilitatwe condition of aggression. There are, however, two ways in which frustration might function as a facilitative instead of a necessary or sufficient condition of aggression. To illustrate, consider an example with three variables: aggressive political participation as the response variable; the functional value of aggressive political behavior—i.e., utilitarian justification for aggression—as one causal variable; and frustration as the other causal variable. The first way that frustration could be a facilitative condi­ tion is by the familiar amplification effect. In this case aggressive political participation would be affected by utilitarian incen­ tive, independent of frustration, but the effect of utilitarian incentive would be enhanced by the presence of frustration. For example, if people in a laboratory experiment were sub­ jected to a frustrating experience, but were taught to attach low utilitarian value to an aggressive response, they would not be expected to behave aggressively. Hence frustration could not be regarded as a sufficient condition of aggression. On the other hand, people who were not subjected to a frustrating experience, but were taught to attach high utilitarian value to an aggressive response, would indeed be expected to behave aggressively. Frustration therefore could not even be regarded as a necessary condition of aggression. However, those experi­ encing frustration plus high utility would be expected to mani­ fest a greater degree of aggressive response than those experi­ encing no frustration and high utility. And thus frustration would be a facilitative condition. The second way this might happen is through an indirect effect. Such an effect occurs when an antecedent variable can be linked to a response variable only by virtue of the mediation of a third, intervening, variable. In this case frustration would have an effect on utilitarian incentive; utilitarian incentive ° Albert Bandura, Aggression: A Social Learning Analysis (Englewood Cliffs, NJ,: Prentice-Hall i 1973), p. 174.

1 24 I Frustration-Aggression Theory

would have an effect on aggressive political participation; in­ dependent of utilitarian incentive, frustration would have no direct causal effect on aggressive participation. In other words, according to this possibility, if frustration were increased, utilitarian justification would increase, producing an increase in aggressive political participation; however, if utilitarian justification were to increase due to some cause other than frustration, aggressive political participation still would in­ crease, even in the absence of a change in frustration. Thus, frustration would be neither a necessary nor sufficient condi­ tion of aggression but could be regarded as indirectly facilitative because of its relationship to a direct cause of aggressive response. Most contemporary social scientists who have attempted to develop general theories relating frustration to political aggres­ sion have recognized that the original frustration-aggression hypothesis was too simplistic. They have proposed revised hypotheses that ascribe one or another of three different roles to frustration. An approach entailing only a slight revision of the basic hy­ pothesis is represented principally by the work of Gurr and appears in the work of Davies and the Feierabends.6 Frustra­ tion, stemming from discrepancy between expectation and 6 The formal model of Gurr's theory, given in Gurr and Duvall, "Civil Conflict in the 1960's: A Reciprocal Theoretical System with Parameter Estimates," Comparative Political Studies, 6 (July 1973), is discussed in Chapter Two. Davies's Rise and Drop hypothesis, given in Davies, "Toward a Theory of Revolution," American Sociological Review, 27 (February 1962), also is dis­ cussed in Chapter Two. Although the Rise and Drop hypothesis is basically a structural hypothesis, Davies emphasizes that the causal mechanism leading directly to collective political aggression is the psychological state of frustra­ tion that results from the rise and drop pattern of socioeconomic change. Implicit in his analysis is the argument that both frustration and belief in the normative justifiability of aggression are important in the determination of collective political aggression. If they both are high, as in the case of Dorr's rebellion in the United States in 1842, the Russian revolution of 1917, and the Egyptian revolution of 1852, revolution is the result. If one or the other is low, as in the case of the depression of the 1930's in the United States where frustra­ tion was high but normative justifiability was low, revolution does not ensue. The Feierabends' application of frustration-aggression theory to collective political aggression is given in Ivo K. Feierabend and Rosalind L. Feierabend 5 "Aggressive Behaviors within Polities, 1948—1962: A Cross-National Study,"

Frustration-Aggression Theory / 125

achievement, is posited to have direct causal influence on aggressive political participation. However, other variables— such as utilitarian and normative justification for aggression— are also important. Their influence is amplified or inhibited depending on whether frustration is high or low. In this in­ stance, frustration is assigned what could be called a DirectAmplification role in the determination of aggressive political participation. A model expressing this Direct-Amplification role for frustration can be written as: (5.1)

A P P = a + b 0 ( F ) + C0 ( Z ^ 1 ) + · · · + c n ( F * Z „ )

+ ε,

where A P P denotes aggressive political participation, F denotes level of frustration, and and Zn denote additional causes of aggressive participation. The facilitative role assigned to frustration by Berkowitz and Bandura, entailing a major modification of the frustrationaggression hypothesis, can be expressed as either a FacilitativeAmplification or a Facilitative-Indirect model. An equation for a Facilitative-Amplification model is: (5.2)

APP = a + bQ(Zx) +••• + K(Zn) + c^F*Zx) + · · · + Cn(F*Zn)+e.

The difference between equations 5.1 and 5.2 is that, in the former, frustration is postulated by the b0 weight to have a direct effect on aggressive political participation independent of its interaction with the variables, whereas, in the latter, it is the variables that are postulated by the b0 to bn weights to have direct impact on aggressive political participation, inde­ pendent of their interaction with frustration. The Utility theory of Korpi postulates that frustration, as indexed by relative deprivation, affects aggressive political participation indirectly through its relationship to an inter­ vening variable, the subjective utility of aggressive action to Journal of Conflict Resolution, 10 (September 1966), 249-271. These authors postulate an inhibitory-amplification interaction between frustration and a variety of conditions, including indicators of normative justification for aggres­ sion (eifectiveness of government and legitimacy of regimes) and utilitarian justification for aggression (coerciveness of government).

126 I Frustration-AggressionTheory

the individual.7 A model for a theory postulating an indirect frustration effect requires two "stages" of equations, where the first stage expresses the relationship between frustration and the intervening variables, while the second stage expresses the subsequent relationship between the intervening variables and aggressive political participation. Equations for a FacilitativeIndirect model are: (5.3a)

Z\

(5.3b)

Zn

(5.3c)

= a

+b

= + a

0

F

+^1(X1) + · • ·

b Q F + ^1(X1)

APP = a + b + · ·

·+

-\-b n (X n )

+ · ·· +

b„(X n )

+ ε, + ε,

+ ε,

where the X1 variables are additional causes of the variables, independent of frustration.8 In the first stage of the model, equations 5.3a and 5.3b, frustration is posited as affecting the 4 to 1 3 to 1 2 to 1 1 to 1 S). Moreover, within the intellectual proletariat type, there exists only a faint trace of high disequilibrium. Thus, the data are biased against the rank disequilibrium hypothesis from the start, since it is precisely high disequilibrium of the intellectual proletariat type that should be most conducive to aggressive political participation. In testing the rank disequilibrium hypothesis, let us begin by ignoring the set of £ variables (the variables in the ExpectancyValue-Norms model) in order to establish a base-line estimate of the effect (if there is any) of RDE on APPln. The parameter estimates are: (5.17)

APPlli = 1.337 + .134(E) + .OCW(S) + .240 (RDE), (.014) (.014) (.043)

where R2 = .299 and M = 2,063. These results lend partial con­ firmation to the rank disequilibrium hypothesis. Since the re­ gression coefficient for the S variable is less than its standard error, S can be eliminated from the equation on the grounds

176 I Frustration-AggressionTheory

that, as predicted, it has zero impact on APP la . The RDE coefficient is positive and substantially larger than its standard error, supporting the rank disequilibrium hypothesis. But the E variable also has an independent positive effect, contrary to expectations. What is it about high education, per se, that is conducive to participation in aggressive political action? The most logical conjecture is that nothing about high edu­ cation itself leads to participation in aggressive political action. Rather, high education probably just reflects the facilitative condition of the social normative beliefs variable. Many of those who score high on education reside in the university milieu. If level of education, independent of residence in a facilitative social context, has no relationship to participation in aggressive political action, then when the UNV variable is brought into the equation, the E coefficient should be reduced to approximately zero. In fact, this is just what happens, as the following equation shows: (5.18)

APP lu = 1.108 + .016(E) + .192(RDE) (.011) (.026) + .262 (UNV), (.018)

where R 2 = .362 and N = 2,063. The regression coefficient for E is now only slightly greater than its standard error, good grounds for eliminating E from the equation. The apparent effect of E observed in equation 5.17 was, indeed, a spurious relationship, due to the fact that E is highly correlated (natu­ rally) with UNV (r =.845), one of the variables from the Expectancy-Value-Norms model. Having established that the ratio interaction formulation of the rank disequilibrium hypothesis appears to be correct, we now obtain a base-line estimate of the impact of RDE on APP ln as given by: (5.19)

APP ln = 1.577 + .505(RDE),

where r2 = .160 and JV = 2,063. Despite its restricted range in this sample, Rank Disequilibrium does show a linear relation-

Frustration-AggressionTheory / 177 ship of modest strength with Aggressive Political Participation. Of course, from equation 5.18 we know that the RDE effect will be substantially reduced (from .505 to . 192) when one of the variables from the Expectancy-Value-Norms model is taken into account. The important question is whether or not RDE will have any direct impact on APPln once all the variables in the Expectancy-Value-Norms model have been taken into account. Just as when the JDF s variable was being considered, it is plausible to hypothesize that the RDE effect might be amplified by the presence of facilitative social norms. Therefore, it is again desirable to include a variable expressing such an ampli­ fication hypothesis, in this instance UNV*RDE. The model to be estimated is the same as that expressed by equation 5.9, except that the RDE variable now replaces JDFs. The param­ eter estimates are: (5.20)

APP l a = 1.371 + .043(UJA) + .0025 (JfJA) (.005) (.0006) + .037 (RDE) + .172 (WVF) (.040) (.053) + .046(4) + .0035 (C/JVF* (.007) (.0007)

)

+ M%(UNV*RDE), (.048) where R 2 = .572 and N = 1,838. Sincetheregressioncoefficient for the RDE*UNV interaction term is equal to its standard error, the amplification hypotheses clearly can be rejected. Re-estimation of equation 5.20 without the UNV*RDE variable gives: (5.21)

APP l n = 1.353 + Mi(UJA) + .0024 (NJA) (.005) (.0006) + .070 (RDE) + .204 (UNV) (.023) (.040) + .046(4) + .0036( UNV*NJA), (.007) (.0006)

178 / Frustration-AggressionTheory

where R 2 = .571 and JV = 1,838. Here the RDE parameter estimate is greater than two and one-half times its standard error. But since R 2 is unchanged at two decimal places and the parameter estimates for the Expectancy-Value-Norms are virtually unchanged, RDE may be regarded as a superfluous variable, and hence deleted from the prediction equation for APP ln . This reduces equation 5.21 to equation 4.14, leaving the Expectancy-Value-Norms model intact. 56 One question that still remains to be clarified is the relation­ ship between rank disequilibrium and the subjective measures of frustration posited to stem, in part, from it. The correlation between RDE and the measure ofjust deserts frustration, JDF s , is in the expected direction (positive) but small: r R D E . J D F s = .284. The correlation between RDE and the measure of refer­ ence group frustration, RGG (ranging low to high, from frus­ tration to gratification), also is in the expected direction (negative), but its magnitude is trivial: r R D E , R G G .064. The connection between frustration (as measured in this study) and rank disequilibrium is surprisingly weak. Unless these frustration measures are unusually deficient, the assumption that people who experience rank disequilibrium are necessarily frustrated, in either the just deserts or reference group sense, must be called into question. 57 This underscores the need to be very cautious, in macro-level investigations, about treating ratio interaction measures similar to that used here as if they were indicators of subjective frustration or relative depriva­ tion. 58 56 Another indication that RDE is a superfluous variable comes from the standardized regression equation: APP l a = AS2(ZJJA) AbQ(NJA) .055 (RDE) + .148(f/JVF) + .158(,4) + .267(ί/JVT*NJA), where the standardized parameter estimate for RDE is a substantively trivial magnitude, quite close to zero. 5 7 Ifpeople who scored high on RDE were necessarily frustrated, while those scoring low on RDE were not necessarily unfrustrated, the relationship be­ tween JDF s , for example, and RDE would conform to a positively decelerated curvilinear function and linear correlation would underestimate the strength of the relationship. However, examination of the scatterplot between JDF s and RDE reveals no tendency other than a slight linear trend. 58 For example, the Feierabends attempt to measure frustration across nations by taking the ratio of want satisfaction (defined by mean standardized level of GNP, caloric intake, telephones, physicians, newspapers, and radios)

Frustration-AggressionTheory / 179

At any rate, whatever may be the precise psychological effects that flow from the intellectual proletariat type of rank disequilibrium, let us presume that the experience of such dis­ equilibrium is at least alienating in some unspecified sense or senses. This structurally-based "alienation of the intellec­ tuals,"59 despite its extremely limited variation in this sample, is related moderately, albeit indirectly via the mediation of the variables in the Expectancy-Value-Norms model, to magnitude of aggressive political response. And it is entirely plausible that a stronger relationship could turn up in a sample containing a much larger proportion of persons with high rank disequi­ librium scores. Therefore, the only safe conclusion to draw from these data is that in societies with little extreme rank dis­ equilibrium, that which does exist bears no direct relationship to aggressive political participation. But a decisive test of the rank disequilibrium hypothesis awaits replication of this re­ search in a society—for example, one of the "third world" or developing nations—where substantial rank disequilibrium can be observed.60 In concluding this section on objective deprivation let us consider the relationship between poverty and aggressive politi­ cal participation. The obvious hypothesis is: the lower the level of personal income, the greater the level of aggressive political participation. The inadequacies of this crude structural hy-

to want formation (defined by literacy or urbanization scores, whichever was higher). But this macro variable very probably has little or nothing to do with frustration and should not be interpreted even as an imperfect indicator of it. See Ivo K. Feierabend and Rosalind L. Feierabend, "Systemic Conditions of Political Aggression: An Application of Frustration-Aggression Theory," in Anger, Violence and Politics, ed. Ivo K. Feierabend, Rosalind L. Feierabend, and Ted Robert Gurr (Englewood Cliffs, N.J.: Prentice-Hall, 1972), pp. 143-149. 59 Not to be confused with Brinton's notion of the alienation or desertion of the intellectuals which refers to their transferral of allegiance from the existing form of government to a radically altered form. See Crane Brinton, The Anatomy of Revolution (New York: Vintage Books, 1957), pp. 41-52 (originally published in 1938). 60 Hibbs carried out a test of the rank disequilibrium hypothesis with nations as the unit of analysis, but found no support for it. However, multicollinearity proved to be a problem at the macro level just as it is with these micro-level data. See Hibbs, Mass Political Violence, pp. 44-50.

180 / Frustration-AggressionTheory

pothesis are just as obvious and require no further comment here. In fact, Income does correlate in the expected direction with APPla at r = —.133, but the magnitude of the correlation is quite trivial.61 5.5

SUMMARY

This chapter has dealt with the frustration-aggression hypothe­ sis. In contemporary theorizing, frustration has been one of the more frequently invoked causal mechanisms proposed to ac­ count for aggressive political participation. Empirical tests of the relationship between frustration and aggressive political participation have raised serious doubts about the tenability of the hypothesis, despite its common-sense appeal. But, there also exists a marked lack of consensus about how actually to define and measure the concept of frustration. And prior tests of the frustration-leads-to-political aggression hypothesis are open to question on the grounds that they have relied upon inappropriate measures—measures insensitive to the kind of frustration that individuals might be most likely to find intolerable. Frustration may be defined as the experience of unrealized expectations. But expectations in regard to what standard of comparison? The level of achievement reached by the best-off members of a person's reference group? The level of achieve­ ment to which a person aspires—represented, for example, by his conception of his best possible level of achievement? The level of achievement to which a person feels he is rightfully entitled—his just deserts? According to the Relative Deprivation theory it is frustration in the last-named sense that people are likely to find hardest to tolerate, that is most likely to anger them and provide incentive for aggressive political participation. Indeed, when frustration 61 The Income variable is a 10-point measure ranging from a low of less than DM 400 net household income per month to a high of greater than DM 3,500 net household income per month. Assuming a conversion rate of 2.5 Deutschemark to the dollar, this would be a range of less than S160 per month to greater than S1,400 per month.

Frustration-AggressionTheory / 181

is indexed explicitly with regard to the degree to which a person feels that his just deserts have not been realized, it turns out that there is a very clear linear relation between magnitude of just deserts frustration and magnitude of aggressive political par­ ticipation. But when frustration is indexed with regard to unrealized expectations pegged to the achievement attained by the best-off members of a person's reference group, or with regard to expectations pegged to a person's conception of his best possible level of achievement, no clear-cut relation emerges between frustration and political aggression. Of course, aggressive political participation is not deter­ mined by just deserts frustration alone. All contemporary theorists have acknowledged the importance of other variables. However, different roles have been ascribed to frustration when taken in conjunction with other determinants of aggressive political participation. Three such roles were identified here. One, the Direct-Amplification role, casts frustration as the central determinant of aggressive political participation. The effect of other variables—in this case utilitarian and normative justification for aggression—is seen to be contingent upon the presence of frustration, while frustration itself has a direct effect on aggressive political participation, regardless of the level of utilitarian and normative justification. Another, the Facilitative-Amplification role, differs from the Direct-Amplification role in that the justification variables are cast as the central determinants of aggressive political participation. Utilitarian and normative justification for aggression are postulated to affect aggressive political participation directly, regardless of level of frustration, while the effect of frustration is seen to be contingent upon the presence of utilitarian and normative justification. Finally, the Facilitative-Indirect role casts frus­ tration in a purely secondary or "supporting" part, with the lead role assigned to the justification variables. Frustration is postulated to affect aggressive political participation only in­ directly, through the mediation of utilitarian and normative justification. The data considered here supported neither the DirectAmplification nor the Facilitative-Amplification role ascribed

182 / Frustration-Aggression Theory

to frustration. Since the Direct-Amplification role corresponds to equation 2.8 of Chapter Two, the Relative Deprivation model expressed by that equation was rejected. Frustration was found to conform to the role ascribed to it by the Expec­ tancy·Value-Norms theory—the Facilitative-Indirect role. Having determined that subjective measures of frustration play no direct role in the explanation of individual differences in level of aggressive political participation, the analysis shifted to consideration of objective conditions that respondents might find to be frustrating. The most theoretically compelling of these, rank disequilibrium, defined as the ratio of education to social status, was found to bear no direct relationship to aggressive political participation independent of the variables in the Expectancy-Value-Norms model. As a direct cause of aggressive political participation, the idea of frustration or relative deprivation has much to recom­ mend it on grounds of prima facie theoretical plausibility. The finding that it has no direct effect on magnitude of aggressive political participation provides support for the assumption of the Expectancy-Value-Norms theory that, apart from the basic determinants of behavior specified by Fishbein (plus the avail­ ability variable), other variables that may influence participa­ tion in aggressive political action do so only indirectly. But to test the completeness of the model for the Expectancy-ValueNorms theory adequately, a variety of additional hypotheses about alternative causal variables still must be considered.

SIX

Left-OutVariabIes

A particularly important phase in the process of theory-testing entails investigation of plausible alternative causes. This is the principal means for establishing the accuracy of a theory— where accuracy refers to whether or not it provides a true causal explanation of some phenomenon. The accuracy of a theory—explanatory accuracy—is to be distinguished from the predictive accuracy of any particular operationalized equation or model for it.1 An operationalization of a theory in the form of an equation relating a response variable, Y, to a set of putative causal variables, X, could be shown to yield very high predic­ tive accuracy and yet still be an inaccurate explanation. For the relationship between T and X could be spurious if the X variables happened to be highly correlated with a set of true causal variables, £, that had been left out of the equation for T. Thus, to determine the explanatory accuracy of a theory, we must be able to show that the relationship between the T and X variables holds regardless of all possible Z variables. Unfortunately, explanatory accuracy can never be proven definitely, since it is never possible to exhaust completely the set of Z variables. But as more and more of the plausible Z variables are ruled out, the likelihood of the theory being a correct explanation is correspondingly increased. Hence, 1 A useful discussion of this topic appears in Jerald Hage 5 Techniques and Problems of Theory Construction in Sociology (New York: Wiley, 1972), pp. 177186. What is referred to here as the "accuracy" of a theory is the requirement of "maximal specificity" elaborated in Carl G. Hempel3 Aspects of Scientific Explanation (New York: Free Press, 1965), pp. 397-403.

184 I Left-OutVariabIes

explanatory accuracy may be visualized as a limit that theories can move toward but never reach. In the last chapter, the Relative Deprivation theory, a major alternative to the Expectancy-Value-Norms theory, was sub­ jected to detailed investigation. In ruling out the Relative Deprivation theory—at least provisionally on the basis of one test—the explanatory accuracy of the Expectancy-ValueNorms theory clearly has been enhanced. Now let us take up a number of hypotheses about left-out variables that represent less comprehensive alternatives to the Expectancy-ValueNorms theory. 6.1

SPECIFIC POLITICAL PERFORMANCE DISSATISFACTION

Recall that in discussing political support and alienation men­ tion was made of the distinction between specific and diffuse support. Specific support involves evaluations of what incum­ bent political authorities do; diffuse support refers to more basic, generalized attachment to political objects. As Easton puts it: Some types of evaluations are closely related to what the political authorities do and how they do it. Others are more fundamental in character because they are directed to basic aspects of the system. They represent more en­ during bonds and thereby make it possible for members to oppose the incumbents of offices and yet retain respect for the offices themselves, for the way in which they are ordered, and for the community of which they are a part. The distinction of roughly this sort I have called "specific" as against "diffuse" support.2 Now the idea of two different kinds of political support, one tied to evaluation of the performance of the authorities, the other representing more general pro-con sentiment for basic aspects of the system, is certainly a plausible notion. But as an operational matter, one needs relatively precise guidelines for 2 Easton, "A Re-Assessment of the Concept of Political Support," British Journal of Political Science^ 5 (October 1975), p. 437.

Left-OutVariabIes / 185

classifying some measures of political affect as instances of spe­ cific support, others as instances of diffuse support. Let us focus on the object of support and the basis for support. Political affect will be said to be an instance of specific support if the following conditions hold: (1) particular incumbent political authorities are the object of support; (2) support is extended on the basis of particular actions taken by the in­ cumbent (s). Political affect will be said to be an instance of diffuse support if: (1) the object of support is a general structure of political authority such as broad categories of authorities (government officials, the police, the courts), political institu­ tions, or the political system or form of government itself; (2) support is extended on the basis of general characteristics of the authority structure (they or it are worthy of respect, uphold basic political values).3 According to these definitions, the Political Support-Aliena­ tion variable could be classified as a measure of diffuse political support because the objects referred to in items S1 to S,A are general structures of authority—the political system, political institutions, courts, leading politicians, and police—and be­ cause the bases of pro-con sentiment for these objects are general characteristics—uphold a person's values, are worthy of respect, are representative, respect basic rights of citizens, have had good intentions. The only exception is item S6, which refers to a general object but a particular action, i.e., guaranteeing a fair trial. Two kinds of specific support will be investigated here. The first involves evaluations of the performance of the incumbent administration as a whole in areas of public policy deemed to be a responsibility of the state and considered personally im­ portant by the respondent. The second involves evaluations of the treatment received by the respondent and/or people like himself during interactions with political authorities. What role might specific support play in the determination of aggressive political participation? Is there good reason for 3 In addition to specific and diffuse support there are two types of mixed support, one defined by support for particular authorities on the basis of general characteristics, the other defined by support for general structures of authority on the basis of particular characteristics.

186 I Left-OutVariabIes

expecting that it might be an important left-out variable, in­ fluencing the APP scale directly? If Easton's rationale for the specific versus diffuse support distinction is correct, specific support should not turn out to be an omitted variable of great consequence. Diffuse support, after all, is conceived as a factor that enables citizens to tolerate dissatisfaction with the performance of incumbent political authorities. So long as a citizen feels a generalized sense of attachment to the broad structures of political authority in his country, discontent and anger that he might feel because of ineffective policy performance or ill-treatment should be muted. Withdrawal of diffuse support for the structure of political authority provides clear-cut justification for par­ ticipating in acts of collective political aggression; withdrawal of specific support does not. Especially in polyarchies such as the Federal Republic of Germany, where citizens can freely contest incumbent administrations, dissatisfaction with the performance of incumbents should afford little incentive for extreme responses such as aggressive political behavior. There­ fore, let us hypothesize that measures of specific support will have no direct relationship to aggressive political participation independent of the variables in the Expectancy-Value-Norms model. Having consigned specific support to the status, at best, of a facilitative condition, which kind of facilitative effect is most likely: facilitative amplification or facilitative-indirect? The answer to this question depends on one's prediction of the rela­ tionship between aggressive political participation and diffuse support when specific support is high. But consideration of that relationship raises another, logically prior question, namely, to what extent are specific and diffuse support likely to be corre­ lated? For if they are very strongly correlated, then few cases of inconsistent combinations will exist, and it will be impossible to estimate reliably the impact on aggressive participation of one kind of support, given different levels of the other.4 4 On this point see Merrill Shanks and Jack Citrin, "The Measurement of Political Alienation: Strategic and Methodological Issues," paper delivered at the conference, "Alienation and System Support," Iowa City, Iowa, Jan­ uary, 1975.

Left-OutVariabIes / 187

Easton argues that diffuse support will vary independently of specific support.5 The primary basis for this proposition appears to be the simple fact of conceptual distinctiveness. But it does not have to follow logically that, because specific and diffuse support are analytically different dimensions of support, they must be unrelated empirically. Indeed, an em­ inently plausible hypothesis is that specific and diffuse support are likely to be related, with diffuse support causally dependent upon specific support. Specific support is tied to experiences people have with the day-to-day performance of political authorities, either in the form of firsthand, direct experiences deriving from personal contact or in the form of secondhand, vicarious experiences deriving primarily from media reports of what the authorities are doing in various areas of public policy and how they treat other citizens in various contexts. It seems reasonable to expect that such experimental learning will have an influence on citizens' generalized sentiment for structures of political author­ ity. Although, as Easton claims, diffuse support surely is in­ fluenced by childhood and continuing adult socialization, independent of experiences with the political authorities, even he, in a passage contradictory to his independence argument, has acknowledged the likely dependence of diffuse support on experiential learning: Members do not come to identify with basic political ob­ jects only because they have learned to do so through in­ ducements offered by others—a critical aspect of socializa­ tion processes. If they did, diffuse support would have entirely the appearance of a nonrational phenomenon. Rather, on the basis of their own experiences, members may also adjudge the worth of supporting these objects for their own sake. Such attachment may be the product of spill-over effects from evaluations of a series of outputs and of performance over a long period of time.6

5 Easton, "A Re-Assessment of the Concept of Political Support," pp. 437, 444. 6 Ibid., p. 446.

188 / Left-OutVariabIes

The crucial point to note here is the suggestion that diffuse support may be influenced by "spill-over effects from evalua­ tions of a series of outputs," that is, by variation in specific support. But to conjecture that diffuse support will be related to specific support is not to say that the two will be inseparable. If specific support is a cause, but not necessarily the cause of diffuse support, then the relationship between them should not be so strong as to preclude reliable testing for the presence of a facilitative-amplification as opposed to a facilitative-indirect effect of specific support on participation in aggressive political action. In fact, prior research generally has turned up positive but moderate correlations between measures of the two kinds of support.7 As to the question of whether, on a priori theoretical grounds, prediction of a facilitative-amplification effect is more or less defensible than prediction of a facilitative-indirect effect, one must acknowledge that a good case can be made for either pre­ diction. Consider the following diagram in which, for the pur­ pose of illustration, all variables are treated as if they were dichotomous: (a) Facilitative-Amplification Pattern Likelihood of Aggressive Political Participation Diffuse Support

Low High

Specific Support High Low Low Low

High Low

7 See, for example, Muller, "The Representation of Citizens," p. 1163 (full citation in footnote 11 of Chapter One); Walter F. Murphy 5 Joseph Tanenhaus j and Daniel L. Kastner5 "Public Evaluations of Constitutional Courts: Alterna­ tive Explanations," Sage Professional Papers in Comparative Politics, 4 (No. 01-045, 1973), p. 46; Samuel C. Patterson, Ronald D. Hedlund, and G, Robert Boynton, Representatives and Represented (New York: Wiley, 1975), p. 52.

Left-OutVariabIes / 189

(b) Facilitative-Indirect Pattern

Likelihood of Aggressive Political Participation Diffuse Support

Low High

Specific Support High

Low

High Low

High Low

In case (a), diffuse support is expected to bear no relation to aggressive political participation when specific support is high, because it is presumed under this condition that normative incentive for aggression stemming from low diffuse support is inhibited. Only when specific support is low, and amplifies the normative incentive for aggression stemming from low diffuse support, is aggressive political participation expected to occur. With quantitative variables, this hypothesis can be expressed by the prediction of a multiplicative interaction effect on ag­ gressive political participation from the product of specific and diffuse support. It captures the basic thrust of the argument by Seymour Martin Lipset to the effect that regimes with high legitimacy are inherently stable, regardless of their effective­ ness; that regimes with low legitimacy but highly effective performance also can survive for rather long periods of time purely on their performance; but that regimes low on both legitimacy and effectiveness of performance will be threatened in short order with serious revolutionary opposition.8 Under case (b) there is no interaction between specific and diffuse support. Variation in diffuse support is expected to be linked to the likelihood of aggressive political participation regardless of whether specific support is high or low. The pre­ sumption here is that effective policy performance of incum­ bents does not provide satisfaction sufficiently compelling to override the incentive for aggression deriving from more funda8 See Seymour Martin Lipset, Political Man, (Garden City, N.Y.: Anchor Books, 1963), pp. 67-69 (originally published in 1960).

190 I Left-OutVariabIes mental feelings of generalized disaffection from the political system. Both the (a) and (b) patterns seem plausible. And it is possible to cite historical examples in support of either one.9 So much for hypotheses about the role of specific support in determining aggressive political participation. We may now begin the analysis of specific support by focusing on how citizens evaluate the performance of the incumbent administration in various areas of public policy. Max Kaase has argued that dissatisfaction with policy per­ formance is likely to be linked to aggressive political participa­ tion only under certain conditions: (1) the policy must have broad relevance in that it affects a sizable portion of the population; (2) the government is assigned responsibility for dealing with the policy;. (3) the policy has high personal im­ portance for the individual.10 In developing a policy evaluation measure, these conditions were taken into account as follows. First, eleven areas of public policy affecting citizens in general or sizable numbers thereof were listed on a deck of cards. The cards were handed to respondents with the request that they be sorted according to whether or not the respondent felt that the policy was a proper task of the government. Second, of those policies deemed to be a respondibility of the government, respondents were asked to state how important each policy was to them personally: very important, important, not very important, absolutely unimportant. Third, again from that set of policies deemed to be a responsi­ bility of the government, respondents were asked to evaluate the performance of the present administration in that area, employing the grading scheme used in the public schools, which ranges from 1 to 5, with a 1 denoting excellent performance 9 Ibid., p. 69. Germany, Austria, and Spain are cited as examples which in the 1930's conformed to what I have called the (a) pattern; Ireland, while part of the United Kingdom, is cited as an example of what I have called the (b) pattern. 10 See Max Kaase, ii Bedingungen unkonventionellen politischen Verhaltens in der Bundesrepubiik Deutschland," Politische Vierteljahressehrift^ 17, Sonderheft 7/1976, p. 182.

Left-OutVariabIes / 191

and a 5 denoting extremely poor performance. The policies were: 1. 2. 3. 4. 5. 6. 7.

8. 9. 10. 11.

Guaranteeing justice for all. Providing welfare services for all who need them. Combating pollution. Creating more opportunities for all citizens to par­ ticipate in making political decisions. Guaranteeing protection and security for individuals. Providing for economic stability. Seeing to it that business corporations, unions, and universities are run in accord with basic democratic principles. Ensuring a strong national defense. Providing for strong and capable political leadership. Ensuring a free-market economy. Providing for peace and order in society.

Policy evaluation scores were defined as the product of the importance of the policy times the quality of the present government's performance in that area, as shown in Chart 6.1. Categories of importance were scored from 0 to 3 such that people with absolutely no interest in the policy received a score of 0 regardless of their rating of government performance. Categories of performance evaluation were scored from — 2 to +2 such that degree of positive and negative evaluation could be weighted by importance relative to the neutral point of mediocre performance. The distribution of responses to each policy evaluation vari­ able is given in Table 6.1. Responses tend to be normally dis­ tributed, with the mode of each distribution at zero.11 The Germans in this sample tend slightly toward negative evalua­ tion of the performance of the Schmidt administration in com­ bating pollution, creating more opportunities for political par­ ticipation, providing for economic stability, and ensuring that 11 With respect to each policy, those receiving a score of 0 are almost always people who give a neutral evaluation of governmental performance; people receiving a score of 0 because the policy area is completely unimportant to them never make up more than 0.5 percent of the total.

(.PG)

Performance of the Government

1 or Excellent (2) 2 or Good (1) 3 or Mediocre (0) 4 or Poor (-1) 5 or Extremely Poor (-2)

Policy Evaluation Score (.IP*PG)

0

0 0 0 0

Absolutely Unimportant (0)

-2

2 1 0 -1

—4

4 2 0 -2

Importance of the Policy (IP) Not Very Important Important (2) (1)

CHART 6.1. Scoring Procedure for Policy Evaluation Variables

-6

6 3 0 -3

Important (3)

Left-Out Variables / 193

major institutions are run in accord with democratic principles. A tendency toward positive performance evaluation is regis­ tered only in regard to the Schmidt administration's actions in the area of ensuring a strong national defense. Ronald Inglehart has advanced the thesis that in con­ temporary industrialized democracies the value priorities of citizens will tend toward bimodality, with one group, labeled "materialists," assigning highest priority to sustenance and security values, and another group, labeled "post-materialists," assigning highest priority to egalitarian, self-realization, and esthetic values.12 Examples of sustenance and security values are: maintain order in the nation; fight rising prices; maintain a high rate of economic growth; make sure that the country has strong defense forces; maintain a stable economy; fight against crime. Examples of egalitarian, self-realization, and esthetic values are: give people more say in the decisions of the government; protect freedom of speech; give people more say in how things are decided at work and in their community; try to make the cities and countryside more beautiful; move toward a friendlier, less impersonal society; move toward a society where ideas are more important than money. If, as Inglehart argues, the materialist and post-materialist types represent very basic, generalized orientations toward societal goals, then it is likely that citizens who adopt a material­ istic perspective will tend to give similar evaluations of govern­ mental performance in policy domains involving sustenance and security values, whereas citizens who adopt a post-ma­ terialistic perspective will tend toward similar evaluations of governmental performance in policy domains involving egali­ tarian, self-realization, and esthetic values. Thus, an implica­ tion of the materialist versus post-materialist distinction is that two empirically separate clusters of policy evaluation variables, 12 See Ronald Inglehart, "The Silent Revolution in Europe: Intergenerational Change in Post-Industrial Societies," American Political Science Review, 65 (December 1971), 991-1017; Inglehart, "Values, Level of Conceptualization, and Protest Potential Among Western Publics," paper delivered at the Tenth World Congress of the International Political Science Association, Edinburgh, Scotland, August 1976; Inglehart, The Silent Revolution (Princeton: Princeton University Press, 1977).

Guaranteeing protection and security for individuals

PE 5

% = 2.8 0.4 16.8 10.3 0.8 43.6 0.5

6.2

6.6

6.1 11.5 0.8

6.4

8.5

8.5 22.7 2.8 16.7

4.3 15.4 0.7

8.9 1.7 40.5 1.7 11.7 14.0 2.3

Creating more oppor­ tunities for political participation

PE 4 8.2

4.2 0.2 30.3 0.5

% = 2.3 0.6 11.2

Combating pollution

PE 3

%= 1-8 0.6

9.7 0.3 37.7 0.2

% = 3.6 1.1 20.8

Providing welfare services for the needy

PE 2

3.7 16.8 0.7

-3 -4 -6

PE Score

-1 -2

6.0 0.1 41.3 0.1

0

% = 3.3 0.4 20.9

+2 +1

Guaranteeing justice for all

+ 6 +4 + 3

PE 1

Policy Evaluation Variables

TABLE 6.1. Distributions of Scores on Policy Evaluation Variables

+ .13

— .04

X

100% (JV= 2099)

100% (JV= 1670)

+ .01

-.70

100% -1.39 (JV= 2173)

100% (JV= 2222)

100% (JV = 2213)

Total

8.9 1.6 40.0 0.8

9.3 0.9 35.3 0.9

% = 3.8 1.2 16.8 10.9 1.1 41.5 0.8

% = 4.2 0.9 19.1

Providing strong and capable political leadership

PEg

9.2 0.9 39.8 1.1

6.1 0.4 31.3 0.4

% = 5.1 2.8 19.7 17.3 4.2 38.5 1.2

8.8

PE 1 1 Providingforpeaceand order in society

Ensuring a strong national defense

PE b

% = 1.7 0.7

% = 4.6 0.7 19.1

Ensuring that major institutions are run in accord with demo­ cratic principles

PE 1

% = 4.1 0.9 15.3

PE 1 0 Ensuring a free-market economy

Providing for economic stability

PE 6

5.0 0.8

4.5 11.3 0.8

6.1 12.9 1.8

7.7 10.9 2.0

4.0

7.9

8.5

3.3

1.4

9.1 15.7 2.5 10.5

5.0 20.0 2.3 14.1

-.80

-.78

100% (JV= 1792)

100% (JV= 1641)

100% (JV = 1686)

+.11

-.03

+.25

100% +1.04 (JV= 1661)

100% (JV = 1665)

100% (JV = 2248)

196 / Left-OutVariabIes

corresponding to these differences in basic value priorities, are likely to turn up when interrelationships between the policy evaluation variables are analyzed. Of the policies presented for evaluation to this sample, the first, third, fourth, and seventh clearly are instances of postmaterial values, as Inglehart has defined them; the fifth, sixth, eighth, ninth, tenth, and eleventh clearly are instances of material values. The fifth, "providing welfare services for all who need them" involves both a material value (welfare) and a post-material, egalitarian value (redistribution of wealth). Let us hypothesize, then, that when subjected to factor analysis, policy evaluation variables 1, 3, 4, and 7 will load on one factor, while policy evaluation variables 5, 6, 8, 9, 10, and 11 will define another factor. As table 6.2 shows, principal factor analysis of the policy evaluation variables provides unambiguous support for the materialist/post-materialist distinction.13 All variables load exactly as hypothesized. The one ambiguous variable, PE 2 , clusters in this sample with the post-material variables, sug­ gesting that this policy has strong redistributive connotations for most respondents. Summary measures of magnitude of dissatisfaction with materialistic and post-materialistic policy performance were constructed by first reversing each policy evaluation variable so as to run from —6 at the most satisfied end of the continuum to +6 at the most dissatisfied end, then computing factor scores.14 The Materialist Policy Dissatisfaction (PDm) variable so created ranges from —2.486 to +2.619; and the PostMaterialist Policy Dissatisfaction (PDpm) variable ranges from — 2.774 to +2.537. In order to eliminate negative numbers (which are undesirable when constructing product terms), an increment of 3 was added to PD m and PD p m . 13 Kaase also found support for the materialist/post-materialist distinction when policy evaluation variables were factor-analyzed. See Kaase, ii Bedingungen unkonventionellen " p. 215. 14 Factor scores were created using the SPSS factor analysis routine. If a respondent was missing scores on less than one-half of the variables defining a factor, the mean was assigned in place of the missing data.

TABLE 6.2. Principal FactorAnalysis of Policy Evaluation Variables Factor Loadings I Material Values

II Post-Material Values

Variance Reproduced

Guaranteeingjustice for all

.338

.660

.550

Providing welfare services for the needy

.286

.535

.369

PE 3

Gombatingpollution

.188

.574

.365

PE i

Creating more oppor­ tunities for political participation

.209

.699.

.532

Guaranteeing protection and security for individuals

.542

.338

.408

PE 6

Providing for economic stability

.630

.193

.434

PE 1

Ensuring that major institutions are run in accord with democratic principles

.276

.686

.547

PE g

Ensuring a strong national defense

.511

.214

.307

Providing strong and capable political leadership

.646

.276

.493

P E i g Ensuring a free-market economy

.740

.253

.612

PE j j Providing for peace and order in society

.637

.252

.470

Policy Evaluation Variables PE 1 PE 2

PE i

PE g

198 / Left-OutVariabIes

Kaase has hypothesized that, of the two dimensions of value priorities, satisfaction-dissatisfaction with government per­ formance in post-materialist policy domains should be most closely linked to aggressive political participation. The reason, he argues, is that governments of industrialized democracies are still largely preoccupied with sustenance and security values, devoting comparatively minor attention to egalitarian, self-realization, and esthetic values. Thus, dissatisfaction with performance in materialist policy domains might be muted by perception that at least the government is properly concerned about these values, whereas dissatisfaction with performance in post-materialist policy domains might be made all the more galling by perception that incumbent politicians have little serious interest in promoting these values.15 Scatterplots of the relationship between the scale of aggres­ sive political participation and the measures of dissatisfaction with government performance in materialist and post-materi­ alist policy domains are given in Figures 6.1 and 6.2. There is no systematic relationship at all between APPin and PDm. Inter­ estingly, those who register high APPxn scores tend to be inter­ mediate on PDm—neither especially satisfied nor dissatisfied with the performance of the Schmidt administration in the area of materialist policies. Intermediate PDm scores do not themselves correspond to any particular levels of APPio, while both low and high PD m scores tend to coincide with low APP ln scores. By contrast, there is a clearly discernible linear trend between magnitude of dissatisfaction with post-materialist policy performance and magnitude of aggressive political par­ ticipation, though the fit to a linear function in this instance is of only modest strength. Using PD pm as the indicator of specific support, PSA 2 as the indicator of diffuse support, and the multiplicative term, 15 Kaase, ii Bedingungen unkonventionellen," p. 195. His analysis of a crosssectional probability sample of West Germans did not, however, disclose support for the hypothesis, at least in regard to a measure of protest potential, since neither post-materialist nor materialist policy dissatisfaction showed correlations with protest potential at other than trivial magnitudes. Note, though, that the protest potential variable encompasses a variety of non-aggres­ sive actions in addition to aggressive actions in the form of civil disobedience.

4.67-

Summary

Statistics Large circles denote 9 or more cases.

M A T E R I A L I S T POLICY DISSATISFACTION (PD m )

FIGURE 6.1 Aggressive Political Participation as a Linear Function of Materialist Policy Dissatisfaction

Summary

Statistics Large circles denote 9 or more cases.

POST-MATERIALIST POLICY DISSATISFACTION

(PDtm)

FIGURE 6.2 Aggressive Political Participation as a Linear Function of Post-Materialist Policy Dissatisfaction

Left-Out Variables / 201

to represent the possible interaction between specific and diffuse support, we may now investigate the role of specific support in the determination of magnitude of aggressive political participation. Recall that the different models are defined as follows: Facilitative-Indirect: (1) a direct effect from to (2) a direct effect from to (3) no directeffect from to (4) no direct effect from to Facilitative-Amplification: (1) no prediction about the relationship between and (2) a direct effect from to (3) no direct effect from to . (4) a direct effect from to . Direct-Amplification: (1) no prediction about the relationship between and , (2) no direct effect from to (3) a direct effect from to (4) a direct effect from to Standardized regression coefficients relevant to testing the above predictions are shown in Figure 6.3. 16 The absence of an effect on from interaction term disconfirms the key prediction of the two amplification models. 17 The Facilitative-Indirect pattern receives the most clear-cut support. Even though the third prediction of that model is disconfirmed, in this instance the estimated direct effect from on is so small as to be of little substantive significance. 16

The prediction equation for .

with unstandardized coefficients is:

where 17 At pages 197 and 198 of "Bedingungen unkoventionellen" Kaase reports what appears to be an interaction between a measure of political distrust, a measure of personal political efficacy, and post-materialist policy dissatisfaction, i.e., distrust and efficacy amplify the relationship between post-materialist policy dissatisfaction and protest potential. However, it is difficult to compare his results with these, since the dependent variable is different and the method of statistical analysis differs from that used here.

202

Left-Out Variables

FIGURE 6.3 Estimated Parameters for a Causal Model of Aggressive Political Participation Involving Interaction between Post-Materialist Policy Dissatisfaction and Political Support-Alienation

When the variable and an interaction term expressing the possibility of an amplification effect due to facilitative social norms are introduced into the Expectancy-Value-Norms model, equation 6.1 shows that this indicator of specific support plays no direct role in the determination of aggressive political behavior:

(6.1)

where

Left-Out Variables / 203

on A P P l n , without controlling for any other variables (see Figure 6.2), now is reduced to virtually zero. Neither the pa­ rameter estimate for the P D p m variable nor that for U N V * P D p m is greater than two and one-half times its standard error. When 6.1 is reestimated without the UNV* PDm interaction term the result is: (6.2)

APPln

= 1.561 + .042 [ U J A ) + .003 [ N J A ) (.005) (.0006) - .006(PDpm) + .230[ U N V ) (.001) (.039) + .048(4) + M M { U N V * N J A ) , (.007) (.0006)

where R 2 = .574 and N = 1,324. Here the parameter estimate for PD pm is greater than two and one-half times its standard error. However, the PD pm coefficient takes on an implausible sign. Since inclusion of PD pm in the equation produces no in­ crease (at two decimal places) in the accuracy with which APP ln is predicted, and the coefficients for the ExpectancyValue-Norms variables are virtually unchanged, PD pm appears to be a superfluous variable.18 Thus PD pm can be trimmed from the equation, reducing equation 6.2 to equation 4.14. Evaluation of the policy performance of the government is one kind of specific support. Another kind arises from people's experiences when they interact with the political authorities. Evaluation of political treatment does not have to be tied only to experiences that people themselves have with agents of the government. In fact, it seems plausible that perceptions of how people similar to oneself are treated, because they represent more generalized evaluations stemming not just from possibly idiosyncratic experiences, may make a particularly important contribution to summary feelings of satisfaction or dissatisfac­ tion with political treatment. 18 The superfluousness of PD p m also can be seen from the standardized regression equation: APP ln = .159(UJA) + .202(NJA) — .0QQ(PDpm) + .167 (UNV) + .167(4) + .283(UNV x NJA), where the standardized PD pm coeffi­ cient is estimated to be a substantively trivial magnitude, i.e., less than 1.

204 I Left-OutVariabIes

To measure evaluation of political treatment, respondents were asked, first, whether they had ever had anything to do with the police, courts, or government agencies. Those who had interacted with these authorities were then asked to give an overall evaluation of their experience: had they generally received good, poor, or partly good/partly poor treatment? Regardless of their personal experiences, all respondents also were asked to give their opinion as to whether people like them­ selves generally received good, poor, or partly good/partly poor treatment from the police, courts, and agencies of the govern­ ment. Political treatment scores for these government authorities were assigned according to the procedure shown in Chart 6.2. Perceived Treatment of Others is given greater weight relative to Personally Experienced Treatment. Perception of good treatment received by others like oneself is assumed to over­ ride the results of direct experiences with agents of the govern­ ment. Differences in personally experienced treatment are given twice as much weight among those who definitely feel that people like themselves receive poor treatment as com­ pared with those who feel that others receive average treatment.

CHART 6.2. Scoring Procedure for Political Treatment Variables Personally Experienced Treatment (PET) Partly Goodj Political Treatment Score (PT0*PET) Perceived Treatment of Others (PTO)

Mo ExperienceA

Good

Partly Poor

Poor

(1)

(2)

(3)

Poor (2)

4

2

4

6

Partly Good/ ,., Partly Poor

9

1

2

3

Good (0)

0

0

0

0

3 People with no personal experience of treatment by political authorities are assigned scores corresponding to those received by the intermediate category of PET.

Left-OutVariabIes / 205

This weighting ensures that, as a necessary condition for scoring in the high or dissatisfaction range (scores greater than 3) on the political treatment variables, one must generally feel that similar others are poorly treated. The Dissatisfaction with Political Treatment (DPT) variable (range: 0-18) is a summation of treatment scores in reference to the police, courts, and government agencies. Using DPT now as the indicator of specific support, Figure 6.4 shows the standardized regression coefficients relevant to testing the pre­ dictions of the different models relating specific support to aggressive political participation.19 When PSA2 alone is con­ sidered, the data support the facilitative-amplification model, since there is an effect estimated from the PSA2*DPT inter­ action term, an effect estimated from PSA2, and no effect estimated from DPT. However, PSA2 is only one component of belief in the normative justifiability of political aggression. When normative incentive deriving both from alienation from the structure of political authority and from ideological justi­ fication—as operationally defined by NJA—is considered, neither amplification model is supported, due to the absence of an effect from the NJA*DPT interaction. Once again, the data best fit the facilitative-indirect model, although there is a small estimated effect from DPT on APPln that runs contrary to the model. Experiences of unsatisfactory treatment by political authori­ ties do not appear to amplify the effect of belief in the normative justifiability of aggression on aggressive political participation. However, dissatisfaction with treatment does appear to play a role in the determination of aggressive political participation, independent of normative justification. Let us now investigate whether this estimated direct effect from political treatment 19

The prediction equations for APP i n with unstandardized coefficients are:

APP l l , = 1.598 + .036(Ρ5Λ 2 ) + .OlS(DPT) + . 0 0 3 ( P S A 2 * D P T ) , (.004) (.007) (.001) where R 2 = .343 and M = 1,633; = 1.559 + -008 ( N J A ) + .024(DPT) + .00006 (MJ A* D P T ) , (.001) (.006) (.00006) where R 2 = .413 and M = 1,507.

FIGURE 6.4 Estimated Parameters for a Causal Model of

Aggressive Political Participation Involving Inter­ actions between Dissatisfaction with Political Treatment and Political Support-Alienation, Nor­ mative Justification for Aggression

DPT

PSA

APP,

DPT

NJA

APP,

Left-Out Variables / 207

will hold up when the full Expectancy-Value-Norms model is considered. As usual, in addition to DPT, the possibility of an interaction effect due to facilitative social norms, defined by UNV*DPT, will be taken into account. The parameter esti­ mates are: (6.3)

APP 1„ = 1.423 + M 2 ( U J A ) + .0025 (JfJA) (.005) (.0006) - .002(DJPT) + .213(UNV) (.005) (.041) + .037(-4) + .0028 (UNV* NJA) (.007) (.0007) + .m(UNV*DPT), (.006)

where R 2 = .574 and N = 1,739. The direct effect estimated for D P T washes out completely with all variables in the APP l n prediction equation, as the DPT parameter estimate is less than its standard error. But the coefficient for the interaction term UNVifDPT, is estimated to be reliably greater than zero. With regard to accuracy of prediction, however, inclusion of the UNVifDPT interaction does not increase R2 appreciably. Also, the parameter estimates for the Expectancy-Value-Norms variables are quite similar to those obtained in equation 4.14. Thus, UNVifDPT appears to be a superfluous variable.20 6.2 INTERNAL-EXTERNAL CONTROL OF REINFORCEMENT Julian Rotter has proposed the concept of internal versus external control of reinforcement as a stable personality trait influencing behavior across a wide variety of situations.21 The idea is that individuals build up generalized expectations about 20 However, the UNV*DPT term does show a non-trivial standardized regression coefficient, as given by: APP i n = .ISQ(UJA) + A65(NJA) —

.012 (DPT) + .154( UNV) +A2T(A) + .2Q%(UNV*NJA) + .\Z\(UNV*DPT). 21 See Julian B. Rotter, "Generalized Expectancies for Internal Versus External Control of Reinforcement," Psychological Monographs, 80 (Whole No. 609, 1966).

208 I Left-Out Variables

the degree to which rewards and punishments in life are con­ tingent upon their own actions (internal control orientation) as opposed to being the result of forces beyond their control such as chance, luck, or powerful others (external control orientation). Rotter argues that people with an internal con­ trol orientation should be more likely to take part in political action than those with an external orientation, and a variety of empirical studies have supported this hypothesis.22 A controversy has arisen, however, with regard to the kind of political action to be expected from internal-external control orientations. H. Edward Ransford has hypothesized that low expectancy of control over events is a form of subjective aliena­ tion likely to lead to radical attacks on the existing social and political structure.23 In attempting to account for civil disorder participation by blacks in the United States, Ransford argues that an external control orientation or feeling of powerlessness "seems to have a logical relationship to violent protest. Briefly, it is reasoned that Negroes who feel powerless to change their position or to control crucial decisions that affect them will be more willing to use violent means to get their rights than those who feel some control or efficacy within the social system."24 And, indeed, Ransford found that, among a sample of Los Angeles blacks interviewed shortly after the Watts riot, those registering an external control orientation were more likely to report that they would be willing to use violence to gain Negro rights than those registering an internal control orientation.25 Thomas J. Crawford and Murray Naditch have attempted to 22 See Thomas J. Crawford and Murray Naditch 5 "Relative Deprivation, Powerlessness, and Militancy: The Psychology of Social Protest," Psychiatry, 33 (May 1970), 208-223; Pearl Mayo Gore and J. B. Rotter, "A Personality Correlate of Social Action," Journal of Personality, 31 (1963), 58—64; Bonnie R. Strickland, "The Prediction of Social Action from a Dimension of InternalExternal Control," Journal of Social Psychology, 66 (1965), 353-358. 23 See H. Edward Ransford 5 "Isolation, Powerlessness, and Violence: A Study of Attitudes and Participation in the Watts Riot," American Journal of Sociology, 73 (March 1968), 581-591, 24 25

Ibid., p. 583.

Ransford's finding is corroborated in David 0. Sears and John B. McConahay, "Racial Socialization, Comparison Levels, and the Watts Riot," Journal of Social Issues, 26 (1970), 121-140.

Left-Out Variables / 209

reconcile the Rotter and Ransford hypotheses.26 These authors propose that both the internal and external control orientations will lead to political action, but of very different kinds, with the internals more likely than the externals to engage in conven­ tional activities—attending rallies and demonstrations or join­ ing organizations intent on working within the system—and the externals more likely than the internals to turn to uncon­ ventional action involving illegal protest and violence. On the other hand, the Ransford thesis, labeled the Alienation-Powerlessness hypothesis, has been challenged by advo­ cates of a Blocked Opportunity hypothesis.2 7 In this view, which corresponds in part to the original Rotter hypothesis, those who participate in any political action—including the kinds of aggressive political behavior displayed in urban civil disorders—will be motivated by a strong sense of internal con­ trol; but specifically aggressive action will be taken only by those among the internals who feel that they have little chance of exercising control by conventional means. Despite claims by its proponents to have found positive results, the Blocked Oppor­ tunity hypothesis has not yet been tested properly, due to a variety of methodological problems.28 To add to the general confusion, it turns out that the InternalExternal Control (IEC) scale developed by Rotter may not, in fact, consistently measure a single, unidimensional attitude. Factor analyses of the IEC items across a sample of U.S. high school students found a single factor.29 But when administered 26 Crawford Militancy."

and

Naditch 5 "Relative Deprivation, Powerlessness, and

27 See Nathan S. Caplan 5 "The New Ghetto Man: A Review of Recent Empirical Studies," Journal of Social Issues, 26 (1970), 59-73; Nathan S. Caplan and Jeffrey M. Paige, "A Study of Ghetto Rioters," Scientific American, 219 (August 1968), 15-21; John R. Forward and Jay R. Williams, "InternalExternal Control and Black Militancy," Journal of Social Issues, 26 (1970), 75-92. 28 Curiously, none of the studies purporting to test the Blocked Opportunity thesis ever measured the degree to which people felt that they had little chance of exercising control by conventional political means. Moreover, the dependent variable in the Forward and Williams article (see footnote 27) had nothing to do with personal participation in aggressive political behavior. 29

See Rotter, "Generalized Expectancies," p. 16.

210 I Left-OutVariabIes to a sample of black college students, the majority of the IEC items separated into three factors, and some did not load on any factor.3 0 Four items from the IEC scale were selected for use in this study. Respondents were given a list that read as follows : Group 1 A. Becoming a success is a matter of hard work; luck has little or nothing to do with it. B. Becoming a success depends mainly on being in the right place at the right time. Group 2 A. As far as world affairs are concerned, most of us are the victims of forces we can neither understand nor control. B. By taking an active part in political and social affairs, the people can control world events. Group 3 A. Many times I feel that I have little influence over the things that happen to me. B. It is impossible for me to believe that chance or luck plays an important role in my life. Group 4 A. In the long run, people get the respect they deserve in this world. B. Unfortunately, an individual's worth often passes un­ recognized no matter how hard he tries. Respondents then were asked to choose the one statement from each group with which they most closely agreed. In case a respondent did not agree with either statement in a group, he was allowed to select a "choose neither" option.31 Internal 30 See Patricia Gurin, Gerald Gurin, Rosina C. Lao, and Muriel Beattie, "Internal-External Control in the Motivational Dynamics of Negro Youth," Journal of Social Issues, 25 (1969), 29-53. 31 "Choose neither" was not a response option offered in the original Rotter items. It was included here to avoid forcing people who might disagree with both statements to take one side or the other.

Left-OutVariabIes / 211

responses are denoted by statement A from Group 1, statement B from Group 2, statement B from Group 3, and statement A from Group 4; in each instance, the opposite statement denotes an external response. These particular items were selected because, out of a total of 23 items, they showed biserial correlations with total IEC scale scores that ranked among the top third of those reported in Rotter's paper on the reliability and validity of the IEC scale.32 The Group 2 and 3 items correspond to those that showed the first and second highest scale-item biserial correla­ tions, while the Group 1 and 4 items correspond to those that showed the sixth and seventh highest such correlations. Thus, given that time constraints did not allow for inclusion of the entire IEC instrument on the interview schedule, it was legiti­ mate to expect that these four items might afford a faithful representation of it. Unfortunately, as Table 6.3 shows, there is virtually no rela­ tionship between any of these items, with only one of the possible six correlations even barely reaching the 0.2 level. People who select an I response to any given item are about as likely as not to select an E response to the others, and vice versa. If there is anything like a consistent, unidimensional attitude of internal versus external control among this heterogeneous sample of West Germans, these particular items just are not able to pick it up. And since they were among the most representative items in the original scale, this evidence calls into question the generalizability of the internal-external control construct beyond the samples studied by Rotter. Interestingly, the IEC items found to be virtually uncorrelated here also turned out to be uncorrelated in the study of black college students by the Survey Research Center. An item corresponding to the Group 1 item here was found in the black student sample to load on a Control Ideology factor, defined by "items which seem to measure the respondent's ideology or 32 See Table 1 at pages 11 and 12 of Rotter, "Generalized Expectancies." Group 1 here corresponds to item 11 from Table 1 in Rotter, Group 2 corre­ sponds to item 17, Group 3 corresponds to item 25, and Group 4 corresponds to item 4.

324 60 380

764

Total

596 129 237

I-E Item #n

539 962 Taub = .207

167 296 76

I-E Item #11 JV E

E JV I

E = External jV = Neither / = Internal

Key

I-E Item #4

I

Total

E JV I

Total

E M I

107 209 72

JV 309 60 266

E

386 631 Tau„ = .028

356 58 217

1234 388 635 Taub = —.018

548 275 411

I

1220

122 190 74

I-E Item #17 E JV

586 233 401

I

I-E Item #17

224 68 540 832

Total

I

829

339 101 389

I

825

409 83 333

466 99 211

413 119 265

E

342 70 369

E

626 781 Taub = ..154

68 243 315

JV

631 797 Taub = . 112

208 326 97

JV

633 776 Taub — . 104

191 305 137

I-E Item #25 JV E

E JV I

Total

E JV I

Total

E JV I

I

TABLE 6.3. Relationships between Internal-External Control of Reinforcement Items

Left-OutVariabIes / 213

general beliefs about the role of internal and external forces in determining success and failure in the culture at large."33 An item corresponding to the Group 2 item here was found in the black student sample to load on a System Modifiability factor, defined by items that refer to "the extent to which racial dis­ crimination, war, and world affairs can be controlled or changed." 3AAn item corresponding to the Group 3 item here was found in the black student sample to load on a Personal Control factor, defined by items phrased in the first person that refer to the degree to which a person "believes that he can con­ trol what happens in his own life." 35 And an item corresponding to the Group 4 item here was found in the black student sample not to load on any factor.36 Thus, three of the four IEC items used here could, in fact, be statistically unrelated because they represent independent dimensions of what may be a complex, multidimensional phenomenon incorrectly conceived as a unidimensional construct by Rotter. In any event, lacking more items relevant to the different dimensions of internal-external control, this possibility cannot be further pursued.3 7

6.3

REJECTION OF INDIVIDUALISM

In addition to the four items from the IEC scale, the interview schedule contained two items pertaining to belief about whether or not people, through their own personal action, determine 33

See Gurin et al., "Internal-External Control," p. 35. Ibid., p. 44. 35 Ibid., p. 35. 36 Ibid., p. 40. 37 The BlumenthaI study (see footnote 46 of Chapter Five) also investigated the concept of internal-external control. Nine items, intended to encompass three sub-dimensions of the internal-external control concept (personal control, control ideology, and system blame) were analyzed. As was the case with these German data, some of the inter-item relationships turned out to be quite weak; but BIumenthal and her associates went ahead and constructed additive indices anyway. All of the sub-dimensions of internal-external control showed correla­ tions (gamma) with the Protest Behaviors Index that were quite trivial in magnitude (< .2). The United States data thus provide evidence suggesting that internal-external control and its sub-dimensions is not an important variable in the determination of aggressive political participation. 34

214 / Left-OutVariabIes their destinies. In the first instance, respondents were asked:38 When it comes to the basic things which most people expect out of life, it seems that some people easily accomplish what they want, whereas others never reach their goals no matter how hard they try. In your opinion, does the responsibility for this lie with the people themselves or with something else? Response options were: (a) people themselves; (b) partly both; (c) something other than people themselves. In the second in­ stance, respondents were asked; Some people believe that they are basically dependent upon others in their life, whereas others believe that they can live their life exactly as they want to have it. How about you? Are you basically dependent upon others or can you live your life exactly as you want to have it? Response options were: (a) live life exactly as I want to have it; (b) partly both; (c) dependent upon others. The (a) re­ sponses clearly denote a belief in the sufficiency of individual­ istic action for reaching life goals. The referent of the (c) response options is less clear, although choice of a (c) instead of an (a) response would seem to imply a rejection of basic tenets of an individualistic philosophy in favor of a more collecdvistic orientation. If these questions are indeed tapping a general acceptancerejection of an individualistic orientation toward life achieve­ ment, they should, at the minimum, show a fair correlation. In addition, one would hope that they do not just reflect dif­ ferences in self-esteem, with confident people selecting the (a) options, insecure people selecting the (c) options. An interpersonal competence aspect of self-esteem39 was measured by a question that asked: 3 8 This question comes from a pilot study for a quality of life survey that was carried out under the direction of Mark Abrams of the Social Science Research Council, London, England. 39 On dimensions of self-esteem see Paul M. Sniderman, Personality and Democratic Politics (Berkeley, Calif.: University of California Press, 1975), pp. 45-54.

Left-OutVariabIes / 215

In general, are you confident of your own opinions, even when others disagree with you, or do you often feel unsure of yourself when others disagree with you ? Response options were: (a) am confident; (b) partly both; (c) am often unsure. The two measures of acceptance-rejection of individualism, Responsibility for Success-Failure and Source of Achievement in Life, do appear to be virtually independent of differences in the interpersonal competence aspect of self-esteem. The corre­ lations between them and the Confidence in Own Opinion Variable are quite small, as shown in Table 6.4. By contrast, the correlation between the two acceptance-rejection of indi­ vidualism measures at least approaches a moderate strength— with a plurality of persons who take the pro-individualism response on one variable also taking it on the other, a plurality of those who take the "partly both" response on one also taking it on the other, and either a plurality or close to it of those who take the anti-individualism response on one variable also taking the same response on the other. How might the degree of individualism be related to aggres­ sive political participation? There seems little reason to expect that differences in acceptance-rejection of individualism would have a direct effect on magnitude of aggressive participation independent of the variables in the Expectancy-Value-Norms model. But think of a person who scores high on the measure of structural just deserts frustration. If, as a general philosophy, he holds an individualistic orientation toward life achievement, might not his frustration be internalized—deflected away from serving as an incentive for aggressive participation? Whereas, if he rejects individualism, might not his frustration then be clearly externalized and correspondingly enhanced as a motiv­ ational incentive for participation in aggressive political action? In short, might not the measure of structural just deserts frus­ tration, through interaction with rejection of individualism, indeed turn out to have a direct effect in the Expectancy-ValueNorms model? To test this interaction hypothesis, it is necessary to have a

Confidence in Own Opinion

1186

766

Responsibility for

917

530

300

87

916

1029

492

113

179 496

People Themselves

Tauh = .271

199 593 418

Partly Both

Total

180

Depends Upon Others

495

297

126

72

144

Partly Both

Taub = .107

Live Life Exactly As I Want It

1016

719

226

71

Depends Upon Others

115

Total

Am Confident

Partly Both

Am Often Unsure

Partly Both

Source of Achievement in Life

Live Life Exactly As I Want It

Something Other Than People Themselves

432

269

100

63

Something Other Than People Themselves

Taub = .062

726

Am Confident 524

Total

360

189

Partly Both

100

53

Partly Both

Am Often Unsure

People Themselves

Responsibility for Success-Failure

TABLE 6.4. Relationships between Interpersonal Competence and Rejection of Individualism

Left-OutVariabIes / 217

quantitative measure of the degree to which a person rejects basic tenets of individualism. In building the Rejection of in­ dividualism (RI) variable, the categories of the Responsibility for Success-Failure (RSF) and Source of Achievement in Life (SAL) variables are assigned scores of 1 for the pro-individualism response, 2 for the "partly both" response, and 3 for the anti-individualism response. Then these scores are multiplied. Chart 6.3 shows the Rejection of Individualism scores resulting from this procedure. The reason for multiplying RSF and SAL scores is to ensure that high scores on RI are received only by persons who reject both individualistic options in favor of at least one anti-individualism response; acceptance of any proindividualism option is sufficient to place a person in the low range of RI. The hypothesis of a direct effect in the Expectancy-ValueNorms model due to an interaction between Rejection of Individualism and Structural Just Deserts Frustration implies a model of the form: (6.4)

APP l n = « + b0 (RI) + b ,(JDF s ) + b2 (RI*JDF s ) + ^i(Zi) + ε·

Ignoring the Expectancy-Value-Norms variable (£() for the moment, the base-line parameter estimates are: (6.5)

APP l n = 1.581 + .081(/2/) + Ml(JDF s ) (.008) (.006) - .0021 (RI*JDF S ), (.0012)

where R 2 = .229 and JV = 2,034. Since the parameter estimate for the RI* JDFs term is not even greater than twice its standard error, the hypothesis of a multiplicative interaction between rejection of individualism and level of frustration must be re­ jected for these data. However, the RI term does have an un­ expected direct effect on APPln independent of frustration. Will this hold up when the variables from the ExpectancyValue-Norms theory are brought into the equation? Once again, the possibility of an amplification effect due to

218

Left-Out Variables

CHART 6.3. Scoring Procedure for the Rejection of Individualism Variable

Source of Achievement in Life (SAL) Live Life Depends Exactly As Partly Upon I Want It Both Others (1) (2) (3)

Rejection of Individualism Score [SAL RSF) Something Other „ t .,.,. Than People Responsibility ^ ^ (3) Success-Failure Partly Both (RSF) People Themselves

(2) (1)

3

6 2

1

9 4

2

6 3

the presence of facilitative social norms must be considered. Therefore, in addition to RI, the social norms amplification variable UNV RI, will be included in the equation. The parameter estimates are: (6.6)

where 570 and N = 1,838. Neither RI nor the UNV RI interaction term is estimated to affect aggressive participation directly, once the Expectancy-Value-Norms variables have

Left-Out Variables / 219

been taken into account. Re-estimation of 6.6 without the in­ teraction term yields: (6.7)

APPla - 1.380 + .043( U J A ) + .0025( N J A ) (.005) (.0006) + .005 ( R I ) + .233 ( U N V ) (.005) (.039) + .045(4) + .0035 ( U N V * N J A ) , (.007) (.0006)

where R 2 = .569 and N = 1,838. The .081 effect of R I estimated from equation 6.5 now is reduced to .005. Rejection of Indi­ vidualism clearly is an irrelevant variable, as the parameter estimate for it is equal to its standard error. 6.4

PERSONAL AND POLITICAL RESOURCES

In the Bay Area Survey carried out as part of the Social Indi­ cators Project of the Survey Research Center at the University of California, Berkeley, Jack Citrin and Merrill Shanks ex­ plored the role of certain social background characteristics as "facilitators" of aggressive political participation, in conjunc­ tion with a measure of alienation from the political system.40 One of these "facilitators," age, is incorporated in the Expec­ tancy·Value-Norms model via the Availability variable. But another, personal and political resources, has not yet been considered. The measure of personal and political resources developed by Citrin and Shanks was built by combining information about a person's level of education, his verbal ability, his political interest, and his organizational involvement. When 40 See Jack Gitrin "Political Alienation as a Social Indicator: Attitudes 5 and Action," paper delivered at the Cambridge Conference on Social Indi­ cators, Cambridge 5 England, September, 1975; Merrill Shanks, "Survey Based Political Indicators: The Case of Political Alienation," paper delivered at the 1975 annual meeting of the American Political Science Association, San Fran­ cisco, California, September, 1975.

220 / Left-Out Variables

their Protest Participation Index (PPT) was regressed against their Political Alienation Index (PAI), their Personal and Political Resources (PPR) measure, and a multiplicative inter­ action term, PAl*PPR, Citrin and Shanks found that only the interaction term was estimated to have a direct effect on PPI, yielding an R 2 of .302. In this study, the verbal ability of respondents was not as­ certained. But items pertaining to organizational involvement, political interest, and education were included on the interview schedule. The first step in constructing a measure of personal and political resources is to take a simple count of the number of different kinds of organizations to which an individual belongs. Respondents were asked if they belonged to any of eight cate­ gories of organizations: (1) social clubs; (2) church-related groups; (3) school service groups such as the West German equivalent of the PTA; (4) business and professional associa­ tions; (5) student associations; (6) labor unions; (7) political parties or clubs; (8) any other organizations. Table 6.5 shows the distribution of scores on the Organizational Involvement (OI) variable. A majority of the sample (54 percent) belongs to at least one kind of organization. But only 23 percent belong to more than one and only 9 percent belong to more than two different kinds of organizations. Of course, with the exception of political parties or clubs, these organizations are not explicitly political. The next step in building the personal and political resources measure is to weight the Organizational Involvement variable by the re­ spondent's political interest. All respondents who reported that they were not at all interested, not very interested, or only some­ what interested in politics were given a score of zero on a Politicized Organizational Involvement (OIp) variable. Those who said that they were very interested in politics were scored according to their number of organizational memberships. Respondents with non-zero scores on the OIp variable thus are those most likely to be involved in any political discussions and activities that take place within the context of their organiza­ tions. From Table 6.5 one sees that most people weeded out

TABLE 6.5. Distributions of Scores on Organizational Involvement, Politicized Organizational Involvement, and Personal and Political Resources

Personal and Political Resources Score N=

Organizational Involvement JV = Score 7 6 5 4 3 2 1 0

1 4 21 62 157 363 827 1227 2662

Politicized Organizational Involvement N= Score 7 6 5 4 3 2 1 0

1 4 19 56 122 209 347 1865 2622

42 35 28 25 24 21 20 18 16 15 14 12 10 9 8 7 6 5 4 3 2 1 0

4 11 24 1 7 44 10 14 6 24 49 43 53 8 10 65 77 106 44 32 71 18 1813 2534

222 I Left-OutVariabIes

because they lack strong political interest are those with only one or two memberships. People with three or more member­ ships are very likely also to show keen interest in political affairs. The final step in the development of the personal and political resources measure is to take education into account. This is done by multiplying the OIp variable by the E variable (recall from Table 5.3 that E ranges from 1 to 7) to define a Personal and political Resources (PPR) variable. Thus, to achieve a high PPR score, one must belong to a variety of organizations, take a strong interest in public affairs, and be well-educated. The right-hand column in Table 6.5 gives the distribution on Personal and Political Resources. Bringing educational attain­ ment into the picture spreads out the OIp scores considerably, resulting in a rather finely discriminated measure of differing amounts of personal and political resources possessed by respondents. Does possession of high personal and political resources am­ plify the effect of political alienation on aggressive political participation among this sample of West Germans? The answer is negative. As Table 6.6 shows, beginning with straight organ­ izational involvement, there is no amplification effect due to the PSA2*OI interaction. When organizational involvement weighted by political interest is considered, a slight amplifica­ tion effect due to the PSA2*OIp interaction does appear. But when the complete Personal and Political Resources variable is taken into account, the amplification effect due to the PSA2*PPR interaction remains slight (also R2 is virtually unchanged). And when NormativeJustification for Aggression replaces Political Support-Alienation as the measure of norma­ tive incentive for aggressive political behavior, the NJA*PPR interaction is estimated to have exactly zero (rounded to two decimal places) direct effect on APP l n . In equation D from Table 6.6, Personal and Political Re­ sources is estimated to have a direct effect on ΑΡΡγη, inde­ pendent of Normative Justification for Aggression. But when the full Expectancy-Value-Norms model is considered, as equa­ tion 6.8 shows, this direct effect does not hold up:

PSA 2 * OI

Political Support-Alienation, PSA 2 Organizational Involvement, OI

PSA 2 * PPR

Political Support-Alienation, PSA 2 Personal and Political Resources, PPR

NJA*PPR

Normative Justification for Aggression, NJA Personal and Political Resources, PPR

Equation D

Equation C

PSA 2 * OI p

Political Support-Alienation, PSA 2 Politicized Organizational Involvement, OI p

Equation B

Equation A

Predictors of APP ln

.603 .163 .003

.471 .115 .114

.471 .111 .113

.513 .111 .037

Standardized Partial Regression Coefficient

.422

.324

(1766)

(1926)

(1982)

(2003)

.301

.322

(JV=)

R2

and Political Resources in Conjunction with Indicators of Normative Incentive for Aggression

TABLE 6.6. Regressive Analysis of the Dependency of Aggressive Political Participation Scores on Personal

224 / Left-Out Variables

(6.8)

APP l n = 1.362 + -041 {UJA) + .0025(JVJ4) (.005) (.0006) + .0066 (PPR) + .180 (UNV) (.0029) (.045) + .056(A) + .0036(UNV*NJA) (.007) (.0007) + .0000(NJA*PPR),

(.0000) where R 2 = .574 and JV = 1,838. With all the the variables in the prediction equation for APPla, the PPR parameter estimate is not greater than two and one-half times its standard error. Also, as to be expected, the parameter estimate for the NJA*PPR interaction is equal to zero. Re-estimation of 6.8 without the NJA*PPR interaction term gives: (6.9)

APP l n = 1.360 + .041 (UJA) + .0025 (NJA) (.005) (.0006) + .009 (PPR) + .163 (UNV) (.002) (.042) + .056(4) + .0036 (UNV* NJ A), (.007) (.0006)

where R 2 = .574 and JV = 1,838. With the worthless NJA*PPR term deleted, the Personal and Political Resources variable now is estimated to affect aggressive participation, independent of the variables from the Expectancy-Value-Norms model. However, inclusion of PPR in the prediction equation does not increase the proportion of variation accounted for in APP i n beyond the 57 percent level reached by the Expectancy-ValueNorms variables; and, with the exception of the UNV coeffi­ cient, the parameter estimates for the Expectancy-Value-Norms variables are quite unchanged. Thus, PPR appears to be a superfluous variable.41 Let us investigate one other PPR variable, an interaction 41 Also, the standardized coefficient for P P R is trivial, as can be seen from t h e e q u a t i o n : APP l t l = A b b ( U J A ) + . 1 6 8 ( Λ 7 Λ ) + .080(ΡΡΛ) + . 1 1 8 ( U N V ) + .194 ( A ) + . 2 6 6 (U J f V * N J A ) .

Left-Out Variables / 225

with availability for collective action. Perhaps the conjunction of high availability and high personal and political resources is a condition especially conducive to aggressive political par­ ticipation. Equation 6.10 gives the test of the hypothesis that availability for collective action, complemented by resources conducive to collective action, will affect aggressive political participation: (6.10)

APP l n = 1.385 + .041 {UJA) + .0025(^)

(.005)

(.0006)

- .0005 (PPR) + .194 (CZ1AT) (.004) (.043) + .047 (Λ) + m^{UNV*NJA)

(.008)

(.0006)

+ .0024 {A*PPR), (.0008)

where R 2 = .576 and N = 1,838. The multiplicative interac­ tion of availability and resources is, indeed, a statistically significant predictor of Aggressive Political Participation. But again, R2 is virtually the same as that obtained for the Expec­ tancy-Value-Norms variables above, and the coefficients for those variables are stable, indicating that A*PPR is a superflu­ ous variable.42 Thus, there seems no compelling reason not to reduce equation 6.10 to equation 4.14. There are a number of possible reasons why the personal and political resources interaction effect found in the Bay Area Survey failed to be replicated with this sample. For one thing, the dependent variables are different. The Gitrin-Shanks Pro­ test Participation Index is a five-point count of participation in acts of protest. A dependent variable with only five response categories is poorly suited for regression analysis and could lead to results that are not highly reliable. Also, two of the four 42 The A * P P R interaction term also appears to be superfluous when the standardized regression coefficients a r e considered: APP l n = . 1 5 6 ( U J A ) + . I W ( N J A ) - .004(PPR) + .140(UMV) + .164(4) + .251( U N V * N J A ) + .096(A*PPR). Here the standardized coefficient for A*PPR is less than .1 (though just barely).

PREDICTED SCORES ( A P P l t l P R E D )

Large circles denote 9 or more cases.

APPW

FIGURE 6.5 Relationship between Residuals and Predicted Aggressive Political Participation Scores

APPln RESIDUALS (APP^RESJD)

226 / Left-Out Variables

actions included in the PPI measure are not aggressive: par­ ticipation in boycotts and peaceful protest. If personal and political resources were a "facilitator" only of such nonaggressive protest actions, it might not show an interaction with political alienation when both non-aggressive and aggressive actions are included in a single measure of protest.

6.5

SUMMARY AND DISCUSSION

Although the Expectancy-Value-Norms model affords a plau­ sible substantive explanation of magnitude of aggressive politi­ cal participation and, in regard to variance accounted for, appears to be a relatively successful statistical explanation, this in itself does not mean that the model provides a correct explanation. Of course the question of the correctness of the model can never be answered definitively. But one can attempt to arrive at a provisional answer by taking into account as many plausible hypotheses about other determinants as pos­ sible. If the model is incomplete, one or more of these left-out variables will have a significant effect in the prediction equa­ tion; if the model is unreliable, one or more of the parameter estimates may well fluctuate markedly when additional var­ iables are brought into the prediction equation—disappearing altogether in cases where the variable in the model is in reality a spurious determinate and the left-out variable is the true cause. A variety of hypotheses about other causes have been con­ sidered in this chapter, encompassing a number of social back­ ground variables, personality variables, and variables reflecting people's experiences with the day-to-day performance of the political system. To the extent that any of these variables are related to aggressive political participation, the relationships appear to be indirect, via the mediation of the variables in the Expectancy-Value-Norms model. When additional variables were entered into the APPln prediction equation, none of the parameters for the variables in the Expectancy-Value-Norms model were ever reduced even close to insignificance. The parameters describing the dependency of Aggressive Political Participation on the two psychological motivation variables, Utilitarian Justification for Aggressive and Normative Justifi-

Left-OutVariabIes / 227

cation for Aggression, fluctuated within a very narrow range. The other coefficients, with only one exception (the coefficient for UNV in equation 6.9), also were quite stable. Thus, the parameter estimates for the Expectancy-Value-Norms model do give the impression of being quite robust. It is never possible to take all potential "left-out" variables into account. Because of this, and because the specification of the form of one or more of the relationships linking aggressive behavior to the variables in the Expectancy-Value-Norms model could always be in error, it is necessary to analyze the residuals; in particular, it is instructive to plot the residuals against the predicted scotes. The residuals (the differences between observed APP i n and predicted APPla scores) are a measure of the ε term in equation 4.12. They index the amount of error associated with each APPin predicted score. An important assumption of regression analysis is that the errors will be independent of the predicted response scores. If this assumption is correct, the distribution of the residuals should appear as a horizontal band when they are plotted against the predicted scores. If this assumption is seriously violated, the distribution of the residuals may appear as a fan-shaped, upwardly or downwardly sloped, or curvi­ linear band when plotted against the predicted scores, meaning that the predictive accuracy of the set of describing variables is not uniform throughout the full range of the response var­ iable. Among other things, lack of uniform predictive accuracy (especially if the residuals plot as a linear or curvilinear func­ tion of predicted scores) can be due to the omission of an impor­ tant describing variable from the prediction equation or to misspecification of the function linking the response variable to one or more of the describing variables. Figure 6.5 gives the plot of the residuals (APP ln RESID) against the predicted scores (APPlnPRED).43 The dashed lines 43 In computing APP RESlD and APP PRED, cases missing on more than in w two describing variables from equation 4.14 were eliminated. Cases missing on two or less describing variables were assigned the mean. This yields a total N of 2,054, slightly higher than the N of 1,838 for equation 4.14, where pairwise deletion of missing data was used. The R 2 obtained by the procedure of assigning the mean to cases with two or less missing scores is .547—quite similar to the R 2 of .569 obtained under the pairwise deletion procedure.

Left-OutVariabIes / 229

intersect the APP ln RESID axis at two standard deviations above the mean and two standard deviations below the mean. It is clear that the vast majority of cases fall within the band of + 2SD on APPhlRESID. But, visually at least, there does appear to be a negative trend. Extreme positive residuals are registered at low-to-intermediate predicted scores, while extreme nega­ tive residuals are registered at intermediate predicted scores. There are so few extreme negative residuals (25 out of 2,054 cases) that these need not concern us. But the extreme positive residuals number 105, an amount (given the distribution of ΑΡΡγη) sufficiently large to suggest that the Expectancy-ValueNorms model may have a tendency to underpredict the likeli­ hood of actual aggressive political participation. The absence of extreme negative residuals means that, as a forecasting device, the Expectancy-Value-Norms model should be rather accurate when it predicts aggressive political participation. But the presence of extreme positive residuals means that the model will be less accurate when it predicts a lack of aggressive political participation. This deficiency of the model deserves further study. There may well be a left-out variable not considered here that provides an impetus for aggressive participation, independent of the Expectancy-Value-Norms variables, such that when the Expectancy-Value-Norms variables take on low values, some people nevertheless will register non-low Aggres­ sive Political Participation scores. To this point the investigation of potential left-out variables has concentrated exclusively on only one side of the ExpectancyValue-Norms equation—the determinants of behavior. But broadening the rubric of "left-out" variables to encompass response as well as explanatory variables leads one to consider the other side of the equation—the behavior itself, specifically, the question of alternative behavioral responses. When the variables in the Expectancy-Value-Norms model take on high values, is participation in aggressive political action the only likely response? Or might some individuals eschew aggressive activities in favor either of more conventional kinds of partici­ pation or of inactivity? This is an important question because if alternative responses were likely to occur when the Expec-

230 I Left-Out Variables

tancy-Value-Norms variables took on high values, the explana­ tory power of the model would be compromised, since it would tend to specify only necessary but not necessary and sufficient conditions for the occurrence of aggressive political behavior. To investigate the question of alternative responses it is useful to consider a typology of political action defined by crossclassification of participation in aggressive political activities with participation in conventional electoral politics.44 The typology consists of five general types of action :(1) Behavioral Withdrawal, or the absence of participation in either aggres­ sive or electoral action; (2) Conformative Participation, or participation exclusively in electoral action; (3) Reformist Action, or participation in medium levels (civil disobedience) of aggressive action as well as in electoral action; (4) Realist Revolutionary Action, or participation in high levels (political violence) of aggressive action as well as in electoral action; (5) Non-conformative Opposition, or participation exclusively in aggressive action.45 Reformist Action and Realist Revolutionary Action are mixed modes, combinations of both aggressive and conventional political behavior. Behavioral Withdrawal and Gonformative Participation are alternatives to aggressive political action. These two responses should not occur when the variables in the Expectancy-Value-Norms model predict participation in ag­ gressive political action. Conventional political behavior was measured by a set of items referring to standard ways by which a person might par­ ticipate in an election campaign. Respondents were asked if within the past five years they had (1) spoken with people and tried to persuade them to vote for a particular candidate or party; (2) worn a campaign button or affixed a sticker to their car; (3) attended political discussions or rallies; (4) spent money 44 Conventional political participation encompasses action that conforms to regime norms about how members should participate in the making of political decisions. In a polyarchy such as the Federal Republic of Germany, campaign activity is a very important dimension of conventional political participation. 45 See Muller, "Behavioral Correlates ofPolitical Support," American Political Science Review, 71 (June 1977).

Left-Out Variables / 231

to help a candidate or a party; (5) worked for a candidate in a campaign. Response options were: No, scored as 0; Yes, Once or Twice, scored as 1; Yes, More than Once or Twice, scored as 2. The Conventional Political Participation index is the sum of scores on the five campaign-related activities. It is fairly strongly correlated with the Aggressive Political Participation variable at r = .520, attesting to the fact that, among politically active people in this sample, mixed modes of behavior are not uncommon.46 To construct the Political-Action Type index, the Conven­ tional Political Participation and Aggressive Political Par­ ticipation variables are trichotomized and cross-classified as shown in Table 6.7. About two-thirds of the respondents (68.3%) manifest the Behavioral Withdrawal type, the ex­ pected mode in most societies under most circumstances. The next most frequent type, as would normally be expected, is Conformative Participation. Almost one-fifth (18.9%) of the respondents manifest this type. Not surprisingly, the least frequest type is Realist RevolutionaryAction (2%). Almost nine-tenths of this sample conforms to one or the other of the political-action types that represent alternatives to aggressive political participation. Will these alternative types be at all frequent when the Expectancy-Value-Norms model predicts aggressive political participation? We may begin the analysis by first looking at the Political Support-Alienation variable, one of the most important elements of the ExpectancyValue-Norms model. In a smaller scale pilot study carried out in West Germany in 1973 it was found that 53 percent of those who scored in the alienation range of a cruder support-aliena­ tion measure manifested either Behavioral Withdrawal or 46 A variety of studies have reported positive correlations between conven­ tional political action and aggressive action. See Max Kaase and Alan Marsh, "The Matrix of Political Action: Protest and Participation in Five Nations," paper delivered at the Tenth World Congress of the International Political Science Association, Edinburgh, Scotland, August, 1976; Muller, "Behavioral Correlates of Political Support;" Mitchell A. Seligson, "Trust, Efficacy, and Modes of Political Participation," paper delivered at the 1977 annual meeting of the Southwestern Political Science Association, Dallas, Texas, March 30April 2, 1977.

232 I Left-Out Variables

TABLE 6.7. Distribution on the Political-Action

Type Index Conventional Political Participation

In Aggressive

Low (0-3)

Political Participation

Medium (4-6)

>

High

Totals

IV

V Political Violence (3.72-4.67)

High (7-10)

7

22

22

51

99

39

229

. III

High Civil Disobedience > Medium (3.04-3.71) , Low Civil Disobedience (2.36-3.03) >

91 I

II

Low

1486

308

102

1896

Totals

1584

429

163

2176

Inactivity (1.33-2.35)

Political-Action Type Index

Type I Type II Type III Type IV Type V

Behavioral Withdrawal, N = 1486 Conformative Participation, JV= 410 Reformist Action, = 138 Realist Revolutionary Action, N = 44 Non-conformative Opposition, N = 98

Conformative Participation. 47 Table 6.8 shows the results for this sample when scores on the PSA variable are trichotomized into equal thirds. Here, a similar phenomenon occurs: fully 57.7 percent of those registering in the alienation range of PSA 47

See Muller, "Behavioral Correlates of Political Support."

(367)

100%

Totals

(63) (295)

1.4 0 1.1 17.2 80.4

Non-conformative Opposition Realist Revolutionary Action Reformist Action Gonformative Participation Behavioral Withdrawal (5) (0) (4)

%

Political-Action Type

100%

4.3 0.4 3.8 19.4 72.1

%

(1240)

(47) (241) (894)

(5)

(53)

JV=

(17-32)

(0-16) TV =

Ambivalence

Support

Political Support-Alienation

100%

(378)

(35) (39) (86) (102) (116)

JV =

(33-48)

9.3 10.3 22.8 27.0 30.7

%

Alienation

TABLE 6.8. Relationship between Political Support-Alienation and Political-Action Type

Totals

Non-conformative Opposition Realist Revolutionary Action Reformist Action Conformative Participation Behavioral Withdrawal

Political-Action Type

100%

(1568)

100%

(350)

100%

(102)

(13) (21) (38) (22) (8) 12.7 20.6 37.3 21.6 7.8

(40) (13) (66) (130) (101)

11.4 3.7 18.9 37.1 28.9

(40) (1) (29) (252) (1246)

High Civil Disobedience (3.04-3.71) °/ N= /0

Low Civil Disobedience (2.36-3.03) °/ JV = /0

2.6 0.1 1.8 16.1 79.5

Inactivity (1.33-2.35) °/ N= /0

16.7 50.0 22.2 11.1 0

(18)

(3) (9) (4) (2) (0)

Political Violence (3.72-4.67) % JV=

100%

Predicted Aggressive Political Participation(APPinPRED)

of the Variables in the Expectancy-Value-Norms Model) and Political-Action Type

TABLE 6.9. Relationship between Predicted Aggressive Political Participation Scores (on the Basis

Left-Out Variables / 235

manifest one of the alternatives to aggressive political par­ ticipation. When all the variables in the Expectancy-Value-Norms model are considered, among those whose combined scores are high enough to predict participation in civil disobedience (the High Civil Disobedience zone of APPlnPRED), only 29.4 percent manifest an action type that represents an alternative to aggression; and among those whose combined scores are high enough to predict participation in political violence (the Political Violence zone of APPlriPRED), only 11.1 percent manifest an action type that represents an alternative to aggres­ sion. Table 6.9 gives the results. When the Expectancy-Value-Norms model predicts polit­ ically aggressive behavior, alternative responses are not very likely to occur. Still, the manifestation of Conformative Par­ ticipation (21.6%) when the model predicts civil disobedience is sufficiently prevalent to pose an intriguing question for future research: is there a set of systematic conditions not considered here that leads some people, despite incentive for civil dis­ obedience reflected by their scores on the variables in the Expectancy-Value-Norms model, to eschew civil disobedience in favor of purely "wi thin-system" kinds of political activity?48 Another question for future research is posed by the large numbers of individuals classified into the Non-conformative Opposition type whose predicted aggressive participation is in the Inactivity or Low Civil Disobedience zones. Why are so many Non-conformative Oppositionists predicted, on the basis of the Expectancy-Value-Norms variables, to be politically non-aggressive? The answer to this question might well be found at the group level. Perhaps these are individuals mobil­ ized to participate in aggressive political action by group directives, regardless of low motivational incentive stemming from personal beliefs in the normative and utilitarian justifi­ ability of political aggression. 48 An initial attempt to develop a model for explanation of why some people engage in both conventional and aggressive action, while others take part in only one or the other, appears in Muller, "Behavioral Correlates of Political Support."

SEVEN

Cross-Validity of the Expectancy-ValueNorms Model

The determination that variables derived from other hypoth­ eses and models appear to be either irrelevant or superfluous attests to the validity of the Expectancy-Value-Norms model for the 1974 sample. But this is only a first step in the direction of validating the model. The critical step tests the reproducibility of the model. How general is it? Do the weights (or causal parameters) reflect general laws or do they pertain only to the particular case of the 1974 sample? To determine the repro­ ducibility of the model, one must first cross-validate it.1 Crossvalidation requires that the model be tested again for either a different sample from the same population or the same sample at a later point in time. If the cross-validity of the model is established, then, of course, it is desirable to carry out a validity generalization study, where validity generalization is determined by testing the model for a sample from a different population. The cross-validation phase of this project entailed re-interviews in the fall of 1976 with 49 percent (jV = 1,310) of the original respondents. Means and standard deviations for the 1974 Full Sample, the 1974 Panel Sample, and the 1976 Panel Sample are given in Table 7.1. 1 An informative review of the literature on cross-validation appears in Bernard M. Finifter, "The Generation of Confidence: Evaluating Research Findings by Random Subsample Replication," in Sociological Methodology 1972, ed. Herbert L. Costner (San Francisco: Jossey-Bass, 1972), 112-175. Also see the papers published under "Symposium: The Need and Means of CrossValidation," Education and Psychological Measurement, 11 (1951), 5-28.

Cross-Validity / 237

TABLE 7.1. Means and Standard Deviations for the Expectancy-Value-Norms Variables, 1974 Full Sample, 1974 and 1976 Panel Samples

Variable

1974

1976

Panel Sample X SD

Panel Sample X SD

.68

(2198)

1.83 .53 (1109)

2.02 .67 (1056)

(Λ" = )

2.28 2.57 (2247)

1.90 2.20 (1089)

1.74 1.88 (1142)

(JV = )

47.25 45.88 (2147)

35.61 35.35 (1045)

29.39 31.71 (1105)

.24 .43 (1310)

.24 .43 (1310)

APP ln

(^ = )

UJA MJA UJVV

1974 Full Sample X SD

(j\r=)

A

(JV = ) UJVV*JVJA

(JV = )

2.01

.42

.49

(2662)

4.08 2.35 (2393) 32.94 50.31 (2147)

(1180)

3.06 1.72 (1289)

16.44 36.40 (1045)

13.52 31.56 (1105)

3.27

1.96

A major difference between the 1974 Full Sample and the two Panel Samples is the proportion of persons residing in university and non-university communities. In the 1974 Full Sample, 42 percent came from university communities. The Panel Samples contain only 24 percent from university com­ munities. The Panel Samples also differ considerably from the 1974 Full Sample in the means and standard deviations of the Expectancy-Value-Norms predictor variables, with the Panel Samples showing substantially reduced variation, as indexed by the standard deviation, and lower means. From the stand­ point of cross-validating the Expectancy-Value-Norms model estimated for the 1974 Full Sample, these differences are not at all unwelcome. Because the distributional characteristics of

238 I Cross-Validity

the variables do differ markedly, the panel samples can provide a more stringent test of the cross-validity of the ExpectancyValue-Norms model than would be the case if the samples showed close similarity. The cross-validity of the Expectancy-Value-Norms model may be determined as follows: (1) Estimate the parameters of the model for the 1974 Full Sample, the 1974 Panel Sample, and the 1976 Panel Sample. These parameters should not fluctuate markedly. (2) Use the parameters from the 1974 Full Sample to predict Aggressive Political Participation in the 1974 Panel Sample and the 1976 Panel Sample. Predictive accuracy (as deter­ mined by R2) should not fluctuate markedly. (3) Use the parameters from the 1974 Panel Sample to predict Aggressive Political Participation in the 1976 Panel Sample and use the parameters from the 1976 Panel Sample to predict Aggressive Political Participation in the 1974 Panel Sample. Predictive accuracy should not fluctuate markedly. Recall that the Expectancy-Value-Norms model estimated for the 1974 Full Sample was: (7.1)

APP l n = 1.390 + .044( U J A ) + .003 ( J f J A ) (.005) (.0006) + .236( U N V ) + .046(4) (.039) (.007) + .004( U N V * N J A ) ,

(.0006) where R 2 = .569 and M = 1,838. The variables in the equation were defined as: APP l n = natural logarithms of the Aggressive Political Participation scale. Range: 1.54 to 4.52. U J A = Utilitarian Justification for Aggression, de­ fined as degree of belief in the efficacy of col­ lective political aggression weighted according to whether a person's political influence cap-

Cross-Validity / 239 ability is regarded as unnecessary (ECA scores reduced to zero), sufficient (ECA scores unchanged) , or insufficient (ECA scores doubled). Range: 0 to 14. Normative Justification for Aggression, defined as the product of the square of a person's degree of alienation from the political system times his degree of leftist ideological commitment. Range: 0 to 230.4. UNV = a dummy variable coded as 1 if the individual lives in a university community, 0 otherwise. A = an index of pure availability for collective action, defined as the sum of the reciprocal of a person's age in years, facilitative marital status, and facilitative employment status, interaction between social norms and personal normative beliefs. The parameters of the Expectancy-Value-Norms model for the 1974 Panel Sample are estimated to be: (7.2)

where .434 and JV= 913. The parameter estimates for the 1976 Panel Sample are:

(7.3)

where

240 I Cross-Validity

As equations 7.1, 7.2, and 7.3 show, the regression weights for the Expectancy-Value-Norms model are remarkably con­ sistent across samples. Apparently, the Expectancy-ValueNorms parameter estimates do, indeed, represent general laws that hold for this population. The predictive accuracy of the model is reduced for the panel samples as compared with the full sample. This is because the standard deviations of the variables in the panel study are all considerably smaller than the standard deviations of the variables in the 1974 Full Sample, and the size of the multiple correlation coefficient will be smaller, the less the variability, everything else being equal. Now let us turn to consideration of the second and third conditions of cross-validation: uniform predictive accuracy regardless of which regression coefficients are inserted into the prediction equation for APP in the 1974 and 1976 Panel Samples. When the weights from the 1974 Full Sample are inserted into the prediction equation for Aggressive Political Participation in the 1974 Panel Sample, the R2 value is .411. When these weights are inserted into the prediction equation for Aggressive Political Participation in the 1976 Panel Sample, the R2 value is .412. This finding is further testimony to the generality of the full sample weights. Also, when the weights from the 1976 Panel Sample are inserted into the prediction equation for Aggressive Political Participation in the 1974 Panel Sample, the R 2 value is .441, as compared with an R 2 value of .389 obtained when the weights from the 1974 Panel Sample are used to predict Aggressive Political Participation in the 1976 Panel Sample. These R2 values are sufficiently similar to indicate that the weights estimated from the panel samples are essentially interchangeable. Table 7.2 gives the results. 7.1

DIRECTION OF CAUSALITY

In addition to providing information about the generality of the Expectancy-Value-Norms model for this population, the panel data can also be used to check on the direction of causality.

Cross-Validity / 241

TABLE 7.2. Prediction of Aggressive Political Participation in the Panel Samples on the Basis of Regression Coefficients Estimated from the 1974 Full Sample and the Other Panel Sample

(a) Prediction o f ^ P P i n 1974 Panel Sample on the basis of regression coefficients estimated from the 1974 Full Sample:

(b) Prediction of APP in 1976 Panel Sample on the basis of regression coefficients estimated from 1974 Full Sample:

(c) Prediction of AW3 in 1974 Panel Sample on the basis of regression coefficients estimated from 1976 Panel Sample:

(d) Prediction of APP in 1976 Panel Sample on the basis of regression coefficients estimated from 1974 Panel Sample:

Is it correct to assume that the flow of causality runs unidirectionally from the Expectancy-Value-Norms variables to Aggressive Political Participation? Or could some of the association between the putative independent and dependent variables be due, in reality, to a reverse flow of causality, with the participation variable exerting a causal influence on the Expectancy-Value-Norms variables ?

242 / Cross-Validity

The question of direction of causality can be investigated by estimating the parameters of the following model:

Here, aggressive political participation in the panel samples is denoted by and . A summary ExpectancyValue-Norms variable for the 1974 Panel Sample, is constructed by adding the scores on and weighted by the regression coefficients from the 1974 Full Sample, which have been established as valid general weights for this population. Similarly, a summary Expectancy-Value-Norms variable for the 1976 Panel Sample, is constructed by adding the scores on and weighted by the regression coefficients from the 1974 Full Sample. The relationship denoted by the double-headed arrow between and is estimated by the correlation coefficient for these two variables. This arrow is double-headed because it represents a cross-sectional correlation for which the direction of causality initially is assumed to be unknown. The relationships denoted by the single-headed arrows are estimated by the standardized regression coefficients from the following equations: (7.4) (7.5) where and are coefficients that describe the degree to which the response variable in each equation is simply a function of itself at a prior time, regardless of the putative causal variable; and the and coefficients are estimates of the

Cross-Validity / 243

degree to which values of the response variable at a later point in time depend on values of the putative causal variable at an earlier time. If the flow of causality is unidirectional from the ExpectancyValue-Norms variables to Aggressive Political Participation, the coefficient should be of a non-trivial magnitude and the coefficient should be close to zero. Also, if the ExpectancyValue-Norms variables exert a relatively strong causal influence on Aggressive Political Participation, and if they afford a relatively complete explanation of it, then the coefficient should outweigh the coefficient. The results are: 1974 Panel

Sample

1976 Panel

Sample

Since the estimated effect o f o n is less than 0.1, while the estimated effect of on . is fairly sizable, we can conclude that the relationship between Aggressive Political Participation and the Expectancy-Value-Norms Variables is, indeed, unidirectional, with the latter causing the former. Moreover, the -leads-toeffect clearly outweighs the -leads-to-. stability coefficient, thus suggesting that the Expectancy-Value-Norms Model is a relatively powerful explanation of Aggressive Political Participation for this population.

EIGHT

Uses and Limitations of the Expectancy-ValueNorms Model

Like any model, the Expectancy-Value-Norms model is a simplification of reality. To the historian, attuned to the unique aspects of events that give rise to mass political aggression, the Expectancy-Value-Norms model certainly will seem an over­ simplification. Indeed, even Crane Brinton, an historian oriented toward the general rather than the particular, was skeptical about the possibility of ever developing models expressed in the form of equations for predicting political revolution. At the conclusion of his search for uniform precon­ ditions or "prodromal" symptoms indicative of the coming of the English, American, French, and Russian revolutions, Brinton observed; "clearly we must infer from what we have just done that in its earlier stages diagnosis of revolution is extremely difficult, and certainly cannot be reduced to a neat formula, a recipe, a set of rules."1 To think that the analysis of revolution (especially the Great Revolutions) could be dis­ tilled to a set of equations, a mere recipe—it does seem implau­ sible at first blush. Yet, with certain qualifications, such is precisely what has been attempted in this book. The most important qualification stems from the distinction between revolution and rebellion. Although definitions have been almost as numerous as theories of revolution, at the core of the concept is the notion that revolution entails substantial 1

Brinton, The Anatomy of Revolution (New York: Vintage Books, 1957), p. 68.

Uses and Limitations / 245

or fundamental social/political change as a consequence of rebellion.2 According to this definition, revolution can be seen as identical to a successful rebellion, where rebellion, following D. E. H. Russell, is defined as "a form of violent power struggle in which the overthrow of the regime is threatened by means that include violence."3 The conditions that determine the outcome of a rebellion are quite different from the conditions that lead to its initiation. Among the most important variables determining the outcome, as Russell has convincingly demon­ strated, is the degree of disloyalty among the armed forces. But the disloyalty of the armed forces is not a precondition of rebellion.4 Thus, when focusing on the early stages of revolu­ tion, one is really looking at preconditions of rebellion. Even if a model could be developed to aid in the forecasting of rebellion, knowledge of the likelihood of rebellion would be of limited use for prediction of revolution. In effect, diagnosis only of a necessary condition of revolution (the outbreak of rebellion) is possible on the basis of "early-stage" variables such as those included in the Expectancy-Value-Norms model. A second qualification stems from the distinction between rebellion and collective political violence. Violent acts of ag­ gressive political behavior engaged in by groups of non-elites qualify as instances of collective political violence. But the occurrence of collective political violence does not a rebellion make. According to Russell's definition of rebellion, the over­ throw of the regime must be threatened. How to determine whether it is threatened or not? Russell suggests the following criteria: (1) the amount of violence; (2) the duration of the conflict; (3) the geographical area involved; (4) the number 2 Useful reviews of definitions of revolution are given in Mostafa Rejai, The Strategy of Political Revolution (Garden City N.Y.: Anchor Press, 1973), pp. 1-9; D. E. H. Russell, Rebellion, Revolution, and Armed Force (New York: Academic Press, 1974), pp. 56-60. 3

Russell, Rebellion, Revolution, and Armed Force, p. 56. Rebellions can begin without the armed forces being disloyal. However, if portions of the armed forces defect during the course of a rebellion, or if they lack willingness to coerce the rebels, then Utilitarian Justification for Aggres­ sion will rise over the long run, and in this sense armed forces disloyalty will contribute indirectly to increasing Aggressive Political Participation. 4

246 / Uses and Limitations

of active rebels.5 To threaten to overthrow the regime, and thereby qualify as an instance of rebellion, collective political violence must be relatively large-scale in terms of amount, duration, scope, and size. Although the Expectancy-ValueNorms model could aid in forecasting the size of conflict, and might have something to say about its potential scope, it will not yield knowledge relevant to the amount and duration of conflict. Thus the Expectancy-Value-Norms model really is limited to diagnosis of the likelihood of potential rebellion—col­ lective political violence that does not necessarily threaten the overthrow of the regime. Having put the Expectancy-Value-Norms model into proper perspective within the broad topic of a theory of revolution, let us now turn to consideration of some of its more specific uses and limitations. First it will be helpful to illustrate just how the model works.

8.1

PROJECTIONS FROM THE MODEL

To appreciate how the Expectancy-Value-Norms model works it is instructive to take a look at the varying projections given as a function of differing combinations of the describing vari­ ables. Recall that the model is estimated as: In Aggressive Political Participation = 1.390 + .044 (Utilitarian Justification for Aggression) + .003 (Normative Justification for Aggression) + .004 (Normative Justification for Aggression— among persons from a facilitative social context, in this case the university) + .266 (Facilitative Social Norms—in this in­ stance exposure to the university milieu) + .046 (Availability) For persons from a context wherein social norms are facilitative, the model may be expressed as: 5

Russell, Rebellion, Revolution, and Armed Force i pp. 68-69.

Uses and Limitations / 247

In Aggressive Political Participation = 1.656 + .044 (Utilitarian Justification for Aggression) + .007 (NormativejustificationforAggression) + .046 (Availability) Since the Facilitative Social Norms term in this case is a dummy variable, its direct effect can simply be added to the intercept to give the projected APP score among people from the uni­ versity milieu, independent of the other variables, while its interaction effect is simply added to the base NJA coefficient to give the estimated effect of Normative Justification for Aggression among people from the university milieu, inde­ pendent of the other variables. Removing the effect due to the presence of facilitative social norms leaves the following model : In Aggressive Political Participation = 1.390 + .044 (Utilitarian Justification for Aggression) + .003 (Normative Justification for Aggression) + .046 (Availability) This expression of the model applies to persons residing in a context where social norms are inhibitory. Projected Aggressive Political Participation scores for an un­ married student in his or her early twenties (Availability score of about 8) as a function of Normative Justification for Aggres­ sion are given in Figure 8.1. If this student were supportive of the political system and centrist ideologically, he might have a PSA score of 12 (which equals a PSA2 score of 1.44, i.e., 12*12/10) and an LIC score of 5, resulting in an MJA score of 7.2 (1.44*5). Assume this student felt that his or her ability to influence government was sufficient and that collective political aggression had been neither especially harmful nor especially beneficial to dissident groups in the past, yielding a Utilitarian Justification for Aggression score of, say, 4. The projected APPln score for this student would be determined by locating where an NJA score of 7.2 falls on slope C of Figure 8.1. In this case an NJA score of 7.2 gives an estimated APP ln score of 2.22, so one would predict inactivity on the part of the student.

NORMATIVE JUSTIFICATION FOR AGGRESSION (NJA)

FIGURE 8.1 Estimated Effects of Normative Justification for Aggression on Aggressive Participation Given Facilitative Social Norms, High Availability, and Varying Utilitarian Justification for Aggression

111 AGGRESSIVE POLITICAL PARTICIPATION (APPln)

Uses and Limitations / 249

Now let us assume that this student became very concerned about contemporary political issues and that with this height­ ened political interest came dissatisfaction with the perfor­ mance of the government in certain policy areas such that his or her alienation from the system began to increase. At the same time, he or she became involved with a left-wing group on campus and this further increased alienation from the system while also increasing commitment to a leftist ideological position such that his or her LIC score went up to 8 while PSA moved into the alienation range at about 36, resulting in an NJA score of 103.7 (3.62*8). Assume further that this student came to feel that his or her ability to influence government was now insufficient, thus raising UJA from 4 to 8. In this case the student's projected APPln score would be determined from slope B instead of slope G of Figure 8.1. It would have increased to a value of 3.07, falling in the High Civil Disobedience zone. Now suppose that this student became even more alienated from the system of government and slightly more intense in leftist ideological commitment such that his or her PSA score went from 36 to 44 and LIC went from 8 to 9, resulting in an MJA score of 174.2 (4.4 2*9). And suppose that at the same time this student came to believe that collective political aggression had been moderately beneficial to dissident groups in the past, registering an increase in UJA from 8 to 12. The student's projected APPln score would then be determined from slope A instead of slope B of Figure 8.1. It would have increased to a value of 3.74, falling in the Political Violence zone. By contrast, consider the graph for a hypothetical Assistent employed at the university (a position roughly equivalent to an assistant professor in the United States), married, in his (for convenience of expression assume male gender) mid-tolate twenties (Availability score of about 4). This is shown in Figure 8.2. If this Assistent were moderately alienated from the political system {PSA score of, say, 38) and moderately left in his ideological views (LIC score of 8) his NJA score would be equal to 103.7. Ifhe believed that collective political aggression had been neither especially harmful nor especially beneficial to challenging groups in the past, but felt that his political in-

FIGURE

2.00

4.67 J

72

96

120

144

168

192

+ .044( 0)

D

+ -044( 4) C

A P P l l l = (1.390 + .236) + [(.003 + .004) ( N J A ) ] + .046(4) + .044(12) A + .044( 8) B

NORMATIVE JUSTIFICATION FOR AGGRESSION ( N J A )

48

8.2 Estimated Effects of Normative Justification for Aggression on Aggressive Participation Given Faeilitative Social Norms, Medium-Low Availability, and Varying UtilitarianJustification for Aggression

Uses and Limitations / 251

fluence capability was insufficient, his UJA score would be equal to 8. Unlike the student with identical values of NJA and UJA, whose projected APPln score would fall in the High Civil Disobedience zone, this Assistent would have a projected APPln score falling in the Low Civil Disobedience zone—an APPln score of 2.89 as determined from slope B of Figure 8.2. Were this Assistent to become even more politically alienated, slightly more intense in his ideological commitment, and moderately positive in his belief in the efficacy of past collective political aggression such that his MJA score went to 174.2 and his UJA score went to 12, his projected APPln score then would be determined from Slope A of Figure 8.2, taking a value of 3.56. This would place the Assistent in the High Civil Disobedi­ ence zone, as contrasted with the hypothetical student, whose projected APPln score on the basis of the same NJA and UJA values falls in the Political Violence zone. Indeed, it would require a combination of UJA scores greater than 4 and very high NJA (scores at least approaching 200) before the Assistent would be expected to score in the Political Violence zone of APPin. Now take the case of a hypothetical Professor Doktor, married, in his late forties (Availability score of about 2). As Figure 8.3 shows, the Expectancy-Value-Norms model predicts that it would be extremely unusual for such a person ever to engage in violent political action. In fact, no matter how high this pro­ fessor were to score on belief in the normative justifiability of political aggression, unless he also were to believe that collec­ tive political aggression had been at least slightly useful in the past and that his own political influence capability was in­ sufficient such that he registered a UJA score at least greater than 8, he would never be expected to score in the Political Violence zone of APPln. Where social norms strongly disfavor aggressive political action, as is the case in the rural and urban communities in this sample, the Expectancy-Value-Norms model predicts that political violence will not occur.6 Figure 8.4 shows that pro6 This should not be interpreted to mean that the model predicts a lack of violence in urban and rural communities generally. It predicts a lack of vio

NORMATIVE JUSTIFICATION FOR AGGRESSION (NJA)

FIGURE 8.3 Estimated Effects of Normative Justification for Aggression on Aggressive Participation Given Facilitative Social Norms, Low Availability, and Varying Utilitarian Justification for Aggression

In AGGRESSIVE POLITICAL PARTICIPATION (APP,„)

Uses and Limitations / 253

jected APP la scores for persons who are highly available for collective action, but reside in communities where social norms are inhibitory, only begin to approach the High Civil Dis­ obedience zone when utilitarian incentive is quite high (UJA = 12) and the scale of normative incentive moves toward the 200 level. APPln scores that actually fall in the High Civil Dis­ obedience zone are expected only when the scale of utilitarian incentive reaches its topmost value (a score of 14—APP ln = MJA slope for this level of UJA not shown) and the scale of normative incentive moves above the 200 level. Among persons with substantially reduced availability for collective action, actual participation in civil disobedience (.APPln scores in the High Civil Disobedience zone) is not expected to occur, so long as social norms are inhibitory. Figures 8.5 and 8.6 show the APPln-MJA slopes for the various levels of UJA, given the condition of inhibitory social norms and Availability scores of 4 and 2, respectively. Only at quite high UJA, MJA above the 200 level, and an Availability score of 4, does expected APPln begin to approach the High Civil Disobedience zone. With Availability down to a score of 2, expected APPln never even approaches the High Civil Dis­ obedience zone, however high the score on the scales of utili­ tarian and normative incentive for participation in collective political aggression. The graphs presented in this section show how differences among individuals in their utilitarian incentive for aggressive

lence generally in any social context where norms are inhibitory, i.e., disfavor aggressive political participation. In this particular sample, the norms in the urban and rural communities are strongly inhibitory; no violence is predicted and, indeed, only 2 of 1,558 persons from the urban-rural context score in the Political Violence zone of APP la . But in urban or rural contexts where social norms were not inhibitory, the model would be expected to predict violence, given appropriate combinations of scores on the utilitarian and normative motivational incentive variables and on availability for collective action. Of course the question of whether or rvot the model will work in the same way across a variety of facilitative social contexts is an important topic for future investigation.

NORMATIVE JUSTIFICATION FOR AGGRESSION ( N J A )

FIGURE 8.4 Estimated Effects of Normative Justification for Aggression on Aggressive Participation Given hibitory Social Norms, High Availability, and Varying Utilitarian Justification for Aggression

In AGGRESSIVE POLITICAL PARTICIPATION (APP,„)

FIGURE 8.5

24

72

96

120

144

168

192

A P P l a = 1.390 +

.OOS( N J A )

+ .044( 0) D

+ -044( 4) C

+ .046(4) + .044(12) A + 044( 8) B

NORMATIVE JUSTIFICATION FOR AGGRESSION ( N J A )

48

216

240

Estimated Effects of Normative Justification for Aggression on Aggressive Participation Given In­ hibitory Social Norms, Medium-Low Availability, and Varying UtilitarianJustification for Aggression

NORMATIVE JUSTIFICATION FOR AGGRESSION (NJA)

FIGURE 8.6 Estimated Effects of Normative Justification for Aggression on Aggressive Participation Given Inhibitory Social Norms, Low Availability, and Varying Utilitarian Justification for Aggression

In AGGRESSIVE POLITICAL PARTICIPATION (APP^)

Uses and Limitations / 257

political action, their normative incentive for aggressive polit­ ical action, their exposure to inhibitory or facilitative social norms, and their availability for collective action combine to produce different expected magnitudes of aggressive political participation, ranging from gradations of inactivity up through gradiations of political violence. If the Expectancy-ValueNorms model is essentially correct, it affords the beginning of a quantitative answer to the engineering question of what it takes to produce aggressive political participation. One must stress "beginning" because the model is "single-stage," encom­ passing only direct determinants of aggressive political par­ ticipation. A complete answer to the question of what it takes to produce aggressive political participation would have to specify a "multi-stage" model encompassing all indirect determinants of aggressive political participation that function as antecedents of the describing variables in the ExpectancyValue-Norms model. This topic will be taken up in the next chapter. 8.2

THE MODEL AS A DIAGNOSTIC TOOL

A person's expected level of Aggressive Political Participation can readily be. determined from personal interview or ques­ tionnaire data that reveal whether he is exposed to inhibitory or facilitative social norms and how he scores on the scale of normative incentive for political aggression, the scale of utilitarian incentive for political aggression, and the scale of availability for collective action. Multiplying the person's score on each of these variables by the appropriate weights from equation 4.14 gives his projected Aggressive Political Participation score. If projected Aggressive Political Participa­ tion falls within the range 1.33 to 2.35, one predicts inactivity; if projected aggressive participation falls within the range 2.36 to 3.03, low civil disobedience; if projected aggressive par­ ticipation falls within the range 3.04- to 3.71, high civil dis­ obedience; and if projected aggressive participation falls within the range 3.72 to 4.67, political violence. How accurate might these projections be? To assess the po-

Reality

1326

1580

Inactivity

Totals

183

70

High Civil Disobedience

Low Civil Disobedience

1

Political Violence

Inactivity

354

83

151

102

18

Low-Civil Disobedience

102

44

26

51

21

High Civil Disobedience

18

0

2

5

11

Political Violence

Expectancy-Value-Norms Model Predicts

TABLE 8.1. Predictive Performance of the Expectancy-Value-Norms Model as Determined from the Relationship between Predicted and Observed Aggressive Political Participation Scores

2054

1413

362

228

51

Totals

191/1934 = .10

88/120 = .73

Political Aggressive

1743/1934= .90

254/1580 = .16

1326/1580 = .84

Inactivity

Politically Non-aggressive

120/354 = .34

151/354 = .43

Low Civil Disobedience

21/102 = .21

51/102 = .50

11/18 = .61

False Negative (> predicted)

High Civil Disobedience

Political Violence

Correct Positive

Predictive Performance

32/120 = .27

83/354 = .23

30/102 = .29

7/18 = .39

False Positive (< predicted)

260 / Uses and Limitations

tential diagnostic performance of the Expectancy-Value-Norms model, it is informative to compare its predictions with reality, that is, to cross-classify predicted APP levels with observed APP levels, as in Table 8.1. Let us first take up the question of diagnosing a person's exact level of APP. The first four rows of the lower sub-table give the probabilities (computed from the upper sub-table) of making a correct diagnosis (Correct Positive), incorrectly diagnosing a level of APP that is too low (False Negative), and incorrectly diagnosing a level of APP that is too high (False Positive). In diagnosing the extremes of the APP continuum, the Expectancy-Value-Norms model does pretty well. The prob­ ability of making a correct diagnosis of political violence is 6 in 10, while the probability of making a correct diagnosis of inactivity is 8 in 10. The model fares less well in diagnosing potential (low) and actual (high) civil disobedience, as the probability of correctly diagnosing the former is 4 in 10, while the probability of correctly diagnosing the latter is 5 in 10. However, if one were concerned only with diagnosing the probability of a person manifesting at least low civil disobedience when this is predicted, or the probability of a person manifesting at least high civil disobedience when this is predicted, then the model does considerably better, since False Negatives would not be counted as errors. Under this relaxed diagnostic cri­ terion, the probability of correctly diagnosing at least low civil disobedience is 0.43 + 0.34 = 0.77, or 8 in 10, while the prob­ ability of correctly diagnosing at least high civil disobedience is 0.50+ 0.21 =0.71, or 7 in 10. In a polyarchy (or democratic polity) one might be especially concerned with the question of predicting not more than low civil disobedience, or not more than high civil disobedience. In these instances, False Positives would not be counted as errors. Given this criterion, the model also fares well, since the probability of correctly diagnosing not more than low civil disobedience is 0.43 + 0.23 = 0.66, or 7 in 10, while the probability of correctly diagnosing not more than high civil disobedience is 0.50 + 0.29 = 0.79, or 8 in 10.

Uses and Limitations / 261

Clearly, the most important diagnostic errors are False Negatives when the prediction is for inactivity, and False Positives when the prediction is for low civil disobedience, high civil disobedience, or political violence. In the former instance the model is predicting that a person will not be politically aggressive when in reality, he might (low civil disobedience) or will (high civil disobedience, political violence) be aggressive. In the latter instance the model is predicting that a person might or will be politically aggressive when, in reality, he will not be aggressive. These diagnostic errors are generally minimal, having a chance of only 2 in 10 of occurring when the prediction is for inactivity or for low civil disobedience, a chance of 3 in 10 of occurring when the prediction is for high civil disobedience, and a chance of 4 in 10 of occurring when the prediction is for political violence. Next, consider just using the model to diagnose whether or not a person will be politically aggressive (actually engage in acts of civil disobedience or political violence). The last two rows of the lower sub-table give the predictive performance of the model relevant to this question. Here one sees that the model performs rather well, yielding a correct diagnosis of politically aggressive individuals in 7 of every 10 cases, a correct diagnosis of politically non-aggressive individuals in 9 of every 10 cases. Pending the results of future replication, it appears at least from these data that the Expectancy-Value-Norms model has respectable diagnostic credentials.7 To what uses might it be put? What are its limitations? 7 Regarding the use of the Expectancy-Value-Norms model as a diagnostic tool, one caveat pertaining to measurement is worthy of note. The single most important element in the model is the Political Support-Alienation variable. PSA is the most important component of Normative Justification for Aggres­ sion—PSA and NJA correlate at r — .920, whereas LIC and NJA correlate at r ~ .624, meaning that PSA and NJA share 85% common variance while LIC and NJA share only 39% common variance, Normative Justification for Aggression, in turn, figures in three different ways in the model: (1) it has a direct effect on APP; (2) together with facilitative social norms it has an inter­ action effect on APP; (3) the presence of facilitative social norms is determined largely by high aggregate Normative Justification for Aggression. Thus, it is especially important that PSA be measured well. Although its components in this sample did show sufficiently high intercorrelation to warrant their additive

262 / Uses and Limitations

One of the obvious limitations of the Expectancy-ValueNorms model is that it relies predominantly on subjective data. The scales of normative and utilitarian incentive for political aggression are subjective measures. Determination of whether the norms in a given social context are facilitative or inhibitory requires information about the aggregate level of normative justification for political aggression in that particular context. Of the five predictor variables in the Expectancy-Value-Norms model, only the Availability term can be derived from objec­ tive social background information. This requirement of sub­ jective data means that the diagnostic capabilities of the model

combination into a composite PSA variable, the agree-disagree format that was used is not the best for general application. This is because the agreedisagree format can be quite susceptible to response set. For general use it might be more desirable to adopt a response format that more directly elicits magnitude ratings. For example, the PSA items could simply be rephrased as: S j: How much difference do you think there is between what people like you value in life and what actually happens in our political system? S 2 : To what extent do you have respect and affection for the political institu­ tions in [blank] ? S 3 : To what extent do you feel that you and your friends are well-represented in our political system? S 4 : To what extent do you feel that the basic rights of citizens are well-respected in our political system? S 5 \ To what extent do you feel very critical of our political system? S 6 : How much of the time do you think that the courts in [blank] guarantee everyone a fair trial ? S 1 : Looking back, to what extent would you say that the leading politicians in [blank] have had good intentions? S Q : Considering everything, how much respect to you think the police in [blank] deserve? Respondents could then give magnitude ratings using a scale consisting of some number of steps ranging from a low of "Not at all" to a high of "A Great Deal." Or, psychophysical scaling based on line production (drawing a line proportional to the magnitude of one's response) or number estimation (giving a number proportional to the magnitude of one's response) could be used in place of categorical scale steps in order to measure magnitude of support. Innovative research on the psychophysical scaling of social opinion, with special emphasis on the development of a psychophysical^ valid political support scale for use in survey research, is reported in Milton Lodge, Josephy Tanenhaus, David Cross, Bernard Tursky, Mary Ann Foley, and Hugh Foley, "The Calibration and Cross-Modal Validation of Ratio Scales of Political Opinion in Survey Research," Social Science Research, 5 (1976), 325-347.

Uses and Limitations / 263

can easily be subverted if respondents decide that it is in their best interests not to reveal their true feelings. Therefore, the model could not long be used by governments to try actually to identify politically aggressive individuals in order to imple­ ment a strategy of preventive detention. The point is that using the model for the purpose of identification would not only be ethically objectionable, it would turn out rather quickly to be self-defeating. If the model were not used to identify particular politically aggressive individuals, it could serve as an early-warning device, a monitor of the "health" of the polity, alerting the government ahead of time to any broad-scale potential for unrest brewing on the horizon, and, if results were made public (as would be necessary to assure respondents that the authorities had no ulterior motive in conducting or condoning such sur­ veys), also alerting dissidents to the presence of opportunities for stepping up efforts to mobilize against the government. But to be used effectively as an early-warning device for diag­ nosing impending increases (or, for that matter, decreases) in aggressive political participation, the model could not rely only on data gathered from simple national probability samples. A necessary supplement to these would be sub-national prob­ ability samples of selected social and institutional contexts that might serve as focal points for the development of social norms favorable to collective political aggression.

NINE

Macro-Micro Linkages

An inherent limitation of the Expectancy-Value-Norms theory derives from its unit of analysis. By focusing on the micro level, the theory is confined to specification of social and psychological attributes that predispose individuals to take part in collective political aggression. But exclusion of macro-level variables from the picture is not necessarily a weakness. Specifically, it is not a weakness so long as relevant macro-level variables can be relegated to the status of unmeasured predetermined var­ iables that affect participation in aggressive political action only indirectly through the mediation of the micro-level varables explicitly taken into account. What are the relevant macro-level variables? The most methodologically sophisticated, rigorous, and comprehensive macro-level study is Hibbs' cross-national causal analysis of determinants of nation-wide variation in the incidence of mass political aggression across 108 polities during the 20-year time period from 1948 to 1967.1 Nation-wide levels of mass political aggression were found to be determined by five 1 Hibbs, Mass Political Violence (New York: Wiley, 1973). From a methodological standpoint, an equally sophisticated and rigorous study is Gurr and Duvall, ti Civil Conflict in the 1960s: A Reciprocal Theoretical System with Parameter Estimates" (See footnote 2 of Chapter Two). However, the Gurr and Duvall work is based on a less comprehensive data set than that used by Hibbs. It should be noted that Gurr and Duvall find results different from Hibbs'. But their strategies of analysis differ substantially. Hibbs' approach is first to evaluate a series of alternative single-equation hypotheses and partial theories, then integrate those hypotheses that survive initial empirical testing into a comprehensive multiequation model. By contrast, Gurr and Duvall begin with a comprehensive multiequation model and then estimate its param­ eters, but they do not consider alternative hypotheses and models.

Macro-Micro Linkages / 265

variables: Negative Sanctions, Average Annual Percentage Change in Energy Consumption Per Capita, Group Discrim­ ination, Communist Party Membership, and Communist Regime (a dummy variable reflecting the presence or absence, of a Communist political system).2 The five-variable equa­ tion yielded a fair degree of predictive accuracy for data at the aggregate level (R2 = .661), and analysis of the residuals did not disclose any serious statistical deficiencies. Is it plausible to assume that these macro-level variables operate as indirect determinants of Aggressive Political Par­ ticipation ? Only if a convincing case can be made linking each macro-level variable to one or more of the micro-level variables in the Expectancy-Value-Norms model. But to do this it is necessary to expand the single-stage Expectancy-Value-Norms formulation into a multi-stage micro model. Consider the antecedents of aggressive political participation that comprise the Expectancy-Value-Norms model. They de­ fine the most proximate stage of causation, the set of unmediated effects that directly produce variation in magnitude of aggres­ sive response. But what causes the causes? Are there systematic micro-level antecedents of the variables in the ExpectancyValue-Norms model? If so, they must be taken into account prior to shifting attention to the macro level. Take the Availability variable first, a non-psychological 2 Recall that Hibbs found two dimensions of mass political violence at the macro level—though the two-dimensional result is disputed in Linehan, "Models for the Measurement of Political Instability," Political Methodology, 3 (Fall 1976). A Collective Protest factor was distinguished by high loadings of Riots, Anti-government Demonstrations, and Political Strikes, whereas an Internal War factor was distinguished by high loadings of Assassinations, Armed Attacks, and Deaths. Micro-level analogues of all the variables that define Collective Protest are included in the Aggressive Political Participation variable used here. Of the variables that define Internal War, a micro-level analogue only of Armed Attacks is included in the Aggressive Political Partic­ ipation variable. (Interestingly, Armed Attacks also consistently showed secondary loadings on the Collective Protest variable, suggesting a fair amount of commonality between it and Riots, Anti-government Demonstrations, and Political Strikes.) Therefore, I take Hibbs' Collective Protest variable as the macro-level analogue of my Aggressive Political Participation variable. For Hibbs' results, see pages 7 to 17 of Mass Political Violence (New York: Wiley, 1973).

266 I Macro-Micro Linkages

variable, defined purely by objective social background attri­ butes of individuals. Through time Availability basically changes in only one direction: down. It reaches its highest levels among unmarried adolescents who have still to complete their schooling. From this point on, since age itself is the dominant component of Availability, its movement is inexorably from high to low. Becoming married and joining the work force will accelerate this decline, while becoming divorced (or widowed) or losing one's job will decelerate—and can even briefly reverse—this decline. But the major antecedent of Availability is time itself. Next take the Facilitative Social Norms variable. This is a socio-psychological variable, since it reflects social contexts in which belief in the justifiability of collective political aggression is relatively widespread. At the micro level, its major deter­ minants are probably those which produce change in the Normative Justification for Aggression variable. In addition, the interaction between Facilitative Social Norms and Norma­ tive Justification for Aggression will be determined by those variables which influence the components of the interaction, meaning those variables which influence Normative Justifica­ tion for Aggression. This leaves the purely psychological variables, Normative Justification for Aggression and Utilitarian Justification for Aggression. The primary determinant of Utilitarian Justifica­ tion for Aggression probably is a macro-level variable, the degree to which government effectively represses dissidents. The more severe the negative sanctions applied to dissidents, the lower is UtilitarianJustification for Aggession likely to be. At the micro level the most important influence on Utilitarian Justification for Aggression is likely to be Normative Justifica­ tion for Aggression. It is plausible to expect that Normative Justification for Aggression may exert some degree of bias on a person's perception of the utility of aggressive actions taken by dissidents. To some extent, regardless of how government responds to dissidents, people who believe that collective political aggression is normatively unjustifiable may want to

Macro-Micro Linkages / 267

believe a priori that it also is undesirable on utilitarian grounds. Similarly, people who believe that collective political aggression is normatively justifiable will have good reason to want to believe a priori that it also is a useful tactic. However, assuming that people are basically rational, the strength of any causal connection between Normative Justification for Aggression and Utilitarian Justification for Aggression due to selective per­ ception bias should be of only modest magnitude.3 In preceding chapters it has been shown that certain indi­ cators of frustration and of specific political performance dis­ satisfaction appear, through the mediation of the Normative Justification for Aggression variable, to have an indirect effect on Aggressive Political Participation. These variables all corre­ late rather strongly with Normative Justification for Aggres­ sion, that is, —Structural Just Deserts Frustration correlates with NJA at r = .514; —Post-Materialist Policy Dissatisfaction correlates with NJA at r = .602; —Dissatisfaction with Political Treatment correlates with NJA at r = .567. This is due primarily to the fact that these variables correlate strongly with the Political Support-Alienation component of NormativeJustification for Aggression, to wit: —Structural Just Deserts Frustration correlates with PSA at r = .492, as compared with a weak correlation of r = .269 between it and Leftist Ideological Commitment, the other component of NJA; —Post-Materialist Policy Dissatisfaction correlates with PSA at r = .587, as compared with a weak correlation of r = .281 between it and LIC; 3 An alternative hypothesis is proposed by Gurr at pages 159 to 160 of Why Men Rebel. He argues that there will be a strong relationship between normative and utilitarian justification for political aggression and that the nature of the relationship will be reciprocal, i.e., both variables causing change in each other.

268 I Macro-Micro Linkages

—Dissatisfaction with Political Treatment correlates with PSA at r — .525, as compared with a moderate correlation of r = .381 between it and L/C. 4 An important unanswered question involves the direction of the relationship. If the indicators of frustration and political performance dissatisfaction do have indirect effects on Aggres­ sive Political Participation, then they must function as deter­ minants of the Political Support-Alienation component of NormativeJustification for Aggression. For were the direction of the relationship to be just the opposite, that is, Political Support-Alienation the cause of Structural Just Deserts Frus­ tration, Post-Materialist Policy Dissatisfaction, and Dissatis­ faction with Political Treatment, then, instead of an indirect effect, frustration and political performance dissatisfaction would have no effect on Aggressive Political Participation. This is because the gross or zero-order correlation (bivariate without controls) between these variables and Aggressive Political Participation would be spurious due to the fact that they and Aggressive Political Participation shared Political Support-Alienation as a common cause. Investigation of the directionality of the relationship between performance dissatisfaction variables and Political SupportAlienation is a complex topic beyond the scope of this analysis. For the purpose of specifying a model of macro-micro linkages, the performance dissatisfaction variables will be postulated to affect Political Support-Alienation, not the other way around. This is predicated on the assumption that a person's generalized affect for the political system will change as a function of his experiences with how incumbent governments perform.5 4 The indicators of frustration and specific support correlate with LIC largely because of their strong relationship with PSA i a variable that correlates moder­ ately with LIC at r = .359. 5 Note that if the direction of influence turns out to be as I hypothesize, the Expectancy-Value-Norms model implies that aggressive political participation is essentially a rational response in the sense that the motivational incentive for aggression arises from normative justification based on aversive experiences with the performance of the political system, as well as utilitarian calculation of success/failure.

Macro-Micro Linkages / 269

The set of performance dissatisfaction variables assumed to have indirect effects on Aggressive Political Participation through the mediation of Political Support-Alienation define a second causal stage of the macro-micro model. A third stage entails specification of determinants of Structural Just Deserts Frustration, Post-Materialist Policy Dissatisfaction, and Dis­ satisfaction with Political Treatment. Assuming that Political Support-Alienation is not a major cause of the performance dissatisfaction variables, it seems plausible to argue that their major antecedents are to be found at the macro rather than the micro level. These performance evaluation variables serve to connect the individual to his en­ vironment. They are likely to vary as a function of changing outputs from macro social, economic, and political institutions. But one micro level variable that surely will have some role to play here is Rank Disequilibrium. Experiencing a condition of rank disequilibrium of the intellectual proletariat type is likely to heighten in some degree feelings of frustration in the just deserts sense.6 It may have some influence as well on feelings of dissatisfaction with political treatment by virtue of the fact that people with high education but low social standing may be predisposed to feel that they will not receive the kind of treat­ ment from political authorities to which they are entitled on the basis of their educational achievement.7 The intellectual pro­ letariat type of rank disequilibrium might even have an indirect influence on policy dissatisfaction through the mediation of Structural Just Deserts Frustration. Here, the premise would be that a person who feels he is not receiving his just deserts would be predisposed to evaluate the policy performance of the government in a negative light. Indeed, for these West German data, StructuralJust Deserts Frustration does correlate moder­ ately with Post-Materialist Policy Dissatisfaction at r = .416. 6 Recall that Rank Disequilibrium correlates with Structural Just Deserts Frustration at r = .284. 7 This hypothesis is implied in Galtung's discussion of the differential treat­ ment to which people in a state of rank disequilibrium may be exposed—see Galtung 5 "A Structural Theory of Aggression," p. 88 (full citation in footnote 50 of Chapter Five). For this West German sample, Rank Disequilibrium cor­ relates with Dissatisfaction with Political Treatment at r = .284.

270 I Macro-Micro Linkages

This completes the elaboration of the micro portion of a pro­ visional multi-stage macro-micro model of determinants of individual participation in aggressive political action. A pic­ torial representation of it is given in Figure 9.1, where the unbroken arrows denote the relationships between micro-level variables just discussed. Note that some of the variables are expressed in more general terms than the particular indicators used here. In this study the UNV variable was used as the indi­ cator of Facilitative Social Norms, but the presence offacilitative social norms obviously is not limited just to university settings, nor is it a necessary feature of the university milieu. Norms change as aggregate levels of Normative Justification for Aggressive change, so that a facilitative context today—be it a university or non-university setting—could become an inhibitory context tomorrow, and vice versa. The LIC variable was used in this study as the indicator of Ideological Approba­ tion of Aggression, but, obviously, leftist ideologies have no corner on the market. At other times and places the appropriate indicator of Ideological Approbation of Aggression might be degree of rightist ideological commitment or, more likely, degree of either rightist or leftist ideological commitment. Finally, the PDpm variable was used in this study as the indicator of Policy Dissatisfaction; but degree of dissatisfaction with governmental performance in post-materialist policy areas is not the only kind of policy dissatisfaction that may influence the Political Support-Alienation variable. Historically, dis­ satisfaction with governmental performance in materialist policy areas certainly has been important. Before discussing the macro-level variables linked to microlevel variables in Figure 9.1 by dashed arrows, the role of "structural" variables in the macro-micro model should be clarified. As stated in the second chapter, micro-structural variables (variables referring to a person's position in the social structure such as education, income, rank disequilibrium, and the like) were not expected to contribute prominently—either as direct or indirect determinants—to explanation of individual differences in level of aggressive political participation. But the term "structural" also is used to refer to aggregate character-

FIGURE 9.1 Example of a Multi-Stage Macro-Micro Model of

Determinants of Aggressive Political Participation (APP) CPM

CR

AKX

\ ^

\

\

A

Governmental Effectiveness

APP

+

JDF' S \ + D P T

RDE AEC

NS t Key A. INDEPENDENT VARIABLES a. Micro-Level A Availability FSN Facilitative Social Norms NJA Normative Justification for Aggression UJA Utilitarian Justification for Aggression IAA Ideological Approbation of Aggression PSA Political Support-Alienation PD Policy Dissatisfaction DPT Dissatisfaction with Political Treatment JDF 8 Structural Just Deserts Frustration RDE Rank Disequilibrium b. Macro-Level C R Communist Regime CPM Communist Party Membership A EC Annual Percentage Change in Energy Consumption per Capita GD Group Discrimination GPR g Gain in Group Power Resources NS t Negative Sanctions NS t . k Negative Sanctions, Lagged

B. SYMBOLS unidirectional relationship interaction effect

>

inestimable parameter reciprocal relationship unoperationalized concept

272 I Macro-Micro Linkages

istics of collectivities of individuals—as a synonym for macro variables. In this sense of the term "structural," such variables are expected to play an important role in the explanation of aggressive participation, albeit not as direct but as indirect determinants. A macro-structural variable included in Figure 9.1, though not yet investigated empirically, is Gain in Group Power Re­ sources. This variable is intended to capture the role that organizational membership plays in the determination of Aggressive Political Participation. The hypothesis, following Korpi (see the discussion in the second chapter), is that mem­ bership in any group gaining power resources will have an indirect effect on APP by virtue of increasing just deserts frus­ tration and utilitarian incentive for aggression. This seems to be a more plausible hypothesis than that advanced by Tilly, namely, that GPRg will bear a direct curvilinear relationship to APP, eclipsing the direct relationship stemming from the psychological variables, NormativeJustification for Aggression and UtilitarianJustification for Aggression. Across nations, Hibbs found that the single most important macro-structural predictor of the Collective Protest dimension of mass political violence was a Negative Sanctions variable, where "negative sanctions are defined as actions taken by political authorities to neutralize, suppress, or eliminate a perceived threat to the security and stability of the govern­ ment, the regime, or the state itself. They include acts of cen­ sorship against mass media, political publications, and the like, as well as restrictions on the political activity and partici­ pation of the general public, or specific persons, parties, or organizations."8 This variable was estimated to have a sub­ stantial causal impact on Collective Protest, as reflected by a standardized regression coefficient of .516. Interestingly, Nega­ tive Sanctions turned out to be a two-edged sword. The immediate effect of repression was not to deter but rather to exacerbate collective political aggression. However, in the long

8

Hibbs, Mass Political Violence, pp. 88—89.

Macro-Micro Linkages / 273

run repression did act as a deterrent.9 This double-edged effect of repression squares well with Tilly's historical survey of political aggression in Europe.10 The scenario runs roughly as follows: (1) dissidents engage in low-level aggressive action, i.e., civil disobedience; (2) government meets civil disobedience with repression; (3) the response of dissidents to repression is to escalate aggressive action, turning to political violence; (4) if government effectively applies repression over a relatively long period of time, collective political aggression subsides. But why does repression cut both ways? The macro-micro model depicted in Figure 9.1 suggests a quite plausible answer. If government takes harsh measures in response to dissidents engaging in civil disobedience, killing or wounding them, jailing them, or otherwise treating them oppressively, the immediate result should be to drive up the Dissatisfaction with Political Treatment variable as represented by the dashed arrow denot­ ing a positive effect OfjViS1t on DPT. Also, Policy Dissatisfaction may be increased, especially if government initiates a concerted program intended to suppress dissidents systematically. As DPTand PD increase, PSA will increase, producing an increase in MJA. The increase in MJA may well be enough to drive APP into the Political Violence zone, assuming that A, FSM, and UJA do not decrease. Thus the short-term effect of repression is likely to be an escalation in the level of aggressive participa­ tion (at least for those dissidents not in prison). Over the longer run, however, if government effectively suppresses most of the original dissidents, they will find it difficult to expand their movement, and others sympathetic to the cause will be unlikely to take part in aggressive action. This is because of the lagged effect of Negative Sanctions, which should substantially reduce belief among others in the society that the actions taken by the original dissidents were helpful 9 Repression has a long-term direct deterrent effect on Internal War in that Negative Sanctions Dl (decade one) has a negative effect on Internal War D2 (decade two). The long-term deterrent effect of repression on Collective Protest is entirely indirect. Ihid., p. 181. lo See Charles Tilly, "Collective Violence in European Perspective," in Violence in America: Historical and Comparative Perspectives, ed. Graham and Gurr, p. 39.

274 I Macro-Micro Linkages

to their cause. The long-term effect of repression is represented by the dashed arrow denoting an inverse relationship between NS t - k and UJA. The second most important macro-structural predictor of Collective Protest was a dummy variable, Communist Regime, scored as 1 if a polity has a Communist form of government, 0 if a polity has a non-Communist form of government. The standardized causal impact of Communist Regime on Collec­ tive Protest was estimated to be — .378. How might this variable be linked to APP at the micro level? To the extent that Com­ munist regimes are more effective than non-Communist regimes in their application of specific acts of repression over the long run, CR would have an effect on UJA through the NS t _ k variable. But the direct impact of CR on APP, independent of NSt-k, is most likely to be transmitted through a negative effect on FSN. In the police-state atmosphere of totalitarian Communism, there will be few if any social settings in which facilitative norms can develop. Because of broad-scale restric­ tions on free speech and the right of peaceful assembly, criticism of the regime will not be able to flourish in any community or social group contexts. Moreover, the constant fear of being reported to the authorities by neighbors or colleagues at work or classmates at the university would further inhibit the expres­ sion of criticism of the regime. Thus, individuals with high Normative Justification for Aggression would be relatively isolated. And when FSN is low, then as shown in Figures 8.4, 8.5, and 8.6, civil disobedience is generally unlikely, regardless of A, NJA, and UJA, while political violence is not expected to occur under any circumstances. The third most important macro-structural predictor of Collective Protest was a Group Discrimination variable, defined as the percentage of the population "which is substantially and systematically excluded from valued economic, political, or social positions because of ethnic, religious, linguistic, or regional characteristics."11 Group Discrimination was esti11 The quotation, cited at page 74 of Hibbs, Mass Political Violence, is from Ted R. Gurr, New Error-Compensated. Measures for Comparing Nations: Some Correlates of Civil Violence, Center of International Studies, Princeton Uni­ versity, May, 1966. The Group Discrimination variable used by Hibbs comes from a data set developed by Gurr.

Macro-Micro Linkages / 275

mated to have a standardized causal impact on Collective Protest of .190. Obviously, Group Discrimination will increase Dissatisfaction with Political Treatment and Structural Just Deserts Frustration among those who are discriminated against. By virtue of the effect of JDFs on PD, Group Discrimination also will have an indirect positive effect on dissatisfaction with policy performance. Note that if the Expectancy-Value-Norms model is correct, it affords an explanation of why groups who are discriminated against so often do not react by taking ag­ gressive action. Group Discrimination will lead to high PSA through its direct effect on DPT and JDFs and its indirect effect on PD. But here the effect of GD stops. Unless IAA is high, high GD will not even lead to high NJA. And even if IAA is high, so that NJA becomes high, low A, low FSN, or low UJA, in certain combinations, could counterbalance high NJA. In short, without knowledge about these relevant variables, there is no reason to expect that grievances resulting from discrimination will lead to collective political aggression. The fourth most important macro-structural predictor of Collective Protest was a Communist Party Membership variable. Communist Party Membership showed a standard­ ized causal impact on Collective Protest of .169. Communist Party Membership is probably related to APP through three micro-level variables: IAA, PSA, and PD. At least in nonCommunist regimes, Communist Party Membership certainly will increase one's degree of leftist ideological commitment and one's alienation from the political system, and prob­ ably will lead one to be dissatisfied with policies of the govern­ ment, regardless of what the incumbents actually do. Of course the relationship probably is reciprocal, with leftist ideological commitment, political alienation, and policy dissatisfaction also providing motivational incentive to join the Communist Party in the first place. Therefore, the CPM variable is shown linked to IAA, PSA, and PD in Figure 9.1 by dashed doubleheaded arrows denoting reciprocal causation. Annual Percentage Change in Energy Consumption Per Capita was the least important variable in the equation for prediction of Collective Protest, carrying a standardized causal weight of —.126. Of the macro-structural variables, it is prob­ ably the most remote determinant of APP, as indicated by its

276 / Macro-Micro Linkages

location in the far lower left-hand corner of the diagram given in Figure 9.1. Annual Percentage Change in Energy Consump­ tion Per Capita is an indicator of economic development. In conventional sociological thought much has been made of the importance of economic development—in terms of both level of development and rate of economic growth—as a precondi­ tion of political stability.12 But Hibbs' cross-national analysis disputes the conventional wisdom. Economic development probably plays such a minor role in the explanation of mass political aggression because it is not systematically related to any of the micro-level determinants of Aggressive Political Participation—except perhaps for the possibility of an inverse effect on level of Rank Disequilibrium, itself a rather remote and not especially powerful cause of APP. Finally, a macro-structural variable not considered by Hibbs, but certainly a determinant of Policy Dissatisfaction, is the somewhat elusive and difficult to measure concept of govern­ mental effectiveness. This concept refers to such things as the quality of incumbent political authorities—how decisively they deal with problems and issues that arise, how flexible and in­ novative they are in formulating policies to resolve societal conflicts—and the resources available to governments in terms of money, skilled administrators, and the like. Here, no attempt will be made to develop a precise definition of the concept, and certainly some facets of it would be very difficult to operationalize, but the differential effectiveness of incumbent ad­ ministrations in managing the government and dealing with the issues of the day is a macro-structural variable that, because it is the primary cause of Policy Dissatisfaction, can have a potentially important effect on Aggressive Political Participa­ tion independent of the other macro variables taken into ac­ count in Figure 9.1. The parameters of the macro-micro relationships depicted by dashed arrows in Figure 9.1 are called "inestimable." Not that they are inherently inestimable: were it possible to carry out surveys in all or most polities of the world for a couple of 12 See the review of the literature in Hibbs, Mass Political Violence, pp. 21-24, 31-36.

Macro-Micro Linkages / 277

decades, the micro-level variables could be aggregated, sep­ arated by decade as per the procedure used by Hibbs with the macro-structural variables, and the parameters of the model could then be estimated. But it is unrealistic even to imagine that such data might ever exist—hence the characterization "inestimable." Nevertheless, it is important to realize that there are important macro-micro linkages. And speculation about the nature of such linkages is therefore a necessary feature of any explanation of aggressive political participation with pretensions of being a complete theory. 9.1

CODA: ON THE EFFECTIVENESS OF REPRESSION

One of the more hotly debated issues in the study of mass political aggression is the question of the effectiveness of repres­ sion. Does repression necessarily sow the seeds of its own destruction? Or, if efficiently applied, can repression alone succeed in suppressing potential rebellion? Gurr, for example, has argued that "coercion alone is demonstrably ineffective, in the long run if not the short, because on balance it is more likely to inspire resistance than compliance."13 Russell, by contrast, counters Gurr's view by noting: "If the thesis that oppression sows the seed of its own destruction is correct, then a rebellion should have occurred in South Africa."14 As has been observed in the preceding section, a repressive strategy by government entailing severe application of negative sanctions actually can be expected, in the short run, to exacer­ bate rather than deter aggressive political behavior. Only if applied concertedly over a relatively long period of time can a repressive strategy be expected to have a significant deterrent effect. But there are other variables than can counterbalance even this long-term deterrent effect. Suppose that long-term application of negative sanctions has resulted in low Utilitarian Justification for Aggression. Given the presence of facilitative social norms, recall from 13 14

Gurr, Why Men Rebel, p. 358.

Russell, Rebellion, Revolution, and Armed Force (New York: Academic Press, 1974), p. 38.

278 / Macro-Micro Linkages

Figure 8.1 that political violence still is projected for people with high Availability and extremely high Normative Justifi­ cation for Aggression, even if Utilitarian Justification for Ag­ gression is in the low range (slope G).15 As soon as Availability moves out of the high range, political violence is not expected to occur if Utilitarian Justification for Aggression is in the low range and NormativeJustification for Aggression remains very high (see Figures 8.2 and 8.3). However, aggressive political behavior in the form of civil disobedience is predicted to occur at all levels of Availability, even with Utilitarian Justification for Aggression in the low range, depending on whether Norma­ tive Justification for Aggression is medium-to-high. Thus the repressive strategy can be expected to deter in the sense of sharply reducing the likelihood of political violence in the long run, but it will not prevent political violence, and it will not deter civil disobedience. Eliminating the possibility of political violence and sharply reducing the likelihood of civil disobedience takes something more than the application of specific acts of negative sanction. Totalitarian repression is necessary. Medium-to-high Norma­ tive Justification for Aggression must be kept isolated, so that social norms are inhibitory, and the projections of the graphs in Figures 8.4 to 8.6 apply across the board to all citizens. Insti­ tutions and social settings in a country where social norms could become facilitative—settings such as universities, labor unions, urban ghettoes—would have to be strictly controlled and indi­ vidual autonomy in general would have to be severely restricted. Thus, the repressive strategy could be fully successful—but only at the price of a totalitarian police state. An alternative strategy for controlling aggressive political participation entails a shift in focus from Utilitarian Justifica­ tion for Aggression to Normative Justification for Aggression. This is a responsive strategy. Ethically, it is the desirable alternative. Practically, it is the preferred alternative if govern2 5 I focus on slope C instead of slope D because the latter applies to people with a score of O on UJAi and it is unlikely that people with high NJA would score O on UJA since this score can be received only by people who feel that political influence is unnecessary.

Macro-Micro Linkages / 279

ment does not wish to opt for the extreme totalitarian repres­ sion strategy. The responsive (or co-optive, depending on one's perspec­ tive) strategy is simple: keep NormativeJustification for Ag­ gression down by seeing to it that the antecedents of Political Support-Alienation remain low. If the model depicted in Figure 9.1 is correct, Political Support-Alienation can be controlled through manipulation of Dissatisfaction with Political Treat­ ment, StructuralJust Deserts Frustration, and Policy Dissatis­ faction. If incumbent administrations attempt to reduce or prevent group discrimination and the application of severe negative sanctions for political dissent, Dissatisfaction with Political Treatment and Structural Just Deserts Frustration should be minimized. If the incumbents attempt to be respon­ sive to the demands and concerns of the non-elite, Policy Dissatisfaction should be minimized. If Political SupportAlienation is thus held at a low level, Normative Justification for Aggression cannot rise very high, the Facilitative Social Norms variable generally will take on a low value, and par­ ticipation in political action will be unlikely to occur. Of course, controlling Political Support-Alienation by seeing to it that Dissatisfaction with Political Treatment, Structural Just Deserts Frustration, and Policy Dissatisfaction do not take on high values is easier said than done. And it is Utopian even to imagine that any incumbent administration could be of such olympian virtue as to be able to keep the performance-dissatisfaction variables at consistently low levels for all groups in the society. But government has the initiative.

APPENDIX A

Comparison of Political Trust-Distrust and Political Support-Alienation as Predictors of Aggressive Political Participation

A debate has arisen about whether trust in government, as measured by the Survey Research Center of the University of Michigan, is a measure especially sensitive to affect for an incumbent administration or to affect for the system of govern­ ment. 1 This issue is important because, on theoretical grounds, there is little reason to expect that low trust in incumbent politicians will provide a strong source of normative justifica­ tion for participation in collective political aggression. In the first place, at least in polyarchies, citizens have legally-guaran­ teed rights peacefully to contest incumbents whom they dis­ like. Second, as Jack Citrin has pointed out, at certain times the current Zeitgeist in a country may make it fashionable to denigrate incumbent politicians, and expressions of cynicism about them may to some extent simply represent a culturallyexpected response whose consequences are purely symbolic in­ stead of behavioral.2 Many researchers have interpreted the trust in government 1 See Arthur H. Miller, "Political Issues and Trust in Government: 1964— 1970," American Political Science Review, 68 (September 1974), 951-972; Jack Citrin, "Comment: The Political Relevance of Trust in Government," and Arthur H. Miller, "Rejoinder to Comment by Jack Citrin: Political Discontent or Ritualism?." both in the same issue of the Review at pages 973-988 and 989-1001, respectively. 2 Citrin, ibid., p. 975.

Appendix A / 281

items as if they were a measure that captured an aspect of diffuse support for the structure of political authority.3 The issue of whether they are or are not especially sensitive to system affect cannot be resolved by conventional dimensional analysis. For even if trust in government were primarily sensitive to affect for an incumbent administration, it would certainly be sur­ prising if such sentiment did not correlate fairly strongly with affect for the system of government, since it is logical to expect that trust in an incumbent administration might have a causal effect on support for the system and/or support for the system might condition a person's trust in the incumbents. The only way to resolve the issue of the empirical meaning of trust in government satisfactorily is to construct separate trust and support measures and then compare how they relate to hypothetical causes and effects of system affect. If the rela­ tionships are different, then there is empirical justification for not combining trust and support items together in a single index, despite high intercorrelation. The results of an extensive analysis of how trust and support relate to other variables indicated that the trust in government items are far more sensitive to affect for an incumbent adminis­ tration than to affect for the system of government, even though the two measures are themselves rather closely correlated.4 Here the two variables will simply be compared as predictors of aggressive political participation. The measure of trust in government is a summated index built from responses to the following items:

T1: In general, one can rely on the federal government to do the right thing. 3 See, for example, Joel D. Aberbach and Jack L. Walker, "Political Trust and Racial Ideology," American Political Science Review, 64 (December 1970), 1199-1219; Arthur H. Miller, "Political Issues and Trust" and "Rejoinder to Comment by Jack Citrin;" Muller, "Correlates and Consequences of Beliefs in the Legitimacy of Regime Structures," Midwest Journal of Political Science, 14 (August 1970), 392-412. 4 See Edward N. Muller and Thomas O. Jukam, "On the Meaning of Political Support," American Political Science Review, 71 (December 1977), 15611595.

282 I Appendix A

T2: How much do you trust the government in Bonn to act as it really should ? T3: When members of parliament or cabinet ministers speak on television or in parliament with journalists, how often in your opinion do they tell the truth? T4: How much do you trust members of the government to put the interests of the people over the interests of their own parties ? Two of the items, T1 and T2, are variants of the Survey Re­ search Center item which asks respondents directly about their evaluation of the trustworthiness of the national government.3 Items T3 and Ti, analogous to Survey Research Center items, were designed to capture evaluation of the integrity of elected representatives and their concern for public interests in general.6 The 5-point items used in the index of political trust were assigned scores from 0 to 4, coded so as to range from positive to negative evaluation.7 When scores are summed across the four items the resulting Political Trust-Distrust (PTD) variable ranges from 0 at the most positive or trusting position to 16 at the most negative or distrusting position. 5 This item is: "How much of the time do you think you can trust the govern­ ment in Washington to do what is right?" 6 These items were developed by Alan Marsh for use in the cross-national study of value change and political behavior (see footnote 8 of Chapter One). ' To attain comparability with the distribution of the other three trust items, the response scores on either side of the neutral value for T1 were collapsed (+1 and — 1 with 0), providing for five response categories in all. Response options for items T v T v and Tt were: "about always," "mostly," "sometimes," "only very seldom," and "never," scored from 0 to 4, respectively (T1 was reflected and scored from 0 to 4). The Reliability Coefficient (KR-20) is .770, yielding an estimated correlation of test scores with true scores of .877. The Reliability Coefficient is slightly below the level of .8, considered desirable to reach because at that level correlations are not appreciably attenuated by measurement error. The inter-item correlation matrix is: T3 T3 7I 3*4

1.00

.45 1.00

.29 .55 1.00

.34 .54 .56 1.00

Appendix A / 283

To what extent does aggressive political participation in­ crease as alienation from the structure of political authority increases? Using OLS regression to estimate standardized weights, here are the results, first with Political SupportAlienation as the independent variable, then with Political Trust-Distrust as the independent variable, and finally with both variables in the prediction equation: (A.l)

APP ln = .501 (PSA), where r 2 = .251 and JV= 2,003.

(A.2)

APP la = .282(PTD), where r2 = .080 and JV= 2,129.

(A.3)

APP la = .538(PSA) - .064(PTD), whereR 2 = .253 and JV = 2,003.

The increase in Aggressive Political Participation estimated to result from a standard unit increase in political alienation is almost twice as large as that estimated to result from a standard unit increase in political distrust. The degree of fit to a linear function rule is more than three times as great when Political Support-Alienation is the indicator of system affect as com­ pared with Political Trust-Distrust. When both variables are in the prediction equation, PTD is estimated to have virtually zero impact on aggressive participation. (The sign of PTD is actually reversed; but one would not want to attach substantive significance to this, since the small weight estimated for PTD is indicative of the absence of any linear relationship, regardless of sign.) Since PSA and PTD are fairly strongly correlated (r = .59), and since the original or gross effect of PTD on APPin is reduced to zero net effect with PSA in the equation, one may interpret the original relationship between P TD and APPla as being due not to any causal influence from PTD but rather to the fact that PTD is associated with PSA. A plot of the relationship between PTD and APP la is given in Figure A.l. The weakness of PTD as a predictor of APPla is evident from the relative flatness of the slope and the many sizable deviations from it. All persons who score in the High Civil Disobedience and Political Violence zones of APP1n devi­ ate markedly from the slope. Even at very high levels of distrust

111 AGGRESSIVE POLITICAL PARTICIPATION (APP, J

Statistics

N = 2129 r- -= .080 APP,n = 1.425 + .07MPTD)

Summary

(PTD)

Large circles denote q or more cases.

POLITICAL TRUST-DISTRUST

FIGURE A.L. Aggressive Political Participation as a Linear Function of Political Trust-Distrust

Appendix A / 285

(scores of 12 to 16 on PTD), there are two large clusters of per­ sons who score at the very lowest position on APP l a , These results suggest that, as an indicator of belief in the normative justifiability of aggressive action stemming from lack of support for the political system, the Political SupportAlienation measure is superior to the Political Trust-Distrust measure. To be sure, on the basis of conventional dimensional analysis criteria, the correlation between the support variable and the trust variable is sufficiently high to warrant combining them into a single index. Yet, in the interest of parsimony, there is no point to this, since equation A.3 indicates that political trust makes no independent contribution to prediction of ag­ gressive political participation.

APPENDIX

B

Summary Characteristics of the Variables

B.l.

SUMMARY STATISTICS

Variable & Chapter Valid Introduced Cases APP - 3 APPln - 3 PIC- 4 ECA - 4 UJA - 4 PSA-4 PSA2-4 LIC-4 NJA-4: URB- 4 UNV - 4 -4 MS - 4 £>S-4 ^4-4 JDF - 5 J^S -5 RGG-5 AGp- 5 -AG,-5 |4G P -5 ~AGp-5

2,198 2,198 2.354 2,370 2,247 2.355 2,355 2,317 2,147 2,662 2,662 2,662 2,662 2,410 2,627 2,393 2,487 2,631 2,384 2,499 2,468 2,468 2,074

Z,ow to High

Mean

Standard Deviation

4.66 - 91.43 1.54- 4.52 0- 2 1 -7 0 - 14 0 - 48 0 - 23.04 1 - 10 0 - 230.4 0- 1 0- 1 18- 87 1.1 - 5.6 0- 1 0- 2 1.1 - 8.6 0 - 15 0 - 30 0 - 10 0 - 10 - 1 0 - 10 0 - 10 - 1 0 - 10

10.0 2.01 1.21 1.77 2.28 24.85 7.06 6.09 47.25 .37 .41 42.01 2.87 .38 .68 4.08 1.86 2.99 6.43 6.34 .61 1.54 .35

10.0 .68 .77 1.27 2.57 9.39 5.04 2.07 45.88 .48 .49 18.04 1.20 .48 .95 2.35 2.45 4.43 1.81 1.89 2.09 1.54 1.72

Appendix B / 287

Variable & Chapter Introduced

Valid Cases

Low to High

|AGf - AGp| - 5 E-5

2,074 2,568 2,548 2,464 2,662 2,188 2,188 2,093 2,436 2,662 2,623 2,534 2,054 2,054 2,607 1,109 1,056 1,089 1,142 1,045 1,105 1,310 1,310 1,180 1,289 1,038 1,027

0 - 10 1 - 7 1 - 7 .14-6 0 - 10 -2.49-2.62 -2.77-2.54 0-18 1 - 9 0-7 0-7 0-42 1.46-4.18 - 1 . 6 5 - 1.85 0 - 10 1.54-3.89 1.6-4.5 0 - 14 0 - 14 0 - 230.4 0 - 230.4 0-1 0- 1 1.19-8.56 1.16-8 1.46-3.83 1.46-3.85

5-5

RDE- 5 Income - 5 PDm- 6 PDpm - 6 DPT- 6 RI-6 01-6 0Ip- 6 PPR- 6 API\nPRED - 6 APPlaRESID - 6 CPB- 6 APPln -7 74

APPlai6

-7

UJA16-7 NJA^-1 •VM16-7 UNV74 - 7 UNV76-7

evm^-1 EVM16-7 B.2.

Mean 1.16 3.42 4.23 .85 2.07 .01 -.02 3.29 3.45 .91 .56 2.77 2.04 0 2.07 1.83 2.02 1.90 1.74 35.61 29.39 .24 .24 3.27 3.06 1.86 1.81

Standard Deviation 1.31 2.15 1.90 .54 2.65 .76 .76 4.02 2.24 1.10 1.07 6.02 .51 .47 2.65 .53 .67 2.20 1.88 35.35 31.71 .43 .43 1.96 1.72 .38 .36

TREATMENT OF MISSING DATA

For the dependent variable of this study, Aggressive Political Participation, respondents were required to have valid scores on all component items. For the independent variables constructed from more than one item, respondents were required to have valid scores on at least more than half of the component

2 8 8 I Appendix B

items. The mean was used as the replacement value for missing data. Special Remarks: 1. In the case of the Political Support-Alienation variable, respondents were required to have valid scores on at least three of the five items with a "system" referent (S p ^2, S3, S i , S5) and on at least two of the three items with an "author­ i t i e s " r e f e r e n t ( S6 , S 7 , S 8 ) . 2. In the case of the Political Trust-Distrust variable, respon­ dents were required to have valid scores on T v T 3 , and T 4 (very few respondents were missing scores on these variables). 3. In the case of the three Political Treatment variables (com­ ponents of Dissatisfaction with Political Treatment), re­ spondents with a missing score on any Perceived Treatment of Others variable but a valid score on the corresponding Personally Experienced Treatment variable were assigned a Political Treatment (PT = PTO*PET) score assuming the middle value (Partly Good/Partly Poor) on Perceived Treatment of Others.

APPENDIX C

German Text of the Variables

C.L.

KEY TO GERMAN TEST OF QUESTIONS USED IN CONSTRUCTING THE VARIABLES

—Aggressive Political Participation: APP constructed from responses to items D, E, F, H, and I as presented in questions 26, 27, and 28. —Political Influence Capability: PIC constructed from questions 9 and 17. —EfRcacy of Collective Aggression: ECA constructed from question 24. —Political Support-Alienation: PSA constructed from responses to items A through H of question 10. —Leftist Ideological Commitment: LIC defined by question 8. —Age in Years: AGE defined by question 31. —Marital Status: MS defined by question 30. —Employment Status: ES defined by question 32. —Structural Just Deserts Frustration: JDFs constructed from questions 2, 3, 4, 5, and 6. —Reference Group Gratification: RGG defined by question 7.

290 / Appendix C

—Aspirational Gratification, experienced, present, future: defined by question 1. —Education: E defined by question 34. —Perceived Social Standing: S defined by question 33. —Income: / defined by question 36. —Materialist Policy Dissatisfaction: constructed from responses to items E, F, H, J, K, and L of question 11 as presented in questions 11, 12, and 13. -Post-Materialist Policy Dissatisfaction: constructed from responses to items A, B, C, D, and G of question 11 as presented in questions 11, 12, and 13. —Dissatisfaction with Political Treatment: DPT constructed from questions 15 and 16. —Rejection of Individualism: RI constructed from questions 18 and 19. —Personal and Political Resources: PPR constructed from questions 23, 34, and 35. —Conventional Political Behavior: CPB constructed from question 14. 1. Wir interessieren uns dafiir, wie die Leute heutzutage iiber ihr Leben denken. Jeder hat eine Meinung iiber das, was ihm eigentlich im Leben zustehen sollte. Skala 1 vorlegen! Hier ist eine Art Leiter. Die oberste Sprosse, die Sprosse mit der Ziffer 10, soil den Leuten wie Ihnen eigentlich zusteht. Die unterste Sprosse, die Sprosse mit der Ziffer 0, soil den allerschlechtesten Zustand kennzeichnen, den Sie haben konnen.

Appendix C / 291

Was meinen Sie, wenn Sie an alles denken, was Ihnen uberhaupt im Leben wichtig ist: A. Alles in allem, wo stehen Sie heute? B. Und wie sah das vor 5 Jahren aus? C. Und was glauben Sie, wie wird es wohl in 5 Jahren sein? 2. Bitte denken Siejetzt einmal an Ihre Wohnverhaltnisse. A. Sind Ihre Wohnverhaltnisse so gut wie es Ihnen eigentlich zusteht? B. Sind Ihre Wohnverhaltnisse sehr viel schlechter, viel schlechter oder

nur etwas schlechter als es Ihnen eigentlich zusteht? G. Und inwieweit ist Ihrer Meinung nach der Staat fur Ihre Wohnverhaltnisse verantwortlich ? Ganz und gar Uberwiegend Wenig Uberhaupt nicht 3. Und wie ist das mit der arztlichen und Krankenhaus-Versorgung die Sie in Anspruch nehmen konnen? A. 1st sie so gut wie es Ihnen eigentlich zusteht ? B. 1st sie sehr viel schlechter, viel schlechter oder

nur etwas schlechter als es Ihnen eigentlich zusteht? C. Und inwieweit ist Ihrer Meinung nach der Staat fur Ihre arztliche und Krankenhaus-Versorgung verantwortlich? Ganz und gar Uberwiegend

292 / Appendix C

Wenig Uberhaupt nicht Α. 1st Ihr Einkommen so hoch wie es Ihnen Ihrer Ansicht nach eigentlich zusteht? B. 1st es sehr viel schlechter,

viel schlechter oder

nur etwas schlechter als es Ihnen eigentlich zusteht? C. Und inwieweit ist Ihrer Meinung nach der Staat fiir Ihr Einkommen verantwortlich? Ganz und gar

Uberwiegend Wenig Uberhaupt nicht Und wie ist das mit Ihrer taglichen Arbeit: A. Verschafft das, was Sie in Ihrer taglichen Arbeit tun, Ihnen so viel Befriedigung wie es Ihnen eigentlich zusteht? B. 1st es

sehr viel schlechter,

viel schlechter oder

nur etwas schlechter als es Ihnen eigentlich zusteht? C. Und inwieweit ist Ihrer Meinung nach der Staat fur die Befriedigung in Ihrer taglichen Arbeit verant­ wortlich? Ganz und gar

Uberwiegend Wenig Uberhaupt nicht Und wenn wir jetzt einmal von einzelnen Dingen absehen. A. Ganz allgemein gesehen:

Appendix C / 293

Geht es Ihnen so gut, wie Sie es eigentlich verdienen wurden? B. Geht es Ihnen sehr viel schlechter, viel schlechter oder

nur etwas schlechter, als Sie es eigentlich verdienen wurden? C. Und inwieweit ist Ihrer Meinung nach der Staat fur all das verantwortlich, was Ihnen im Leben wichtig ist? Ganz und gar Uberwiegend Wenig Uberhaupt nicht 7.

Skala 1 erneut vorlegen! Hier haben wir wieder die Leiter. Diesmal stellen Sie sich bitte vor, daB alle Leute^ die Sie naher kennen auf die verschiedenen Sprossen dieser Leiter verteilt sind. Die Bekannten, denen es am besten geht, stehen auf Sprosse 10: die Bekannten, denen es am schlechtesten geht, stehen auf Sprosse 0. Was meinen Sie, wo stehen Sie?

8. Viele Leute verwenden die BegrifTe "Links" und "Rechts," wenn es darum geht, unterschiedliche politische Einstellungen zu kennzeichnen. Wir haben hier einen MaBstab, der von links nach rechts verlauft. Wenn Sie an Ihre eigenen politischen Ansichten denken, wo Wiirden Sie diese Ansichten auf dieser Skala einstufen? Kugelschreiber uberreichen! Skals 2 vorlegen und vom Befragten selbst ausfiillen lassen! Listen- und Lfd. -Nr. ubertragen! Es ist darauf zu achten, daB der Befragte ein Kreuz innerhalb eines Kastchens macht!

294 I Appendix C

Wie wiinschenswert ist es Ihrer Meinung nach, daB Leute wie Sie sich an der Politik beteiligen: 1st es sehr wiinschenswert, wiinschenswert, nicht sehr wiinschenswert oder

ganz und gar nicht wiinschenswert? Gelben Kartensatz mischen und vorlegen! Hier ist eine Reihe von Ansichten. Bitte sagen Sie mir anhand dieser Skala, wie stark Sie den einzelnen Ansichten zustimmen oder sie ablehnen. Skala 3 vorlegen! + 3 bedeutet: "Sie stimmen voll zu"; — 3 bedeutet: "Sie lehnen entschieden ab." Die Werte dazwischen dienen zur Abstufung Ihres Urteils. A. Es macht mir Sorge, wenn isch an den Unterschied zwischen dem denke, was Leute, wie ich, im Leben wollen, und was tatsachlich in unserem politischen System geschicht. B. Die politischen Einrichtungen der Bundesrepublik sind mir Iieb und wert und ich achte sie hoch. C. Meine Freunde und ich fiihlen uns in unserem politischen System eigentlich doch sehr gut vertreten. D. Ich bin immer wieder erschrocken und betroffen dariiber, daB die wesentlichen Rechte der Biirger in der deutschen Politik so wenig beachtet werden. E. Heutzutage bin ich gegeniiber unserem politischen System sehr kritisch cingestellt. F. Die Grundeinstellung der Leut die bisher in der Bundesrepublik politisch tonangeber waren, war immer in Ordnung. G. Die Gerichte in der Bundesrepublik gewahren jedermann einen fairen ProzeB—es spie dabei keine Rolle, ob er arm oder reich, gebildet oder ungebildet ist.

Appendix C / 295

Η. Alles in allem genommen, verdient die Polizei in der Bundesrepublik groBen Respekt. I. Man kann sich im allgemeinen daraufverlassen, dafi die Bundesregierung das Richtige tut. 11. Auf diesen Kartchen stehen verschiedene politsche Ziele. Bitte schauen Sie sich diese Kartchen einmal an. Sagen Sie mir bitte fur jedes Ziel, ob es die Aufgabe einer Bundesregierung ist oder nicht. A. Gerechtigkeit fur alle gewahrleisten B. Bessere Vorsorge fiir Alte, Schwache und Bediirftige treffen C. Wirksame Bekampfung der Umweltverschmutzung D. Mehr Moglichkeiten fiir alle Biirger schaffen, an politischen Entscheidungen mitzuwirken E. Schutz und Sicherheit fiir den einzelnen gewahrlei­ sten F. Fiir wirtschaftliche Stabilitat sorgen G. Durchsetzung demokratischer Grundsatze in Wirtschaft, Verbanden und Universitaten H. Sicherstellung der nationalen Verteidigung J. Fiir eine starke und handlungsfahige politische Fiihrung sorgen K. Sicherung der freien Marktwirtschaft L. Fiir Ruhe und Ordnung im Lande sorgen 12. Nun mochte ich noch gerne wissen, fur wie wichtig Sie die Ziele halten, die Ihrer Ansieht nach Aufgaben einer Regierung sind. 1st das Ziel . . . sehr wichtig, wichtig, nicht sehr wichtig oder absolut unwichtig? 13. Nun geben Sie der jetzigen Regierung bitte fiir jedes dieser Ziele eine Note zwischen 1 und 5, je nachdem, wie die Regierung sich Ihrer Ansicht nach dabei bewahrt oder wie sehr sie sich dafur eingesetzt hat. Die Noten sind wie in der Schule: 1 ware "sehr gut", das hieBe, "die Regierung hat sich ausgezeichnet bewahrt";

296 / Appendix C

5 ware "mangelhaft", das hieBe, "die Regierung hat sich iiberhaupt nicht bewahrt." Wie ist es mit Ziel . . . ? Welche Note hat die Regierung da verdient? 14. Und nun zu dem, was die Leute tan, um einer Partei oder einem Kandidaten zum Wahlsieg zu verhelfen. Bitte sagen Sie mit, ob Sie innerhalb der letzieg. Jahre eine der folgenden Tatigkeiten ein- oder zweismal unternommen haben oder ofter. INT: Vorgaben bitte vorlesen!

Haben Sie mit Leuten gesprochen und sie zu iiberzeugen versucht, welche Partei oder welche Kandidaten sie wahlen sollten? Haben Sie eine Plakette getragen oder einen Aufkleber an Ihr Auto geldebt? Haben Sie irgendeine politsche Diskussion oder eine Kundgebung besucht ? Haben Sie Geld gespendet, um cinem Kandidaten oder einer Partei zu helfen? Haben Sie irgendeine Arbeit verrichtet, um einen Kandidaten bei zeinem Wahlkampf zu unterstutzen ? 15. Hatten Sie in irgendeinem Zusammenhangjemals mit der Polizei, mit einem Gericht oder mit anderen Behorden zu tun? Furjede It. Frage 39 angekreuzte Stelle fragen: A., B., C. Sagen Sie mir doch bitte, ob man Sie bei (vor) . . . im allgemeinen eher gut, teils gut/teils schlecht oder

eher schlecht behandelt hat. 16. Und wie behandelt man Ihrer Meinung nach Leute wie Sie im allgemeinen bei der Polizei? A. Eher gut, teils gut/teils schlecht oder eher schlecht?

Appendix C / 297

Β. Und wie ist das vor Gericht? C. UndbeidenBehorden? 17. Sind Sie der Ansicht, daB Sie einen groBen Einflufj, einen nennenswerten EinfluB, kaum EinfluB

oder keinen EinfluB darauf haben, wie die Bundesrepublik regiert wird ? 18. Wenn es um die grundlegenden Dinge geht, die die meisten Leute vom Leben erwarten, dann scheint es so, daB einigen Leuten alles sehr leicht fallt, wahrend andere nichts erreichen, wie sehr sie sich auch bemiihen. Liegt das Ihrer Ansicht nach irgendwie an den Leuten selbst oder an etwas anderem? 19. Einige Leute glauben, daB sie weitgehend von anderen abhangig sind. Andere Leute wieder glauben, daB sie ihr Leben genauso gestalten, wie sie es haben wollen. Wie ist das mit Ihnen: Sind Sie weitgehend von anderen abhangig

oder leben Sie genauso wie Sie es haben wollen? 20. Wie sehr vertrausen Sie darauf, daB die Regierung in Bonn so handelt. wie sie es eigentlich sollte? Eigentlich immer, meistens, manchmal, nur sehr selten

oder niemals? 21. Wenn Bundestagsabgeordnete oder Minister im Fernsehen oder im Parlament mit Journalisten sprechen, wie oft sagen sie Ihrer Meinung nach die Wahrheit ?

298 / Appendix C

Eigentlich immer, meistens, manchmal, nur sehr selten oder

niemals ? 22. Wie sehr vertrauen Sie darauf, daB Regierungsmitglieder das Wohl des Volkes iiber die interessen ihrer eigenen politischen Partei stellen? Eigentlich immer, meistens, manchmal, nur sehr selten oder

niemals ? 23. Wie stark interessieren Sie sich fur offentliche Angelegenheiten und Politik ? Sind Sie sehr interessiert, etwas interessiert, nicht sehr interessiert oder iiberhaupt nicht interessiert? 24. Die nachste Frage bezieht sich auf einige Aktivitaten in der Bundesrepublik. Bitte sagen Sie mir dazu jeweils, ob diese Gruppen Ihrer Meinung nach mit ihrem Verhalten ihrer Sache im allgemeinen geniitzt haben, im allgemeinen geschadet, oder ob dieses Verhalten den Zielen dieser Gruppen weder geniitzt noch geschadet hat. A. Um Veranderungen an den Universitaten herbeizufuhren, haben Studenten Lehrveranstaltungen gestort und Raume und Mobiliar beschadigt. Was glauben Sie, hat dies den Gruppenzielen im allge­ meinen geniitzt, im allgemeinen geschadet oder weder noch? B. Einige Gruppen haben offentliche Gebaude gesturmt und sich mit der Polizei geschlagen, um gegen Tatigkeiten der Regierung zu protestieren.

Appendix C / 299

Was glauben Sie, hat dies den Gruppenzielen im allgemeinen genutzt, im allgemeinen geschadet oder weder noch? C. Einige Geheimorganisationen—etwa die von BaaderMeinhof—haben zur Guerilla-Taktik gegriffen, um gegen Staat und Gesellschaft in der Bundesrepublik zu protestieren. Was glauben Sie, hat dies den Gruppenzielen im allgemeinen geniitzt, im allgemeinen geschadet odor weder noch ? 25. Hier ist ein Kartensatz mit verschiedenen Moglichkeiten, die Regierung zu beeinflussen, Sagen Sie mir bitte zu jeder Moglichkeit, ob Sie sie billigen oder ablehnen, oder ob dies von bestimmten Dingen abhangt. 26. Wir haben diese Verhaltensweisen auch nochmal auf dieser Liste hier aufgefuhrt. A. Beteiligung an einer Unterschriftensammlung B. Zeit opfern, um ftir eine politische Partei oder einen Kandidaten im Wahlkampf zu arbeiten C. Wehrdienst verweigern D. Besetzung von Fabriken, Amtern und anderen Gebauden E. Sich weigern, Mieten, Raten oder Steuern zu bezahlen F. Beteiligung an Schlagereien (Kampf mit Polizisten, Kampf mit anderen Demonstranten) G. Teilnahme an einer genehmigten politischen Dem­ onstration H. Beteiligung an einer Gruppe, die die Regierung mit Gewalt stiirzen will I. Beteiligung an einem wild en Streik J. Versuchen, Freunde fur die eigenen politischen Ansichten zu gewinnen Stellen Sie sich nun bitte einmal vor, wir wiirden die Bundesbiirger uber diese Verhaltensweisen befragen:

300 / Appendix C

Bitte sagen Sie fiiir jede dieser Verhaltensweisen, wieviel Prozent aller Bundesburger jede dieser Verhaltensweisen Ihrer Ansicht nach wohl billigen werden. 27. Nehmen Sie bitte wieder den Kartensatz. Sagen Sie mir bitte zu jedem Kartchen, ob Sie grundsatzlich bereit waren, dies zu tun oder nicht, oder ob Sie dies ganz davon abhangig machen wurden, worum es dabei im einzelnen geht. 28. Was von dem auf diesen Kartchen haben Sie schon innerhalb der letzten zwei Jahre zu tun versucht? Was ist dabei herausgekommen ? Bitte legen Sie die Kartchen in das jeweils zutreffende Feld auf diesem Blatt. 29. Geschlecht der Zielperson: Mannlich Weiblich 30. Familienstand der Zielperson: Verheiratet Ledig Geschieden/getrennt lebend Verwitwet 31. Alter der Zielperson: Jahre 32. Sind Sie berufstatig? Voll berufstatig (einschlieBlich mithelfende Familienangehorige) Teilweise berufstatig Voriibergehend arbeitslos Nicht berufstatige Rentner, Pensionare, im Ruhestand In Berufsausbildung (einschlieBlich Fachschulen fiir gewerbliche Betriebe) In Schulausbildung (einschlieBlich Universitat, Akademien, Hochschulen) Nicht berufstatig, z.B. Hausfrauen ohne Berugsausiibung

Appendix C / 301

33. Es wird heute viel iiber die verschiedenen Bevolkerungsschichten gesprochen. A. Welcher Schicht rechnen Sie selber sich zu, der Arbeiter-schicht, der Mittelschicht, der oberen Mittelschicht oder der Oberschicht? B. Rechnen Sie sich eher zum Durchschnitt oder zum oberen Teil der . . . oder zum unteren Teil der . . . ? 34. Welche Schule haben Sie zuletzt besucht, ich meine, welchen SchulabschluB haben Sie? Volksschule (und Berufsschule) ohne abgeschlossene Lehre oder Berufsausbildung Volksschule mit abgeschlossener Lehre oder Berufsausbildung Mittelschule ohne AbschluB/mehrjahrige Fachschule ohne AbschluB/Handelsschule ohne AbschluB Hohere Schule bis Obertertia Mittlere Reife Hohere Schule ohne Abitur (langer als Untersekunda) Hohere Fachschule mit AbschluB Abitur Hochschule/Universitat ohne AbschluB Hochschule/Universitat mit AbschluB 35. Sind Sie selbst Mitgliea in einem Verein oder gehoren Sie zu einem Berufsverband, einer gewerkschaftlichen, weltanschaulichen, kirchlichen oder politischen Organisation, dem Elternbeirat einer Schule oder irgendeiner sonstigen Vereinigung, die es hier in . . . (Ort/bei GroBstadten Stadtviertel und/oder Stadt vorlesen) gibt? Ich bin Mitglied— A. eines Vereins, Klubs, einer gesellschaftlichen Vereinigung, Organisation B. einer kirchlichen, weltanschaulichen Vereinigung C. des Elternbeirates einer Schule D. eines Berufsverbandes

302 I Appendix C

Ε. einer studentischen Vereinigung F. einer gewerkschaftlichen Organisation G. einer sonstigen Vereinigung 36. Wenn Sie einmal alles zusammenrechnen: Wie hoch ist dann etwa das monatliche Netto-Haushaltseinkommen, das Sie (alle zusammen) haben, nach Abzug der Steuern und der Sozialversicherung ? Bitte nennen Sie mir anhand dieser Liste den Buchstaben, der auf Sie zutrifft. A. B. G. D. E. F. G. H. J. K.

unter 400 DM 400 bis unter 600 DM 600 bis unter 800 DM 800 bis unter 1,000 DM 1,000 bis unter 1,200 DM 1,200 bis unter 1,500 DM 1,500 bis unter 2,000 DM 2,000 bis unter 2,500 DM 2,500 bis unter 3,500 DM 3,500 DM und mehr

INDEX

Aberbach, Joel, 281 aggressive political participation: de­ pendent variable, 3-8; definition, 37; "collective protest" and "in­ ternal war," 46-48; dimensionality of, 42-47; zones of, 54-65; alterna­ tives to, 231-32 Ajzen, I., 26, 27, 95 alienation-powerlessness hypothesis, 209 Amsel, A., 128 Atkinson j John W., 21 auxiliary theory, 69-70 Azrin, N. H., 128 Bandura, Albert, 7, 71, 72, 73, 123, 125, 126, 127 Barnes, Samuel, 47 Beattie, Muriel, 210, 213 Bendix, Reinhard, 135 Berkowitz, Leonard, 121, 122, 125, 127, 128, 145 von Beyme, Klaus, 132 Blalock, Hubert M. and Ann B., 69 blocked opportunity hypothesis, 209 Blumenthai, Monica, 156 Bowen, Don R. and Elinor, 153 Boynton, G. Robert, 188 Brinton, Crane, 179, 244 Burnstein, E., 131 Calder, BobbyJ., 25 Cantril, Hadley, 130, 133, 134 Caplan, Nathan S., 209 causes of aggressive political partic­ ipation: gap between expectation and experience, 11-14; "relative" hardship, 15-17; change in group control of power resources, 18-20; utilitarian incentive, 20-23; atti­ tude about the action, personal normative beliefs, and social norms, 23-31 Chadiha, Letha B., 156 Citrin, Jack, 186, 219, 220, 225, 226, 280, 281

Cohen, A. R., 131 Cole, Gerard A., 156 collective political participation, 4-5 control of power resources, 18-20, 272-74 Costner 1 Herbert L., 236 Crawford, ThomasJ., 208, 209 Cross, David, 262 cross-validation, 236-38 Dahl, Robert, 5, 6, 133, 134 Davies 1 James C., 13, 14, 124, 135 de Jouvenel, Bertrand, 5 Dennis j Jack 1 9, 81 de Tocqueville, Alexis, 135, 153 Devinney, L. C., 129 Dollard j John j 17, 121, 122 Doob, Leonard W., 17 Durall, Raymond, 12, 24, 264 Easton j David, 79, 80, 81, 184, 186, 187 Eckstein, Harry, 7, 11, 12 Edwards, Allan L., 51 Edwards, W., 28 expectancy-value-norms theory, 2331 Feather, N. T., 28, 113 Federal Republic of Germany, 3-4, 89, 186 Feierabend j Ivo K. and Rosalind L., 7, 12, 124, 125, 167, 178, 179 Finifter, Ada W., 88 Finifter, Bernard M., 236 Fishbein, Martin, 24, 25, 26, 27, 28, 29, 31, 69, 95, 96, 116, 182 Foley, Mary Ann and Hugh, 262 Forward j John R., 209 frustration-aggression theory, 17, 121-23 Galtung, Johan, 167, 168, 169, 170, 171, 174, 269 Gawiser, Sheldon, 153 general theory of behavior, 23-26

304 I Index Geschwender j James 5 135 Gore, Pearl Mayo, 208 Graham, Hugh Davis, 12, 273 Greenstein, Fred I., 20 Grofman, Bernard N., 8, 9, 128, 130, 153,156 Gurin, Patricia and Gerald, 210, 213 Gurr, Ted Robert, 7, 12, 16, 17, 18, 27, 79, 89, 124, 129, 131, 137, 144, 159, 162, 167, 179, 264, 267, 273, 274, 277 Hage,Jerald, 183 Hake, D. F., 128 Hedlund, Ronald D., 188 Hempel, Carl G., 183 Hibbs, Douglas A., Jr., 6, 12, 46, 48, 53, 179, 264, 265, 272, 274, 276, 277 Hill, R.J., 24 Hoselitz, Bert, 131 Hudson, Michael C., 12 Hutchinson, R. R., 128 Infratest, Munchen, 8-9 Inglehart, Ronald, 193 interview schedule, 8-9 Jayaratne, Toby Epstein, 156 Jukam, Thomas O., 281 Kaase, Max, 47, 132, 190, 196, 198, 201,231 Kaplan, Morton A., 131 Kastner, Daniel L., 188 Klingemann, Hans D., 90 Korpi, Walter, 21, 22, 28, 32, 125, 126, 272 Lao, Rosina C., 210, 213 Laupheimer, Yola, 8 Linehan, William, 47, 48, 265 Lipset, Seymour Martin, 135, 189 Lodge, Milton, 262 Marsh, Alan, 47, 153, 231, 282 Marx, Karl, 135 Masotti, Louis H., 153 McConahay, John B., 208 McPhail, Clark, 16, 116, 137, 138 Miller, ArthurH., 280, 281

Miller, Neal, 121 Miller, Roger LeRoy, 97, 166 Mowrer, O. H., 17 Muller, Edward N., 8, 9, 46, 47, 65, 88, 128, 130, 153, 156, 188, 230, 231, 232,235, 281 Murphy, WalterF., 188 Naditch, Murray, 208, 209 Nesvold, BettyA., 12, 130 Nie, Norman, 5 Nunnally, Jum C., 34, 51 Oberschall, Anthony, 18 Paige, Jeffry M., 88, 209 Pastore, N., 131 Patterson, Samuel C., 188 political behavior: conventional and unconventional, 4-5; strong and weak, 5-6; civil disobedience and political violence, 6-7; typology of political action, 230-35 Polsby, Nelson W 7 ., 20 Pool, Jonathan, 8 Portes, Alejandro, 16, 32 Przeworski, Adam, 16 rank disequilibrium theory, 167-74 Ransford, H. Edward, 208-09 Rao, Potluri, 97, 166 Rejai, Mostafa, 245 relative deprivation theory: defini­ tion, 16-20; expansion and modi­ fication of expectancy-value-norms theory, 32-36 Reppart, Dorothea, 8 reproducibility of a model, 236 research design, 7-10 Rise and Drop or J-Curve hypothesis, 13-14, 135 Ross, Michael, 25 Rothaus, P., 131 Rotter, Julian, 207, 208, 209, 210, 211, 213 Russell, D.E.H., 245, 246, 277 Russett, Bruce M., 12 sampling sites, 9 Schleth, Uwe, 8

Index / 305 SchmidtAdministration, 191-93 Schwartz, David, 80, 88 Sears, David O., 208 Sears, Robert R., 17 Seligson, Mitchell A., 231 Shanks, Merrill, 219, 225 Sherif, Muzafer and Carolyn W., 8 Skinner, B. F., 71 Sniderman, Paul M., 214 Snyder, David, 18, 19 social learning theory, 71-73 socioeconomic change, 134-36 Star, S. A., 129 StoufFer, S. A., 129 Strickland, Bonnie R., 208 support, diifuse and specific, 79-80, 184-90 Suchman, Ε. A., 129 Tanenhaus, Joseph, 188, 262 Tanter, Raymond, 46 Teune, Henry, 16 Thomas, Kerry, 23 Tilly, Charles, 16, 18, 19, 20, 273 Tilly, Louise, 16 Tilly, Robert, 16

Tittle, C. R., 24 Tufte, Edward R., 34, 53 Tursky, Bernard, 262 unrealized expectations, 125-27 utilitarian justification theory: defini­ tion, 20-23; contraction of the expectancy-value-norms theory, 33 variables: micro and macro, 12-14, 264-65; predetermined, 15; ex­ planatory, irrelevant, and left-out, 32; dummy, 97-98 V-Curve relationship, 153-54 Verba, Sidney, 5 Wahlke 1 John C., 5 Walker j Jack L., 281 Weede, Erich, 16 Wicker, A. W., 23 Wildenmann, Rudolf, 8 Williams j Jay R., 209 Williams, R. M., Jr., 129 Wilner, Ann, 131 Worchel, P., 131

LIBRARY OF CONGRESS CATALOGING IN PUBLICATION DATA

Muller, Edward N Aggressive political participation. Includes index. 1. Political participation. 2. Political psychology. I. Title. JA74.5.M84 301.5'92 ISBN 0-691-07605-7

78-70309