The contingency theory of organizations 9780761915737, 0761915737, 9780761915744, 0761915745

Table of contents :
Core paradigm and theoretical integration --
Organic theory and research --
Bureaucracy theory and research --
Causality and contingency in bureaucracy theory --
Controversies in contingency theory research --
Challenges from other theories --
Fit concept and analysis --
Fit affects performance --
Neo-contingency theory --
Future opportunities.

Citation preview

THE CON TIN GENCY THEORY OF OR GA NI ZA TIONS

[p. i; half ti tle] The Con tin gency The ory of Or ga ni za tions

[p. ii; se ries page] [Barbara: Jan elle has on disk]

[p. iii; ti tle page] THE CON TIN GENCY THEORY OF OR GA NI ZA TIONS By Lex Donaldson

Copyright © 2001 by Sage Pub li ca tions, Inc. All rights re served. No part of this book may be re pro duced or uti lized in any form or by any means, elec tronic or me chan i cal, in clud ing pho to copy ing, re cord ing, or by any in for mation storage and re trieval sys tem, with out per mis sion in writ ing from the pub lisher. For information: Sage Pub li ca tions, Inc. 2455 Teller Road Thou sand Oaks, Cal i for nia 91320 E-mail: or [email protected] Sage Pub li ca tions Ltd. 6 Bonhill Street Lon don EC2A 4PU United King dom Sage Pub li ca tions In dia Pvt. Ltd. M-32 Mar ket Greater Kailash I New Delhi 110 048 In dia

Printed in the United States of Amer ica

Li brary of Con gress Cat a loging-in-Pub li ca tion Data Donaldson, Lex The con tin gency the ory of or ga ni za tions / by Lex Donaldson p. cm. — (Foun da tions for or ga ni za tional science) In cludes bib lio graph i cal ref er ences and in dex. ISBN 0-7619-1573-7 (cloth: alk. pa per) ISBN 0-7619-1574-5 (pbk.: alk. pa per) 1. Con tin gency the ory (Man age ment) I. Ti tle. II. Se ries. HD30.4 D65 2000 302.3’5--dc21 00-010586

01

02

03

04

Acquiring Editor: Editorial Assistant: Production Editor: Typesetter/Designer:

05

10

9

8

7

Marquita Flemming MaryAnn Vail Denise Santoyo Denyse Dunn

6

5

4

3

2

1

To Derek Pugh, Leader and Su per vi sor

Contents

In tro duc tion to the Se ries Acknowledgments Preface 1. Core Paradigm and Theoretical Integration A Formal Definition of Contingency Theory The Core Contingency Theory Paradigm Conceptual and Theoretical Integration of Structural Contingency Theory Plan of the Book

2. Organic Theory and Research Organic Theories and Research Underlying Connections in Organic Theory Conclusions

3. Bureaucracy Theory and Research Bureaucracy Theory Structural Differentiation

xi xiii xv 1 5 7 16 31

35 36 55 58

61 62 69

Divisionalization Structures of Multinational Corporations Matrix Structures Other Possible Contingencies Synthesizing the Size and Task Contingencies Conclusions

4. Causality and Contingency in Bureaucracy Theory The Causal Model of Size and Bureaucratic Structure The Curvilinear Relationship of Size and Bureaucratic Structure Functionalist Explanations A Contingency Theory Causal Model of Size and Bureaucracy Conclusions

5. Controversies in Contingency Theory Research Technology Versus Size Determinism Versus Choice Size as Cause of Bureaucratic Structure Reverse Causality Configurationalism Versus Cartesianism Contingencies of Organizational Power Conclusions

6. Challenges From Other Theories Divisionalization Is Not an Institutional Effect Not Population-Ecology Effect Rational Not Political Explanation Not Threat-Rigidity Conclusions

7. Fit Concept and Analysis Philosophical Issues in Contingency Theory Research The Relationship Between Fit and Performance The Identification of the Fit Line Validation of the Fit Model Conclusions

8. Fit Affects Performance Studies of Contingency Fit and Performance The Empirical Estimation of the Effect of Fit on Performance Lessons for Future Research Conclusions

9. Neo-Contingency Theory Problems in Traditional Contingency Theory Disequilibrium and Ongoing Change Performance-Driven Organizational Change

77 81 86 88 91 98

101 102 106 110 119 122

125 126 131 137 138 141 153 157

161 163 168 170 176 178

181 182 185 200 202 212

215 216 226 239 242

245 246 247 249

Quasi-Fit Fit as Hetero-Performance Discussion Conclusions

10. Future Research Opportunities Fit and Bureaucracy Theory Researching Neo-Contingency Theory Conclusions

257 263 268 269

273 274 275 288

References

291

Index

315

About the Au thor

325

Introduction to the Series

T

he title of this se ries, Foundations for Organizational Science (FOS), de notes a dis tinc tive fo cus. FOS books are ed u ca tional aids for mas ter ing the core the o ries, es sen tial tools, and emerg ing per spectives that constitute the field of organizational science (broadly con ceived to include organizational behavior, organizational theory, hu man resource management, and business strategy). Our am bitious goal is to assemble the “essential li brary” for mem bers of our pro fessional com mu nity. The vision for the series emerged from conversations with several col leagues, in clud ing Pe ter Frost, Anne Huff, Rick Mowday, Benjamin Schneider, Su san Tay lor, and Andy Van de Ven. A num ber of com mon interests emerged from these sympathetic encounters, including: en hanc ing the qual ity of doc toral ed u ca tion by pro vid ing broader access to the master teachers in our field, “bottling” the experience and in sights of some of the found ing schol ars in our field be fore they re tire, and pro vid ing pro fes sional de vel op ment op por tu ni ties for col leagues seeking to broaden their understanding of the rapidly expanding subfields within or ga ni za tional sci ence. xi

xii

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

Our unique learn ing ob jec tives are re flected in an un usual set of instruc tions to FOS au thors. They are en cour aged to: (a) “write the way they teach”—framing their book as an extension of their teaching notes, rather than as the ex pan sion of a hand book chap ter; (b) pass on their “craft knowledge” to the next generation of scholars—making them wiser, not just smarter; (c) share with their “vir tual stu dents and col leagues” the in sider tips and best bets for re search that are nor mally reserved for one-on-one mentoring sessions; and (d) make the com plexity of their subject matter comprehensible to nonexperts so that readers can share their puz zle ment, fas ci na tion, and in trigue. We are proud of the group of highly qual ified au thors who have em braced the unique ed u ca tional per spec tive of our “Foun da tions” series. We en cour age your sug ges tions for how these books can better satisfy your learn ing needs—as a new comer to the field pre par ing for prelims or de vel op ing a dis ser ta tion pro posal, or as an es tab lished scholar seek ing to broaden your knowl edge and pro fi ciency. —D

AVID

—

A. WHETTEN

SE R I E S

EDITOR

Acknowledgments

THE Ac knowl CON edg TIN ments GENCY THEORY OF OR GA NI ZA TIONS

I

should like to thank those who have helped me with this project. First, thanks to Dave Whetten who asked me to write this book. I am glad Dave gave me the op por tu nity to lay out con tin gency the ory and its future. I should like to thank those who taught me and provided an early introduction to the intellectual excitement of contingency theory: John Child, Jerald Hage, and David Hickson. Also thanks to Derek Pugh, who headed the research group at London Business School where I con ducted my first con tin gency anal y ses. Other help ful con tingency research colleagues there were Roger Mansfield and Malcolm Warner. More latterly I have received support in my contingency en deavors from American colleagues, including Al fred Chandler, Paul Law rence, Nitin Nohria, and Mike Tushman. Thanks also to my gen er ous hosts, es pe cially to Janine Nahapiet, at Templeton College, Oxford University, where part of this book was written. Templeton College combines high levels of efficiency and friend li ness and proves to us all that this is pos si ble. Other parts of this book were written at London Business School and I should like to thank my hosts there: Nigel Nicholson and Paul xiii

xiv

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

Willman. London Business School combines professionalism with so cia bil ity, so thanks also to my com pan ions there: Pino Audia, Pat rick Barwise, Julian Birkinshaw, Ian Coop er, Sumantra Ghoshal, Rob Goffee, Lynda Gratton, John Hunt, Andrew Likierman, Constantinos Markides, Peter Moran, Tim Morris, Anand Narasimhan, Maury Peiperl, Henri Servaes, Ken Simmonds, John Stopford, and Chris Voss. Thanks to Peter Dodd, former Dean of the Aus tralian Graduate School of Man age ment, and Greg Whittred, the Acting Dean, for sup port throughout the writing of this book. Thank you again to June Ohlson, my ever-enthusiastic editor, whose efforts have resulted in a clearer book. —LEX DONALDSON AU S T R A L I A N G R A D U A T E SC H O O L O F M A N A G E M E N T, UN I V E R S I T I E S O F N E W S O U T H W A L E S A N D S Y D N E Y

Preface

T

his book be gan when Dave Whetten ap proached me to con trib ute a book to the Foundations for Organizational Sci ence se ries. He wanted each book to make its topic ac ces si ble to new com ers to the field, to stimulate their interest in the possibility of joining in its research. The read ers would in clude doc toral stu dents and also fac ulty and research colleagues who were unfamiliar with the topic. Each book would pro vide an over view of the lit er a ture, the crit i cal is sues, and the fu ture pos si bil i ties. I en thu si as ti cally agreed to write a book on con tingency the ory that would match that agenda. I have been in volved in con tin gency re search for al most thirty years now and so felt equipped for the task. In deed, since I have now writ ten four books on organizational the ory, the reader might well ask: “What’s new here?” Most of the book is new and does not re peat ma terial from my earlier books. Of course, there is a thematic continuity with my previous books on organizational theory in that throughout them I have ar gued for con tin gency the ory. While I have written about contingency the ory fairly extensively, much of this has been in the context of critiques of other organiza tional theories that challenge con tin gency the ory. In con trast, in this xv

xvi

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

book there is lit tle (though some) cri tique of other the o ries. This book is an exposition and crit i cal dis cus sion of contingency theory and its research. The present book is a far more comprehensive treat ment of con tin gency the ory than any I have pre vi ously at tempted. It is broader in the range of the o ries con sid ered un der the con tin gency um brella. It is also much deeper in its anal y sis of the the o ries, ev i dence, and meth odological issues. This allows more appreciation of the coherency of contingency the ory over all. It also in volves frank rec og ni tion of some of the deficiencies in contingency theory research. The coherent un der ly ing model pro vides the plat form from which to make good some of the deficiencies through a series of improvements in theory and method that chart a course for fu ture re search. In keeping with a book intended to explain con tingency theory, Chap ters 2 and 3 lay out the foun da tions by re view ing the pi o neer ing contributors to theory and empirical research. I have tried to give an ac cu rate ac count of the re ceived body of con tin gency the ory as it co mes down to us, in order to pass on this tra di tion to schol ars who are new to it. However, more per sonal views are of fered also at many places. The opening chapter presents a the oretical integration to provide the reader with an over view that makes sense of what is a large lit er a ture. It also argues that there is an under ly ing core par adigm that renders contin gency the ory co her ent. Chap ter 4 makes an in-depth ex am i na tion of the causal mod els in the re ceived bu reau cracy re search lit er a ture and finds them to be deficient and attempts to put them on a more truly con tin gency the ory ba sis. Chap ters 7 and 8 ex am ine in de tail the con cept of fit and its re la tion ship with per for mance, in clud ing the em pir ical re search stud ies. Chapter 9 pres ents possi ble new developments for contin gency the ory, to make it more co her ent and, it is hoped, valid. These new developments include the concepts of dis equilibrium, quasi-fit, and hetero-performance. All three are novel concepts that, I be lieve, sub stantially re vise and imp rove contingency theory. In particular, the hetero-performance concept may be of particular in terest to those seek ing to ad vance con tin gency the ory. Chap ter 10 of fers sug ges tions on how to operationalize the ideas in this book in terms of hy poth e ses for fu ture em pir i cal re search. Thus most of this book is mak ing ar gu ments that I have not pre sented pre vi ously.

Preface

xvii

In or der to pro vide the reader with a rounded treat ment of con tingency theory in this one book, however, I have restated some ar guments made in some ear lier books. Thus sec tions of Chap ters 5 and 6 draw upon de fenses of con tin gency the ory made in Donaldson (1995a, 1996a). To have omitted this material in its entirety could have left a reader con fused about whether con tin gency the ory can re ply to cer tain crit i cisms. Sim i larly, a sec tion of Chap ter 9 draws upon or ga ni za tional port fo lio the ory from Donaldson (1999) to pres ent the im pli ca tions of the idea that or ga ni za tional change is per for mance driven. Hope fully, the con densed sum ma ries of fered in these sec tions may help make this material more accessible to a wider au di ence. Through out much of this book a uni fy ing idea is that the dy nam ics of organizational change in contingency theory are best captured by the struc tural ad ap ta tion to re gain fit (SARFIT) model. This is in te gral to the core par a digm ar tic u lated in the open ing chap ter. The SARFIT model re curs in the dis cus sion at many places, such as about or ga ni za tional change, bu reau cracy the ory, func tion al ism, and in the operationalization of fit for fu ture re search. Thus this book has a tri ple agenda: to pay hom age to a rich tra di tion and pass it on, to ad vance a co her ent in ter pre ta tion of the ar ray of the ories and re search within it, and to set signposts to what may be fruit ful av e nues for fu ture re search. In these ways it seeks to ful fill the in ten tion of the Foun da tions for Or ga ni za tional Sci ences se ries.

THE CON Core Par a digm TIN GENCY and TheTHEORY o ret i cal InOF te gra ORtion GA NI ZA TIONS

1

Core Paradigm and Theoretical Integration

T

he con tin gency the ory of or ga ni za tions is a ma jor the o ret i cal lens used to view organizations. It yields many insights and has substantial em pirical sup port. Con tin gency the ory con tains much of im por tance in the his tory of or ga ni za tional sci ence. Its re search forms the ba sis of much that is taught to day. More over, con tin gency re search is actively pur sued by scholars in the contemporary era and it is being pro jected into the fu ture in a se ries of ex cit ing the o ret i cal and em pir ical de vel op ments. The aim of this book is to in tro duce the reader to the rich tra di tion of con tin gency the ory re search, in di cate the po ten tial for fu ture de vel op ments, and of fer guid ance on how to pro ceed in re search both in the ory and em pir i cal prac tice. The essence of the con tingency the ory par adigm is that organizational ef fec tive ness re sults from fit ting char ac ter is tics of the or ga ni zation, such as its struc ture, to con tin gen cies that re flect the sit u a tion of the or ga ni za tion (Burns and Stalker 1961; Law rence and Lorsch 1967; Pennings 1992; Wood ward 1965). Con tin gencies in clude the en vi ronment (Burns and Stalker 1961), or ga ni za tional size (Child 1975), and 1

2

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

organizational strategy (Chandler 1962). Because the fit of organiza tional char ac ter is tics to con tin gen cies leads to high per for mance, or gani za tions seek to at tain fit. For this rea son, or ga ni za tions are mo ti vated to avoid the mis fit that re sults af ter con tin gen cies change, and do so by adopting new organizational characteristics that fit the new levels of the contingencies. There fore the or ga ni za tion be comes shaped by the contingencies, be cause it needs to fit them to avoid loss of per for mance. Or ga ni za tions are seen as adapt ing over time to fit their chang ing con tin gen cies so that ef fec tive ness is main tained. Thus con tin gency the ory con tains the con cept of a fit that af fects per for mance, which, in turn, impels adaptive organizational change. This results in organizations mov ing into fit with their con tin gen cies, so that there is an align ment be tween the or ga ni za tion and its con tin gen cies, cre at ing an as so ci a tion between contingencies and or ga ni za tional char ac ter is tics (Burns and Stalker 1961; Wood ward 1965; Van de Ven and Drazin 1985). Some of the more im por tant con tin gency the o ries of or ga ni za tional struc ture in volve the three con tin gen cies of the en vi ron ment, or ga nizational size, and strategy. The environmental stability contingency affects mech a nis tic struc ture (Pennings 1992). The rate of tech no log ical and mar ket change in the en vi ron ment of an or ga ni za tion af fects whether its struc ture is mech a nis tic (i.e., hi er ar chi cal) or or ganic (i.e., par tic i pa tory; Burns and Stalker 1961). The mech a nis tic struc ture fits a sta ble en vi ron ment, because a hierarchical ap proach is efficient for rou tine op er a tions. Given the rou tine na ture of op er a tions, the man agers at upper lev els of the hi er ar chy pos sess suf fi cient knowl edge and information to make de ci sions, and this cen tral ized con trol fos ters ef ficiency. In con trast, the or ganic struc ture fits an un sta ble en vi ron ment, because a participatory ap proach is required for innovation. Knowl edge and information re quired for in no va tion are dis trib uted among lower hi er ar chi cal lev els and so de cen tral ized de ci sion mak ing fos ters in no va tion. An or ga ni za tion that has the mis fit ting, mech a nis tic structure in an un sta ble en vi ron ment is un able to in no vate and so be comes ineffective (Burns and Stalker 1961). The size contin gency affects bureaucratic structure. The size of an or ga ni za tion, that is, the num ber of its em ploy ees, af fects the de gree to which its struc ture is bu reau cratic, for ex am ple, rule-gov erned, and is decentralized (Pugh and Hickson 1976; Pugh and Hinings 1976). The bu reau cratic struc ture fits a large or ga ni za tion, be cause large size leads to repetitive operations and administration so that much decision

Core Par a digm and The o ret i cal In te gra tion

3

making can be by rules, rendering decision making in expensive and efficient (Child 1975; Weber 1968). In con trast, an unbureaucratic, or sim ple, struc ture, which is not rule-gov erned and is cen tral ized, fits a small or ga ni za tion, be cause top man age ment can make al most all the de ci sions per son ally and ef fec tively (Child 1972a). A large or ga ni za tion that seeks to use the mis fit ting, sim ple struc ture will find top man agement over whelmed by the num ber of de ci sions it needs to make, so that the or ga ni za tion be comes in ef fec tive. The strat egy con tin gency af fects di vi sional struc ture. The func tional structure fits an un diversified strategy, because all its activities are focused on a sin gle prod uct or ser vice so that ef fi ciency is en hanced by specialization by function (e.g., departments of production, market ing, etc). In con trast, the di vi sional struc ture fits a di ver si fied strat egy, because it has diverse activities serv ing vari ous product-markets, so effectiveness is en hanced by co or di nat ing each prod uct or ser vice in its own di vi sion (Chan dler 1962; Galbraith 1973). An or ga ni za tion with a diversified strat egy that seeks to use the mis fit ting, func tional struc ture will find top management overwhelmed by the number of decisions and also suf fer lack of re spon sive ness to mar kets, so that the or ga ni zation be comes in ef fec tive. Thus structural con tingency theory argues that organizational struc ture needs to fit the three con tin gen cies of the en vi ron ment, size, and strategy. As seen, each of these contingencies affects a particular aspect of struc ture: or ganic, bu reau cratic, and di vi sional, re spec tively. Change in any of these contingencies tends to produce change in the corresponding structural aspect (Burns and Stalker 1961; Chandler 1962; Child 1973a). In this way the organization moves its structure into align ment with each of these con tin gen cies, so that struc ture and con tin gency tend to be as so ci ated (Child 1973a; Hage and Aiken 1969; Rumelt 1974). There are other contingencies of or ga ni za tional structure, as we shall see, but these preliminary re marks serve to ex em plify contingency theory. Con tin gency the ory is to be dis tin guished from uni ver sal ist ic the o ries of or ga ni za tion, which as sert that there is “one best way” to or ga nize, mean ing that max i mum or ga ni za tional per for mance co mes from the maximum level of a structural variable, for example, specialization (Tay lor 1947). Clas si cal man age ment is an ear lier or ga ni za tional theory that argues that maximum organizational performance results from max i mum for mal iza tion and spe cial iza tion (Brech 1957), and it

4

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

is therefore a universalistic type of theory. Similarly, neo-human re lations is also an earlier universalistic type of organizational theory, which ar gues that or ga ni za tional per for mance is max i mized by max imizing participation (Likert 1961). Contingency theory differs from all such uni ver sal ist ic the o ries in that it sees max i mum per for mance as re sult ing from adopt ing, not the max i mum, but rather the ap pro pri ate level of the struc tural vari able that fits the con tin gency. There fore, the op ti mal struc tural level is sel dom the max i mum, and which level is opti mal is de pend ent upon the level of the con tin gency vari able. Much of contingency theory research has studied organizational struc ture (Donaldson 1995a, 1996a; Law rence 1993), and this tra di tion is referred to as structural contingency the ory (Pfeffer 1982). In this book we will fo cus on struc tural con tin gency the ory re search be cause it is a large and com plex body of work. There are, how ever, con tin gency the o ries of many dif fer ent or ga ni za tional char ac ter is tics, such as leader ship (Fiedler 1967), hu man re source man age ment (Delery and Doty 1996), and strategic decision-making pro cesses (Frederickson 1984). There are many common issues that run across the various con tingency the o ries of the dif fer ent or ga ni za tional char ac ter is tics. Hence by pro vid ing a dis cus sion that fo cuses on struc tural con tin gency the ory we may also illuminate con tingency the ories of other organizational characteristics. Con tem po rary or ga ni za tional re search ers are seek ing to build upon the struc tural con tin gency tra di tion and make new con tri bu tions to it. For example, some contemporary contingency researchers are con cerned to show the ef fect of fit on per for mance that con tin gency the ory pos tu lates (Gresov 1990; Ham il ton and Shergill 1992, 1993; Hill, Hitt, and Hoskisson 1992; Jennings and Seaman 1994; Keller 1994; Kraft, Puia, and Hage 1995; Mahoney 1992; Marsden, Cook, and Kalleberg 1994; Palmer, Jennings, and Zhou 1993; Powell 1992; Schlevogt and Donaldson 1999). Other re search ers are in ves ti gat ing the con tin gency fac tors and iden ti fy ing which as pects of or ga ni za tion they ef fect (e.g., Anderson 1996; Birkinshaw 1999; Jarley, Fiorito, and Delaney 1997). Oth ers again are study ing the con tin gency ad ap ta tion pro cesses (e.g., Priem 1992, 1994). Still other re search ers, whose pri mary al le giance is to a the ory other than con tin gency the ory, nev er the less find in con tingency work el e ments that may help fully be in cor po rated into their own pro jects (e.g., Fligstein 1985).

Core Par a digm and The o ret i cal In te gra tion

5

The foregoing, brief definition of structural con tingency theory raises sev eral ques tions. Is there a uni fy ing struc tural con tin gency the ory par a digm, or just a loose ag glom er a tion of vaguely re lated the o ries? Is it correct to say that contingencies cause organizational struc ture or are there other causal patterns, such as the structure caus ing the con tin gency? Do con tin gency fac tors de ter mine or ga ni za tional struc ture or is there choice? Is there really an underlying fit of or ganizational char ac ter is tic to con tin gency that drives cau sal ity, as con tin gency the ory ar gues, or is the whole idea of fit just an ex er cise in tau tol ogy? These are some of the ques tions we must con sider in dis cuss ing the con tin gency the ory of or ga ni za tions. In or der to be gin our dis cus sion, let us first clar ify ex actly what we mean by con tin gency the ory by of fer ing a def i ni tion in for mal terms. The next step will be to con sider whether there is a con tin gency the ory paradigm that unites the dis pa rate con tin gency the o ries. We will seek to show that there is for structural contingency theory. We will then move to con sider the more spe cific con tin gen cies and struc tural vari ables in struc tural con tin gency the ory. It will be shown that many of them fall into two groups: organic theory and bureaucracy theory. These are seen as be ing the o ries that are to some de gree in con flict, but that may be brought to gether, ren der ing struc tural con tin gency the ory co her ent—and not overly com plex. The re main der of this chap ter provides a con cep tual and the o ret i cal in te gra tion, and an in tro duc tion to the issues and material in the body of this book. In this way it puts in place some ma jor build ing blocks and pro vides an over view for what is to come. The pro cess there fore is to move from con tin gency logic, to the most gen eral con tin gency the ory, to con tin gency the o ries, and then in the en su ing chap ters to con crete find ings and meth ods.

A Formal Definition of Contingency Theory The con tin gency the ory of or ga ni za tions is a sub set of the con tin gency ap proach in sci ence, so let us first de fine the con tin gency ap proach and then the con tin gency the ory of or ga ni za tions within it. At the most abstract level, the contingency approach says that the effect of one vari able on an other de pends upon some third vari able, W. Thus the ef fect of X on Y when W is low dif fers from the ef fect of X on Y

6

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

when W is high. For ex am ple, it might be that when W is low, X has a pos i tive ef fect on Y, whereas when W is high, X has a neg a tive ef fect on Y. Thus we can not state what the ef fect of X on Y is, with out know ing whether W is low or high, that is, the value of the vari able W. There is no valid bivariate relationship between X and Y that can be stated. The relationship be tween X and Y is part of a larger causal sys tem in volv ing the third vari able, W, so that the valid gen er al iza tion takes the form of a trivariate re la tion ship. A bivariate re la tion ship is too sim ple to cap ture the lawlike regularity con nect ing X and Y. There fore a more complex causal statement is required. However, the contingency statement is just one step more com plex in that it con sists of only one more vari able, going from two to three variables. The third vari able, W, mod er ates the re la tion ship between X and Y and can there fore be called a mod er a tor of the re la tion ship or a con dition ing vari able of the re la tion ship (Galtung 1967). How ever, while a contingency fac tor is a mod er a tor or con di tion ing vari able, it plays a more spe cific role, so that not all mod er a tors are con tin gen cies. In the con tin gency the ory of or ga ni za tions, the re la tion ship is be tween some characteristic of the organization and effectiveness. Thus the con tingency factor determines which characteristic produces high levels of effectiveness of the or ga ni za tion (or some part of it, such as a depart ment or individual mem ber). For example, a mechanistic structure pro duces high ef fec tive ness when the task un cer tainty contingency is low, and an organic structure produces high effectiveness when the task un cer tainty con tin gency is high. The reason for the focus on ef fectiveness in contingency theory is that or ga ni za tional the ory has been con cerned to ex plain the suc cess or failure of organizations. However, organizational effectiveness can have a broad meaning that includes efficiency, profitability (Child 1975), employee satisfaction (Dewar and Werbel 1979), innovation rate (Hage and Dewar 1973), or patient well-being (Alexander and Randolph 1985; see also Pennings 1992). Or ga ni za tional ef fec tive ness can be de fined as the abil ity of the or ga ni za tion to at tain the goals set by itself (Parsons 1961), or by its ability to func tion well as a system (Yuchtman and Sea shore 1967), or by its abil ity to sat isfy its stake holders (Pfeffer and Salancik 1978; Pickle and Friedlander 1967). Organiza tional effectiveness and performance are sim i lar con cepts and will be used in ter change ably in this book.

Core Par a digm and The o ret i cal In te gra tion

7

The Core Contingency Theory Paradigm As just seen, a contingency is any vari able that mod er ates the ef fect of an organizational characteristic on or ganizational performance. Given this def i ni tion of a con tin gency, it is clearly quite en com pass ing, so the open-ended na ture of such a def i ni tion can prompt un ease that there is no sin gu lar, con tin gency the ory of or ga ni za tions, only a mul tiplicity of contingency theories and hence no unifying paradigm. There fore the crit i cism is some times made that there is no struc tural contingency the ory of or ga ni za tions, only a loose group ing of dis pa rate theories, each of which makes a connection be tween its con tingency and some as pect of or ga ni za tional struc ture, for ex am ple, task uncertainty and the organic structure, or size and bureaucracy. The question, therefore, is whether there exists some more overarching framework within which the various struc tural con tin gency the o ries come to gether. Thus, at a more ab stract level, are there com mon al ties be tween the struc tural con tin gency the o ries? We be lieve that there are. It is possible to create a theoretical framework that con nects con tingencies in the abstract with organizational structure in the abstract. The ar gu ment is highly gen eral so that it ap plies to all the con tin gen cies and their cor re spond ing struc tural as pects. In this sense, there is a uni fy ing con tin gency paradigm. Struc tural con tin gency the ory con tains three core el e ments that together form its core paradigm. First, there is an association between contingency and the or ganizational structure. Sec ond, con tingency de ter mines the or ga ni za tional struc ture, be cause an or ga ni za tion that changes its contingency then, in consequence, changes its structure. Third, there is a fit of some level of the or ga ni za tional struc tural vari able to each level of the contingency, which leads to higher per formance, whereas mis fit leads to lower per for mance. This fit-per for mance relationship is the heart of the contingency theory par a digm. It pro vides the the o ret i cal ex pla na tion of the first two points: the as so ci a tion between con tin gency and struc ture, and con tin gency change caus ing struc tural change. An or ga ni za tion that changes the level of its con tingency tends to have been in fit when it made the change, and thence to move into mis fit so that its per for mance de creases. The or ga ni za tion then changes its or ga ni za tional struc ture to fit the new level of the contingency vari able, in or der to avoid fur ther per for mance loss. There fore,

8

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

be cause of the per for mance lost by be ing in mis fit, or ga ni za tions tend over time to move to ward fit. Thus any or ga ni za tion tends to adopt the struc ture that fits its level of the con tin gency. This means that a change in contingency leads to a change in structure, so that contingency determines structure. In this way the contingency and the organiza tional structure move into alignment and so arises the association between the con tin gency and the or ga ni za tional struc ture. Let us now examine structural con tingency theory research to see that these three el e ments are com mon al i ties that ex ist across the dif ferent contingencies and theories. We will consider each of three core commonalties in turn: (1) the association between contingency and or ga ni za tional struc tural variable or at trib ute; (2) the change pro cess that con tin gency change causes or ga ni za tional struc tural change; and (3) the fit of struc ture to con tin gency that af fects per for mance. 1. As so ci a tion Be tween Con tin gency and Or ga ni za tional Struc ture Con tin gency the ory re search into or ga ni za tional struc ture fre quently shows an as so ci a tion be tween con tin gency and struc ture. Some times these as so ci a tions are pre sented as cross-tab u la tions (e.g., Wood ward 1965) and at other times as correlations (e.g., Holdaway, Newberry, Hickson, and Heron 1975, p. 48). Such cor re la tions are seen for the size con tin gency and var i ous as pects of bu reau cratic struc ture (e.g., Child 1973a, p. 170, Ta ble 2). They are seen for the strat egy contingency and divisional structure (Grinyer, Yasai-Ardekani, and Al-Bazzaz 1980, p. 198). They are seen also for technology and structure (Child and Mansfield 1972, pp. 378-379). Thus there is a clear com mon al ity across the di verse con tin gen cies of organizational structure in that, despite differences in contingencies and their corresponding structural as pects, the contingency theories postulate theoretically an association be tween con tin gency and struc ture, and dem on strate this as a cen tral part of their empirical research. Thus a bivariate association is expected be tween any contingency factor and its structural variable. Of ten these bivariate re la tion ships are lin ear, that is, as the con tin gency fac tor in creases in value, so does the struc tural vari able. How ever, as so ci a tions do not have to be lin ear and may be curvilinear, as, for ex am ple, are sev eral of the con tin gencystruc ture as so ci a tions in Wood ward’s anal y sis of tech nol ogy and struc-

Core Par a digm and The o ret i cal In te gra tion

9

ture. For ex am ple, as the tech nol ogy con tin gency in creases, the span of control of the first-line supervisor first increases and then decreases (Wood ward 1965). Whether lin ear or curvilinear, there is an as so ci a tion be tween the con tin gency and the struc ture as a core com po nent of the structural contingency research paradigm that provides a com mon al ity across the di verse struc tural con tin gency the o ries. 2. Con tin gency Change Causes Or ga ni za tional Struc tural Change The use of cross-sectional methods in empirical stud ies that show cor re la tions be tween con tin gency and struc ture is taken in some commentaries to imply that structural contingency the ory is static (Galunic and Eisenhardt 1994). However, structural contingency theory deals with or ga ni za tional change. It contains a theory that is dynamic, which is sup ported by stud ies of or ga ni za tions chang ing over time. While there are differences in the contingency and structural factors, there is a sim i lar view of or ga ni za tional change. Con tin gency causes struc ture in that change in con tin gency leads to change in structure (e.g., Burns and Stalker 1961; Chan dler 1962). The changes in con tin gency lead the or ga ni za tion out of fit with the old struc ture, which lowers per formance. Eventually the organization resolves this by adopt ing a new and better struc ture that fits the new level of the con tin gency, thereby res toring per formance. This model of organizational change is seen in Burns and Stalker (1961), where in creas ing tech nolog i cal and mar ket change even tu ally cause the or ga ni za tion to change from a mech a nis tic to an or ganic struc ture in their em pir i cal case studies of changes over time. Using Thomp son’s con cepts of con tin gency and structure, Van de Ven, Delbecq, and Koenig (1976) show that change in the task con tin gency leads to change in co or di na tion modes, so that the task con tin gency causes struc ture. Again it is seen in the way that strat egy leads to struc ture, that is, di ver si fy ing causes the or ga ni zation to adopt a divisional structure (Chandler 1962; Channon 1973, 1978; Dyas and Thanheiser 1976; Rumelt 1974). Thus a model of or ga ni za tional change ex ists in struc tural con tin gency the ory, in which con tin gency changes cause change in or ga ni za tional struc ture. This uni fied the ory of or ga ni za tional change that goes across the con tin gen cies (e.g., en vi ron ment and strat egy) is one more sense in which there is the con tin gency the ory of or ga ni za tions. There fore the

10

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

dy nam ics that con tin gency changes cause struc tural changes is a sec ond core com po nent of the con tin gency par a digm. Contingency theory depicts organizational change as an organization ally ra tio nal pro cess of re stor ing ef fec tive ness, so that is a func tionalist type of sociological theory (Burrell and Morgan 1979). Several other organizational theories challenge this rationalist, functionalist ac count and they will be dis cussed fur ther (see Chap ter 6). Much of the argument that change is func tional for the or ga ni za tion hangs on the idea that organizations changing their structures are doing so to move from misfit into fit and thereby restoring their performance, which takes us to our next point. 3. Fit Af fects Per for mance Con tin gency the o ries hold that there is a fit be tween the or ga ni za tional struc ture and con tin gency that has a pos i tive ef fect on per for mance. In formal terms, there is a trivariate relationship be tween structure, contingency, and per for mance. Where the struc tural vari able is at the level that fits the level of the contingency, high performance results. Where the struc ture is at a level that does not fit the level of the con tin gency, low per for mance re sults. This con tin gency-struc ture-fit ex plains the association be tween con tin gency and struc ture. Thus the de ci sive proof of a con tin gency the ory in volves em pir i cally dem on strat ing that there is a fit of struc ture to con tin gency that pos i tively af fects per formance. This re quires show ing that the com bi na tion of con tin gency and struc ture that is held to be a fit causes high per for mance and that the combination held to be a mis fit causes low per for mance. Each con tin gency the ory spec i fies the struc tures that fit its con tin gency, so that the fits and mis fits are unique to that the ory (e.g., di vi sional struc ture fits di ver si fi ca tion, bu reau cratic struc ture fits large size). De spite such di versity about the ex act def i ni tion of fit, these dif fer ent views all con tain the con cept of fit. Thus, at the more ab stract level, the con cept of fitpro vides a the o ret i cal com mon al ity across dif fer ent con tin gency theories. Contingency the ory re search looks for a re la tion ship be tween fit and per for mance, that is, searches for a trivariate re la tion ship among con tin gency, struc ture, and per for mance. For ex am ple, Wood ward (1965) dem on strates that fit of span of con trol of the first-line su per vi sor to the technology contingency is associated positively with organiza tional per for mance. Other ex am ples are Child (1975) and Khandwalla

Core Par a digm and The o ret i cal In te gra tion

11

(1973), who show that fit of bu reau cratic struc ture to the size con tin gency is as so ci ated pos i tively with or ga ni za tional per for mance. In Chap ter 8 we will re view stud ies that em pir i cally sup port fits that pos i tively af fect organizational performance for a range of different contingencies. Hence the fit-performance relationship is a common el ement found among the di verse con tin gency the o ries and so con sti tutes a third core com po nent of the con tin gency the ory par a digm. Structural Ad ap ta tion to Re gain Fit The core con tin gency the ory par a digm is il lus trated by re search on the strat egy con tin gency and its re la tion ship with di vi sional struc ture. The research supports each of the three elements of the paradigm. More over, it sup ports a more par tic u lar the o ret i cal model of how contin gency change leads to struc tural change, namely, struc tural ad ap tation to re gain fit (SARFIT). There is an as so ci a tion be tween the con tin gency vari able of stra te gic di ver si fi ca tion and divisionalization (i.e., the dis tinc tion be tween the functional and the di vi sional struc tures; e.g., Grinyer and Yasai-Ardekani 1981). Strategy contingency change causes divisional structural change. This is shown by stud ies across time. In the case his to ries of Chan dler (1962), strat egy leads to struc ture, that is, di ver si fy ing causes the or ganization to adopt a di vi sional struc ture. Quantitative s tud ies of large corporations also show that diversification precedes and causes divisionalization (Channon 1973, 1978; Dyas and Thanheiser 1976; Fligstein 1985; Rumelt 1974; Suzuki 1980). The pro cesses in volved in chang ing the or ga ni za tional struc ture in response to changes in the level of the contingency variable can be spec i fied in the the o ret i cal model of struc tural ad ap ta tion to re gain fit (SARFIT; Donaldson 1987). This is shown in Figure 1.1. The SARFIT model states that an organization is initially in fit, hav ing a struc ture that fits its ex ist ing level of the con tin gency vari able. Fit pos i tively af fects performance. However, the organization then changes its level of the contingency variable while retaining its existing structure, which thereby be comes a mis fit with its new con tin gency level (thus the ef fect of the con tin gency vari able on fit is shown as neg a tive in Fig ure 1.1). In turn, the misfit leads to lower performance (reversing the positive effect of fit on per for mance shown in Fig ure 1.1). When per for mance becomes so low (be cause of the ef fects of the mis fit and other causes)

12

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

Fig ure 1.1. The Con tin gency The ory of Struc tural Ad ap ta tion to Re gain Fit (SARFIT)

that it becomes less than the satisficing (i.e., satisfactory) level, the organization then makes an adap tive change. The adap tive change is to adopt a new or ga ni za tional struc ture that fits its new con tin gency level (shown as a positive ef fect of or ga ni za tional struc ture on fit in Fig ure 1.1). The new fit re stores per for mance. Thus struc tural change oc curs in response to contingency change and is triggered by the feedback effect from the low performance caused by misfit. The organization adapts its struc ture to changes in the con tin gency in or der to main tain ef fec tive func tion ing.

Core Par a digm and The o ret i cal In te gra tion

13

To in ves ti gate whether the SARFIT the o ret i cal model holds for the phe nom ena of strat egy and struc ture, we must first operationalize fit. From Chandler (1962) it can be seen that functional structures fit undiversified firms and mis fit di ver si fied firms, while di vi sional structures fit diversified firms and misfit undiversified firms (Donaldson 1987, pp. 8-9). This operationalization was empirically validated by show ing that firms in fit had sig nif i cantly higher sub se quent fi nan cial per for mance (growth in profit on sales and on cap i tal), so that fit was a cause of performance (Donaldson 1987, pp. 16-17). This operational def i ni tion has been in de pend ently rep li cated on other firms (Ham il ton and Shergill 1992, 1993). Initially, firms are undiversified and tended to have a functional structure, so that empirically most of them were in fit (Donaldson 1987; see also Figure 1.2). Movement away from this initial position was overwhelmingly through strategic change, by diversification, while re tain ing their func tional struc ture, so en ter ing mis fit. Of firms that moved from fit to mis fit, 83 per cent did so by in creas ing their strategy con tin gency (Donaldson 1987, p. 14). Once in mis fit of struc ture to strategy, the firms would consequently experience re duced per formance. Then, when fi nan cial per for mance be came low, so that a cri sis of poor performance occurred, firms moved from misfit into fit by adopting a new, divisional structure that fit ted their di ver si fied strategy, so mak ing a struc tural ad ap ta tion. Firms in mis fit were more than four times more likely to change only their struc ture (Donaldson 1987, p. 14), showing that misfit led to structural change. Of firms that changed only their struc ture, 72 per cent moved from mis fit to fit and only 5 per cent moved from fit to mis fit (Donaldson 1987, p. 14), showing that when firms made struc tural change it was over whelm ingly a move from misfit into fit, that is, structural adaptation. Of the firms that moved from misfit to fit by changing only their structure, 90 percent did so by adopting the di vi sional struc ture (Donaldson 1987, p. 15). Thus firms moved from misfit to fit by divisionalizing. The fit would be beneficial for their performance. Hence, overall, the firms moved from mis fit into fit by adapt ing their struc ture to their strat egy contingency. The ad ap ta tion of struc ture to strat egy is sub ject to time lags, however. There was no pat tern of firms that di ver si fied in one de cade be ing more likely to change their struc tures in the sub se quent de cade, so that,

14

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

Fig ure 1.2. Change in Strat egy Causes Change in Struc ture: Dy nam ics of Struc tural Adaptation

on average, structural change did not follow strategic change within ten years (Donaldson 1987, p. 13). Indeed, 77 percent of the firms in mis fit had been so for ten years or more (Donaldson 1987, pp. 15-16), pointing to the lengthiness of lags of structural change to strategic change. Strat egy slowly leads to struc ture. The rea son is that per for mance mediates struc tural change, so that it oc curs only when per for mance be comes low. For di ver si fied firms in mis fit, we can com pare those that moved into fit, by struc tur ally chang ing from func tional to di vi sional, with those re tain ing the mis fit ting func tional struc ture. The divisionalizing firms had significantly lower sales and profits, and especially earnings p er share, at the start of the period than those firms that retained the func tional struc ture (Donaldson 1987, pp. 17-18). These results show that poor performance and failure to satisfy powerful stake holders (i.e., own ers) trig ger struc tural ad ap ta tion. Over all, con tin gency change (di ver si fi ca tion) caused struc tural change (divisionalization). The ini tial con tin gency change cre ates dysfunctions that are cor rected by struc tural change, so that the or ga ni za tion starts

Core Par a digm and The o ret i cal In te gra tion

15

out in fit and ends in fit. The firms went from an old fit to a new fit through the intervening stage of misfit. The effect of mis fit on struc tural change is, how ever, in di rect, run ning through per for mance. It is the feedback effect of low performance that is the more immediate cause of structural change. Change in the strategy con tingency leads eventually to struc tural change, be cause of the need to re store per for mance, as the pioneers of structural contingency theory stated (e.g., Chan dler 1962). While the idea that divisionalization leads to di ver si fica tion is fa mil iar, the im por tant point is that it is through the pro cess of mis fit and low per for mance. This con firms the func tion al ist con tin gency theory of organizational change and, more specifically, the model of struc tural ad ap ta tion to re gain fit (SARFIT). Fur ther sup port for the SARFIT model co mes from em pir i cal stud ies by Ezzamel and Hilton (1980) and Hill and Pickering (1986). Ezzamel and Hilton (1980) found that be fore divisionalization, firms tended to have de te ri o rat ing per for mance (as mea sured by share price), and that af ter divisionalization, per for mance tended to rise. Sim i larly, Hill and Pickering (1986, pp. 34-35) found that structural changes (pre dom inantly divisionalization) of firms were seen by their top man age ment as about equally attributable to strategic change and responding to prob lems. The stra te gic change re ferred to ma jor new ac qui si tions and di ver si fi ca tion (Hill and Pickering 1986, p. 35) and so is con sis tent with the theory that diversification leads to divisionalization. Decline in company performance was the most frequent type of problem (Hill and Pickering 1986, p. 35). An other prob lem was the “need to in crease accountability,” which prob a bly also re flects con cerns about un sat is factory per for mance (Hill and Pickering 1986, p. 35). These two prob lems are con sis tent with the the ory that low per for mance trig gers struc tural change. Yet another problem was “intra-organizational communica tions,” which is consistent with internal disorganization from misfit. Thus the results corroborate the theory that diversification leads to mis fit that leads to low per for mance and divisionalization. As we have just seen, while mis fit pro duces a neg a tive ef fect on or ga niza tional per for mance, struc tural change to a fit ting struc ture is not usually immediate. The pioneering empirical study shows that firms often remain in misfit for some time, consequently in curring performance loss (Chan dler 1962). Struc tural change into fit is caused by a cri sis of poor performance. Chandler (1962) documents how performance crises pre cip i tated by events such as eco nomic re ces sions were re quired

16

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

to trig ger cor po ra tions to adopt the di vi sional struc tures they needed in order to fit their diversification. This low level of performance results in part from the ef fect of the mis fit, but in part also from other causes that depress organizational performance (as will be discussed further in Chapter 9). Thus performance has to drop to a low level before the organization’s management takes the corrective action needed. Or ga ni za tions fail to make needed adap tive change un til their performance has deteriorated substantially, so that there is a clear problem to be solved. This is consistent with the theory of Simon (1976) that mana gers are boundedly rational (through incomplete knowl edge, etc.) so that, rather than max i mize, they satisfice. As long as the or ga ni za tion main tains a level of performance that is at or above the satisficing level (i.e., the level judged to be sat is fac tory), man a ge rial decision making is not engaged. Managerial de ci sion mak ing con sists of prob lem solv ing, which is ini ti ated when a prob lem oc curs, as man ifest in subsatisficing performance. It takes the form of looking for a solution good enough to return performance to the satisficing level. There fore an or ga ni za tion in mis fit adopts a new or ga ni za tional structure and moves into fit only when per for mance be comes low. In these ways the SARFIT model specifies the processes whereby contingency change causes struc tural change, add ing to the core par a digm. In summary, structural con tingency theory contains a core par adigm that gen er al izes across any con tin gency. There is an as so ci a tion be tween con tin gency and or ga ni za tional struc ture. There is or ga ni zational change, that is, con tin gency change caus ing change in the structure. And there is a fit of the struc ture to the con tin gency that pos i tively affects per for mance. The fit of struc ture to con tin gency ex plains why organizations change their structure in response to changing their contingency and thus why con tin gency and struc ture are as so ci ated.

Conceptual and Theoretical Integration of Structural Contingency Theory The con tin gency the ory of or ga ni za tional struc ture may be in te grated by stat ing that there are two main con tin gen cies, task and size, with the task con tin gency be ing com posed of task un cer tainty and task in ter depen dence. There are two main con tin gency the o ries of or ga ni za tional struc ture: or ganic the ory and bu reau cracy the ory. Task un cer tainty is

Core Par a digm and The o ret i cal In te gra tion

17

the main con tin gency of the or ganic the ory, with task in ter de pen dence play ing the role of a mi nor con tin gency. Size is the main con tin gency of bu reau cracy the ory with task in ter de pen dence, once again, play ing the mi nor role. There are a number of different structural contin gency theo ries, each connecting a par ticular contingency to some specific aspects of organizational structure (e.g., Blau 1970; Burns and Stalker 1961; Chan dler 1962; Child 1973a; Thomp son 1967; Wood ward 1965). However, many of the con tin gency and struc tural vari ables can be re duced to a few un der ly ing con structs that com pose those un der ly ing the o ries. Sim i larly, there is con sid er able sim i lar ity among many of the theories so that theoretical integration is possible. We will offer a conceptual and the o ret i cal in te gra tion of struc tural con tin gency the ory. This provides an over view of many of the is sues that will be dis cussed in de tail later, in or der to give the reader a map of the in tel lec tual ter rain to be nav i gated in the en su ing chap ters. We will be gin by show ing how many different contingency variables can be reduced to a few underlying con structs. In the sub se quent sec tion we shall pres ent two the o ries, the organic and the bureaucracy theories, that, between them, capture much of the contingency the ory of or ga ni za tional struc ture. Contingencies Con tin gencies of or ga ni za tional struc ture in clude some that are within the or ga ni za tion and some that are out side it. Con tin gency fac tors such as task uncertainty and task in ter de pen dence are aspects of the work be ing per formed and so lie in side the or ga ni za tion. Or ga ni za tional size is how many peo ple are in the or ga ni za tion and so is also an in ter nal organizational char ac ter is tic. Other con tin gency fac tors are char ac teris tics of the en vi ron ment, such as en vi ron men tal un cer tainty. How ever, they af fect the in ter nal con tin gen cies, which in turn shape other in ternal or ga ni za tional char ac ter is tics, for ex am ple, or ga ni za tional struc ture. For in stance, environmental uncertainty affects task uncertainty, which causes the adoption of an or ganic struc ture. Thus, struc ture is caused by need ing to fit the intraorganizational con tin gen cies, some of which are, in turn, caused by the en vi ron men tal contingencies. In this way, the or ga ni za tion is shaped by the need to fit its en vi ron ment. En vironmental contingencies indirectly shape the organization through the intervening variables of the intraorganizational contingencies.

18

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

Thus, while en vi ron men tal con tin gen cies may be more ul ti mate causes of organizational structure, the intraorganizational contingencies through which they work are the more immediate and di rect causes of struc ture, so they are em pha sized in this book. Research into organizational structure has identified a number of contingencies. Task un cer tainty (Gresov 1990), tech nol ogy (Woodward 1965), in no va tion (Hage and Aiken 1967a), en vi ron men tal change (Child 1975), tech no log i cal change (Burns and Stalker 1961), size (Blau 1970), pros pec tor strat egy, de fender strat egy (Miles and Snow 1978), di ver sifi ca tion, ver ti cal in te gra tion (Rumelt 1974), and task in ter de pen dence (Thompson 1967) are some of the better-established contingencies. These con tin gency fac tors can be re duced to a few com mon, un der lying con cepts. We shall ar gue that there are three un der ly ing con tin gencies: task un cer tainty, task in ter de pen dence, and size.

Task Un cer tainty The set of contingencies made up of task uncertainty, tech nology, tech no log i cal change, innovation, and en vi ron men tal in sta bil ity have an un der ly ing con cept of uncertainty. En vi ron men tal and tech no log ical change lead to un cer tainty for the or ga ni za tion and its man ag ers, creating uncertainty in the tasks conducted inside the organization. This task uncertainty is reinforced by the need for innovation that is part of the response to environmental and technological change (Burns and Stalker 1961; Hage and Aiken 1970). The tech nol ogy used by the organization to transform its inputs into outputs also reflects dif fer ences in task un cer tainty and feeds back to af fect task un cer tainty (Wood ward 1965). More ad vanced tech nol ogy re quires greater pre dictability of the tasks, while also in creas ing task pre dict abil ity (Wood ward 1965). Thus task un cer tainty is the core con cept un der ly ing this set of con tin gency fac tors. More over, the dis tinc tion be tween de fender ver sus prospector strategy (Miles and Snow 1978) is, in large degree, that between cost re duc tion through rou tine op er a tions ver sus in no va tion, and so it also re lates to the un der ly ing con tin gency of task un cer tainty. Hence a parsimonious treatment of contingencies can reduce these contingency vari ables to task un cer tainty, for many pur poses.

Core Par a digm and The o ret i cal In te gra tion

19

Task In ter de pen dence A second set of contingency factors, including as pects of strategy, clusters around the underlying contingency of task interdependence. Task inter de pen dence clas si fies in what way activities in an organiza tion are con nected with each other: pooled (in di rect con nec tion only), sequential (di rect, one-way con nec tion), and re cip ro cal (di rect, twoway con nec tion; Thomp son 1967). Some of the ma jor con tin gen cies of strat egy can be sub sumed un der task in ter de pen dence. Di ver si fi ca tion— whether of products, ser vices, or customers—is an aspect of the strategy of a firm (Rumelt 1974). It is achieved strategy that is man ifested in a set of concrete activities, rather than be ing strat egy as an intention. Sim i larly, ver ti cal in te gra tion is an as pect of achieved strat egy (Rumelt 1974). Diversification and vertical integration describe how far the ac tiv i ties of a firm are closely con nected, or not, in the hor i zon tal (diversification) and vertical (vertical integration) dimensions, respectively. Diversification and vertical in tegration may therefore both be grouped under the concept of task interdependence. High diversification, such as in a firm that makes unrelated prod ucts, is pooled in ter de pen dence among the prod ucts (Thomp son 1967). Ver tical in te gra tion is se quen tial in ter de pen dence be tween the stages of the value-added chain within the firm. An un di ver si fied firm, such as one mak ing a sin gle prod uct, will have close con nec tions among the functional de part ments be cause all are in volved with the same prod uct, so that its in ter de pen dence is also se quen tial. How ever, there may also be reciprocal in ter de pen dence among the func tional de part ments, if there is innovation, requiring in teraction between the research and other departments (Lorsch and Law rence 1972). Hence the strat egy con tin gencies of diversification and vertical in tegration can be subsumed under task in ter de pen dence, as can the in ter ac tions be tween func tions stem ming from in no va tion. Sup port for the idea that many con tin gen cies are re duc ible to task uncertainty and task interdependence may be taken from Dess and Beard (1984). They re view a num ber of con cep tu al iza tions of en vi ronmen tal con tin gen cies and ar gue that they can be re duced to three di men sions: dy na mism (which sub sumes sta bil ity-in sta bil ity and turbulence), complexity (which subsumes homogeneity-heterogeneity and concentration-dispersion), and munificence (which subsumes

20

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

capacity). They show that twenty-three en vi ron men tal vari ables that com pare across fifty-two dif fer ent in dus tries re duce in a fac tor anal y sis to the three un der ly ing fac tors of dy na mism, com plex ity, and mu nif icence. Several authors have identified dy namism as one of the major environmental contingencies of organizations (Child 1975; Duncan 1972; Thomp son 1967). Dess and Beard (1984) em pha size that dy na mism is not sim ply the rate of change, which it self could be con stant, thereby rendering the environment pre dict able, but rather the de gree of un pre dict abil ity. As they state, “Dy na mism should be re stricted to change that is hard to predict and that heightens uncertainty for key organizational members” (p. 56). This confirms the importance of uncertainty as a key element of dynamism. Therefore dy namism can largely be subsumed under un cer tainty. Sim i larly, the homogeneityhet er o ge ne ity as pect of their en vi ron men tal com plex ity con cept relates to the de gree of di ver si fi ca tion of the or ga ni za tion, be cause di ver sifi cation across diverse product-markets renders the environment more heterogeneous. Given that diversification relates to task interdependence, en vi ron men tal com plex ity re lates, in turn, to the task in terde pen dence con tin gency. Thus two of the en vi ron men tal con tin gency con cepts of Dess and Beard, dy na mism and com plex ity, map onto the intraorganizational con tin gen cies of task un cer tainty and task in terdepen dence, re spec tively. The third environmental con tingency con cept of Dess and Beard (1984), munificence, re lates to the quan tity of slack re sources, which we discuss below (in Chapter 9) mainly as a mod erator, whereby eco nomic stringency fosters organizational change. Similarly, the con centration-dispersion aspect of their complexity concept taps the de gree of com pe ti tion in the en vi ron ment of an or ga ni za tion, which again moderates organizational change (as will also be discussed in Chap ter 9). Thus the three di men sions iden ti fied by Dess and Beard in their par si mo ni ous model of the en vi ron ment are con sis tent with our pro posal that task un cer tainty and task in ter de pen dence are ma jor under ly ing con tin gen cies and that com pe ti tion and lack of mu nif i cence moderate the organizational change that occurs in response to these two con tin gen cies. Sim i larly, Law rence and Dyer (1983) ar gue that indus trial en vi ron ments form two di men sions, in for ma tion com plex ity, which taps un cer tainty stem ming from vari a tions, and re source scar city; these two dimensions re late to the task un cer tainty con tin gency and re source mu nif i cence mod er a tor be ing used herein.

Core Par a digm and The o ret i cal In te gra tion

21

Thus many of the con tin gency vari ables can be clas si fied un der ei ther task un cer tainty or task in ter de pen dence. Clearly, these two con cepts can them selves be sub sumed un der the con cept of a task con tin gency. Size A dif fer ent con tin gency is size. Size has turned out to be a ma jor contingency factor that affects many different aspects of structure and many of them quite strongly (Blau 1972; Child 1973a; Pugh, Hickson, Hinings, and Turner 1969). The size contingency is the number of organizational mem bers who are to be or ga nized (Blau 1970), de ter mining the structure that is required. Size is therefore appropriately operationalized in empirical stud ies by the number of employees (Pugh et al. 1969; Pugh and Hinings 1976). However, the number of em ploy ees is con cep tu ally and em pir i cally closely re lated to other as pects of or ga ni za tional mem ber ship, such as the num ber of mem bers in a la bor union (Donaldson and Warner 1974). The num ber of employ ees is also of ten closely cor re lated with other as pects of the scale of an or ga ni za tion, such as sales or as sets, so that these vari ables may be used as in di ca tors for size (see Child 1973a, p. 170, Ta ble 1; Donaldson 1996b, pp. 147-156; Hopkins 1988; Lioukas and Xerokostas 1982). How ever, they are not al ways highly cor re lated, so that they are, at best, mere proxi es for the num ber of em ploy ees, which re mains the op er a tional mea sure of size. Or ganic and Bu reau cracy The ories There are two contrasting contingency the ories of organizational struc ture that are in flu en tial in the struc tural con tin gency lit er a ture: the or ganic and the bu reau cracy the o ries. Each the ory has a dif fer ent model of or ga ni za tional struc ture that needs to be dis tin guished. Or ganic The ory Or ganic the ory con sid ers the fun da men tal di men sion of or ga nizational struc ture to be a con tin uum that runs from the poles of mech anistic to or ganic struc ture (Burns and Stalker 1961). The mech a nis tic struc ture is top-down so that top man age ment seeks to con trol lowerlevel employees in every way possible. Top management cen tralizes

22

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

decision mak ing, so that it makes de ci sions about what should hap pen (Burns and Stalker 1961). It also gives sub or di nates de tailed job as signments that de fine their re spon si bil i ties (Burns and Stalker 1961). Fur ther, top man age ment lays down rules that em ploy ees are to fol low and doc u ments they must use, such as forms that they must fill out (Weber 1968). Thus the mech a nis tic struc ture can be de fined as cen tral ized in de ci sion mak ing, spe cial ized in roles and for mal ized (much use of rules and documents; Hage 1965; Pennings 1992). In con trast, the or ganic structure is decentralized (so that lower-level employees exercise au ton omy in de ci sion mak ing) and also low on func tional spe cial iza tion and formalization, so that how employees should do their jobs is not prescribed by top management (Burns and Stalker 1961). Thus the organic structure nei ther con trols employees through centralization nor through func tional spe cial iza tion and for mal iza tion. In stead, the or ganic struc ture re lies on the ini tia tive and ex per tise of mid dle-level and lower-level employees. Hence in the organic model of structure, lower-level em ploy ees are only lightly con trolled by their or ga ni za tion. In sum, the mech a nis tic struc ture is cen tral ized and high on both functional spe cial iza tion and for mal iza tion, whereas the or ganic struc ture is de cen tral ized and low on both func tional spe cial iza tion and for maliza tion (Hage 1965; Pennings 1992). The con trast be tween mech a nis tic and or ganic struc tures de fines a single structural di mension. Thus organizations that are centralized are also specialized and formalized (i.e., mechanistic). Con versely, if the organization is de centralized, it will also be low on specialization and formalization (i.e., organic). Mechanistic and organic are two poles of a con tin uum of mech a nis tic and or ganic, with or ga ni za tions distributed at points along that con tin uum. But if an or ga ni za tion is mid dling on cen tral iza tion, then the or ga ni za tion will also be likely to be middling on specialization and formalization, because all three struc tural vari able go to gether as part of the same di men sion. Ac cord ing to the organic the ory, the mech a nis tic-or ganic struc ture fits the con tingency of task uncertainty (Hage and Aiken 1969). Specifically, the mecha nis tic struc ture fits low task un cer tainty, whereas the or ganic struc ture fits high task un cer tainty. Much of the task un cer tainty co mes from the environment of the or ganization, caused by high levels of technological and mar ket change re quir ing the organization to in novate in or der to re main ef fec tive and com pet i tive, that is, to have high per for mance (Burns and Stalker 1961). Where task un cer tainty is low,

Core Par a digm and The o ret i cal In te gra tion

23

the knowl edge and in for ma tion pos sessed by se nior man ag ers en ables them to ex er cise high con trol over all op er a tions and em ploy ees so that the or ga ni za tion is ef fec tive. How ever, where task un cer tainty is high, much expertise and information is distributed among employees, so that they have to be em pow ered to use their ini tia tive and make de ci sions in a participatory manner, in order for the organization to innovate and be ef fec tive (Burns and Stalker 1961). In creasing task un cer tainty over time re quires an or ga ni za tion to change its struc ture from mech anis tic to ward or ganic in or der to main tain fit (Burns and Stalker 1961; Hage and Aiken 1970). Thus the struc tural vari ables of cen tral iza tion, specialization, and formalization go together be cause high levels of each pro vide the greater lev els of hi er ar chi cal di rec tion that fit highly certain tasks, while low levels of each also go together because they provide the participation that fits highly uncertain tasks. In this way arises the con cept of the or ga ni za tional struc ture as a sin gle di men sion from mech a nis tic to or ganic. This struc tural con cept has been used in much contingency theory research (Burns and Stalker 1961; Dewar and Hage 1978; Hage 1965, 1974, 1980, 1988; Hage and Aiken 1967a, 1967b, 1969; Hage and Dewar 1973). Bu reau cracy The ory In con trast, bu reau cracy the ory con sid ers the fun da men tal di men sion of organizational structure to be a different continuum, which runs from the poles of unbureaucratic, that is, sim ple, struc ture to bu reau cratic struc ture. There ex ists a sim ple type of or ga ni za tional struc ture that is cen tral ized as well as be ing low on func tional spe cial iza tion and formalization (Mintzberg 1979). Conversely, there is a bureaucratic type of organizational structure that is decentralized as well as being high on functional specialization and formalization (Child 1972a; Weber 1968). Thus top man age ment ei ther con trols em ploy ees di rectly through making de ci sions (cen tral iza tion) in the sim ple struc ture, or indirectly through tight job definitions (specialization) and rules (formalization) in the bureaucratic structure (Blau and Schoenherr 1971). Top man age ment sub sti tutes be tween such di rect and in di rect controls, b ut does not try to use high levels of both simultaneously. Thus top man age ment is ap ply ing suf fi cient con trol, rather than max imiz ing con trol. There is a con tin uum of de grees of bu reauc ra ti za tion, and an or ga ni za tion can lie at any point along it (Child 1972a).

24

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

A sim ple struc ture, while be ing low on spe cial iza tion and for malization like the or ganic struc ture, is, nev er the less, not or ganic be cause the sim ple struc ture is cen tral ized, rather than be ing de cen tral ized like the organic structure. Similarly, a bureaucratic organization, though highly specialized and formalized like the mechanistic structure, is decentralized and so is not mechanistic. Thus the or ganic and bu reau cratic mod els of or ga ni za tional struc ture dif fer ac cord ing to their view about the dimension that underlies organizational structure. This dimension is composed of the two structural el ements of cen tral ization and spe cial iza tion-for mal iza tion. These el e ments can be com bined in two dif fer ent ways to yield the or ganic and bu reau cracy the o ries (see Fig ure 1.3). Or ganic the ory sees cen tral iza tion as pos i tively cor re lated with specialization-formalization. Low centralization goes with low spe cial iza tion-for mal iza tion in the or ganic struc ture. High cen tral iza tion goes with high spe cial iza tion-for mal iza tion in the mech a nis tic struc ture (Hage 1965; Pennings 1992). In contrast, bureaucracy theory sees central iza tion as neg a tively cor re lated with specialization-formalization. High centralization goes with low specialization-formalization in the simple structure. Low centralization goes with high specializationfor mal iza tion in the bu reau cratic struc ture (Child 1972a). According to bureaucracy the ory, the level of bureaucratization of the struc ture fits the con tin gency of size (i.e., the num ber of or ga ni za tional em ploy ees; Child 1975). Spe cifically, a low level of bu reaucrati za tion fits a small or ga ni za tion. Thus, for a small or ga ni za tion, the fit ting struc ture is high on cen tral iza tion and low on spe cial iza tion and for mal iza tion (i.e., a sim ple struc ture). The lim ited com plex ity of de cision mak ing, re sult ing from small size, al lows top man age ment to di rectly con trol the orga ni za tion through making the deci sions. Conversely, a high level of bureaucratization (i.e., a bu reaucratic structure) fits a large or ga ni za tion. Thus, for a large or ga ni za tion, the fit ting struc ture is low on cen tral iza tion and high on spe cial iza tion and formalization. Increasing complexity and a tall hierarchy, re sulting from size, re quires top man age ment to del e gate many de ci sions. Also, size fos ters the di vi sion of la bor, and the re cur rent na ture of many de cisions al lows them to be for mal ized. In creasing size of an or ga ni za tion requires that its struc ture change from sim ple to ward bu reau cratic, to main tain fit and ef fec tive ness. There is some differ ence be tween the or ganic and bu reau cratic the o ries in their con cept of de cen tral iza tion. Bu reau cracy the ory sees

Core Par a digm and The o ret i cal In te gra tion

25

High

Simple

Mechanistic

Low

Organic

Bureaucratic

Low

High

Centralization

SpecializationFormalization

Fig ure 1.3. Or ga ni za tional Struc tures in Or ganic and Bu reau cratic The ory

increased de centralization as be ing mainly delegation of au thority down to mid dle man ag ers, with some oc ca sional del e ga tion to work ers (Chandler 1962; Child 1973a). Organic theory includes managerial del e ga tion, but also au ton omy and par tic i pa tion in de ci sion mak ing by tech ni cal ex perts at low lev els in the hi er ar chy (Hage 1980) and even by shop-floor work ers (Wall, Corbett, Mar tin, Clegg, and Jack son 1990). Thus the real ex tent of de cen tral iza tion is greater un der or ganic the ory than un der bu reau cracy the ory. Sim i larly, the or ganic struc ture in cludes also lateral forms of coordination such as cross-functional project teams and ad hoc com mu ni ca tion (Law rence and Lorsch 1967). These nonhierarchical co or di na tion mech a nisms pro vide ad di tional fo rums in which par tic i pa tion oc curs. Organic theory sees the degree to which or ganizational structures are organic as being driven by the need to fit not only the situational con tin gency of task un cer tainty but also the hu man needs and as pi ra tions of or ganizational members (Lorsch and Morse 1974). Thus the argument about replacing mechanistic with organic structures is ad vo cat ing more par tic i pa tory struc tures that not only of fer ef fec tiveness ben e fits through fit but also ac cord better with hu man is tic val ues. This cre ates a ten sion with the bu reauc ra ti za tion the ory that tends to ar gue the need for bu reauc ra ti za tion ac cord ing to a Weberian logic of subjugation of the organizational member in pursuit of ef fectiveness (Weber 1968).

26

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

The organic and bureaucracy theories tend to dif fer in their views about the di rec tion in which or ga ni za tional struc tures are headed over time. Organic theory sees the trend as being for task uncertainty to increase, be cause of in creases in sci en tific knowl edge and in no va tion rates, so that or ga ni za tions in crease over time the de gree to which their structures are organic (Burns and Stalker 1961; Hage 1988). In con trast, bureaucracy theory sees a tendency toward increasing levels of bu reauc ra ti za tion, be cause of in creas ing con cen tra tion in fewer, larger or ga ni za tions of global reach, fa cil i tated by mod ern com mu ni ca tions technologies (Weber 1968). In a formal sense both theories logically imply that an organization could move count er to the trend, if its contingencies changed in the opposite di rection, such as task un certainty reducing or size de creas ing. Clearly a num ber of or ga ni za tions have down sized in the 1980s and 1990s, so that there could be said to be some thing of a countertrend away from growth in cer tain quar ters (Budros 1997; Cascio 1993; Littler and Bramble 1995). Nevertheless, bureaucracy theory tends to see increasing use of the bureaucratic struc ture (Blau and Meyer 1987), while organic the ory de picts in creasing use of the or ganic struc ture (Burns 1963; Flan ders, Pomeranz, and Woodward. 1968; Hage 1974, 1980, 1988). Thus organic theory fore sees decreasing specialization and formalization, while bureaucracy the ory foresees them increas ing. Both the o ries foresee in creas ing de cen tral iza tion, though the ex tent is greater for organic than for bu reau cratic theory. Hence organic and bu reau cracy theo ries are op timis tic or pes si mis tic, re spec tively, about the like li hood of in creasing employee participation and freedom from regulation. Thus the divergence be tween the organic and bureaucracy theories con nects with different pol icy prescriptions, value positions, and predictions about the fu ture that help to an i mate the de bate between them. As just seen, or ganic and bu reau cracy the o ries are to some de gree in conflict, making ri val analyses about structural trends. This conflict be comes played out as a con tro versy in the struc tural con tin gency lit erature about the role of task versus the size contingency. Or ganic theory seeks to promote task as the major con tingency, so that task strongly and per va sively shapes or ga ni za tions, lead ing them to take on the organic structure, espe cially as task uncertainty in creases (e.g., Burns 1963; Hage 1988). Bu reau cracy the ory seeks to show, in con trast, that size is the more im por tant con tin gency, in terms of the num ber

Core Par a digm and The o ret i cal In te gra tion

27

of structural aspects it affects and the strength of the relationships, lead ing or ga ni za tions to take on the bu reau cratic struc ture as their size increases (Child 1973a). At the extremes, bureaucracy theory argues that the task has only lim ited ef fects (Child and Mansfield 1972) or that its effects are the opposite to those that organic theory asserts (Blau, Falbe, McKinley, and Tracy 1976). These controversial issues will be dis cussed in the en su ing chap ters (es pe cially Chap ters 3 and 5). Thus far we have con trasted the or ganic and bu reau cracy the o ries on the struc tural el e ments of cen tral iza tion andspecialization-formalization. How ever, the con cept of bu reau cracy is wider than just these el e ments and entails two other aspects of structure: structural differentiation and divisionalization. Structural differentiation refers to the extent to which the organization is split into separate parts, both horizontally (the num ber of di vi sions, the num ber of job ti tles, the span of con trol of the CEO, etc.) and ver ti cally (the num ber of lev els in the hi er ar chy; Blau 1970). Con cep tually, a bu reau cratic struc ture has a high level of struc tural dif fer en ti a tion, for ex am ple, many de part ments and hi er archi cal lev els (Blau and Schoenherr 1971; Pugh, Hickson, Hinings, and Turner 1968; Weber 1968). Divisionalization can also be thought of as an other as pect of bu reau cracy, be cause it fea tures decentral ization and in creased func tional spe cial iza tion and for mal iza tion (Chenhall 1979; Grinyer and Yasai-Ardekani 1981). Divisionalization also correlates positively with struc tural dif fer en ti a tion, both horizontally (span of con trol of CEO) and ver ti cally (hi er ar chi cal lev els; Grinyer and YasaiArdekani 1981), thereby ce ment ing the con nec tion be tween division alization and bu reau cratic struc ture. The in clu sion of struc tural dif ferentiation and divisionalization un der the con cept of bu reau cracy is a fur ther way in which it dif fers from the or ganic struc tural con cept. Task interdependence is a minor con tingency of both the organic and the bu reau cratic struc tures. As seen, task un cer tainty is the main con tin gency de ter min ing whether struc ture is mech a nis tic or or ganic. However, task interdependence determines whether coordination mechanisms are of the mech a nis tic or or ganic type, so that task in terde pen dence is a sec ond ary con tin gency of or ganic struc ture. Ac cord ing to Thomp son (1967), where task in ter de pen dence is pooled, then the fit is stan dard iza tion, that is, rules and pro ce dures set by the hi er ar chy, that is, a mech a nis tic co or di na tion mech a nism. Where task in ter de pen dence is se quen tial, then the fit is plan ning by the hi er ar chy, that

28

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

is, another mechanistic coordination mechanism. In contrast, where task interdependence is reciprocal, then the fit is mutual adjustment between organizational members, that is, an organic coordination mech a nism (Thomp son 1967). Hence lower (i.e., pooled and se quen tial) task interdependence re quires mech a nis tic co or di na tion mech a nisms, whereas high (i.e., reciprocal) task interdependence re quires organic coordination mechanisms. Com bining both task contingencies, low task uncertainty and low task in ter de pen dence are fit ted by mech anistic structures, while high task uncertainty and high task inter dependence are fit ted by or ganic struc tures. The task interdependence contingency also affects bu reaucratic struc ture in that task in ter de pen dence af fects divisionalization. Increasing diversification de creases the task in ter de pen dence be tween prod uct-mar kets, so that the fit ting struc ture is di vi sional rather than functional (Chan dler 1962; Rumelt 1974). Given that division alization is a minor component of overall bureaucratic structure, then task interdependence is a minor contingency of bu reaucracy. Task inter dependence is a mi nor con tin gency rel a tive to size, which strongly affects many of the other as pects of bu reau cratic struc ture (e.g., formalization). Over all, in creas ing task in ter de pen dence leads from divisions co or di nated mech a nis ti cally to functions co ordinated or gan i cally. Thus, whereas or ganic and bu reau cracy the o ries dif fer in em phasiz ing the task uncertainty and size contingencies, re spectively, both theories can be ex tended in a way that incorporates task interdepen dence as a contingency, thereby constituting a commonality be tween the organic and bu reau cratic mod els of or ga ni za tional struc ture. Using the three contingencies and the two theories, we can crossclas sify them, as shown in Ta ble 1.1. The rows are the three con tin gen cies: task un cer tainty, task in ter de pen dence, and size. The col umns are the two the o ries: or ganic and bu reau cracy. The cross-clas si fi ca tion yields six cells and into each cell is placed the names of some of the major au thors. Syn the sis of Or ganic and Bu reau cracy The ories It is pos si ble to syn the size the or ganic and bu reau cracy the o ries in an in te grated model of or ga ni za tional struc ture. This shows that struc tural con tin gency the ory, while it con tains ten sions (as dis cussed above), is

Core Par a digm and The o ret i cal In te gra tion

29

TA BLE 1.1 Key Au thors in Struc tural Con tin gency The ory Re search

CONTINGENCIES

THEORIES Organic

Task Uncertainty

Burns & Stalker Hage Lawrence & Lorsch Perrow Woodward

Task Interdependence

Lawrence & Lorsch Thompson

Size

Bureaucracy

Chandler Grinyer & Ardekani A ston Group Blau

not frag mented or in cho ate. Com bining the size and task con tin gen cies we can say the fol low ing (see Fig ure 1.4). As or ga ni za tions grow in size they increase specialization-formalization, structural differentiation, and decentralization. As task in ter de pen dence de creases because of diversification, this causes divisionalization, so that decentralization in creases, as does spe cial iza tion-for mal iza tion (e.g., cre ation of profit reporting) and struc tural dif fer en ti a tion (e.g., the number of levels), be yond that which would ex ist for its size alone. Con versely, in creas ing task in ter de pen dence leads to a func tional struc ture, which im plies less specialization-formalization, structural differentiation, and decentral ization, as shown by the negative effects of task interdependence on these three struc tural vari ables in Figure 1.4. As task un cer tainty inc reases, for example, through increased innovation in products or ser vices, there is a re duc tion in for mal iza tion and an in crease in de cen traliza tion. The re duc tion in for mal iza tion from in creas ing task un cer tainty off sets to a de gree the in crease in formalization com ing from size in crease. However, the re duc tion in formal ization from in creas ing task un cer tainty reinforces the decrease in for mal iza tion that co mes from any inc rease in task in ter de pen dence. The de cen tral iza tion from in creas ing

30

THE CON TINGENCY

Contingencies

THEORY OF OR

GANIZATIONS

Organizational Structure SpecializationFormalization

Structural Differentiation

Decentralization

Size

+

+

+

Task Interdependence

-

-

-

Task Uncertainty

-

+

Fig ure 1.4. Over all Con tin gency The ory Model of Or ga ni za tional Struc ture

task un cer tainty adds to that com ing from in creas ing size. How ever, the in crease in de cen tral iza tion from in creas ing task un cer tainty off sets to a de gree the de crease in decentralization com ing from any in crease in task in ter de pen dence. As will be seen below (Chap ters 2 and 3), much of the re duc tion in formal iza tion and in crease in de cen tral iza tion re sult ing from task uncertainty is lo cal ized in spe cific parts of the or ganization, but for the present we are dis cuss ing over all struc tural lev els of the or ga ni za tion as a whole. Syn the sizing the bu reau cratic and or ganic the o ries in this way gives a sense that it is possible to integrate structural contingency re search into a model that is the o ret i cally co her ent and only mod estly com plex. However, t he view of the two the o ries is highly con densed and so requires a great deal of ex pla na tion and jus ti fi ca tion re gard ing the the ory and the supporting empirical research. We will consider first the organic the ory and its sup port ing re search (in Chap ter 2) and sec ond, bu reau cracy the ory and its re search (in Chap ter 3). While task un cer tainty and size are treated as independent con tingencies in the literature, the possibility nevertheless arises that they may be brought together causally, so that one causes the other. This would constitute a causal synthesis in structural contingency theory, which would further increase its in tegration. A de rived question is whether this causal in te gra tion would lead to the o ret i cal in te gra tion of the or ganic and bu reau cratic the o ries. These is sues will be discussed in Chap ter 3.

Core Par a digm and The o ret i cal In te gra tion

31

Plan of the Book In Chap ter 2 we dis cuss or ganic the ory and its sup port ing em pir i cal research. In Chapter 3 we discuss bu reaucracy the ory and its supporting empirical re search. We also attend to the is sue of rec on cil i a tion between the or ganic and bu reau cratic the o ries in two ways. A model that syn the sizes both the or ganic and the bu reau cratic the o ries is presented. The pos si bil ity of a causal re la tion ship between task and size, the two ma jor con tin gen cies of or ganic and bu reau cratic the o ries, is then discussed. In Chapter 4 we consider the prevailing causal models among size and the bureaucratic structural variables. These are seen to be prob lem atic from the view point of con tin gency the ory. They are re for mu lated so as to have causality work through fit, as contingency theory holds. Also discussed are reasons why some size and structural re lation ships are curvilinear rather than lin ear. A fur ther is sue that re ceives attention is the relationship between the systems functionalism of contingency the ory and the ac tions by peo ple that cre ate cau sa tion at the sys tems level. In Chapter 5 we consider some of the con troversies in and around contingency theory that reflect the ongoing tension be tween the or ganic and bu reau cratic the o ries. The ri val claims of these two the o ries re gard ing the tech nol ogy and size con tin gen cies have led to con tro versy about their rel a tive strengths as de ter mi nants of struc ture and also to what is the ef fect of tech nol ogy on struc ture. These is sues will be crit ically dis cussed. The idea that bu reau cracy is made in ev i ta ble be cause of de termination by contingencies such as size has been considered repellant enough by some schol ars to lead to ques tion ing of de ter minism. Critics have as serted that there is stra te gic choice, open ing up the possibility of choice in fa vor of more par tic i pa tory, or ganic struc tures. The ar gu ments for choice will, in turn, be crit i cally ex am ined. Again, the causal theory that contingencies cause structure has been ques tioned by pro pos ing that con tin gency is not a cause but just a cor re late of struc ture, or that struc ture causes con tin gen cies. Ei ther way again escapes from structures, such as bureaucracy, as be ing inevitable, so opening the door to a choice of alternatives, such as participatory,

32

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

organic structures. Therefore we shall need to critically examine the issues of cau sal ity and re verse cau sal ity. Con cepts and variables from con tin gency the ory re search are some times used within the frame work of con fig u ra tions. This pos its that there are only a few types and that they need not fit any contingencies, so removing contingency de terminism. Moreover, organizational change is held to be problematic and there fore in fre quent and sharply dis con tin u ous. We will crit i cally ex am ine con fig u ra tions. There is a stream of con tin gency the ory concerned with or ga ni za tional power, which has at tracted some crit i cism, so this is also dis cussed in Chap ter 5. In Chapter 6 we examine ex planations alternative to contingency theory that come from other organizational the ories such as in sti tutional, organizational economics, political, and population ecology the o ries. We show that con tin gency the ory of fers a sound ex pla na tion de spite the chal lenge posed by these the o ries. A cen tral idea in con tin gency the ory is that the fit of or ga ni za tional char ac ter is tic to the con tin gency fac tor leads to higher or ga ni za tional performance. The ques tion thus arises as to whether this is just a tautol ogy. This raises the is sue of what is a fit, how to iden tify it, and how to show em pir i cally that it raises per for mance. These philo soph ical, con cep tual, and meth od olog i cal is sues are dis cussed in Chap ter 7. In Chapter 8 we review stud ies of the relationship be tween fit and performance. We show that there is a body of em pir i cal work that val idates the idea that con tin gency fits pos i tively af fect performance. However, there are also technical problems that can obscure the strength of the ef fects of fit on per for mance and so need to be cor rected in fu ture re search. The dis cus sion iden ti fies eight les sons that can help make fu ture re search more valid and re veal more fully the im por tance of con tin gency fit. In Chap ter 9 we iden tify some the o ret i cal prob lems within con tingency the ory and sug gest how they may be over come. This new the ory con struc tion makes con tin gency the ory more co her ent. The pro posed the o ret i cal re for mu la tion also makes con tin gency the ory more dynamic, through the concept of a disequilibrium theory, so that it becomes a more com pre hen sive the ory of or ga ni za tional ad ap ta tion and growth. A cen tral idea is that or ga ni za tional change is per for mance driven, and this insight is formalized through organizational port folio theory. Again, prob lems in struc tural ad ap ta tion to changes in the con tin gency are addressed through the concept of quasi-fit. Furthermore, it is

Core Par a digm and The o ret i cal In te gra tion

33

suggested that the concept of fit itself may be improved through reconceptualizing fit in a way that is con sis tent with the model of organizational adaptation dynamics that is being used in this book. Some of the new theorizing in this chapter draws upon ideas from economics and finance and so benefits from cross-fertilization from those dis ci plines. Finally, in Chap ter 10 we iden tify some pos si ble fu ture op por tu ni ties and chal lenges for con tin gency the ory in or ga ni za tional sci ence. The pre ced ing the o ret i cal dis cus sion is drawn upon to sug gest hy poth e ses that might be used by schol ars in fu ture or ga ni za tional re search. The new concepts and theories pre sented as part of neo-contingency theory, such as disequilibrium, organizational portfolio theory, and quasi-fit, are dis cussed and sug ges tions are made on how to test them in future em pir i cal re search. A new op er a tional def i ni tion of the ef fect of fit on or ga ni za tional per for mance is also pre sented as a guide to fu ture in ves ti ga tion into this key topic.

THE Or ganic CON The TIN oryGENCY and Re search THEORY OF OR GA NI ZA TIONS

2

Organic Theory and Research

I

n the pre vi ous chap ter we in tro duced the or ganic the ory of or ga nizational structure. The organic the ory states that the mechanistic (centralized, formalized) structure fits situations of low task un certainty, whereas the organic struc ture (de centralized, unformalized) fits situations of high task uncertainty, such as innovation. In this chap ter we pres ent the the ory in greater depth to gether with the em piri cal re search. The studies that orig inated the organic the ory of organizational struc ture mostly in volved field stud ies of ac tual or ga ni za tions, so that this branch of struc tural con tin gency the ory had a strong ground ing in empirical re ality at its in ception. This chapter will review the main stud ies and then bring out their theoretical connections to pro vide a discussion of how closely they can be in te grated into a co her ent the oret i cal model. It will be seen that or ganic the ory con sti tutes a co her ent the ory that is sup ported by em pir i cal re search.

35

36

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

Organic Theories and Research The point at which the contingency the ory ap proach was invented can not be stated with any ex ac ti tude. The first the ory of or ga ni za tional structure to use the term contingency theory is that by Lawrence and Lorsch in 1967 (p. 156), which made an early contribution to organic theory. This contribution in macro-Organizational Behavior (OB), however, was at least simultaneous with a contribution in micro-OB that bore the title of a contingency the ory of leadership by Fiedler (1967). This in turn was pre dated by re search of a con tin gency type, in which social psychologists conducted exper i ments on decision making in groups. The ex per i ments con trasted groups us ing a hi er ar chical struc ture of com mu ni ca tions with those us ing a participa tory structure (Leavitt 1951). Where the task involved in novatory prob lem solv ing, the par tic i pa tory struc ture pro duced su pe rior de cisions, but where the task was routine, the hierarchy was superior. Hence task un cer tainty was a con tin gency fac tor that mod er ated the effectiveness of group structures. Thus the idea of contin gency re lationships was established in leadership and group studies by the late 1960s and this may have con di tioned its ac cep tance in or ga ni za tional struc tural re search. More spe cif i cally, the so cial psy cho log i cal ex per iments on group de ci sion mak ing con tained within them the key idea of or ganic the ory: Tasks low on un cer tainty are most ef fec tively man aged hi er ar chi cally and tasks high on uncer tainty are most ef fec tively man aged participatorily. Thus ef fec tive group struc ture var ies from hi er ar chi cal to par tic i pa tory and each fits a level of the con tin gency of task un cer tainty. The organic the ory of organizational structure was laid down in a series of pi o neer ing con tri bu tions. They can be di vided into two sets: those that related organizational structure to the contingency of task uncertainty, often as influenced by the environment, and those that related or ga ni za tional struc ture to the con tin gency of tech nol ogy. Task Un cer tainty There are three key contributions that show how organizational struc ture re lates to the con tin gency of task un cer tainty and the en vi ron ment of the or ga ni za tion: Burns and Stalker (1961), Hage (1965), and Lawrence and Lorsch (1967).

Or ganic The ory and Re search

37

Burns and Stalker The most fun da men tal con tri bu tion to or ganic the ory was pro vided by Burns and Stalker (1961), who offered an el egant the ory that has proved to be compelling for many subsequent scholars. They dis tinguish be tween two or ga ni za tional struc tures, the mech a nis tic and the or ganic (Burns and Stalker 1961, pp. 119-122). The mech a nis tic structure emphasizes hi erarchy, with the task of the organization being divided into specialized roles whose occupants remain dependent upon their superordinates, who retain much of the knowledge and information. The im age of a mech a nis tic or ga ni za tion struc ture is of a hi er archy with centralized de cision making that tightly prescribes lower-level roles. In contrast, in the organic struc ture, un der stand ing of the task is widely shared among em ploy ees who use their ini tia tive, ac cept joint re spon si bil ity, and work flex i bly. The im age of an or ganic or ga ni za tion is a net work in which ex perts col lab o rate in fluid and ad hoc ways. Burns and Stalker (1961, pp. 119-122) iden tify sev eral other distinguishing characteristics of the organic structure, such as com mit ment to the “tech no log i cal ethos” (p. 121), but these wider fea tures were not often used by subsequent scholars, so that the mechanistic ver sus or ganic dis tinc tion has come to re fer to struc tural char ac ter istics in ter nal to the or ga ni za tion. Burns and Stalker (1961) make the contingency argument that mechanistic struc tures are ef fec tive only in con di tions of low rates of technological and market change, whereas high rates of such change re quire the or ganic struc ture for the or ga ni za tion to be ef fec tive. With low rates of change, the top man ag ers pos sess ad e quate knowl edge to spec ify the work roles of their sub or di nates. In con trast, with high rates of change, the top man ag ers lack much of the knowl edge and so must rely on the ex per tise of their sub or di nates, who can or ga nize the work among themselves. The mechanistic or ga ni za tion is not only a struc ture but also a culture, in which subordinates are psychologically dependent upon their superordinates (Burns and Stalker 1961). The mechanistic structure corresponds to the structure prescribed by the clas si cal man age ment school, with its em pha sis on cen tral ized con trol, job specialization, and clear job descriptions (Brech 1957; Taylor 1947). In con trast, an or ganic struc ture en cour ages and le giti mates the ex er cise of dis cre tion by em ploy ees. The or ganic struc ture cor re sponds to the struc ture pre scribed by the neo-hu man re la tions school, with its

38

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

em pha sis on em ployee em pow er ment, self-di rected teams, and par tic ipation in de ci sion mak ing (Argyris 1964; Likert 1961). In stead of the uni ver sal claims to a “one best way,” Burns and Stalker ar gued that each model was valid in its own place. Its place was given by the con tin gency factor of technological and market change. Accommodating the two rival, pre ced ing the o ret i cal schools of clas si cal man age ment and neohu man re la tions into a broader frame work was an at trac tive fea ture of the con tin gency the ory of Burns and Stalker. Burns and Stalker (1961, p. 122) pres ent the mech a nis tic and or ganic structures as two polar ex tremes. The mech a nis tic and or ganic types each mark the poles of a con tin uum that reg is ters dif fer ences in de grees of mechanisticness and organicness. Organizations lie at any point along the continuum. Burns and Stalker (1961) ex tend their anal y sis to ar gue that the po liti cal system is different between mechanistic and organic structures, with the looser roles def i ni tions of the or ganic struc ture fos ter ing more po lit i cal be hav ior than in the more tightly de fined mech a nis tic struc ture. Their ex po si tion of the over all the ory is pur sued in a book that re counts ex ten sive, com par a tive field stud ies of con trast ing or ga ni za tions (Burns and Stalker 1961). Their method is qualitative and anthropological. Some of their em pir i cal claims are prob lem atic, such as that the fail ure of in no va tions in some tex tile firms was due to their mech a nis tic structures, whereas the failure seems to have been caused by lack of top management sup port (Burns and Stalker 1961). Nev er the less, the mech a nistic ver sus or ganic dis tinc tion has proven to be an en dur ing con tri bution, pro vid ing schol ars with a ready way to char ac ter ize the dif fer ing mod els of the over all or ga ni za tional struc ture. This helped raise con tingencyanal y sis from just the in di vid ual and group lev els to the or ga ni za tional level. Fur ther, the con tin gency fac tor of tech no log i cal and mar ket change moved the con tin gency out to the en vi ron ment of the or ga ni zation. The elegant simplicity of their model has helped propagate the con tin gency ap proach so that this is prob a bly, even today, the sin gle most widely known contingency theory of or ga ni za tional struc ture. Hage Hage (1965) pro vided an early con tri bu tion that was sim i lar to the theory of Burns and Stalker, thereby reinforcing it (see Hage 1980). Hage (1965) ar gues that ef fi ciency is max i mized by a struc ture that is

Or ganic The ory and Re search

39

cen tral ized in de ci sion mak ing, for mal ized (e.g., us ing rules), and low on com plex ity. This struc ture is sim i lar to the mech a nis tic struc ture of Burns and Stalker (1961). In contrast, in novation is maximized by a struc ture that is de cen tral ized, low on for mal iza tion, and high on complex ity. This struc ture is sim i lar to the or ganic struc ture of Burns and Stalker (1961). “Com plex ity” is a mea sure of the amount of knowl edge avail able within the or ga ni za tion. Com plex ity is high when the or ga nization employs highly educated people and fosters their knowledge through occupational spe cialization, so that they gain added depth. This type of specialization is to be distinguished from narrow task specialization, for ex am ple, of jobs on an au to mo bile as sem bly line, or specialization by func tion, which is part of the mech a nis tic struc ture. The use of the term complexity to have this mean ing is some what id io syn cratic to Hage and his fol low ers. The the ory of Hage (1965) is pre sented as a se ries of ab stract for mal propositions. The theory holds in es sence that cen tral iza tion and for malization al low man age ment to co or di nate op er a tions closely so that efficiency re sults through tight control of a workforce that includes many em ploy ees with low ed u ca tion (i.e., low com plex ity). In con trast, high com plex ity, to gether with de cen tral iza tion and low for mal iza tion, means that professionally trained persons with high expertise (i.e., high com plex ity) seek to im plement their ideas and val ues, which in cludes an emphasis on adopting the latest developments in their pro fession, so that innovation en sues. Innovation is the rate of new programs (such as new ther a pies in a med i cal or ga ni za tion) adopted in a unit of time (Hage and Aiken 1967a). Thus the organic structure provides the ideas for new innovations, while also freeing per sonnel from bu reau cratic con straints so that they can im ple ment them, thereby let ting “a thou sand flow ers bloom.” Stated as such it is a uni ver sal ist ic the ory, in that each prop o si tion gives a sim ple, main ef fect of struc ture on out come. It is com pat i ble with con tin gency think ing, how ever, by saying that if efficiency is desired, because of an environment that is competing on cost, then a centralized and formalized structure must be adopted to re main com pet i tive (Hage and Aiken 1970). Sim i larly, it can be said that, if innovation is desired, because of an environment that is competing on new, differentiated products or ser vices, then a decentralized and lowly formalized structure must be adopted to re main competitive (Hage and Aiken 1970). In this way the argument takes on the logic of a fit to the environment as a contingency.

40

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

Clearly, the Hage (1965) model parallels that of Burns and Stalker (1961): Their mechanistic struc ture is his cen tral ized and for mal ized structure, and their organic structure is his decentralized and unformalized struc ture. By re stat ing the struc tural con cepts in the abstract structural variables of centralization and formalization, Hage ad vances con tin gency the ory in two ways. First, he pro vides a com pact sum mary of the main ways in which the mech a nis tic and or ganic types differ. Sec ond, his use of vari ables in both the o ret i cal state ments and empirical re search makes it clear that these are differences of degree. Organizations vary in their amount or level of centralization and formalization rather than existing only at the two ends of these vari ables, that is, as be ing ei ther mech a nis tic or or ganic. By us ing vari ables (de rived from Hall 1963) to mea sure cen tral iza tion and for mal iza tion, Hage also pro vided a meth od olog i cal model for sub se quent re search ers. This helped to broaden the em pir i cal re search of or ganic the ory from the anthropological, qualitative methods used by Burns and Stalker (1961) to quantitative meth ods. In a se ries of em pir i cal stud ies, Hage and his col leagues (Dewar and Hage 1978; Hage and Aiken 1967a, 1967b, 1969; Hage and Dewar 1973) showed that struc ture was re lated to con tin gen cies and pro duced the expected causal outcomes. Centralization and formalization were as sociated pos i tively with task routineness (Hage and Aiken 1969), that is, low levels of the task uncertainty contingency. Hage and Dewar (1973) showed that having a decentralized, low formalization struc ture and high com plex ity leads to higher sub se quent in no va tion rates, thereby sup port ing the causal the ory that or ganic struc ture causes in novation. The con trary ar gu ment can also be made, how ever, that cen tral ization pro motes in no va tion, rather than pre vent ing it, be cause cen tral ized power allows top management to force through changes despite any resistance from other employees. Thus radical changes or those that require sys temswide co or di na tion to be suc cess fully im ple mented may be more likely to occur where a directive approach is taken by a top man age ment that per son ally makes all or many de ci sions (Dunphy and Stace 1988), that is, cen tral iza tion. How ever, while cen tral iza tion may fos ter the mag ni tude of an in no va tion, de cen tral iza tion fos ters the rate of in no va tion, that is, de cen tral iza tion leads to a higher rate of in novation, that is, innovations per unit time, the as pect of in no va tion in the work of Hage and his colleagues. Decentralization al lows many

Or ganic The ory and Re search

41

initiatives to be taken and hence many dif fer ent in no va tions to occur simultaneously. This distinction between the mag ni tude and the rate of in no va tion may go some way to ac count for the mixed find ings in organizational re search about the re la tion ship be tween cen tral iza tion and in no va tion (Slappendel 1996). Law rence and Lorsch Law rence and Lorsch (1967, p. 156) pro duced a the ory that is al lied to that of Burns and Stalker, and of Hage, but is more com plex. Be cause of its greater com plex ity, we shall have to de scribe it at greater length. Law rence and Lorsch (1967) con cep tu al ize or ga ni za tional struc ture in terms of dif fer en ti a tion and in te gra tion. The in ter nal sub units of an organization, such as the functional departments of a firm (e.g., pro duc tion, re search, sales), dif fer from each other but also must co op er ate with each other to ac com plish the over all task of the or ga ni za tion, so the differentiated units must be in te grated. Differentiation refers concretely to differences between departments in goal orientations, time ori en ta tions, for mal ity of struc tures, and interpersonalorientations (Law rence and Lorsch, 1967, pp. 30-38). Dif fer en ti a tion be tween de partments arises be cause de partments dif fer in their tasks. Law rence and Lorsch (1967, p. 36, Fig ure II-1) found that task cer tainty is re lated to for mal ity of struc ture, con sis tent with the find ings of other con tin gency the ory re search. More over, per for mance was higher where greater task un cer tainty was as so ci ated with less struc tural for mal ity and with less centralization (Lawrence and Lorsch 1967, p. 43, Table II-3; p. 129, Figure V-2; and p. 143, Figure VI-1). Thus departments, such as research, that had highly un cer tain tasks, had struc tures low on for mality, whereas de part ments, such as pro duc tion, that had highly cer tain tasks had structures high on formality (Lawrence and Lorsch 1967, p. 36, Fig ure II-1). In such ways there is cre ated high dif fer en ti a tion of structures between departments. The greater the differentiation, the more in te gra tion that is re quired for ef fec tive ness of the over all or ga nization (Lawrence and Lorsch 1967). Integration is achieved by using integrative de vices, with higher lev els of in te gra tion be ing achieved by the more sophisticated de vices, which in or der of in creas ing so phis ti cation are: hi erarchy, rules, integrating in dividuals, and integrating depart ments (Law rence and Lorsch 1967, p. 138, Ta ble VI-1). Co or di na tion is im proved where cen tral iza tion, the ver ti cal dis tri bu tion of

42

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

in fluence in Lawrence and Lorsch’s study, fits task uncertainty. De cisions are taken by upper hierarchical levels where tasks are certain, whereas in flu ence is shared more with lower hi er ar chi cal lev els where tasks are uncertain (Law rence and Lorsch 1967, p. 128, Table V-5; p. 129, Fig ure V-2; and p. 143, Fig ure VI-1). For Law rence and Lorsch, like Burns and Stalker (1961), the con tingency fac tor that de ter mines the re quired or ga ni za tional struc ture comes from the environment: It is the rate of new product innovation and/or changes in the mar ket or pro cess tech nol ogy that con trib utes to task un cer tainty in the in dus try. Low rates of prod uct in no va tion, for example, mean that the firm needs lit tle or no re search and de vel opment and this would, in turn, reduce its required differentiation and also its need for com plex in te gra tion mech a nisms, none be yond hi erar chy and rules (Law rence and Lorsch 1967). Thus suc cess ful firms in low-change in dus tries have low dif fer en ti a tion (Law rence and Lorsch 1967, p. 103, Ta ble IV-6) and em ploy sim pler in te grat ing mech a nisms (Lawrence and Lorsch 1967, p. 138, Ta ble VI-1). Con versely, high rates of en vi ron men tal change and innovation mean that the firm needs at least some de part ments that deal with highly un cer tain tasks (such as R and D). The re sult is that their over all de gree of dif fer en ti a tion is high, and this would need to be matched by more complex integration mecha nisms (such as re li ance upon an in te grat ing de part ment). Thus suc cess ful firms in high-change (un cer tain) indus tries have high dif fer en ti a tion and also achieve high in te gra tion by us ing a full ar ray of in te grat ing mech a nisms (Law rence and Lorsch 1967, p. 103, Ta ble IV-6; p. 138, Ta ble VI-1). Law rence and Lorsch (1967) showed these re la tionships em pir i cally in field stud ies of sev eral firms from each of three indus tries: con tain ers, pro cessed food, and plas tics. Subsequent study by Lorsch and Allen (1973) of large, diversified corporations failed to find that the diversification between di visions was accompanied by the use of complex integration mechanisms in successful corporations. Thus the simple, bivariate relationship be tween dif fer en ti a tion and in te gra tion did not hold. This led Lorsch and Law rence (1972) to clar ify their find ings, us ing the typology of inter dependence from Thomp son (1967). In the first study (Law rence and Lorsch 1967), the firms were less di ver si fied and so their sub units were functional departments that de pended upon each other, that is, mar keting sells what production makes. However, in the second study (Lorsch and Al len 1973) of di ver si fied cor po ra tions, the sub units were

Or ganic The ory and Re search

43

divisions, which for the highly diversified corporations had very lim ited direct interdependence between the divisions. This clarified that in tense in te gra tion is re quired only where there is in tense in ter de pendence. The higher the interdependence, the higher the sophistication of the in te gra tion needed for the firm to be successful. Dif fer en ti a tion be comes less im por tant in this re for mu la tion (Lorsch and Lawrence 1972). The first study had found that differentiation across func tional de part ments of un di ver si fied firms was nec es sary for suc cess in in no va tory in dus tries, be cause they re quired novel prob lem solving (Law rence and Lorsch 1967). How ever, dif fer en ti a tion across di vi sions was not as so ci ated with suc cess in highly di ver si fied cor po rations, be cause each di vi sion was self-con tained and so high dif fer en tiation across them had no im pact on their ef fec tive ness or that of the corporation as a whole (Lorsch and Al len 1973). Thus dif fer en ti a tion made a con tri bu tion to or ga ni za tional suc cess only if the or ga ni za tion was an in ter de pen dent sys tem, such as a sin gle busi ness that needed to in no vate. There was no gen eral pos i tive ef fect of dif fer en ti a tion on orga ni za tions. Sim i larly, whereas the ear lier for mu la tion had in te gra tion needing to match differentiation, the reformulation had integration need ing to match in ter de pen dence (Lorsch and Law rence 1972). Differ en ti a tion played a sec ond ary role in that it added to the in te gra tion needed by in ter de pen dence. As stated, in te gra tion is achieved be tween in ter de pen dent de part ments through having the appropriate co ordination de vices (Lawrence and Lorsch 1967). For highly in ter de pen dent de part ments these co or dina tion de vices in clude de part ments of inte gra tors, that is, per sons who work to integrate the functional departments. Successful integrators do not rely sim ply on hi er ar chi cal au thor ity to in flu ence de ci sions, but have to match it with their ex per tise (Law rence and Lorsch 1967, p. 66). This is achieved by hav ing the inte gra tors re side in a de part ment sep a rate from those they are integrating, in order to make them independent and im par tial. More over, the inte gra tors need to have an ori en ta tion in goals, time frame, and struc ture that is in ter me di ary be tween the de partments that they are seek ing to in te grate, so that they can trans late be tween the dif fer ent de part men tal sub cul tures (Law rence and Lorsch 1967, pp. 58-62). Fur ther, the inte gra tors need to use the right con flict resolution approach: neither forcing (by power) nor smoothing (by being evasive), but confronting (by facing the facts and solving the prob lems; Law rence and Lorsch 1967, p. 78). Con fronting is the most

44

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

effective way to resolve conflict in all the situations stud ied, so it is a uni ver sal ist ic fea ture, contrasting with the contingency nature of the rest of the the ory (Miner 1982). We may formalize the theory of Lawrence and Lorsch, in its inter dependence vari ant, into a causal model (Fig ure 2.1). The in no va tion, that is, de gree of nov elty and num ber of new prod ucts per unit of time, that is in tended by the man age ment in creases the in ter de pen dence between the func tional de part ments in volved in mak ing the in no va tion oc cur. The novel prob lem solv ing en tailed by in no va tion re quires that the func tional de part ments trans fer in for ma tion back and forth be tween them selves in an un pre dict able way. Thus the greater the in no va tion intended, the more intense the interdependence between functional de part ments. The in ter de pen dence raises the de gree of in te gra tion that is re quired. If the ac tual in te gra tion matches the req ui site in te gra tion, then fit results and, consequently, there is high achieved integration and so high per for mance. A high level of ac tual in te gra tion is pro vided by hav ing inte gra tors. Their req ui site char ac ter is that they be in ter me di ary in their ori en ta tion be tween the func tional de part ments. The degree of difference between departments that the integrators must span is greater, the greater the differentiation between departments. Therefore the requisite character of the integrators is de termined by dif fer en ti a tion. The dif fer en ti a tion is in turn de ter mined by the de gree of task un cer tainty as caused by the in tended in no va tion. Greater in nova tion causes greater task un cer tainty in some func tional de part ments (e.g., research) but not others (e.g., production), thereby creating greater differentiation. T he greater gap between departmental sub cultures that the integrators have to span increases their work and so tends to in crease the num ber of inte gra tors re quired, so that dif fer enti a tion adds also to req ui site in te gra tion. To il lus trate the the ory of Law rence and Lorsch, con sider their two most con trast ing in dus tries: con tain ers and plas tics. The con tainer indus try had low rates of new prod uct in no va tion (Law rence and Lorsch 1967, p. 86, Ta ble IV-2), lim it ing the amount of novel prob lem solv ing required, so that firm differentiation was low. The interdependence between de part ments was me dium, in that mar ket ing col lects cus tomer or ders that are passed to pro duc tion, which de liv ers to the cus tomer. In Thomp son’s (1967) terms this is “se quen tial in ter de pen dence” that can be managed by planning and was accomplished by the CEO, that is, coordination by the simple device of hi erarchy, plus some rules and

Or ganic The ory and Re search

45

Fig ure 2.1. A Causal Model of In ter de pen dence, In te gra tion, and Dif fer en ti a tion of Func tional De part ments

procedures as additional, only slightly more sophisticated, coordina tion devices. Thus suc cess ful firms in the low-in no va tion in dus try of containers could be run hi er ar chi cally be cause of the sim plic ity of their structures and medium level of their interdependencies. The more successful firms conformed to this pattern, provid ing ev idence that match ing their struc ture to the re quire ments of low in no va tion led to higher per for mance. In con trast, the plas tics in dus try had high rates of new prod uct in novation (Lawrence and Lorsch 1967, p. 86, Table IV-2), so successful firms re quired re search de part ments pur su ing novel sci en tific ideas in projects with long time frames using per missive cultures with low structure. These research departments differed from the sales and produc tion de part ments (Law rence and Lorsch 1967, p. 48, Fig ure II-2), contributing to high firm differentiation. The interdependence be tween de part ments was high, in that much novel prob lem solv ing was required, and so a considerable amount of information had to pass back and forth between the func tional departments. In Thompson’s terms this is “re cip ro cal in ter de pen dence,” which re quires mu tual adjust ment, that is, ad hoc co or di na tion be tween the func tional de partments, achieved through cross-functional pro ject teams fa cil i tated by integrators. Thus the greater interdependence within plastics firms called for more so phis ti cated co or di nat ing de vices that in volved cre ating the ad di tional roles of inte gra tors. In turn, these inte gra tors had to have ori en ta tions (in goals, etc.) in ter me di ary be tween the func tional departments and to use confrontation to resolve con flicts be tween these departments (Lawrence and Lorsch 1967, pp. 58-62, 78). Thus

46

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

successful firms in the high-innovation industry of plastics were not able to run purely hierarchically because of the complexity of their struc tures and high level of their in ter de pen den cies. To be suc cess ful in this in no va tory type of in dus try, firms had to adopt more par tic i pa tory structures (Law rence and Lorsch 1967, p. 138, Ta ble VI-1; and p. 143, Figure VI-1). The more successful firms conformed to this pattern, pro vid ing ev i dence that matching their struc ture to the re quire ments of high innovation led to higher performance (Law rence and Lorsch 1967, p. 80, Ta ble III-6). Thus in the the ory of Law rence and Lorsch (1967), the con tin gency is the intended level of innovation that the organizational structure needs to fit in its differentiation and integration. In no va tion leads to dif fer en ti a tion and in ter de pen dence, which in turn lead to in te gra tion. The language of Lawrence and Lorsch (1967) stresses the idea of an optimal de fined by the con tin gency to which the ac tual must be fit ted to at tain high per for mance. This is ex pressed by talking about the con trast between “requisite” (i.e., re quired) and “achieved” (i.e., ac tual), for ex am ple, the req ui site in te gra tion and the achieved in te gra tion. For low in no va tion the most ef fec tive struc ture is cen tral ized, us ing planning and formal procedures. Thus where the environmentally given task has low un cer tainty, a struc ture of the type pre scribed by clas si cal management is the most effective (Brech 1957). In con trast, for high in no va tion the most ef fec tive struc ture is de cen tral ized, us ing par tic ipation (Likert 1961). Thus where the environmentally given task has high un cer tainty, a struc ture of the type pre scribed by neo-hu man rela tions is the most ef fec tive. Each of clas si cal man age ment and hu man re la tions are cor rect in their own place, thus mak ing a syn the sis of these two the o ries. The con tin gency fac tor, in no va tion, is sim i lar to the techno log i cal and mar ket change con tin gency of Burns and Stalker (1961): both orig i nate in the en vi ron ment and con di tion the un cer tainty of the or ga ni za tion’s task. Struc turally, Law rence and Lorsch (1967) note that the high par tic ipation in organizational decision making in their innovatory firms is sim i lar to the or ganic struc ture of Burns and Stalker (1961). How ever, unlike the co he sion of the or ga ni za tion through shared values in the or ganic type of Burns and Stalker, Law rence and Lorsch have the more sophisticated con cept of dif fer en ti a tion be tween de part ments. Only in the research and development depart ment is the structure low on formalization and per mis sive in the man ner of the or ganic con cept.

Or ganic The ory and Re search

47

By contrast, the production de partment is struc tured and di rec tive, that is, mechanistic. These differences between functional depart ments in innovatory firms are shown empirically by Lawrence and Lorsch (1967). The cross-func tional pro ject teams would be ex pected to be more or ganic, as is con sis tent with their inte gra tors be ing fa cil i tators rather than au thor ity fig ures. Thus Law rence and Lorsch’s the ory is sim i lar to Burns and Stalker’s but is more com plex and more re al is tic. Prob a bly be cause of its com plex ity it is not as widely known as Burns and Stalker’s; how ever, Law rence and Lorsch’s the ory is well known. It was very im por tant in the de vel op ment of con tin gency the ory and re mains a ma jor con tri bu tion. Subsequently, Lorsch and Morse (1974) further developed the con tin gency the ory by an em pir i cal study that brought in the pre dis posi tions of or ga ni za tional mem bers as an ad di tional con tin gency of an individual psychological kind. They showed that, for high organiza tional per for mance, fit is re quired among en vi ron ment, or ga ni za tion, and member predispositions (Lorsch and Morse 1974, p. 114). More specifically, where the environment was low on uncertainty, this was fit ted by a mech a nis tic type of struc ture and short-term time ori en ta tion with em ploy ees who were com fort able be ing con trolled, liked to work with others, and had low tolerance for ambiguity (Lorsch and Morse 1974, p. 112). Con versely, where the en vi ron ment was high on un cer tainty, this was fit ted by an or ganic type of struc ture and long-term time orientation with em ployees who preferred autonomy and had high tol er ance for am bi gu ity (Lorsch and Morse 1974, p. 112). Fit led to higher feelings of competence and higher unit performance (Lorsch and Morse 1974, pp. 114-115). The in ter pre ta tion was that fit de pended upon mem bers’ psy cho log i cal pre dis po si tions match ing the struc ture required by the environment, which produced higher performance that cre ated feel ings of com pe tence that in turn mo ti vated con tin u ing higher per for mance (Lorsch and Morse 1974, pp. 114-115). Tosi, Aldag, and Storey (1973) crit i cized the mea sures of the en vi ronment used by Lawrence and Lorsch (1967). Lawrence and Lorsch’s the ory about dif fer en ti a tion and in te gra tion be ing shaped by the contin gency of in no va tion has re ceived few rep li ca tion at tempts, prob a bly be cause it is so com plex. Miner (1982) crit i cized it as not hav ing been replicated. However, elements of it have been rep li cated. Tung (1979, p. 691) investigated environments and structure and concluded in support for Law rence and Lorsch (1967). A study in Hol land rep li cates

48

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

portions of their model, thereby showing that aspects of the theory generalize geographically (Faas 1985). An unpublished study by Crawford (1983) of twenty diverse Australian organizations, which improved on the methods of Law rence and Lorsch, cor rob o rated several of their findings. Thus there is some support for Lawrence and Lorsch’s the ory from sub se quent re search. Nohria and Ghoshal (1997) have de vel oped a con tin gency model of the struc ture of mul ti na tional corporations that uses the terms of differentiation and integration; how ever, there are some dif fer ences in the mean ing of these terms, and their work will be dis cussed be low (Chap ter 3). Technology Part of or ganic the ory con sists of schol ars who ar gue that the need for or ga ni za tions to adopt or ganic struc tures fol lows from the con tingency of technology. Three main contributors may be distinguished: Perrow, Thomp son, and Wood ward. Perrow Perrow (1967) was a pioneering the orist of technology as a con tingency. For him, tech nol ogy re fers not to hard ware or lay out of equipment but to the cog ni tive pro cess ing in volved in com plet ing a task, that is, the per ceived na ture of the raw ma te ri als and the search be hav iors involved i n treating it. He distinguishes two dimensions: task analyzability and the num ber of ex cep tions. The cross-clas si fi ca tion of these two dimensions cre ates four different situations (Perrow 1967, p. 196, Fig ure 1). An a lyz able tasks with few ex cep tions are termed routine technology, exemplified by mass production steel mills. Analyzable tasks with many exceptions are termed engineering technology, for example, heavy ma chin ery. Unanalyzable tasks with few ex cep tions are termed craft technology, for example, specialty glass. Unanalyzable tasks with many exceptions are termed nonroutine technology, for example, aero space. The two sit u a tional di men sions (analyzability and number of ex ceptions) re late to the uncertainty of a task, and so the the ory may best be thought of as a re fine ment of the task un cer tainty contingency theory. Thus the tech nol ogy con tin gency of Perrow may be sub sumed un der the pre vi ously dis cussed task un cer tainty con tingency rather than constituting a sep a rate con tin gency.

Or ganic The ory and Re search

49

The four dif fer ent tech nol o gies, cre ated by the cross-clas si fi ca tion of the two dimensions, each require a particular type of organizational struc ture to fit it. An a lyz able tech nol o gies re quire cen tral ized struc tures, while unanalyzable struc tures re quire de cen tral ized struc tures (Perrow 1967, p. 199, Fig ure 3). The greater the num ber of ex cep tions, the more flex i bil ity that is re quired (Perrow 1967, p. 199, Fig ure 3). Thus rou tine technologies (analyzable, few exceptions) require formal, centralized structures (Perrow 1967, p. 199, Figure 3). Engineering technologies (analyzable, many exceptions) re quire flexible, centralized structures (Perrow 1967, p. 199, Fig ure 3). Craft tech nol o gies (unanalyzable, few ex cep tions) re quire de cen tral ized struc tures (Perrow 1967, p. 199, Fig ure 3). Nonroutine tech nol o gies (unanalyzable, many ex cep tions) re quire flex i ble, poly centralized struc tures (Perrow 1967, p. 199, Fig ure 3). Perrow (1967, p. 199) states that the for mal, cen tral ized struc ture is clos est to the mech a nis tic struc ture (Burns and Stalker 1961) and the flexible, polycentralized structure is closest to the organic structure. Thus his struc tural con cept can be clas si fied as part of or ganic the ory. In rou tine tech nol ogy, re li ance is placed on plan ning. Tech ni cal ex perts direct those supervising production operations, and there is little discretion over coding of cases (Perrow 1967, p. 200). In engineering tech nol ogy, the many ex cep tions are han dled by the tech ni cal ex perts, who exercise considerable dis cre tion and ad just to feed back, whereas the operational su per vi sors receive di rec tion from them and run op erations to plan (Perrow 1967, p. 200). In craft technology, the unanalyzable na ture of the ma te rial means that the work ers have to ex er cise dis cre tion and power, re ly ing on feed back (Perrow 1967, p. 200). In nonroutine technology, the per ceived na ture of the raw ma te rial is such that there are many ex cep tions that have to be dealt with through “experimentation and ‘feel’” (Perrow 1967, p. 199). Re li ance has to be placed on ad just ment to feed back rather than on plan ning. Tech ni cal ex perts and those su per vis ing pro duc tion op er a tions ex er cise dis cretion over coding of cases and use high power in close collaboration with each other (Perrow 1967, pp. 199-200). Thus the the ory of Perrow may be sub sumed un der the or ganic the ory that low task un cer tainty re quires a mech a nis tic struc ture and high task un cer tainty an or ganic struc ture. The four tech nol o gies also lead to differences in identifica tion and goals (Perrow 1967, pp. 200-202). Perrow (1967, p. 198, Fig ure 2) il lus trates his dis cus sion with ex amples drawn not only from in dus try but also from peo ple chang ing

50

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

orga ni za tions. Rou tine tech nol ogy is ex em pli fied by cus to dial in sti tutions. En gi neering tech nol ogy is ex em pli fied by a pro grammed learn ing school. Craft technology is exemplified by socializing institutions. Nonroutine technology is exemplified by elite psychiatric agencies. Perhaps be cause the con tingencies refer to generalizable cognitive pro cesses rather than spe cific types of in dus trial equip ment, the tech nology contingency constructs of Perrow have been used widely, especially to study peo ple-pro cess ing or ga ni za tions, such as hos pi tals (e.g., Al ex an der and Randolph 1985). This brings out the ex tent to which “tech nol ogy” is some thing of a mis no mer, be cause it car ries con no tations of hardware, whereas Perrow (1967) is distinguishing different approaches to problem solving based on the perceived nature of the case sit u a tions and the de gree of de vel op ment of knowl edge. Thus it is best clas si fied as a re fine ment of the task un cer tainty con tin gency idea. Thompson Thomp son (1967) ar gues that task and tech nol ogy are ma jor con tingency fac tors of or ga ni za tional struc ture. He of fers a typology of types of tech nol ogy and their re spec tive or ga ni za tional struc tures. Three differ ent types of tech nol o gies are dis tin guished: me di at ing, long-linked, and in ten sive. These cor re spond to three types of task in ter de pen dence between organizational subunits: pooled, sequential, and reciprocal, respectively (Thomp son 1967). Me di ating tech nol ogy re fers to the link ing of cus tom ers, such as a bank link ing lend ers and bor row ers, and in volves pooled in ter de pendence. Pooled interdependence means that two organizational subunits (e.g., branches of a bank) have no di rect con nec tion, so that their in ter de pendence is in di rect, re sid ing in their both drawing resources from some cen tral pool. An other ex am ple of pooled in ter de pen dence is the re la tionship among di vi sions in a con glom er ate: there is no di rect con nec tion be tween them be cause each busi ness is un re lated to the oth ers, but they all draw funds from the common, corporate pool of new capital (Lorsch and Al len 1973). This low de gree of in ter de pen dence can be effectively struc tured by rules and pro ce dures (Thomp son 1967). Long-linked tech nol ogy refers to sequential interdependence where task A is the in put to task B, for ex am ple, stages along an as sem bly line. Sequential interdependence means that the subunits have a direct connection, so that the out put of one sub unit is an in put to the other

Or ganic The ory and Re search

51

sub unit. Another example of sequential interdependence is the re lation ship among di vi sions in a ver ti cally in te grated com pany, such as in for est prod ucts: One di vi sion pro duces the raw ma te ri als of pa per and cardboard that are then pro cessed by the pack ag ing di vi sion and sold to cus tom ers (Lorsch and Al len 1973). This me dium de gree of in ter depen dence can be ef fec tively struc tured by plan ning (Thomp son 1967). In ten sive tech nol o gies use varying techniques ac cording to feedback from the object w orked upon, for example, a hospital using various diagnostic and treat ment tech niques ac cord ing to the con di tion of the patient, and involve reciprocal in ter de pen dence. Reciprocal interdependence means that the subunits have a two-way connection, in which the output of each sub unit is an in put to the other sub unit, so that they trans act back and forth in an un pre dict able man ner. An other ex am ple of re cip ro cal in ter de pen dence is the re la tion ship be tween the re search and the pro duc tion de part ments in a func tion ally struc tured com pany that is in no vat ing new prod ucts: the novel, com plex prob lem solv ing re quires much in for ma tion to be passed back and forth be tween the two de part ments in an un pre dict able fash ion (Lorsch and Law rence 1972). Reciprocal interdependence can be effectively structured by mutual adjustment, that is, each organizational subunit will try to attain its ob jec tives, but then re ceive feed back about how suc cess ful it has been, and other sub units have been, lead ing to re for mu la tion of its ob jec tives for the next time pe riod (Thomp son 1967). Thus each of the three types of interdependence requires a particular coordinating structure. (This is the theory that Lorsch and Lawrence, 1972, draw upon in their re for mu la tion of their orig i nal study, as dis cussed above.) Thomp son (1967) also the o rizes about or ga ni za tional pol i tics, dis cretion, assessment, and cooperation between the organizations and its environment. There are the o ret i cal re la tion ships run ning across these top ics con nect ing them with technology. In terms of linkages with other structural contingency theo ries, coordination by rules and procedures is associated with the classical management school (e.g., Weber 1968) and may be classified as a mechanistic structural device. Coordination by planning would also seem to be quite mech a nis tic, al beit some what more flex i ble be cause it al lows for the plan to be mod i fied as the task sit u a tion changes. Co ordination by mutual adjustment is associated with the neo-human relations school (e.g., Likert 1961) and entails employees agreeing among themselves their next actions, and so may be classified as an

52

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

organic structural device. Hence the three types of interdependence (pooled, se quen tial, and re cip ro cal) are each fit ted by vary ing de grees of mech a nis tic or or ganic struc tures. Thus task in ter de pen dence can be con sid ered to be a con tin gency of or ganic struc ture. Thomp son’s tech nol ogy the ory led to sub se quent em pir i cal re search, such as the study by Gerwin and Christoffel (1974) of Thompson’s the ory of task in ter de pen dence. Thomp son had illustrated his the ory by his own study of the main te nance func tion of a USAF bomber wing. Gerwin and Christoffel (1974) found that a com puter pro gram based on Thomp son’s the ory was able to pre dict cor rectly the struc ture of an engineering factory from the task interdependencies. This replicated Thomp son and showed that his the ory gen er al ized em pir i cally from a governmental to a private-sector organization. Other research has con firmed Thompson’s theory of the relationship between task inter dependence and coordinating structures, including in social service agencies (Van de Ven, Delebecq, and Koenig 1976; Van de Ven and Ferry 1980). Woodward An other path-break ing con tin gency the ory study was by Wood ward (1958, 1965). She ex am ined the struc tures of one hun dred com pa nies distributed across different industries. She studied various aspects of the organizations, quantitatively and qualitatively. The quan titative mea sures in cluded the num ber of lev els of man age ment, the spans of control of the chief executive and first-line supervisor, the ratio of man ag ers and su per vi sors to to tal per son nel, the per cent age of costs allo cated to wages, and the ra tio of di rect to in di rect work ers (Wood ward 1965). She interpreted her measures as indices of the extent to which each com pany was fol low ing the pre scrip tions of clas si cal man age ment. Woodward concluded that the classical management prescriptions were be ing fol lowed in some but not all the firms and that there was an association with operations technology. She distinguished stages of advance in technology, of which the three main stages are: unit and small batch pro duc tion, large batch and mass pro duc tion, and pro cess pro duc tion (ex am ples would be the man u fac ture of mu si cal in stru ments, automobiles, and oil, respectively). As firms move to more advanced tech nol ogy their pro duc tion be comes smoother and more con tin u ous, associated with more au to ma tion and cap i tal in vest ment.

Or ganic The ory and Re search

53

Wood ward’s (1965) key finding was that the classical man agement model was found in large batch and mass pro duc tion but not in unit and small batch nor in process pro duc tion. Thus she iden ti fied a historical sequence whereby advance in technology from small batch to mass pro duc tion re quired the adoption of clas si cal management pre scriptions, but fur ther tech no log i cal ad vance re quired their abandon ment. Be fore and af ter the clas si cal man age ment phase, the pre scriptions of human relations, with their emphasis on par ticipation and flexibility, were appropriate and tended to be followed. Drawing on Burns and Stalker (1961), she stated that the tech no log i cal im per a tive led first to the or ganic struc ture, then to the mech a nis tic, and fi nally to the organic. Woodward’s book contains many quantitative findings, show ing that the means on many struc tural vari ables vary ac cord ing to technology. It also included evidence that firms that were away from the mean had poorer performance (Wood ward 1965, p. 69, Ta ble 4). This pro vided the first quan ti ta tive de pic tion of the con tin gency idea that the firm had to fit its struc ture to the con tin gency fac tor in or der to have high per for mance. Woodward (1965) lacked a well-de vel oped the ory of why all of the structural variables are associated with technology. However, she has the broad argument that mass produc tion ren ders the task more pre dictable so that it can be for mal ized and proceduralized, whereas process tech nol ogy furthers task predict abil ity but embod ies the predict able el e ments in au to ma tion, leav ing the un pre dict able el e ments to the hu man op er a tors who must be so phis ti cated prob lem solv ers. Woodward’s contingency fac tor was internal to the organization, whereas Burns and Stalker’s was environmental, but both involved tech nol ogy. They mu tu ally sup ported each other in the the sis that the con crete task fac ing the or ga ni za tion de ci sively af fected the struc ture that it needed. Both Woodward (1965) and Burns and Stalker (1961) argued that the mechanistic structures were appropriate for firms in traditional industries with older technologies but that modern tech nol o gies forced the adop tion of or ganic struc tures in or der to main tain effectiveness. Partly be cause Wood ward’s model is more com plex than Burns and Stalker’s model, it is not quite as well known. However, its influence upon con tin gency the ory re search ers has prob a bly been greater than Burns and Stalker’s because of the quantification. It has become an ex emplar of contingency theory research at the macro-level, that is,

54

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

where the unit of anal y sis is the or ga ni za tion. The ar gu ment that technology is a contingency of organizational structure has stimulated much subsequent research, with mixed findings. Woodward (1965) found re la tion ships be tween op er a tions tech nol ogy and many or ga niza tional struc tural vari ables, and some sub se quent stud ies have rep licated some of these re la tion ships (e.g., Zwerman 1970); how ever, other stud ies have found weaker and fewer re la tion ships (e.g., Hickson, Pugh, and Pheysey 1969; Child and Mansfield 1972) or op po site re la tion ships (Blau, Falbe, Mc Kin ley, and Tracy 1976). These is sues will be dis cussed in de tail be low (see Chap ter 5). Woodward (1970) and her colleagues con ducted further research that led to somewhat different concepts to analyze the management systems of firms. They argued that the technological imperative (mean ing the firm has to obey it or suf fer per for mance loss) was tem pered in some situations where managerial ide ol ogy had some in flu ence (see also Dawson and Wedderburn 1980; Reeves and Turner 1972). How ever, this later work has had less in flu ence than the ear lier work and so Wood ward’s name is syn on y mous with the idea of a tech nological im per a tive. Tech nol ogy may be imag ined to be a hard, con crete ma te rial fac tor of the organization. Im agery invoked may be of smokestacks, as sembly lines, and oil re fin er ies. How ever, the tech nol ogy vari able mea sured by Wood ward (1965) con sists of two as pects: type of prod ucts pro duced and how many. The type of prod ucts pro duced is about whether they are in te gral, that is, dis crete items that can be counted (e.g., 12 trum pets), or in a form, such as a liq uid or gas, that has to be mea sured by vol ume (e.g., li ters) or weight (e.g., tons). Within the man u fac tur ing of in te gral prod ucts, the dis tinc tion is drawn as to how many prod ucts are produced in a batch: unit, small batch, large batch, and mass. This captures batch size and relates to the con tinuity of the production process. Thus Woodward’s technology variable ranges from single product to many prod ucts to con tin u ous flow, that is, from lumpy to flowing, and has been termed production continuity (Hickson, Pugh, and Pheysey 1969). While the products produced are a material fact, they are the re sult of nu mer ous fac tors, such as avail able hard ware and skills, including managerial decisions about layout and batch size (Ayoubi 1981). The production continuity may be a given for lowerlevel em ploy ees, but is to some de gree a choice for man age ment. Hence pro duc tion con ti nu ity is par tially the re sult of a de ci sion about so cial

Or ganic The ory and Re search

55

or ga ni za tion, not wholly a ma te rial fac tor. Thus the tech nol ogy the sis of Wood ward is to some ex tent about how one as pect of or ga ni za tional struc ture af fects other as pects (Bur rell and Mor gan 1979, p. 223 n. 30). This places a qual i fi ca tion against any ten dency to in ter pret Wood ward’s technology work as being about how hard impersonal technology molds the soft hu man or ga ni za tion.

Underlying Connections in Organic Theory Burns and Stalker (1961), Hage and Aiken (1970), Lawrence and Lorsch (1967), Perrow (1967), and Woodward (1965) all argue that structure is contingent upon the uncertainty of the task facing the organization, and that this un cer tainty re lates to tech nol ogy. The task uncertainty results from either the environmentally induced in no vation, for Burns and Stalker, Hage and Aiken, and Law rence and Lorsch, or raw ma te ri als for Perrow, or the in ter nal or ga ni za tional pro duc tion set-up for Woodward. The uncertainty of the tasks facing organiza tional mem bers is raised by in no va tion, unanalyzable raw ma te rial with many ex cep tions, or au to ma tion. Thus these stud ies have in com mon the con tin gency fac tor of task un cer tainty re lated to tech nol ogy. Moreover, the stud ies ar gue that low task un cer tainty re quires a centralized, for mal ized struc ture of the mech a nis tic type, whereas high task un certainty re quires a par tic i pa tory struc ture of the organic type. Thus the stud ies all con clude that high task un cer tainty re quires an organic structure,so that they sub scribe to the or ganic the ory. The task un cer tainty contingency and required struc ture of these or ganizational-level stud ies ech oes that of the group de ci sion-mak ing stud ies in mi cro-OB (Leavitt 1951), so pro vid ing a con nec tion be tween the macro-OB and mi cro-OB con tin gency the o ries. However, not all contingency theories of organizational structure con nect task un cer tainty and or ga ni za tional struc ture. As seen above, Thomp son (1967) argues that the link ages among the tasks be ing per formed by the organization, that is, the task interdependence, con stitute another contingency. There is, however, a com monality between Thomp son and these other au thors. For all of them the con tin gency fac tor is the task: task un cer tainty for Burns and Stalker, Hage and Aiken, Lawrence and Lorsch, Perrow, and Woodward, and task interdependence for Thomp son. Thus there is con sid er able unity here.

56

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

These con tri bu tions to the con tin gency the ory of or ga ni za tional structure are all dis cussing one contingency factor, the task, that has two dimensions, uncertainty and interdependence. In this way, far from being a pleth ora of the o ries with no uni fy ing con nec tions, the stream of contingency the ory re search flowing from the seminal work of Burns and Stalker, Hage and Aiken, Lawrence and Lorsch, Perrow, Thomp son, and Wood ward can be seen as a highly co her ent the ory. Indeed one might go farther with this conceptual syn thesis by suggesting that task in ter de pen dence may be mainly re duc ible to task un cer tainty. A closer ex am i na tion of each of three types of task inter dependence re veals that they re quire their cor re spond ing co or di na tion mech a nism be cause of task un cer tainty. The rea son why re cip ro cal interdependence can not be co or di nated through plan ning and re quires mu tual ad just ment (i.e., ad hoc co or di na tion) would seem to be im plic itly be cause the fu ture inter ac tions can not be pre dicted (Thomp son 1967). The cru cial as pect is thus not that the in ter de pen dence is re cipro cal (i.e., two way) or that it may be fre quent, but that it is un cer tain. Hence a re cip ro cal in ter ac tion re quires mu tual ad just ment be cause it is highly un cer tain rather than be cause it is re cip ro cally in ter de pen dent. Similarly, the distinction between pooled and sequential interdepen dence is re ally just whether there is an in di rect or a di rect re la tion ship between two organizational subunits (Thompson 1967). If there is a di rect re la tion ship be tween them, then that flow needs co or di na tion by plan ning, so that one sub unit can ef fi ciently deal with in puts it re ceives from the other. But this as sumes that the flow is fairly pre dict able and that task uncertainty is at least moderate. Again, Thompson (1967) states that pooled interdependence requires co ordination through rules and stan dard iza tion, but this as sumes that the (in di rect) re la tionship be tween sub units is highly pre dict able, so that task un cer tainty is low. If the re la tion ship is un pre dict able, then co or di na tion by rules is not possible and so other means of coordination would be required, such as discussion, so the the ory breaks down. Thus the three modes of in ter de pen dence and co or di na tion mech a nisms that Thomp son pos its are based on the assumption that the three interdependencies corre spond to a scale of in creas ing task un cer tainty. If they do not, then the three coordination modes (pooled, sequential, and reciprocal) spec ified by the the ory would not be fits. Hence the the ory is better stated as progressively greater un cer tainty in the task interaction between two or ga ni za tional sub units that leads to co or di na tion by stan dard iza tion,

Or ganic The ory and Re search

57

plan ning, and mu tual ad just ment, re spec tively. Thus the the ory of task in ter de pen dence and co or di na tion mech a nisms is better for mu lated as a theory about task uncertainty and coordination mechanisms. This com po nent of the task in ter de pen dence the ory can then be sub sumed un der task un cer tainty the ory. An other com po nent of the theory of task in ter de pen dence is that each type of task in ter de pen dence should be lo cated at dif fer ent lev els in the hi er ar chy, thus giv ing the def i ni tion and group ing of or ga ni zational sub units: pooled at the top, se quen tial in the mid dle, and re ciprocal at the bottom (Thompson 1967). Yet this conclusion can be reached by stating that the types of interdependence correspond to degrees of intensity of in teraction: pooled (nil or low), sequen tial (medium), and re cip ro cal (high). The the ory is that ac tiv i ties that need in tense in ter ac tion should be placed near to each other or ga ni za tion ally. The more in tense the in ter ac tion, the or ga ni za tion ally closer the ac tiv ities should be in or der to fa cil i tate their in ter ac tion. There fore ac tiv i ties with low interaction are placed in different divisions, which become the pri mary or ga ni za tional sub units at the top of the hi er ar chy. Ac tivities with me dium in ter ac tion are placed within the same di vi sions, as adjacent de partments or sections, which become the subunits in the middle of the hierarchy. Activities with high interaction are placed within the same sec tions, as ad ja cent jobs, which be come the sub units at the bottom of the hierarchy. Thus intensity of in teraction governs position in the hierarchy. Hence task interdependence can be reconceptualized into task un cer tainty and in ten sity of in ter ac tion. For in stance, consider again the argument of Lorsch and Lawrence (1972) that greater task interdependence between de part ments in creases the coordination required and so leads to the use of coordination mechanisms additional to hi erarchy, contacts between departments, and planning. In the container firms, the se quen tial task in ter de pendence be tween func tional de part ments was han dled ef fec tively through planning by the hierarchy (Lorsch and Law rence 1972). But this rests on the un cer tainty of the task be ing low: cus tomer de mand is pre dictable for the up com ing pe riod, man u fac tur ing re sources are sta ble, and upper management possesses ad equate knowledge of technical pro cesses be cause of low in no va tion. Sim i larly, in the plas tics firms the re cip ro cal task in ter de pen dence be tween func tional de part ments was handled effectively by mutual adjustment, through cross-functional project teams facilitated by integrators (Lorsch and Law rence 1972).

58

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

But this rests on the uncertainty of the task be ing high, so that many novel problems had to be solved through discussion between the functional departments, involv ing in tense in ter ac tions be tween them that could not be com pletely sched uled be cause in ter ac tions could not be predicted. Therefore cross-functional pro ject teams were re quired so that rep resentatives of each func tional de part ment were brought together to fos ter in tense in ter ac tion. In ternally these teams or ga nized them selves by mu tual ad just ment be cause the un cer tainty pre cluded planning of their task interactions. Over all, where task un cer tainty is low (e.g., con tainer firms), the in ter ac tion be tween func tions is of low intensity, so that their ef fec tive co or di na tion is achieved by hi er ar chi cal planning. Conversely, where task uncertainty is high (e.g., plastics firms), the interaction be tween functions is of high intensity, so that effective coordination is achieved by mutual ad justment. Thus for Lorsch and Law rence (1972), like Thomp son (1967), the task in ter de pen dence con tin gency de com poses into task un cer tainty and in ten sity of in ter ac tion. De com posing task in ter de pen dence into task un cer tainty and in tensity of in ter ac tion gives a clearer the ory. It also means that task in ter dependence becomes largely subsumed under task uncertainty, thus increasing theo ret i cal co her ency and par si mony. Task un cer tainty is a contingency of co or di na tion mech a nisms. In ter ac tion in ten sity is the contingency of lo ca tion in the hi er ar chy.

Conclusions The key idea in the organic the ory of or ga ni za tional struc ture is that mech a nis tic struc ture, which em pha sizes hi er ar chy, is ef fec tive for low task un cer tainty; and or ganic struc ture, which em pha sizes par tic i pation, is effective for high task uncertainty. The pioneering organic theories have this idea as an underlying theoretical commonality, which their empirical stud ies sup port. A major source of task uncertainty is innovation, much of which comes ultimately from the environment of the organization, such as technological and market change. The mechanistic or ganizational struc ture is shown to fit an en vi ron ment of a low rate of mar ket and tech no log i cal change. Con versely, the or ganic or ga ni za tional struc ture is shown to fit an en vi ron ment of a high rate of mar ket and technological

Or ganic The ory and Re search

59

change. In creasing en vi ron men tal un cer tainty in creases different iation, so that some parts of the organization deal with much of the un certainty that is entailed by innovation and be come or ganic, while other parts are mechanistic. Sim i larly, con tin gency re search into op er a tions tech nol ogy held that advances in operations technology lead from organic to mechanistic and then or ganic structures. At the high est lev els of task cer tainty, oper a tions could be au to mated so that the hu man or ga ni za tion mainly dealt with the re sid ual un cer tain el e ments and had to be par tic i pa tory. The main other contin gency factor to emerge from or ganic the ory research into or ga ni za tional struc ture was a sec ond di men sion of the task: task in ter de pen dence. In no va tion leads to re cip ro cal in ter de pendence be tween func tional de part ments, which re quires in te gra tion by means of cross-functional pro ject teams us ing open prob lem solv ing, that is to say, organic el e ments added to the over all struc tures. While task in ter de pen dence is usu ally treated as a sep a rate con tin gency from task uncertainty, a closer examination reveals that much in the task interdependence con tin gency is re duc ible to task un cer tainty, thereby mark ing an even greater co her ency to the or ganic the ory. The or ganic the ory of or ga ni za tional struc ture pro duces a con sis tent the o ret i cal model. There is con sid er able em pir i cal re search sup port for or ganic the ory over all. The link be tween task un cer tainty and or ganic struc ture is well es tab lished. The con nec tion of or ganic struc tures with higher innovation rates is empirically supported. The connections among task un cer tainty, dif fer en ti a tion, in te gra tion, and per for mance are also empirically supported. The relationship between task inter dependence (or interaction intensity) and position of organizational sub units in the hi er ar chy is also em pir i cally sup ported. Only for op er ations tech nol ogy is the em pir i cal ev i dence such as to en gen der doubt about the va lid ity of the the ory (as will be dis cussed in Chap ter 5). In sum, or ganic the ory is a co gent and co her ent the ory that is supported by em pir i cal re search, so that it con sti tutes a ma jor com po nent of the con tin gency the ory of or ga ni za tional struc ture and a sig nif i cant part of or ga ni za tional sci ence.

THE Bu reau CON cracy TINThe GENCY ory and THEORY Re search OF OR GA NI ZA TIONS

3

Bureaucracy Theory and Research

D

e spite the suc cess of or ganic the ory re search, it is chal lenged by bureaucracy the ory, which has a dif fer ent model of or ga ni za tional struc ture and its con tin gen cies. Bu reau cracy the ory ar gues that structural formalization is accompanied not by centralization, but by decentralization. For bu reau cracy the ory the main con tin gency is not task but organizational size. Bu reau cracy the ory re search has been at least as suc cess ful as or ganic the ory. In this chap ter we will dis cuss bu reau cracy the ory and its at ten dant re search. The suc cess of this challenge to or ganic the ory poses the ques tion of the re la tion ship between these two the o ries. We will ar gue that both may be brought to gether in a par si mo ni ous model of or ga ni za tional struc ture. More over, the key con tin gen cies of each, task uncer tainty and size, may be caus ally con nected, thereby fur ther ing the co her ence of struc tural con tin gency the ory as a whole. As we saw in Chapter 1, bu reaucracy the ory holds that there are three interrelated as pects of struc ture: bu reau cratic struc ture (spe cialization-formalization, decentralization), structural differentiation, 61

62

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

and divisionalization. We will now examine in greater detail bu reaucracy the ory and re search. We will be gin with out lin ing the the ory and re search on bu reau cratic struc ture, then that on struc tural dif fer en ti ation and on divisionalization. Through out much of this dis cus sion size will be the main con tin gency, with task (i.e., task in ter de pen dence) a mi nor con tin gency. We will then dis cuss whether cer tain other causes of organizational structure are contingencies or not, in the process defining the cri te rion for a cause to be a con tin gency. Finally, we will attend to the issue of synthesizing the bureaucracy with the organic the ory, in a way that fol lows on from, but goes far ther than, the brief syn the sis of fered at the end of Chap ter 1.

Bureaucracy Theory Bu reau cracy the ory (Blau, 1970, 1972; Child 1973a) in its mod ern form emerged from a se ries of em pir i cal stud ies. How ever, from the out set of modern research, thinking about bureaucracy was strongly influ enced by the Weberian model of bu reau cratic or ga ni za tional struc ture, which omit ted the idea of par tic i pa tion that is so cen tral to or ganic theory. Weber (1968) held that there was a general, historical ten dency for administration to move toward the bureaucratic type. The bu reaucratic struc ture fea tures full-time, sal a ried, ca reer ad min is tra tors who are ap pointed on merit, tech ni cally qual i fied, ar ranged in a hi er ar chy, and sub ject to rules and dis ci pline (Weber 1964, pp. 333-334). Bu reau cracy pos sessed sev eral ad van tages in clud ing ef fi ciency, pre dict abil ity, reliability, and the “strin gency of its dis ci pline” (Weber 1964, p. 337). The de vel op ment of bu reau cratic struc ture is pro moted by a num ber of factors, including size and communications technologies (Weber 1964, pp. 338-339). Bureaucratic theory led to a num ber of em pir i cal stud ies that used qual i ta tive meth ods to make case stud ies of or ga ni zations (Crozier 1964; Gouldner 1954; Selznick 1949). In contrast to Weber, these stud ies iden ti fied neg a tive con se quences of bu reau cracy, thereby ques tion ing the su pe rior ef fec tive ness that Weber at trib uted to bureaucracy. Weber (1968) advanced his theory in terms of ideal-types, that is, comparison of real organizations by contrasting them with types so

Bu reau cracy The ory and Re search

63

pure that they do not exist. The ideal-types of authority structures included b ureaucracy, charisma, and traditionalism. The ideal-types yield imag i nary con tinua that run be tween each of two ideal-types so that an actual organization is located somewhere along this line. Given more than two ideal-types, with imag i nary lines run ning be tween each pair, each actual or ganization is really positioned in a multi dimensional space. How ever, in the ab sence of mea sure ment of these dimensions, the lo cation of the organization in the space cannot be stated pre cisely and anal y sis be comes vague. For or ga ni za tional anal ysis to be come sci en tific it needed to de velop mul ti di men sional mea sures (Price 1972, 1997). There fore re search ers sought to de velop mea sures of each of the aspects of Weberian bureaucracy (Hinings, Pugh, Hickson, and Turner 1967). The Aston Group (named af ter the Uni ver sity of Aston, in Bir ming ham, Eng land) iden ti fied four main as pects of or ga ni za tional struc ture, for each of which they de vel oped a multi-item mea sure ment scale (Pugh et al. 1963; see also Donaldson 1997). The variables are special ization, standardization, f ormalization, and centralization (Pugh, Hickson, Hinings, and Turner 1968). Centralization is how far up the hi er ar chy de ci sions are taken. Specialization is the di vi sion of la bor in the ad ministrative work of the organization; it is measured by a scale of func tional specialization that assesses the degree to which administrative work has been spe cial ized by func tions. There is also a scale of role specialization that examines specialization within each function, which cor re lates with func tional spe cial iza tion. Standardization is the de gree to which work is gov erned by rules, pro ce dures, and sci en tific meth ods. Formalization is the extent of pa perwork, such as for de fin ing pro cedures and jobs, recording performance, and passing information. Child (1973a) later called this vari able “doc u men ta tion,” which is more de scrip tive. More over, other re search ers, es pe cially in the United States, use formalization to refer to documentation and standardization together (e.g., Hage and Aiken 1967a). There fore we will call the Aston for mal iza tion vari able by the termdocumentation , and use formalization to re fer to both doc u men ta tion and stan dard iza tion. The Aston Group also gath ered data on what they termed configuration , which is a variegated assort ment of mea sures of reporting rela tionships, such as the num ber of lev els in the hi er ar chy, which they termed ver ti cal span, and the span of con trol of the CEO, to gether with

64

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

the proportion of em ployees in each specialism. Most vari ables o f configuration are mea sured by sin gle in di ca tors, rather than the multiitem scales used for the main struc tural vari ables (Pugh et al. 1968). The Aston Group took each of the aspects of bu reaucracy in the Weberian ideal-type, measured it, and then looked to see which vari ables went together empirically to form the dimensions of a multi dimensional space (Pugh et al. 1968). The Group com pared the struc tures of a sam ple of or ga ni za tions through a quan ti ta tive anal y sis. The Aston Group and their later col leagues con ducted nu mer ous em pir i cal stud ies, of which the orig i nal one is called the Aston Study. This study compared the orga ni za tional struc tures of fifty-two di verse work or ganizations, including man ufacturing, re tail, and gov ernmental (Pugh et al. 1968). It found that spe cial iza tion, stan dardization, doc umen ta tion, and ver ti cal span were strongly, pos i tively cor re lated. In a fac tor anal y sis these three vari ables formed one fac tor called struc tur ing of ac tiv i ties. Centralization was in a second fac tor, concentration of au thority, which was independent of structuring (see also Pugh, Hickson, and Hinings 1969). However, in a subsequent, nationally representative sam ple (the Na tional Study), Child (1972a) found that cen tral iza tion and struc tur ing were not in de pend ent di men sions but rather op po site poles of the same dimension. He stated that Weber’s theory pos ited bureaucracy as con tain ing del e ga tion so that struc tur ing and de cen traliza tion should be ex pected to be pos i tively cor re lated and form part of the same fac tor in fac tor anal y sis. Studies sub se quent to the Aston and Na tional Stud ies have in ves ti gated the un der ly ing fac to rial di men sions of or ga ni za tional struc ture (e.g., Reimann 1973, 1974), though Starbuck (1981) has noted that this question re ceived less im por tance af ter the initial stud ies (see also Pugh 1981a, 1981b). But even without factor anal y sis, the pos i tive cor re la tions between spe cial iza tion-for mal iza tion and decentralization in later studies (e.g., Donaldson and Warner 1974; Grinyer and Yasai-Ardekani 1980; Hinings and Lee 1971; Routamaa 1985), as is confirmed by a meta-analytic re view (Wag ner, Buchko, and Gooding 1988), tend to support t he view of Child that struc tur ing and de cen tral iza tion are part of the same di men sion. In sum mary, the Aston Group found that bu reau cratic struc ture is composed of specialization, formalization, decentralization, and ver ti cal span.

Bu reau cracy The ory and Re search

65

The Aston Group stud ied the con text of the or ga ni za tion, which was a large num ber of vari ables in clud ing con tin gency fac tors, some in side the organization and oth ers in its environment (Pugh and Hickson 1976; Pugh and Hinings 1976; Pugh and Payne 1977). The main find ing to emerge from the Aston Group stud ies of or ga ni za tional con text is that size (meaning the number of employees) is a contingency of organizational struc ture. Size is pos i tively cor re lated with the struc turing of ac tiv i ties (i.e., spe cial iza tion and for mal iza tion; Pugh, Hickson, Hinings, and Turner 1969). Size is also pos i tively cor re lated with ver tical span, that is, the number of levels in the hierarchy. The major replication of the Aston Study, the National Study by Child (1973a), also found size to be an important correlate of organizational struc ture, as did numerous subsequent studies (Pugh and Hinings 1976). Bureaucratic structuring, such as specialization, formalization, and decentralization, in creases as size increases, but at a decreasing rate with respect to size (Child 1973a). Thus while size growth increases bureaucratization, the rate at which it does so de creases with suc ces sive size in cre ments. Hence while large or ga ni za tions are more bu reau cratic than small ones, they are much less so than would be the case if the difference was pro por tion ate to their size dif fer ences. This curvi linear relationship becomes lin ear if size is transformed logarithmically, which is a reason why size is usually transformed in bureaucracy research (other rea sons are to re duce the ef fects of out li ers and to make the size distribution less skewed and more normal; see Blau and Schoenherr 1971; Child 1973a). Child (1972a) argued that larger organizations have taller hi er archies down which au thor ity is del e gated, but such del e gated au thor ity is cir cum scribed by the in creas ing struc tur ing of ac tiv i ties. Size growth weak ens the di rect con trol over lower lev els by top man age ment, which instead relies more on the indirect controls provided by bureaucratic struc ture, such as writ ten job de scrip tions, rules, and pro ce dures. Thus mid dle- and lower-level man ag ers ex er cise more au thor ity in large, than in small, or ga ni za tions, but they are hedged around by tight job def i nitions, rules, procedures, and paperwork that limit their discretion. Thus centralization and structuring may be thought of as substitute means of con trol (though Child 1972a pre ferred to call them “al ter na tive means of con trol”). There is a trade-off be tween the two means of con trol: cen tral iza tion and spe cial iza tion-for mal iza tion (Zeffane 1989a).

66

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

In sum, size leads to a taller hierarchy, de cen tral iza tion, and struc turing, thereby shifting top management control over employees from centralization (i.e., di rect means) to spe cial iza tion-for mal iza tion (i.e., in di rect means). The relationship of size and most of the major Aston structural variables, that is, spe cial iza tion, stan dard iza tion, and doc u men ta tion, were replicated in subsequent, empirical field studies (Donaldson 1996b, pp. 137-143) that covered organizations in diverse countries and of di verse types. (The data from many of these stud ies are avail able in the Aston Data Bank; Richards 1980.) For example, size has been shown to cor re late pos i tively with spe cial iza tion in forty stud ies that used the Aston scale. These in clude six teen coun tries: Al ge ria, Can ada, Egypt, Finland, France, Germany, Hong Kong, India, Iran, Japan, Jordan, Po land, Singapore, Sweden, the United Kingdom, and the United States. They include also the following different types of organizations: banks, churches, col leges, gov ern men tal, la bor un ions, man u fac tur ing firms, and hos pi tals (for details see Donaldson 1996b, pp. 138-139). These stud ies are re ported in Ayoubi (1981), Azumi and McMillan (1981), Badran and Hinings (1981), Blau, Falbe, McKinley, and Tracy (1976), Bryman, Beardsworth, Keil, and Ford (1983), Child and Kieser (1979), Clark (1990), Conaty, Mahmoudi, and Miller (1983), Donaldson and Warner (1974), Glueck (unpublished, in Aston Data Bank), Green wood and Hinings (1976a), Grinyer and Yasai-Ardekani (1981), Heron (unpublished, in Aston Data Bank), Hickson, Hinings, McMillan, and Schwitter (1974), Hinings and Lee (1971), Hinings, Ranson, and Bryman (1976), Kuc, Hickson, and McMillan (1981), Payne and Mansfield (1973), Pugh and Hickson (1976), Reimann (1977), Routamaa (1985), Shenoy (1981), Tauber (1968), Wong and Birnbaum-More (1994), and Zeffane (1989b; some sources report more than one study). A meta-an a lytic re view con cluded that the size-spe cial iza tion re la tionship generalizes and has a positive, mean correlation of .53 across twenty-seven stud ies, to tal ing 1,066 or ga ni za tions (Miller 1987, p. 317, Ta ble 2). An other meta-anal y sis of a slightly dif fer ent set of stud ies (35, to tal ing 1,241 or ga ni za tions) cal cu lated a sim i lar, pos i tive cor re la tion of .61 and showed that, correcting for the reliability of the spe cial ization scale, the true cor re la tion is .82 (Donaldson 1986, p. 90, Ta ble 9),

Bu reau cracy The ory and Re search

67

which is con sid er able. For stan dard iza tion, its mean cor re la tion was .55, across nine studies (309 organizations; Donaldson 1986, pp. 80-81), which corrected is a true correlation of .69 (Donaldson 1986, p. 90, Table 9). For documentation, its mean correlation was .42, across twenty-four stud ies (834 or ga ni za tions) in Miller (1987, p. 317, Ta ble 2) and .51 across twenty stud ies (595 or ga ni za tions) in Donaldson (1986, pp. 82-83), which corrected is a true correlation of .59 (Donaldson 1986, p. 90, Ta ble 9). In sum mary, the gen eral re la tion ships are: size and spe cial iza tion are correlated over .8, size and standardization are correlated almost .7, and size and doc u men ta tion are cor re lated al most .6. Size and spe cializa tion are very closely re lated, while size and for mal iza tion (i.e., stan dardization and documentation) are closely related. Thus inc reasing organizational size leads to greater bureaucratic structuring, that is, specialization, rules and pro ce dures, and pa per work, in or ga ni za tions of many types around the world, con sis tent with Weber (1964). The main ex cep tion is cen tral iza tion that is not con sis tently cor re lated with size in the replication studies (Donaldson 1986; Miller 1987). There prob a bly is a gen eral re la tion ship be tween size and de cen tral ization, but technical problems in the measurement of decentralization obscure it in many studies (Green wood and Hinings 1976b; Grinyer and Yasai-Ardekani 1980; Lincoln, Hanada, and McBride 1986; Mansfield 1973; Marsh 1992). More over, size ef fects on de cen tral iza tion can easily be conf ounded by public accountability because that can have an effect of similar magnitude as size. A more thorough un derstanding of the re lationships among size, public accountability, and de cen tral iza tion needs to be at tained through fu ture re search. Public account abil ity measures the distinction be tween privatesector firms and public-sector (i.e., governmental) or ganizations. Pri vate-sector firms are traditionally free to pur sue prof it abil ity, whereas public-sector or ga ni za tions are traditionally ex posed to de mands for compliance with a broad range of pub lic con cerns for which they are held ac count able through their board of di rec tors (Pugh et al. 1969). Thus pub licly ac count able or ga ni za tions are more cen tral ized, whereas private-sector firms are more decentralized be cause they give more dis cre tion to the man ag ers as long as they pro duce ac cept able per for mance on a lim ited range of mea sures, for ex am ple, prof it abil ity. The

68

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

pos i tive cor re la tion be tween pub lic accountability and cen tral ization is quite im por tant and is at least as im por tant as size across a num ber of stud ies (Pugh et al. 1969; Rich ards 1980). The main scales used by the Aston Group, such as spe cial iza tion and standardization, were constructed from items that were conceptually closely re lated. These are the most use ful scales and are those in volved in the main find ings for which their re search has be come known, such as the positive correlation between size and structuring. However, some of the scales de rived af ter data col lec tion by fac tor anal y sis are not always so conceptually meaningful. For example, they identified a factor of con text that they called de pend ence, which was com posed of an as sort ment of scales (Pugh et al. 1969). The con cep tual co her ence of this fac tor has been crit i cized (Mindlin and Aldrich 1975).

Fit and Per for mance The Aston Study is writ ten in a pri mar ily de scrip tive way, re port ing as so ci a tions among variables; how ever, there is a sense that the struc tures stud ied are play ing a con struc tive role, with out claim ing them to be the most effective (Pugh and Hickson 1976; Pugh and Hinings 1976). The interpretation is vaguely functionalist but without much causal theory. Child (1972a, 1973a) provides more of a theoretical interpretation. This in cludes the ory of the con tin gency type. Or ga nizations of larger size needed to adopt more structuring in order to provide more in di rect con trol over em ploy ees, be cause di rect con trol was be ing di luted by de cen tral iza tion (Child 1972a, 1973a). This led to the expectation that organizations of larger size would have higher performance if they had greater struc tur ing, whereas they would have lower performance if they had less struc tur ing. The un der ly ing logic was of the contingency theory type. There is a fit between size and structuring such that larger organizations require more structuring than smaller or ga ni za tions (Child 1975). Or ga ni za tions that are in fit (e.g., a large organization with high structuring) will consequently have higher per for mance than those in mis fit (e.g., a large or ga ni za tion with only me dium struc tur ing). Child (1975) showed em pir i cally this relationship be tween fit (of struc tur ing to size) and per for mance (see also Khandwalla 1973). Thus the Aston Group revealed that size is a contingency of organizational structure.

Bu reau cracy The ory and Re search

69

Structural Differentiation The idea of size as a con tin gency of or ga ni za tional struc ture re ceived considerable support also from the work of Blau and his coll eagues. Blau ad vanced an ar gu ment that size in crease led to per va sive changes in or ga ni za tional struc ture. Some of these are that in creas ing size leads to more bureaucratic structuring such as specialization, rules, and documents, and more decentralization (Blau and Schoenherr 1971). This as pect of his the ory closely par al lels the ar gu ments of the Aston Group and helps sup port the idea that size causes bu reau cracy, un der stood as structuring with decentralization. Blau and Schoenherr (1971) ar gue that size causes a shift from di rect to more in di rect forms of con trol, again sim i lar to the Aston Group. How ever, in re gard to the specialization as pect of struc tur ing Blau made a much more in-depth ar gu ment. The spe cial iza tion of roles, as stud ied by the Aston Group, is part of the more encompassing concept of structural differentiation ad vanced by Blau (1970). Structural dif ferentiation refers to the breaking up of the organization along any dimension (Blau 1970). As instances, the number of divisions, the num ber of sec tions in each di vi sion, the num ber of oc cupa tional job ti tles, and the num ber of lev els in the hi er ar chy are all aspects of structural differentiation. Structural differentiation refers to the shape of the hi er ar chy rather than to the dif fer ences in the at ti tudes of the em ploy ees and con trols in dif fer ent de part ments that Law rence and Lorsch (1967) term differentiation (which was dis cussed in the pre vi ous chap ter). Blau (1970) pres ents a highly ab stract for mal the ory of the ef fect of size on struc tural dif fer en ti a tion. The the ory is rigorously deduced from a few basic ax ioms. The main ideas are as follows. Struc tural dif fer en ti a tion in creases with size, but at a de creas ing rate with respect to size (Blau 1970). Thus structural differentiation in creases as size in creases, but its slope be comes shal lower at larger size, so that the curve flattens out somewhat from initially being steep at smaller sizes to being shallow at larger sizes (see Figure 3.1). Initial increments in size pro duce greater in creases in struc tural dif fer en ti ation than in cre ments of the same mag ni tude when the or ga ni za tion is larger in size. The rate at which struc ture be comes more dif fer en ti ated as size in creases, de creases with in creas ing size. This re la tion with size is shown empirically for many aspects of structural differentiation,

70

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

Fig ure 3.1. Size and Struc tural Differention

such as the number of di visions, the number of job titles, and the number of lev els in the hi er ar chy (Blau and Schoenherr 1971). An other re lated the o ret i cal prop o si tion is that the average size of an organiza tional component, for example, a division, grows larger as the size of the overall organization in creases. Again, the average span of control (the number reporting di rectly to a man ager) in creases as organiza tional size increases (Blau 1970), at a decreasing rate with respect to size. This prop o si tion leads in turn to one of the most un ex pected parts of Blau’s the ory. The con ven tional wis dom is that as or ga ni za tions grow in size they be come top heavy. The in creas ingly tall hi er ar chy vis i ble as or ga ni zations

Bu reau cracy The ory and Re search

71

grow in size, leads, it is widely be lieved, to a rapid growth in managers and their as so ci ated ad min is tra tive staff rel a tive to the in crease in op er ating per son nel. Critics as sert that there is a dis pro por tion ate growth in man ag ers and ad min is tra tors to work ers as or ga ni za tions grow in size (Par kin son 1957), lead ing to ex ces sive over head cost and loss of speed and ef fec tive ness. Thus, in this view, the pro por tion of total em ploy ees who are man ag ers and ad min is tra tors would be expected to rise with increasing orga ni za tional size. How ever, Blau (1970) showed em pir i cally the op posite: that the propor tion of total employees who are manag ers and ad min is tra tors de creases with in creas ing or ga ni za tional size. Thus, far from being top heavy, large or ga ni za tions en joy econ omies of scale in ad min is tra tion. This chal lenges the stan dard eco nomic doc trine, which is that large or ga ni za tions have econ o mies of scale in pro duc tion and pro cure ment, but dis econo mies in ad min is tra tion. Blau’s theory explains the economies of scale in administration in the following way (Blau 1970). Because the number of levels in the hierarchy in creases with in creas ing size, if the span of con trol re mained constant with increasing size, then the ratio of man ag ers to work ers would rise as size in creases. How ever, as noted above, the span of con trol ac tu ally in creases with in creas ing or ga ni za tional size, suf fi ciently that the ra tio of managers to workers ac tually decreases as size in creases, pro duc ing the econ o mies of scale in ad min is tra tion. The span of con trol in creases with in creas ing size for two rea sons. First, or ga ni zations use more bureaucratic structuring (e.g., rules) as they grow larger and this re lieves man ag ers of hav ing to make de ci sions, so per mit ting their spans of con trol to widen. Sec ond, as or ga ni za tions grow larger, jobs be come more spe cial ized, hence jobs within a work group be come more ho mog e nous, mak ing su per vi sion eas ier, so that, again, the spans of control of supervisors and managers widen. These two factors are only partially countered by an opposite tendency: the increasing complexity of the more structurally differentiated larger organization makes its coordination more difficult, so consuming managerial time and re duc ing their spans of con trol. The for mer two factors combined are stronger t han the latter single factor, so that increasing or ga ni za tional size leads to wider spans of con trol and thus pro duces econ o mies of scale in ad min is tra tion. The pro por tion of to tal em ploy ees who are man ag ers and ad minis tra tive staff may be re ferred to as the ad min is tra tive in ten sity of an organization. As just discussed, the relationship between size and

72

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

Fig ure 3.2. Size and Ad ministrativeIntensity

administrative in ten sity is neg a tive. More over, it is curvilinear, rather than linear (see Figure 3.2). Administrative intensity decreases with size at a de creas ing rate with re spect to size (Blau 1970). Thus, while economy of scale is greater for larger than smaller organizations, its rate of increase ta pers off as size in creases (Blau 1970). This neg a tive curvilinear relationship be tween size and administrative intensity is rendered linear by transforming size logarithmically (see Blau and Schoenherr 1971). As just seen, ad min is tra tive in ten sity de creases with in creas ing size be cause for mal iza tion and span of con trol in crease with

Bu reau cracy The ory and Re search

73

increasing size. However, as also seen, both formalization and span of con trol in crease at a de creas ing rate with in creas ing size. There fore the rate at which economies of scale in administration in crease as size increases, de creases as size in creases. There are other as pects and sub tle ties to Blau’s the ory of struc tural differentiation. Here we have discussed some of the main points for which his the ory is best known. Blau’s the ory of struc tural dif fer en ti ation gives us a clear the o ret i cal model of how the hi er ar chy de vel ops as the organization grows in size. The theory encompasses many aspects, both ver ti cal (e.g., hi er ar chy) and hor i zon tal (e.g., num ber of divisions), that are applicable across work organizations of all types. Thus the theory gives us the fundamental anatomical structure (i.e., reporting re la tion ships such as shown on an or ga ni za tion chart) onto which other variables can be added, such as the distribution of de cisions (cen tral iza tion) or the use of rules and pro ce dures (struc tur ing). The combination of this model of struc tural dif fer en ti a tion together with bureaucratization cap tures many of the important features of how or ga ni za tional structure changes with increasing or ganizational size. The re la tion ships be tween or ga ni za tional size and nu mer ous as pects of structural differentiation that Blau and Schoenherr (1971) found have been rep li cated in many em pir i cal stud ies (see Blau 1972), in cluding pri vate-sec tor or ga ni za tions (e.g., Goldman 1973) and la bor un ions (Jarley, Fiorito, and Delaney 1997, p. 852, Table 2). Also, the re la tionships hold in countries other than the United States, for example, Britain (Child 1973a) and Ger many (Child and Kieser 1979). Thus the size-struc tural dif fer en ti a tion re la tion ship gen er al izes. Argyris (1972) ar gues that the re la tion ships be tween size and struc tural dif fer en ti a tion found by Blau are created by U.S. civil service regulations. Yet, as we have just seen, these same re la tion ships are found in com mer cial or ganizations and in those outside the United States, supporting Blau’s orig i nal in ter pre ta tion that there is a gen eral causal con nec tion be tween size and struc ture. Blau’s the ory has at tracted a con sid er able criti cal lit er a ture (Childers, Mayhew, and Gray 1971; Hummon 1971; Mayhew, James, and Childers 1972; Meyer 1971; Sprecht 1973), some of which has stimulated response from Blau (1971). Meyer (1979) found in a study over time that there was also a sub stan tial ef fect of the time of founding of the or ganization, ad ditional to that of size. Daft and

74

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

Bradshaw (1980) show that while administrative differentiation in uni ver si ties is caused by size, dif fer en ti a tion of ac a demic de part ments is much less so and is more caused by the rise of new specialisms in the wider academic community related to the growth of knowledge and stu dent de mand. Thus there are lim its to the generalizability of Blau’s the ory. There have been other crit i cisms of Blau’s the ory (e.g., Marsh and Mannari 1989), some of which have been re but ted (see Donaldson 1996b, pp. 58-107), while others (e.g., Freeman and Hannan 1975; Free man and Kronenfeld 1973) will be re but ted in Chapter 6. The neg ative relationship be tween size and ad min is tra tive in ten sity that Blau and Schoenherr (1971) found em pir i cally and ex plained theoretically is found also in subsequent empirical stud ies (Goldman 1973; Hickson, Pugh, and Pheysey 1969; Hinings and Lee 1971; Van de Ven and Ferry 1980). There is some vari a tion in findings, with a study find ing no re la tion ship (Holdaway et al. 1975) and some stud ies a pos i tive re la tion ship that con tra dicts the theory (Al-Jibouri 1983; Ayoubi 1981; Marsh and Mannari 1981). How ever, a metaan a lytic re view finds that the av er age cor re la tion be tween or ga ni za tional size and the pro por tion of man ag ers is –.45 (for de tails see Donaldson 1996b, pp. 104-105). Thus on balance the bulk of evidence sup ports Blau’s theory that in creasing size leads to a lower pro por tion of man ag ers and staff to total em ployees (see also Marsden, Cook, and Kalleberg 1994). While this en cour ages the view that bureaucratization is functional, this re ally requires some closer ex am i na tion of the economics. For de clin ing pro por tions of man ag ers and staff to di rectly raise ef fi ciency, the cost of these per son nel must also be de clin ing. Yet larger or ga ni za tions have more hi er ar chi cal lev els and each level tends to re ceive higher sal ary and other ben e fits, so that the av er age cost per man ager could pos si bly rise with in creas ing size. Therefore the question arises as to whether the de cline in the pro por tion of man ag ers as size in creases is suf fi cient to more than off set the rise in their av er age costs. Yet per son nel costs are not the ul ti mate yard stick, be cause even if the average costs of managers were increasing as organizations grew, this could be more than coun ter acted if man ag ers in creased or ga ni zational performance by enhancing effectiveness. Thus the question becomes whether the performance per manager is rising as size in creases. Thus there is a need for fu ture in quiry to ex am ine in more de tail is sues of per son nel costs and per for mance rel a tive to the size of

Bu reau cracy The ory and Re search

75

the managerial component, to address more definitively the issue of economies of scale in administration. As we have seen, the Aston Group found that bu reau cracy was composed of specialization, formalization, decentralization, and vertical span. Clearly, ver ti cal span is a ver ti cal as pect of struc tural dif fer en ti ation in Blau’s terms. Moreover, conceptually the Aston functional spe cial iza tion vari able would be sub sumed un der struc tural dif fer enti a tion, as a hor i zon tal as pect (Child 1973a). Thus the Aston model of bureaucracy can be said to be composed of formalization, de cen traliza tion, and struc tural dif fer en ti a tion. This ac cords with the model of bu reau cracy that emerges from Blau’s re search (Blau and Schoenherr 1971). Thus both the Aston Group and Blau find that bureaucracy is composed of formalization, de cen tral iza tion, and struc tural dif fer enti a tion, with the complementarities be ing that the Aston Group had a fuller set of mea sures of for mal iza tion, while Blau mea sured a fuller set of as pects of struc tural dif fer en ti a tion. The work of the Aston Group and Blau has been con firmed in a large study of a di verse range of es tab lish ments in the United States (num ber ing more than 600). This found that size was pos i tively re lated to the number of departments and vertical levels (Marsden et al. 1994, p. 914), that is, to structural differentiation. Also, size was positively related to formalization and decentralization (Marsden et al. 1994, p. 914). Fur ther more, the struc tur ing clus ter was ex tended by show ing that in ter nal la bor mar kets cor re late pos i tively with for mal iza tion, as Weber the o rized, and with size (Mar sden et al. 1994, p. 914). Size was also neg a tively re lated to ad min is tra tive in ten sity, show ing econ o mies of scale. The find ings are syn the sized into a causal model (Mar sden et al. 1994, p. 914, Fig ure 1), in which (sim pli fy ing) size causes de part ments that cause formalization, which reduces ad min is tra tive in ten sity; size also causes vertical lev els that cause de cen tral iza tion, which reduces administrative in ten sity (the causal model will be fur ther dis cussed in Chap ter 4). In this way the Aston and Blau el e ments of struc ture are integrated into one causal model that empirically val idates the theory that size causes for mal iza tion, de cen tral iza tion, and struc tural differentiation. Both the Aston Group and Blau ar gue that size is an im por tant de termi nant of or ga ni za tional struc ture. More over, both are at least broadly compatible with contingency theory in that effective structures for

76

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

large organizations are different from effective structures for small organizations, so that size moderates the relationship between struc ture and per for mance. Thus the con tin gency the ory of or ga ni za tional structure in cludes also a recognizable school of thought in bu reaucracy the ory, which ar gued the im por tance of size as a con tin gency of organizational structure. This bureaucracy theory school had a unity to its work in the same way that the or ganic the ory had a unity. Clearly each school cham pi oned a dif fer ent model of or ga ni za tional struc ture and a different con tingency factor (size or task) as being the more important. There was a fur ther the o ret i cal sig nif i cance of the Aston Group and Blau. As noted above, Weber (1964) as serted the pos i tive func tions of bu reau cracy; how ever, later stud ies ar gued that bu reau cratic struc ture pro duced dysfunctions, such as ri gid ity, low out put, and even sub ver sion of organizational goals (Crozier 1964; Gouldner 1954; Merton 1949; Selznick 1949). This of fered a coun ter to Weber and ar gued the recalcitrance of hu man beings subject to attempts at top-down controls and that human organizations to some extent have a life of their own, that is, the emer gent view of or ga ni za tion (Scott 1992). The dysfunctions of bureaucracy were shown empirically by case study methods (Crozier 1964; Gouldner 1954; Selznick 1949). In contrast, the Aston Group and Blau instead used quantitative comparisons across organizations and produced a far more positive account of bureaucracy. They lent sup port to Weber by ar gu ing that bu reau cracy, a top-down or ga ni za tional struc ture, is ef fec tive for large or ga ni za tions and even pro duces econ o mies of scale in ad min is tra tion. While there may well be dysfunctional consequences to bureaucracy, as the case studies revealed, bureaucracy has positive functions that more than compensate for them. By comparing w hole organizations and their overall per formances, the bureaucracy the ory research of the Aston Group and Blau was able to show the over all posi tive ef fect of bu reaucracy that the case studies could not examine. Fur ther, while bureaucracy the ory mea sured cen tral iza tion and for mal iza tion, that is, some as pects of or ganic struc ture, there is no mea sure ment in their research of other as pects of the or ganic struc ture, such as par tic i pa tion or lat eral com mu ni ca tions (Pugh et al. 1968). This dis plays the lack of ad her ence to the or ganic model by the fol low ers of bu reau cracy the ory and the extent to which they fo cused on Weberian bu reau cracy the ory.

Bu reau cracy The ory and Re search

77

Divisionalization As argued in Chapter 1, divisionalization is a com ponent of bu reaucratic struc ture. It re fers to the struc ture at the apex, that is, the way the work of the organization is divided among the senior man agers who report to the CEO. In a func tional struc ture the work is di vided into dif fer ent func tions, such as man u fac tur ing and sales, so that these are the ma jor group ings in the or ga ni za tion (Child 1984). In con trast, in a divisional structure the work is divided into different divisions, such as by prod ucts or cus tom ers or re gions, so that these are the ma jor groupings in the organization (Child 1984). Divisionalization is mainly caused by the need to fit strat egy, such as di ver si fi ca tion, which is its major contingency. The diversification con tingency is, in turn, part of the more ab stract con tin gency of task in ter de pen dence. Re lated struc tural types, such as mul ti na tional and ma trix struc tures, are also com po nents of bu reau cratic struc ture that are af fected by the task inter dependence con tin gency. Divisionalization should be considered to be a part of the bu reaucratic type of struc ture. It is pos i tively re lated to bu reau cracy. Divisiona lization is associated with greater levels of bureaucratic structuring: functional specialization, pro fes sional qual i fi ca tions, doc u men ta tion, and ver ti cal span (Grinyer and Yasai-Ardekani 1981, p. 478). More over, divisionalization is also associated with decentralization (Chenhall 1979; Grinyer and Yasai-Ardekani 1981, p. 478), empirically con firm ing that the di vi sional struc ture is a de cen tral ized or ga ni za tional struc ture (Chandler 1962). Thus when a large firm divisionalizes, this is not a move away from bureaucracy but rather increases bu reau cracy. The de cen tral iza tion of the divi sional structure does not im ply that it be comes less bu reau cratic, rather it is dis play ing the fa mil iar gen eral pat tern of the bu reau cratic model of be com ing more bu reau crat i cally structured while be com ing more de cen tral ized. Thus divisionalization is just a fur ther stage in the pro cess of grad ual change over time from a simple (i.e., unbureaucratic) structure to a highly bu reaucratic structure. Chan dler (1962) ar gued that strat egy leads to struc ture so that di versi fi ca tion leads to divisionalization. He showed that cor po ra tions that diversified needed to adopt the di vi sional struc ture that matched their diversified operations, so that there is a fit between divisionalization

78

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

and strat egy as a con tin gency. In this way Chan dler con trib utes to contin gency the o ries of or ga ni za tional struc ture, though be ing a his to rian he did not state it as a con tin gency the ory. Chan dler (1962) in ten sively an a lyzes the his to ries of four large U.S. cor po ra tions. He con cludes that there is a pat tern whereby af ter chang ing their strat egy they even tu ally changed their or ga ni za tional struc ture, so that strat egy led to struc ture. In par tic u lar, firms di ver si fied and so changed from func tional to di visional struc tures be cause the lat ter al lowed more ef fec tive man age ment of di ver si fied cor po ra tions. In a broader sur vey of sev enty other large U.S. firms he finds the same pat tern (Chan dler 1962). Thus, for Chan dler, un di ver si fied firms can be ef fec tively man aged by a func tional struc ture, but diversified firms need a divisional struc ture to op er ate ef fec tively. Subsequent writers have formalized these insights to argue that the greater information-processing requirement posed by diversified operations makes im per a tive the adop tion of a struc ture in which each sep a rate prod uct-mar ket op er ates as a self-con tained busi ness by be ing an au ton o mous di vi sion (Galbraith 1973). This marks a for mal con tingency theory of the connection be tween structure at the apex of the firm and strat egy as the con tin gency fac tor. Whereas de part ments of a func tion ally struc tured firm are treated as cost and rev e nue cen ters, the di vi sions in a divisionalized firm are profit cen ters. This is a shift from con trol over means to con trol over ends (Mintzberg 1979). The sep a ration of the cor po rate head of fice from di vi sional man age ment sep a rates stra te gic from op er a tional de ci sion mak ing, which helps make stra tegic de ci sions more ob jec tive and ben e fi cial for the long-run in ter ests of the overall firm (Chan dler 1962). Chan dler’s sem i nal study has led to a con sid er able body of work that confirms the original the sis, shows its generality beyond the United States and ex tends it in var i ous ways. Whereas Chan dler (1962) con ducted case his to ries and clas si fied them to re veal pat terns, sub se quent re search ers have mea sured stra te gic and struc tural vari ables and used sta tis ti cal anal y ses to test for con nec tions. Such stud ies have con firmed Chandler’s insight that diversification is a major cause of division alization among large U.S. corporations (Fligstein 1985; Mahoney 1992; Palmer, Friedland, Jennings, and Powers 1987; Palmer, Jennings, and Zhou 1993; Rumelt 1974). Studies of large corporations in other coun tries have shown that strat egy leads to struc ture in those coun tries also. This holds true for Aus tra lia, Can ada, France, Ger many, It aly, Japan, New Zealand, and the United King dom (Ca pon, Christodolou,

Bu reau cracy The ory and Re search

79

Farley, and Hulbert 1987; Channon 1973, 1978; Chenhall 1979; Dyas and Thanheiser 1976; Hamilton and Shergill 1992, 1993; Khandwalla 1977; Pa van 1976; Suzuki 1980). Thus the causal con nec tion be tween di ver si fi ca tion and divisionalization is well es tab lished. The con tin gency the ory ex pla na tion of the con nec tion be tween di versification and divisionalization is the fit between these two variables that af fects per for mance. For an un di ver si fied firm a func tional structure is the fit, whereas for a di ver si fied firm a di vi sional struc ture is the fit. Conversely, for an undiversified firm a divisional structure is the misfit, and for a diversified firm a functional structure is the misfit. This theoretical model receives support from empirical studies of strategy-structure fit and performance (Donaldson 1987; Hamilton and Shergill 1992, 1993). The contingency fit effect on performance holds when con trol ling for pos si ble uni ver sal ist ic ef fects of di vi sional struc ture (Ham il ton and Shergill 1993). There is a cor re spon dence be tween the anal y ses of Chan dler (1962) and Thomp son (1967). In Chan dler’s case his to ries, the un di ver si fied firm is composed of functional departments that are interdependent with each other; indeed Chandler (1977) documents the rise of the modern busi ness firm by the ver ti cal in te gra tion of sup ply, man u facture, sales, and distribution functions. Subsequently, diversification creates multibusi ness firms in which the ac tiv i ties of one di vi sion are connected only loosely or not all with the other divisions. This corre sponds to the decrease in task interdependence in moving from se quen tial to pooled in ter de pen dence found in Thomp son’s schema. For Thompson (1967), the organizational subunits connected only by pooled interdependence become the major organizational units, that is, divisions, at the highest level in the hierarchy, cor responding to Chan dler’s the sis that di ver si fi ca tion leads to divisionalization. Hence both Chan dler (1962) and Thomp son (1967) dis cuss the con tin gency fac tor of task in ter de pen dence, al beit im plic itly in the case of Chan dler, and see it as lead ing to divisionalization. In dis cuss ing task interdependence in the context of or ganic the ory (Chap ter 2), we con cluded that it could be de com posed into task un certainty and interaction intensity. Here, too, diversification leads to divisionalization through decreasing task interdependence because intensity of in ter ac tion be tween the ac tiv i ties deal ing with each of the different product-markets decreases. The shift from sequential to pooled interdependence as the organization diversifies highly means

80

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

that the activities of the product-markets go from being directly connected to hav ing no di rect con nec tion, that is, from mod er ate to nil interaction intensity, and so from functional to di vi sional struc tures. Thus the reformulation of task interdependence as interaction intensity re gard ing the ex pla na tion of group ing of ac tiv i ties and their colocation applies also to divisionalization and its vari ants. Hence the task in ter de pen dence explanation of divisionalization can be re in terpreted as interaction int ensity, consistent with our reformulation. How ever, in dis cuss ing the lit er a ture we will con tinue to use the term task in ter de pen dence be cause that is more fa mil iar and is em bed ded in the literature. Nev er the less, the in flu ences at trib uted to the task in terde pen dence con tin gency can be re in ter preted, with some gain in clarity and pre ci sion, as be ing due to the in ten sity of task in ter ac tion. The task in ter de pen dence con tin gency can also be ap plied to dis tinguish different types within the divisional structure. The greater the task in ter de pen dence, the more cen tral ized the co or di na tion. The distinc tion is drawn be tween un re lated prod uct or ser vice com pa nies and those where the prod ucts or ser vices are re lated or are ver ti cally in te grated (Rumelt 1974). Where divisions are unrelated (i.e., low task inter dependence) they are highly autonomous, have little con tact with the head office, and few managers transfer between them (Lorsch and Allen 1973; Pitts 1974, 1976, 1977). Also the head of fice is small, and the di vi sion and its man ager are ap praised and re warded ac cord ing toprofit abil ity of the di vi sion (Lorsch and Al len 1973; Pitts 1974, 1976, 1977). In contrast, the more the divisions are related, the less autonomous they are, they have more con tact with the head of fice and more man ag ers transfer between them (Lorsch and Allen 1973; Pitts 1974, 1976, 1977). Also the head of fice is large, and the di vi sion and its man ager are appraised according to their contribution to overall company profit abil ity (Lorsch and Al len 1973; Pitts 1974, 1976, 1977). While task in ter de pen dence caused by stra te gic di ver si fi ca tion is the main contingency of divisionalization, size also leads to division alization, additional to the effect of di versification (Grinyer and Yasai-Ardekani 1981). This size ef fect may be in ter preted as due to the need to curb managerial op por tun ism, which in creases as func tional organizations grow larger and more complex (Williamson 1970). How ever, the size ef fect may also be in ter preted as due to op ti mal plant size plac ing lim its on the size of func tions (Khandwalla 1977). Again,

Bu reau cracy The ory and Re search

81

the size ef fect may be in ter preted as due to the need to de com pose the organization into small enough pieces that a manager can supervise each di vi sion with out ex ceed ing his or her cog ni tive ca pac ity (Jaques 1976). Each of these different interpretations is a contingency theory, because divisionalization needs to fit size in order to pro duce higher performance, albeit for dif fer ent rea sons in each case. Studies of the relative effects of strategy and size have been mixed (Chenhall 1979; Donaldson 1982a; Fligstein 1985; Grinyer and Yasai-Ardekani 1981; Grinyer, Yasai-Ardekani, and Al-Bazzaz 1980; Khandwalla 1977; Palmer et al. 1987; for discussion see Child 1982; Donaldson 1986; Grinyer 1982). However, stud ies of organizational change support strategy more than size as a cause of divisionalization (Donaldson 1987; Fligstein 1985; Hamilton and Shergill 1992, 1993; Mahoney 1992; Palmer, et al. 1987; Palmer, Jennings, and Zhou 1993), so that, over all, size is sec ond ary to strat egy (for de tails see Chap ter 6). Hence, while divisionalization is a com po nent of bu reau cratic structure, it dif fers in its cau sa tion. Whereas size is the ma jor con tin gency of bu reau cratic struc tur ing, par tic u larly of spe cial iza tion and for mal ization, and also of struc tural dif fer en ti a tion, task in ter de pen dence is the major contingency of divisionalization. Nevertheless, there is some over lap of cau sal ity be tween divisionalization and bu reau cratic structure in that size has a mi nor ef fect on divisionalization, while di ver si fication increases bureaucratic structure (Grinyer and Yasai-Ardekani 1981). Also, decentralization, which is part of both divisionalization and bureaucratic structure, is positively affected by both size (Pugh and Hinings 1976) and diversification (Grinyer and Yasai-Ardekani 1981). De spite these com mon al ties, the dis tinc tive fea ture of divisionalization is that it is a component of bureaucratic structure whose cause is mainly task in ter de pen dence, rather than size, whereas size is the main cause for the rest of bu reau cratic struc ture and is the cen tral contingency in bu reau cracy the ory.

Structures of Multinational Corporations One area of or ga ni za tional struc ture that has re ceived ex ten sive study is multinational cor po ra tions (MNCs). This has followed a task contin gency ap proach by ar gu ing that as firms go from just op erating

82

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

do mestically to operating in ter na tion ally, they need to change their or ga ni za tional struc ture (Stopford and Wells 1972). In par tic u lar, more geographically extensive operations can lead from the appropriate struc ture be ing func tional to it be ing an area or geo graph ical di vi sional struc ture in which di vi sions are spe cial ized by re gions or coun tries of the world (Davis 1972). Greater diversification by product or service can also lead to product or service divisions that control operations worldwide (Da vis 1972). The re search sup ports the con tin gency theory that task in ter de pen dence de ter mines the re quired or ga ni za tional structure: diversification of product or service tends to lead to worldwide product or ser vice divisions, whereas diversification by area tends to lead to world wide area di vi sions (Channon 1973; Rumelt 1974; Stopford and Wells 1972). The decisive point is on what ba sis self-sufficient di visions can be formed, given the task interdependencies (Galbraith 1973; Thomp son 1967). An in ter me di ate stage be tween purely do mestic and extensively multinational operations may see the use of an international division to which overseas subsidiaries report, leaving the domestic operations organized separately (Davis 1972). This is something of an anomalous structure and there is evidence that U.S. corporations have used this structure more than European cor po rations (Franko 1974), and in a more pro longed man ner (Daniels, Pitts, and Tretter 1984, 1985). Egelhoff (1982, 1988a, 1988b) has developed quite a comprehensive con tingency theory of MNC structures from the in for ma tion-pro cess ing re quire ment posed by var i ous con tin gencies, build ing upon Galbraith (1973). Bartlett and Ghoshal (1989) an a lyze MNCs in terms of the ten sion be tween lo cal re spon sive ness and global inte gra tion. Their con clu sion is that in creas ingly MNCs will con verge on the trans na tional strat egy, which has some tinges of a uni ver sal ist ic ar gu ment ap plied to strat egy. However, Ghoshal and Nohria (1993) offer a sophisticated con tingency the ory of the or ga ni za tional struc ture of mul ti na tional cor po rations. The main idea is that each strategy re quires a cer tain de gree of con trol over the global units in the MNC and also of lo cal re spon siveness, which requires that subsidiaries vary in their structures. These con trols and com plex i ties are ex pen sive so that they should be avoided if not re quired. Strat egy re fers spe cif i cally to the mul ti na tional strat egy be ing pur sued by the cor po ra tion. It var ies on two di men sions: forces for global integration and forces for local responsiveness. Each of these two dimensions is dichotomized, and their cross-classification

Bu reau cracy The ory and Re search

83

yields four strategies: international (low global, low local), global (high, low), multinational (low, high) and transnational (high, high). The struc ture re quired to fit each strat egy is: ad hoc vari a tion struc ture fits international strategy, structural uniformity fits global strategy, dif fer en ti ated fit struc ture is the fit to multi na tional strategy, and in te grated variety struc ture fits trans na tional strat egy. Structureis the de gree of con trol ex er cised by the mul ti na tional over its parts through centralization, formalization, and normative inte gra tion (shared val ues of man ag ers; Ghoshal and Nohria 1993). In the in ter na tional strat egy,the MNC is at tempt ing nei ther global in te gra tion of its sub sid iar ies nor to make them lo cally re spon sive (in stead it may emphasize exporting from the domestic plants). Therefore it needs little con trol over its sub sid iar ies and hence is low on all three means of con trol (i.e., cen tral iza tion, for mal iza tion, and nor ma tive in te gra tion), so that its sub sid iar ies can vary in their in ter nal struc tures in an ad hoc man ner, thus the MNC over all has the “ad hoc vari a tion” struc ture. In the global strategy, the MNC is attempting global in tegration of its sub sid iar ies with out en cour ag ing them to be lo cally responsive. Therefore it needs sub stan tial con trol over its sub sid iar ies and thus it is high on at least one of the three means of control (i.e., centralization, formalization, and normative integration), so that all its subsidiaries have uni form in ter nal struc tures, hence the MNC over all has “struc tural uni for mity.” In the mul ti na tional strat egy, the MNC is not at tempt ing global in te gra tion but rather encourages its subsidiaries to be lo cally re spon sive. There fore it needs each of its sub sid iar ies to fit its en vi ronment, spe cif i cally the lo cal re sources and level of com plex ity, which is termed “dif fer en ti ated fit.” In thetrans na tional strat egy, the MNC is attempt ing si mul ta neously global in te gra tion of its sub sid iar ies and to make them locally responsive. Therefore it needs subs tantial control over its sub sid iar ies of struc tural uni for mity, to gether with fit ting each subsidiary to its en vi ron ment, as in dif fer en ti ated fit, hence the over all struc ture of the MNC is “in te grated va ri ety.” In an empirical study of forty-one MNCs, sev en teen were found to be in fit be tween their strat e gies and struc tures, while twenty-four were in mis fit (Ghoshal and Nohria 1993). The MNCs in fit had higher perfor mance on av er age in re turn on as sets, growth in re turn on as sets, and growth in revenues, thus providing validation of the fit model (Ghoshal and Nohria 1993). The fact that most organizations in the study were found to be in mis fit, and so had re duced per for mance,

84

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

helps dis pel any illusion that contingency research always describes or ga ni za tions as being in fit and so is ir rel e vant to prac tice. Ex am ples of MNCs in fit are: Baker In ter na tional (ad hoc vari a tion struc ture fit ting international strat egy), Air Prod ucts and Chem i cals (struc tural uni formityfit ting global strat egy), Brit ish-Amer i can To bacco (dif fer en ti ated fit match ing multinational strategy), and Volvo (integrated vari ety fit ting trans national strat egy; Ghoshal and Nohria 1993). An ex am ple of a mis fit is Ko dak, which had an ad hoc vari a tion struc ture, fea tur ing weak integration and dif fer en ti a tion, that failed to pro vide the strong con trols and sys tem atic lo cal struc tural vari a tions needed to suc cessfully pursue its transnational strat egy. An other ex am ple of a mis fit is Siemens, which had an integrated variety structure, fea tur ing strong controls through centralization, formalization, and normative inte gra tion, as well as dif fer en ti ated fits of sub sid iar ies to their lo cal en vironments, which was excessive given its international strategy (Ghoshal and Nohria 1993). Thus MNCs need to invest in structures that provide the degree of central control and local variation that match their strategy, while avoiding the unnecessary expense and structural complexity of structures that are overelaborate for their strat egy (Ghoshal and Nohria 1993). In review, Ghoshal and Nohria (1993) of fer a so phis ti cated con tingency the ory that has fits at two lev els: the over all MNC struc ture to its worldwide strat egy and the sub sid iary to its lo cal en vi ron ment. While using the Aston concepts of centralization and formalization, the theory adds to these elements of structure the concept of nor mative integration, which is control through cor po rate cul ture, a di men sion of or ga ni za tional struc ture dif fer ent from the more tra di tional as pects and one that fo cuses on the shared val ues of man ag ers. This re flects an increasing interest among organizational theorists in or ganizational cul ture (Mar tin 1992) and a long-stand ing ques tion in so ci ol ogy about how far social systems are normatively integrated (Durkheim 1964) and how far this substitutes for bureaucratic controls (Lockwood 1964). Ghoshal and Nohria (1993) adopt the view that nor ma tive in tegra tion can sub sti tute for cen tral iza tion and for mal iza tion and hence that cul tural con trols are ef fec tive sub sti tutes for these struc tural controls (Ghoshal and Nohria 1993). Moreover, the study places cultural con trols in a con tin gency frame work, in that shared val ues are seen as beneficial only where strategies require structural in tegration (i.e., global and trans na tional strat e gies).

Bu reau cracy The ory and Re search

85

Ghoshal and Nohria (1993) state that the four structures provide varying levels of integration and dif fer en ti a tion, be cause global in tegration requires structural integration, while local re sponsiveness requires dif fer en ti a tion. Cross-clas si fy ing the four struc tures on in tegration and differentiation, respectively, they are: ad hoc variation (low, low), structural uniformity (high, low), differentiated fit (low, high), and integrated variety (high, high). Thus the study by Ghoshal and Nohria (1993) draws on the language of Law rence and Lorsch (1967), that is, differentiation and integration, but uses the terms in ways that are dif fer ent enough to re quire com ment. As we have seen, for Lawrence and Lorsch (1967), an organization dif fer en ti ates and then integrates be cause the dif fer en ti a tion splits up the or ga ni za tion while the integration makes these parts work in con cert. Thus for Law rence and Lorsch integration must accompany differentiation, so that a successful or ga ni za tion is either low on both dif fer en ti a tion and in tegration or high on both. In contrast, in Ghoshal and Nohria (1993), their four cell model makes clear that a suc cess ful or ga ni za tion can be high on in te gra tion, while also low on dif fer en ti a tion (struc tural uniformity) and also can be low on integration while also being high on differentiation (differentiated fit). Thus two of the four cells in their model are contrary to Lawrence and Lorsch. This shows that for Ghoshal and Nohria, dif fer en ti a tion and in te gra tion are not two sides of the same coin that need to match each other, as they are for Law rence and Lorsch. For Ghoshal and Nohria, differentiation and integration re fer to struc tural as pects at dif fer ent lev els of the or ga ni za tional hi erar chy. Dif fer en ti a tion is struc ture at sub sid iary level (and the dif fer ent re la tion ship of each sub sid iary to the head of fice), whereas in te gra tion re fers to the con trol achieved by the MNC over its sub sid iar ies through its overall structure (centralization, formalization, and normative integration). Thus dif fer en ti a tion here is the dif fer ence be tween subsidiaries that may or may not need to be in te grated, de pend ing upon the strategy. The multinational strat egy re quires differentiation, but not integration. Similarly, the global strategy requires integration, despite the lack of differentiation between subsidiaries. Therefore while the study uses the con cepts of Law rence and Lorsch, they are not embedded in the same theory. The study by Ghoshal and Nohria (1993) has elements of both the bureaucracy and organic theories. It draws upon bureaucracy theory for the struc tural vari ables of cen tral iza tion and for mal iza tion (Pugh

86

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

et al. 1968). It draws upon or ganic the ory for in te gra tion through normative integration, that is, through shared values, communications and in ter per sonal net works (Burns and Stalker 1961). In deed, Nohria and Ghoshal (1997) use the termdif fer en ti ated net work to re fer to the modern MNC, which displays af finities with the organic structure. High lo cal re spon sive ness re quires fit to the con tin gen cies of re sources and complexity in the environment of the subsidiary. This in some ways echoes the theory of Lawrence and Dyer (1983) that en vi ronments vary on two dimensions: resource scarcity and information complexity, which re flects re source de pend ence (Pfeffer and Salancik 1978) and or ganic the o ries (Burns and Stalker 1961), re spec tively. The main con tin gency of Ghoshal and Nohria (1993) at the cor porate level is strat egy, with the de gree of global in te gra tion be ing sought implicitly determining the degree of task interdependence be tween sub sid iar ies. Over all, the re search on MNCs sup ports the idea that task is a con tin gency of or ga ni za tional struc ture.

Matrix Structures Many MNCs use a matrix structure (Egelhoff 1988a; Galbraith and Kazanjian 1988), though this struc ture may also be used by or ga ni za tions that oper ate purely do mestically (Burns 1989; Da vis and Law rence 1977; Kolodny 1979). Three main types of matrix may be distinguished: functional-product (or project), functional-area, and product-area (instead of product there may be a diff erentiation by service or cus tomer). The es sence of ma trix is that an em ployee within the ma trix has two bosses, who in turn re port to a com mon boss. Be low this com mon boss there are two di men sions, for ex am ple, one of var i ous functions and one of various products, with each function and prod uct headed by a sep a rate man ager. The ma trix is an in ter me di ate struc ture, so that, for in stance, the func tional-prod uct ma trix com bines el e ments of both a func tional and a prod uct di vi sional struc ture. There is differentiation both of functions and products, with managers in charge of each at the same level. There fore the prod uct unit is less than a com plete di vi sion, be cause func tional spe cial ists within it are re porting also to func tional man ag ers out side it, so that the prod uct unit has less au ton omy than it would if it were a di vi sion.

Bu reau cracy The ory and Re search

87

Matrix structures are appropriate where there is dual fo cus (Da vis and Law rence 1977). This cor re sponds to an in ter me di ate level of task interdependence (Donaldson 1985, pp. 167-171). For instance, each prod uct or pro ject is a sep a rate en tity in or der to fa cil i tate its in ter nal co or di na tion, but certain resources are drawn from central functions (Corey and Star 1971, pp. 61-107). Thus the task interdependence between the products is less than for a func tional struc ture yet more than for a product divisional structure. Similarly, a functional-area ma trix fits a sit u a tion where ar eas are di verse enough to be re quired to be man aged sep a rately, but some shar ing of cen tral func tions is sought, so that again task interdependence is medium. Again, a product-area matrix fits a situation where both products and areas are diverse enough to be re quired to have each their own man ag ers, with re source units that re port to both, so that again task in ter de pen dence is me dium. Thus the task interdependence contingency de termines where the matrix structure, and what type of matrix, is a fit (Donaldson 1985, pp. 167-171). The issue arises as to whether matrix structures should be seen as part of or ganic the ory or of bu reau cracy the ory. Ma trix struc tures are, in the main, just a more com plex form of hi er ar chy, rather than be ing in her ently or ganic struc tures. They have been dis cussed by pro po nents of or ganic the ory such as Law rence (Da vis and Law rence 1977) and are some times de scribed as be ing de cen tral ized. Yet hav ing two di men sions of man ag ers at a level in the hi er ar chy in creases the num ber of managers at that level rather than delegating decision making down from that level. Managers from each side of the matrix may have to confer be cause nei ther pos sesses com plete au thor ity over a sub or di nate, but this is really just complexity in a hierarchy, rather than being a nonhierarchical structure. Again, the subordinate has two superiors, and this vi o lates the clas si cal school of man age ment pre scrip tion about unity of com mand (Brech 1957). How ever, hav ing two bosses is not less hi er ar chi cal than hav ing one boss—indeed it could considered more hi er ar chi cal—and so is compatible with a mech a nis tic structure. The functional-project sub type of the ma trix struc ture fea tures a crosssectional pro ject team, which is tem po rary and which is in tended to have an or ganic struc ture in ter nally, in or der to fa cil i tate in ter ac tion among the var i ous spe cial ists to pro mote prob lem solv ing. How ever, the other sub types of the ma trix, such as func tional-area or prod uct-func tional,

88

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

have sub or di nate units that are per ma nently re port ing up to their two bosses and so are not tem po rary ar range ments (Child 1984) nor nec essar ily or ganic in their in ter nal op er a tions. Research identifies the enhanced po tential for conflict in a ma trix be cause of sub or di nates hav ing two bosses (Fisher and Gitelson 1983; Joyce 1986; Knight 1977), so Davis and Lawrence (1977) recommend that matrix structures need to be supported by a matrix cul ture of openness and problem solving, which is organic in na ture. How ever, not all struc tures will nec es sar ily im ple ment such a cul ture. More over, a cul ture of open ness and prob lem solv ing could be adopted by or ga ni zations that lack a matrix structure. Therefore, overall, matrix struc tures in gen eral (the func tional-pro ject sub type ex cepted) do not seem to be inherently organic. They seem to be just a more complex type of hierarchy. Hence they can be subsumed within bureaucracy the ory as part of the sub sec tion of that the ory that deals with the task interdependence contingency, as a vari ant of the dis tinc tion be tween functional and divisional structures. Thus the stream of re search con nect ing strat egy with struc tures such as functional, divisional, and matrix is compatible with the con tingency the ory of task in ter de pen dence.

Other Possible Contingencies Are there other pos sible contingencies of bureaucratic struc ture in ad di tion to or ga ni za tional size? All con tin gen cies are causes of or ganizational structure, in that organizations change their structures to attain fit and performance. However, a variable could be a cause of organizational struc ture with out it be ing a con tin gency. In such a case the vari able causes change in or ga ni za tional struc ture but not be cause there is some un der ly ing fit that the or ga ni za tion seeks to at tain in ord er to gain performance. There are two steps involved in showing that a vari able is a con tin gency of or ga ni za tional struc ture. The first step is to show that the vari able is a cause of struc ture. The sec ond step is to show that the cause is a con tin gency. We will dis cuss here four pos si ble contingencies of bureaucratic structure: public accountability, en vironmental hos til ity, CEO per son al ity, and na tional cul ture. As seen above, public accountability is pos itively correlated with centralization (Pugh et al. 1968) and appears to be one of its causes.

Bu reau cracy The ory and Re search

89

Accepting that it is a cause, is it a con tin gency? If pub lic ac count abil ity raises cen tral iza tion be cause the gov ern ment re fuses to del e gate certain decisions be low the board of di rec tors (or equiv a lent body), then the orga ni za tion is sub ject to a causal force. How ever, if there is no fit be tween public accountability and centralization that raises per for mance, then public accountability is not a contingency. In this way pub lic ac count abil ity could be a cause of cen tral iza tion with out cau sation being of the functionalist kind, that is, driven by the search for some ben e fi cial out come. Whether a factor is a con tingency of organization de pends on whether aligning the structure and the contingency produces higher performance. Hence a contingency factor is always em bedded in a con tin gency the ory that spec i fies per for mance as re sult ing from a fit of the or ga ni za tional struc ture to the con tin gency. Dem on strating that a cause is a contingency in volves showing that the fit posited by the theory is em pir i cally valid, that is, that fit af fects per for mance. In the ab sence of the ory and ev i dence about why cen tral iza tion fits pub lic ac count abil ity and with what ben e fi cial con se quences for per for mance, then public accountability would not be a contingency of cen tral ization. Similar considerations apply to environmental hostility, CEO personality, and na tional cul ture. Khandwalla (1977) identified a factor of environmental hostil ity that is positively correlated with centralization of decision mak ing. Environmental hostility would be a contingency of centralization if centralization fits hos tile sit u a tions, in that, for ex am ple, top man agement can act more quickly and de ci sively to meet a threat by cen traliz ing. Here the outcome of centralization un der hos tility would be functional and positive, that is, performance would be raised. In contrast, hos til ity could be just a cause of cen tralization, due, for example, to threat-rigidity (Staw, Sandelands, and Dutton 1981). In this case, hostility causes senior managers to feel threatened and become de fen sive, so they move to mak ing all de ci sions them selves and no lon ger trust their sub or di nates to ex er cise del e gated au thor ity. The enhanced centralization result ing from hostility cre ates ri gidity, which is dysfunctional, negatively affecting performance. Therefore there would be no un der ly ing func tion al ist type of fit be tween hos til ity and cen tral iza tion, so hos til ity would not be a con tin gency. Which of these views is sounder needs to be established through empir i cal research.

90

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

Turning to CEO personality, consider, for instance, the need for achieve ment of the CEO. This is pos i tively cor re lated with cen tral iza tion (Miller, Droge, and Toulouse 1988). If there is some fit between hav ing a CEO with high need for achieve ment and cen tral iza tion that raises per for mance, then that CEO per son al ity need is a con tin gency of cen tral iza tion. For ex am ple, it might be that high need for achieve ment causes CEOs to work very hard and ef fec tively, but only if they can see them selves as the cause of the re sult ing gains in or ga ni za tional per formance. In such a case, centralization fits a CEO with high need for achievement, so that need for achievement is a contingency. In stead, how ever, it could be that a CEO with high need for achieve ment has a need al ways to be cen ter stage in the or ga ni za tion and so uses his or her power to with hold del e ga tion down to sub or di nates, even though this is coun ter pro duc tive. In such a sce nario, CEO need for achieve ment is a cause of cen tral iza tion but not a con tin gency of cen tral iza tion. Hence again, whether the causal factor, CEO need for achieve ment, is also a con tin gency de pends on whether there is a fit of it to cen tral iza tion that raises performance. Lewin and Stephens (1994) argue that CEO per son al ity causes or ga ni za tions to adopt a struc ture that is a mis fit. In this case, CEO personality is a cause of structure but not a contingency fac tor that can im prove per for mance. Na tional cul ture has be come a fac tor that has at tracted an in creas ing number of contingency the ory re searchers (Hickson and McMillan 1981; Lammers and Hickson 1979). Early structural contingency theory fo cused on con tin gen cies that were pre sumed to have uni ver sal ef fect, such as tech nol ogy or size (Blau 1970; Pugh et al. 1969). Later re search exam ined whether some na tional cultures led to dif fer ent re la tion ships be tween con tin gency and or ga ni za tional struc ture (e.g., Hickson, Hinings, McMillan, and Schwitter 1974; McMillan, Hickson, Hinings, and Schneck 1973; Zeffane 1989a). Ini tially the method was to see whether relationships found in countries such as the United King dom and the United States held else where, with con trary find ings being explained by the nationality of the organizations studied. This ap proach has been crit i cized as the “so ci ol ogy of re sid ual vari ables” ap proach, be cause it pre sumes that na tion al ity is the cause of any differ ences in find ings, whereas there are many po ten tial other causes (Lammers and Hickson 1979). A prefer a ble ap proach is to deduce from the ory about na tional cul tures, hy poth e ses that are stated prior to the re search (Tayeb 1987), or to show that vari a tions in or ga ni-

Bu reau cracy The ory and Re search

91

za tional characteristics are correlated with variables that directly measure na tional cul ture. Na tional cul ture could be ex tended in fu ture to in clude lan guage dif fer ences and dif fer ences in na tional po lit i cal structure that may af fect or ga ni za tional struc ture. There could be other new con tin gency vari ables added in this way. To be es tab lished as a con tin gency of struc ture, it needs to be shown that there is a fit of struc ture to national cul ture that affects performance and that this is the reason why na tional cul ture causes struc ture. There is ev i dence that the cen trally planned econ o mies of the for mer So viet block coun tries, such as Poland, had orga ni za tions that were more cen tral ized than or ga ni za tions in the West (Kuc, Hickson, and McMillan 1981). How ever, this could be due to the pub lic ac count abil ity ef fect, that is, to the gov ern ment cen tral iz ing or ga ni za tional de ci sion mak ing, so that the res er vation made ear lier ap plies, that is, it could be a cause but not necessarily a con tin gency. Thus one should not pre sume that ev ery vari able that is a cause of orga ni za tional struc ture is also a con tin gency. An un der ly ing fit model has to be specified theoretically. Every time some empirical re search shows that a factor is a cause of an organizational structural variable and this is in ter preted as it be ing a con tin gency, then fur ther re search opens up to show whether the hy poth e sized fit af fects per for mance in the way claimed.

Synthesizing the Size and Task Contingencies Thus far, in this and the pre ced ing chap ter, we have kept both the bu reaucracy and the organic theories separate, as theories that are independent of each other. Some schol ars have sought to syn the size the two the o ries as es sen tially sep a rate but equal (Col lins and Hull 1986; Dewar and Hage 1978; Hage 1980; Hull 1988; Hull and Col lins 1987). How ever, there are other schol ars who seek to as sim i late ei ther task or size into the other, thereby leaving size or task as the dominating contingency. Astley (1985) sought to show a per va sive ef fect of size on organizational structure, reinterpreting in terms of size some aspects of struc ture that are nor mally at trib uted to task. In con trast, some propo nents of task seek to as sert a larger role for it. In par tic u lar, Gerwin (1979a, 1981) has argued that the task interdependence contingency has pervasive ef fects across many as pects of or ga ni za tional struc ture.

92

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

Doubtless the size and task contingencies will each continue to have cham pi ons who will push their fron tiers, to some ex tent at cost to the other con tin gency. Is it pos si ble, in stead, to rec on cile the or ganic and bureaucracy theories of or ganizational structure, so that each is ac cepted as il lu mi nat ing dif fer ent as pects of struc ture and thereby comple ment ing the other? Rec on ciling Bu reau cracy and Or ganic The ories As dis cussed above, the bu reau cracy and or ganic the o ries are somewhat opposed in their structural models. Bureaucracy theory states that both specialization-formalization and decentralization in crease with size. Or ganic the ory states that both spe cial iza tion-for mal iza tion and centralization decrease with task uncertainty. As we have seen, each theory has supporting bodies of em pirical research. Both the bureaucracy and or ganic the o ries deal in a broadly sim i lar set of structural varia bles of specialization, formalization, and centralization, thereby cre at ing the con tra dic tion in their two views. However, there are dif fer ences in their mea sures that re solve the seem ing con tra dic tion be tween their find ings. Bu reau cracy re search mea sures of spe cial iza tion, for mal iza tion, and centralization ad dress the over all or ga ni za tional struc ture rather than local variations. Moreover, they tend to focus on the administrative core of the or ga ni za tion, such as ac count ing and per son nel struc tures (Pugh et al. 1968). In order to be comparable across widely differing sorts of organizations, the structural measures avoid the work flow, that is, the operations whereby in puts are trans formed into out puts, which would in volve de tails that are spe cific to technologies and competencies used in particular in dustries (McKelvey and Aldrich 1983; Pugh et al. 1963). It is this overall structure, relating to the ad ministrative core, that becomes in creasingly bureaucratized with in creas ing size, so that func tional spe cial iza tion (e.g., dif fer ent kinds of accountants), formalization (e.g., procedures for hiring employees), and de cen tral iza tion (e.g., delegation of fi nan cial de ci sions down the hierarchy) in crease with size (see Figure 3.3). Moreover, the Aston Group studied organizations that vary in size, which brings out the true strength of the re la tion ships be tween size and bu reau cratic structure (e.g., Child 1972b; Pugh et al. 1969).

Bu reau cracy The ory and Re search

93

Fig ure 3.3. Rec on ciling Bu reau cracy and Or ganic The ories Or ga ni za tional Level Administrative Core

Theory Bureaucracy Theory

Contingency Size

Bureaucratic Structure: • Functional specialization • Formalization • Decentralization Workflow

Organic Theory

Task Uncertainty

Organic Structure: • Occupational specialization • Lack of formalization • Participation

In contrast, or ganic theory measures tend to focus on the actual work done in the or ga ni za tion, so that they re flect the task con tin gency more. Further, some of the organic theory research is conducted on sam ples of smaller or ga ni za tions in which the in flu ence of the task on struc ture would be more per va sive, as well as re strict ing the size range and so attenuating the correlation between size and structure (e.g., Dewar and Hage 1978; Hage and Aiken 1967a, 1967b, 1969). There are also dif fer ences in the con cep tual def i ni tions of some of the struc tural measures. Specialization in bu reau cracy re search re fers to func tional spe cial iza tion (Pugh et al. 1968), whereas in or ganic research it is occu pa tional spe cial iza tion (Hage and Aiken 1967a). Again, de cen traliza tion in bu reau cracy re search re fers to del e ga tion down the hi er ar chy mainly among man ag ers (though su pervisors and workers are not ex cluded; Pugh et al. 1968), whereas in or ganic re search it is par tic i pation, including that by employees low in the hierarchy (Hage and Aiken 1967a). Thus bu reau cracy re search cap tures how the over all ad min is tra tive su per struc ture elaborates with size. Organic research, in contrast, captures how, as task uncertainty in creases, specific jobs become less bound by rules, pro ce dures, and pa per work and be come more subject to the dis cre tion of their tech no log i cally expert in cumbents, who also ex er cise some up ward in flu ence through par tic i pa tion.

94

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

Thus both the bu reau cracy and the or ganic the o ries are valid in their own domains, which are different arenas within the organization, macro and micro, respectively. Size pri mar ily de ter mines the macro-struc ture in the sense of the level of bu reau cratic struc tur ing and de cen tral iza tion of the over all or ga niza tion. Task pri mar ily de ter mines the mi cro-struc ture within this picture, that is, for each de part ment, sec tion, or role. This mi cro-struc ture can vary across the or ga ni za tion so that some sec tions are mech a nis tic and oth ers are or ganic. Thus size de ter mines the macro-struc ture and task determines the mi cro-structure. As organizations grow larger they be come more bu reau cratic (higher on struc tur ing and de cen traliza tion of the over all struc ture). If the or ga ni za tion needs to in no vate, it will be come more or ganic in those roles in volved in in no va tion, such as in the R&D department or in cross-functional project teams, with the re main ing roles be ing more mech a nis tic. If the in no va tory ac tiv ity is large in the or ga ni za tion, then it will be come in ter nally dif fer en ti ated into units that will be partly co or di nated through for mal means such as hierarchy and for mal iza tion. This reconciliation re ceives empirical sup port in that the re la tionships be tween size and bu reau cratic struc tur ing hold across in dus tries that are in no va tive such as elec tron ics, sim i lar to in dus tries that are not in no va tive, such as con fec tion ery (Child 1973a). The re search on in novation shows effects at the more micro levels and variations within organizations in organicness (Lawrence and Lorsch 1967). While or ganic structures may be found in specific parts of the organization, hierarchy still exists as an overarching coordination mechanism (Lawrence and Lorsch 1967). In a study of research units, Tushman (1979) found that, while in ter de pen dence of work within units led to more organic structures, interdependence between units led to more mech a nis tic struc tures be cause su per vi sors co or di nated through hi erarchy. Hence the re quire ment for co or di nat ing units led to hierarchy superimposed over the organic research units (see also Tushman 1977). Product de velopment of IBM mainframe computers involved large num bers of peo ple who were or ga nized into sev eral de part ments. The pro ject flowed from one de part ment to another with formal sign-offs sig nal ing sat is fac tory at tain ment of pro ject tar gets and transfer of au thor ity (Corey and Star 1971).

Bu reau cracy The ory and Re search

95

A Par si mo ni ous Model Hav ing rec on ciled both bu reau cracy and or ganic the ory as each be ing valid at particular or ga ni za tional lev els, we now need to bring both theo ries to gether into one model. There are two struc tural di men sions and three contingencies in the contingency theory of or ganizational struc ture, mak ing it a par si mo ni ous model. Or ga ni za tional struc ture is com posed of two main di men sions: bu reau cratic struc ture and grouping. Organizational structure is molded by the contingencies of size, task un cer tainty, and task in ter de pen dence. As we have seen, bureaucratic structure includes structuring (specialization, formalization, and decentralization) and structural differentiation. Bu reau cratic struc ture is strongly af fected by the con tingency of or ganizational size. As organizations increase in size, they in crease their structuring (including de centralization) and structural differentiation. However, the total level of the structur ing vari ables is af fected to a de gree by task un cer tainty. In no va tion re duces formalization and cen tral iza tion, in de pend ent of size ef fects, thereby affecting structuring. Grouping defines the or ganizational subunits and their colocations (Child 1984). It in volves whether dif fer en ti a tion is by functions, products, services, customers, or geog raphical areas. Task in ter de pen dence is the con tin gency that af fects group ing, in cluding divisionalization. Divisionalization is conceptually re lated to bu reaucratic structure and has here been sub sumed un der bu reau cratic struc ture. How ever, while bu reau cracy is de ter mined pri mar ily by size, divisionalization is determined primarily by task interdependence. Therefore divisionalization is somewhat sep a rate from bu reauc ra tization. Moreover, task interdependence de termines also the type of divisional structure (e.g., by product or area), so that again it has an effect on divisionalization that is sep a rate from size. In summary, bureaucratic structuring is increased by size and reduced by task un cer tainty. Task in ter de pen dence in creases divisionalization and de fines the na ture of the or ga ni za tional sub units. In this way, a highly parsimonious model can be abstracted of how contin gencies shape or ga ni za tional struc ture, help ing to stem the fear that the contingency theory literature is so com plex that it de fies com prehen sion (Bour geois 1984).

96

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

Causal Con nec tions Be tween Size and Task Con tin gencies As we ar gued above, bu reau cracy and or ganic the ory are each valid in their own do mains, which cor re spond to the macro-struc tures and micro-structures, respectively. This reconciliation leaves each of size and task as independent contingency variables, each with their own sep a rate ef fects. The next ques tion is whether size and task are caus ally connected so that their ef fects are re lated. It may be the case that size causes task. Size is the an te ced ent cause of structure, with task the intervening variable. Thus the immediate effects on struc ture of task are ac tu ally caused by size. Size is the dis tal cause of structure and task is its prox i mal cause. Let us now con sider some of the re la tion ships dis cussed above and see if they can be placed into the causal model that size causes task, which causes struc ture. We will ex am ine, in turn, the for mal iza tion, divisionalization, and or ganic as pects of or ga ni za tional structure. As has been seen, two ma jor ef fects of or ga ni za tional size are spe cialization and formalization. Size causes specialization, in that as size increases the di vi sion of la bor, so each job be comes nar rower in scope. The same few operations are done repeatedly by the incumbent and this com bines with the nar row scope of the job to re duce its task un certainty. This al lows the job to be cod i fied into rules and stan dard op er ating pro ce dures, which can be writ ten into de tailed job de scrip tions and manuals. Thus size leads to specialization, which leads to lower task uncertainty and hence to formalization. Size is the ultimate cause of formalization, but task uncertainty is the intervening variable that con nects size and for mal iza tion. Hence task un cer tainty is the im mediate cause of for mal iza tion. More over, task un cer tainty ex plains why increasing size leads to increasing formalization. Thus a major size effect on struc ture in bu reau cracy the ory can be ex plained by the task contingency. One may go farther and say that formalization, through job de scriptions and manuals of procedures, reduces the uncertainty for incumbents, so that for mal iza tion feeds back to lower task un cer tainty, fostering fur ther formalization, tightening the causal connection between task uncertainty and formalization. Hence a more extended causal model is that size leads to spe cial iza tion that re duces task un certainty that increases formalization that re duces task un cer tainty that in creases for mal iza tion.

Bu reau cracy The ory and Re search

97

We can also bring di ver si fi ca tion into the syn the sis. As seen ear lier, diversification leads to divisionalization. However, diversification by business or ga ni za tions usu ally oc curs when they are large in size, because the in dus try has ma tured, leav ing the firm large but with lim ited expansion opportunities in its industry. Therefore, while diversification is the im me di ate cause of much divisionalization, size can of ten be the more ultimate cause of diversification. Here size causes task interdependence (i.e., di ver si fi ca tion) that causes divisionalization. There may also be an ef fect of size on in no va tion. There is some ev idence that size is pos i tively re lated to the ex tent of in no va tion ac tiv ity, though not nec es sar ily to more in no va tive out puts (Co hen and Levin 1989, p. 1071; Damanpour 1992). Given that the the ory be ing dis cussed here is pri mar ily about in no va tive ac tiv ity, size is an ul ti mate cause of such in no va tive ac tiv ity and thereby of the organicness of parts of the or ga ni za tional struc ture that re sult from the task un cer tainty en tailed by in no va tion. Hence size is the more ultimate cause of many aspects of organizational structure: formalization, divisionalization, and organicness. These effects of size are be cause size af fects task that, in turn, pro duces the effects on organizational structure. As we have seen, size affects bureaucratization through reducing task uncertainty. Size can affect division alization through re duc ing task in ter de pen dence by di ver sification. Size can in crease the level of organicness through in creas ing the level of in no va tion that is sought and thereby the task un cer tainty. In all these cases size causes task that causes the organizational struc ture. Thus size and task can be brought together in the same causal process, so that the bureaucracy and organic theories are causally connected. Hence con tin gency the ory is re vealed to have a deeper co her ence in that the dis pa rate ef fects claimed of size and task can be rec on ciled in a model in which they are causally linked. To the extent that they are connected, their effects on structure are no longer independent, because the level of size is not independent of the level of the task contingency. Thus or ganic and bu reau cracy the ory merge. Size causes increasing bureauc ra ti za tion, which means in creas ing struc tural differentiation, for malization, and de centralization. It also causes in creasing di versification, which causes divisionalization, that is, ad ditional de cen tral iza tion and formalization. Fur ther, size also leads to innovation being sought that creates organic structures in some

98

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

parts of the organization, thereby off set ting to a de gree the ten dency in duced by large size to ward min ute spe cial iza tion and for mal iza tion in those parts and also prompt ing par tic i pa tion in them. This is a merger in which bu reau cracy the ory is the dom i nant partner in the sense that the overall structure conforms to bu reaucracy theory, and size is the ultimate cause with pervasive in fluence. How ever, the key ar gu ments of or ganic the ory are re tained, so that or ganic and par tic i pa tory struc tural el e ments are given a role, al beit lo cal ized to certain parts of the organization, with task being the immediate cause of many of the structural elements, both organic and bu reaucratic. Such a synthesis of two of the ma jor streams of contingency theory research, or ganic and bu reaucracy, is a fur ther step to ward in te gra tion of con tin gency the ory. It con tin ues the move to ward in tegrat ing contingency the ory that we be gan in Chap ter 1.

Conclusions The bu reau cracy the ory of or ga ni za tional struc ture holds that bu reaucratization is com posed of struc tur ing (for mal iza tion and de cen tral ization), structural dif ferentiation, and divisionalization. The level of bureauc ra ti za tion is strongly af fected by size. Bureaucracy theory took as its point of departure the model of bureaucracy that was originally advanced by Weber. Contingency theory re search moved from his ideal-type to a multivariate frame work that was used in quan ti ta tive stud ies com par ing across or ga ni za tions. The resulting view is that a bureaucratic structure has many depart ments and hi er ar chi cal lev els, is highly spe cial ized and for mal ized, but also highly decentralized. Bureaucracy increases with size, but at a decreasing rate with re spect to size, so that bu reau cracy grows pro por tionately less than size. Furthermore, the ratio of managerial and administrative staff to total employees decreases as size increases, thereby cre at ing econ o mies of scale in ad min is tra tion. In this way, and through improving organizational performance, bureaucracy theory ar gues that bu reauc ra ti za tion is func tional (i.e., ef fec tive), in con trast to earlier case studies of bu reaucratic dysfunctions. The level of bureaucratization needs to fit size, and larger organizations with more bureaucratic structures have performance superior to those

Bu reau cracy The ory and Re search

99

with less bu reauc ra ti za tion, con sis tent with the fit idea of con tin gency theory. Divisionalization is a com po nent of the bu reau cratic type of or ga niza tional struc ture in that divisionalization in creases for mal iza tion and de cen tral iza tion. Strat egy af fects this as pect of struc ture in that di versi fi ca tion in creases (and ver ti cal in te gra tion de creases) division alization. Thus diversification reduces the level of the task interdependence contingency and in this way causes divisionalization. Ma trix struc tures fit in ter me di ate lev els of the task inter de pen dence con tingency due to me dium lev els of ei ther prod uct or area di ver sity. Thus the structures of func tional, di vi sional, and ma trix, to gether with their sub types, can be assimilated within bureaucracy theory. Size also affects division alization, both directly, and indirectly t hrough diversification. Thus divisionalization is subsumable un der bu reau cratic the ory in that it is part of the bureaucratic structure and has some affinity in causation with bu reauc ra ti za tion more gen er ally. Public accountability also affects bu reaucratization by in creasing centralization and for mal iza tion, though whether it is a con tin gency is open to dis cus sion at the pres ent time. Sim i larly, en vi ron men tal hos tility and CEO need for achievement increase centralization, though whether they are contingencies is also open to discussion. National culture ef fects struc ture, though whether it is a con tin gency is also an open ques tion. For any con tin gency, it needs to be shown that there is some fit of some or ga ni za tional char ac ter is tic to the con tin gency that positively affects per formance. Oth erwise it is just a cause without being a con tin gency. The main ac com plish ment of bu reau cracy the ory has been to iden tify how macro-or ga ni za tional struc ture in many as pects is shaped by needing to fit the contingencies, principally size. The organic theory school can be rec on ciled with this pic ture. In no va tion, and hence task un cer tainty, leads cer tain parts of the or ga ni za tion (e.g., de part ments, sections, or roles) to be more organic (i.e., less formalized and more decentralized), while the other parts are mech a nis tic. Thus size de ter mines bureaucracy at the macro-structural level, whereas task has more lo cal ized ef fects on parts of the or ga ni za tional mi cro-struc ture. Much of the empir i cal lit erature on the contin gency theory of organizational struc ture can be dis tilled into a par si mo ni ous model of three con tin gen cies and two struc tural di men sions. More over, the size

100

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

and task contingencies can be brought together in the causal model that size causes task, which in turn causes structure. Thus much of contin gency the ory and re search can be com bined into a highly co herent theory.

THEsal Cau CON ity/Con TINtin GENCY gencyTHEORY inBureaucracy OF OR The GA ory NI ZA TIONS

4

Causality and Contingency in Bureaucracy Theory

W

hile bureaucracy research has established numerous em pir i cal reg u lar i ties and of fers the o ret i cal in ter pre ta tions of them, there are some thorny theoretical issues pertaining to bureaucracy theory. What exactly are the causal relationships between the components of bureaucratic struc ture, such as formalization, decentralization, and struc tural dif fer en ti a tion? What is the the o ret i cal rea son why so many of the ob served re la tion ships be tween vari ables in the causal model are not linear? Causality is being dis cussed in terms of an impersonal, functionalist or ga ni za tional sys tem, but how do peo ple ac tu ally make this happen—what is the role of human action? Is the causal model really a con tingency theory? We will deal with such ques tions in this chapter. By way of over view, the fol low ing sev eral is sues may be dis tin guished. The effect of one variable on another, such as size on an element of bureaucratic structure, could be direct or indirect. Thus, rather than assuming that size directly causes structure, we need to consider the pos si bil ity that size af fects some other vari ables that then cause 101

102

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

struc ture. The sec ond is sue is that of func tion al ism, which would hold that the struc tural vari ables in the model raise organizational performance. Such an effect of structure on organizational performance could be ei ther a main ef fect, so that struc ture in creases per formance, or a con tin gency ef fect, so that it is the fit of struc ture to the con tin gency that increases per for mance. Whether a main or con tin gent func tion al ist effect, these ben e fits could be an tic i pated so that struc ture is in creased immediately following a size increase, or there could be a process whereby struc tural ad just ment lags and so low ers per for mance, leading even tu ally to struc tural change. If there is such a lag, then the ef fect of size on struc ture is in di rect through mis fit and lower per for mance. The structural adaptation to regain fit (SARFIT) model holds that structure is functionalist, that is, has positive performance conse quences that re sult from fit of struc ture to con tin gency and that change is lagged so that the ef fect of size on structure is indirect. This is a com plex the o ret i cal model, and in deed the most com plex of the the oret i cal al ter na tives just briefly out lined. In or der to make the case for it we will go through the sim pler al ter na tives and show their lim i ta tions. In so do ing we shall move from sim ple to com plex mod els of cau sal ity. We shall at tend first to di rect causal mod els of size and bu reau cratic struc ture, which are not func tion al ist in that they make no pos tu la tion that the structure is beneficial in its consequences. Subsequently, we shall ex am ine func tion al ist the o ries of size and bu reau cratic struc ture, beginning with universalistic variants that posit functionalist conse quences of struc ture that are main ef fects, not in volv ing con tin gen cies. Next we shall ex am ine con tin gency func tion al ist the o ries that none the less do posit di rect cau sa tion of struc ture by size. Then we shall consider a con tin gency func tion al ist the o ret i cal model that pos its in di rect cause of struc ture by size, with lags and feed back ef fects of mis fit on per for mance, in the SARFIT man ner. Finally, we shall at tempt to bring all the lessons of the discussion together in a synthesis that is based on SARFIT but also draws se lec tively on other ap proaches.

The Causal Model of Size and Bureaucratic Structure As seen in the previous chapter, the bureaucratic model contains three main as pects of or ga ni za tional struc ture: formal iza tion, de cen tralization, and structural differentiation. What is the causal

Cau sal ity/Con tin gency in Bu reau cracy The ory

103

Fig ure 4.1. A Causal Model of the Re la tion ship of Size and Bu reau cratic Struc ture

relationship among these aspects of structure? Child (1973a) con cludes from his empirical study that (simplifying) size causes spe cial ization and the level of specialist qualifications (that to gether he calls “complexity”), which cause formal iza tion that, in turn, causes de cen tralization, while size also, directly, causes decentralization. The explanation is that size leads to more specialists, who are trained to bring in new man age ment sys tems that in crease for mal iza tion (Child 1973a, pp. 181-183), which pro motes re lin quish ing of di rect con trol by the center, while large size makes direct personal con trol dif ficult, forcing de centralization. Similarly, Marsden, Cook, and Kalleberg (1994) find the causal model that (sim pli fy ing) size causes de part ments that cause for mal ization, which reduces administrative in tensity, while size also causes ver ti cal lev els that cause de cen tral iza tion, which also reduces administrative intensity. Thus size affects the horizontal aspect of structural differen tiation that affects in turn formalization, which reduces ad ministrative in tensity. Size also af fects the ver ti cal aspect of struc tural differen tiation that af fects in turn cen tral iza tion, an other ver ti cal aspect of struc ture, which also re duces ad min is tra tive in ten sity. Thus in seek ing a more exact causal re la tion ship among the

104

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

com po nents of bu reau cratic struc ture, it may be fruitful to decom pose the e f fect ofsize on the hor i zon tal as pects from the ef fect of size on the ver tical aspects. Draw ing se lec tively on both causal mod els, to gether with the causal the ory of Blau (1970, 1972), we may posit the causal model of the relation ship among size and the bu reau cratic struc tural vari ables in Fig ure 4.1. Size positively affects horizontal structural differentiation in all its as pects (di vi sions, de part ments, spe cial iza tion, etc). Hor i zon tal dif ferentiation nar rows the scope of each job and so in creases task cer tainty, which al lows jobs to be cod i fied into writ ten job de scrip tions and manu als of pro ce dures, so that for mal iza tion in creases. The spe cial iza tion aspect of hor i zon tal dif fer en ti a tion also fos ters for mal iza tion be cause greater ad min is tra tive ex per tise leads to the adop tion of new man agement sys tems. The nar row ing of the scope of jobs also in creases the span of con trol at each level in the hi er ar chy, be cause it is eas ier to su per vise a less di verse set of jobs. This causes a neg a tive ef fect of hor i zon tal dif ferentiation on vertical dif fer en ti a tion, that is, the num ber of hi erar chi cal lev els is less ened (Fig ure 4.1). Size also increases the verti cal aspect of structure: the number of hierarchical lev els and de cen tral iza tion. In creasing size leads to greater hi er ar chi cal lev els in order to avoid ex ces sively wide spans of con trol. This pos i tive ef fect of size on hi er ar chi cal lev els is off set to some ex tent by the in creas ing span of con trol (part of hor i zon tal dif fer en ti a tion),so that hierarchical lev els in crease as size in creases, but at a more mod est rate than they would if span of control remained constant. Having many hierarchical levels isolates top management from the lowest, operational, level, so that man age ment lacks knowl edge about op er a tions. More over, the elon gated hi er ar chy pre vents speedy and ef fec tive com mu ni ca tions. There fore, top man age ment is forced to del e gate. Also, the increasing number of middle managerial and supervisory levels pro vides an in creas ing depth of man ag ers to whom de ci sion authority can be del egated. Top management can have increased confidence that del e ga tion will not be abused, be cause of the in di rect controls over lower-level per son nel pro vided by for mal iza tion, so that for mal iza tion has a positive effect on decentralization. Hence for mal iza tion sub stitutes for centralization, so that indirect, impersonal control re places di rect, per sonal con trol by top man age ment.

Cau sal ity/Con tin gency in Bu reau cracy The ory

105

Ad min is tra tive in ten sity is raised by the com plex ity of co or di na tion re quired through hav ing more de part ments and other sub units, so that horizontal differentiation positively affects administrative in tensity. How ever, ad min is tra tive in ten sity is re duced by for mal iza tion, so that formalization negatively af fects ad min is tra tive intensity. In addition, more hi er ar chi cal lev els in crease ad min is tra tive in ten sity, so that there is a pos i tive ef fect of ver ti cal dif fer en ti a tion on ad min is tra tive in ten sity. How ever, the in creas ing span of con trol neg a tively af fects the num ber of hi er ar chi cal lev els, so that span of con trol (i.e., hor i zon tal dif fer en tia tion) has an in di rect neg a tive ef fect on ad min is tra tive in ten sity through ver ti cal dif fer en ti a tion. Thus ad min is tra tive in ten sity is af fected pos i tively by differentiation (both horizontal and vertical) and neg atively by formalization. While size increases hor izontal differentiation, this indirectly i n creases for mal iza tion and also in creases the span of con trol, which re duces the in crease in ver ti cal dif fer en ti a tion due to size. Thus the pos i tive ef fect of hor i zon tal dif fer en ti a tion on ad min is tra tive intensity is off set by the in di rect, neg a tive ef fect of for mal iza tion. Sim i larly, the pos i tive ef fect of ver ti cal dif fer en ti a tion on ad min is tra tive intensity is off set by the mod er at ing ef fect of span of con trol on hi er ar chi cal lev els. Jointly these two offsetting causations are stronger than the effects of hor izontal and vertical differentiation on administrative in tensity, so that the over all ef fect of size on ad min is tra tive in ten sity is neg a tive. In sum, size pos i tively af fects both hor i zon tal and ver ti cal dif fer entiation. Horizontal differentiation positively af fects formalization. Horizontal dif ferentiation neg atively affects verti cal differentiation. Verti cal differ en ti a tion pos i tively affects de cen tral iza tion. Formal ization positively af fects decentralization. Horizontal and vertical differentiation both pos i tively af fect ad min is tra tive in ten sity. For malization also negatively affects ad ministrative in tensity. This causal model specifies the relationships between each of the components o f bureaucratic structure: formal iza tion, decentralization, struc tural dif ferentiation, and administrative in tensity. It is consistent with the causal models and empirical re sults of Blau (1970, 1972), Child (1973a), and Marsden, Cook, and Kalleberg (1994). However, while this causal model of size and bureaucratic structure is informative, it leaves open some deeper the o ret i cal is sues. The first is sue is ex plain ing why the re la tion ships in the model are not lin ear.

106

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

The Curvilinear Relationship of Size and Bureaucratic Structure A the o ret i cal is sue in bu reau cracy the ory is that the em pir i cal re la tionship between size and bureaucratic structure is curvilinear, so that structure in creases with size at a decreasing rate with respect to size. For smaller sizes, struc ture in creases steeply with in creases in size, but at larger sizes structure in creases only modestly with size increases. Thus the slope of structure is initially steep where size is small, then steadily de creases, so that it is shal lower at large size. Blau (1970) shows that this holds for all the as pects of struc tural dif fer en ti a tion he stud ied. The Aston Group shows sim i lar curves for its mea sures of hi er ar chi cal lev els and spe cial iza tion (Child 1973b, p. 172-173) and Mar sden, Cook, and Kalleberg (1994, p. 916, Fig ure 2) show them for some, though not all, of their struc tural vari ables. Blau’s (1970) the ory of struc tural dif fer en tiation treats the curvilinearity explicitly. The curvilinear relationship be tween size and the many dif fer ent as pects of struc tural dif fer en ti a tion is de duced from the ba sic prop o si tion that size causes struc tural dif ferentiation a t de creas ing rates with re spect to size (Blau 1970, p. 204). The rea son why the rate of struc tural dif fer en ti a tion de creases with size is be cause of in creas ing pres sure caused by the greater dif fi culty of co or di na tion re sult ing from dif fer en ti a tion (Blau 1970, pp. 207, 212-214; but see also Meyer 1972, p. 439). This in creased pres sure means that there is a neg a tive feed back on struc tural dif fer en ti a tion from the cost of increasing admin is tra tive over head (Blau 1970, p. 217). However, in creas ing ad min is tra tive over head is ac com pa nied in the model of Blau (1970, p. 218, Fig ure 6) by in creas ing ad min is tra tive econ omy (as discussed in the pre vi ous chap ter), so the in creas ing struc tural dif fer en tiation produces economies as well as costs. There fore any eco nomic con strain t should be driven by the net ef fect of increas ing ad min istrative economy and overhead (i.e., savings and costs). Hence the feed back ar row in the causal model (Blau 1970, p. 218, Fig ure 6) should come from the joint ef fect of in creas ing administrative economy and overhead, not overhead alone. Moreover, given that, empirically, size increases are continuing to create economies even at large size (Blau 1970), there seems no reason for organizations to reduce the rate of struc tural dif fer en ti a tion as size in creases. Thus the logic of this as pect of Blau’s theory is unclear. The rea son for the curvilinearity be tween size and struc tural dif fer en ti a tion must be sought elsewhere. We will

Cau sal ity/Con tin gency in Bu reau cracy The ory

107

con sider first ver ti cal struc tural dif fer en ti a tion and then its hor i zon tal aspect. Abdel-khalik (1988) ar gues that there is a math e mat i cal re la tion ship between size and the number of lev els in a hi er ar chy, which ac counts for the negative geometric relationship. Assuming that the span of control is the same at each level of the hi er ar chy, let us say seven, then in a two-level hi er ar chy there is a boss and seven work ers, so that size of the or ga ni za tion is eight. In a three-level hi er ar chy, there is a boss, 7 su per vi sors, and 49 work ers (7 x 7), so size is 57. In a four-level hi er archy, there is a boss, 7 man ag ers, 49 su per vi sors (7 x 7) and 343 work ers (7 x 7 x 7), so size is 400. Hence, ev ery new level that is added in creases size by a num ber that takes the num ber of ex ist ing work ers and mul ti plies it by seven. Thus, math e mat i cally speak ing, ev ery ad di tional level adds to size the number seven raised to another power. This means that size in creases by the span (of control) being raised to its next power. Thus size of the or ga ni za tion is equal to a se ries com posed of 1 plus the span plus the span squared plus the span cubed, and so on, to span to the number of hi er ar chi cal lev els mi nus one. Thus size is an increasing geo met ric se ries. Hence in creases in the num ber of hi er ar chi cal lev els correspond to geometrically in creas ing size. This means that geo met ri cal in creases in size are re quired to pro duce suc ces sive in creases in the num bers of lev els. There fore each suc ces sive level of the hierarchy re quires a greater size in crease than the pre vi ous level. Thus hierarchical lev els are a neg a tive geo met ric func tion of size, in that size must in crease geo met ri cally for lev els to be added. The relationship is geo met ric, but neg a tively so, in that the curve of lev els on size be comes shal lower as size in creases. Hence hi er ar chi cal lev els in crease with size at a de creasing rate with respect to size, the observed em pirical re lation ship (Blau and Schoenherr 1971). Hence the in sight of Abdel-khalik (1988) is es sen tially that or ga ni zational hierarchies are pyramidal structures, so that adding one more level at the base al ways in volves add ing many more em ploy ees than the pres ent num ber of work ers. The sce nario is that all the old work ers be come supervisors and each su per vises the num ber of new work ers given by the span of cont rol. Thus the number of new workers is the num ber of old work ers mul ti plied by the span of con trol. Abdel-khalik (1988) assumes that hierarchies are symmetric in his mathematical model, so that all spans are the same and all man ag ers and su per vi sors have the same number of subordinates. Nei ther of these con di tions

108

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

ap plies to real or ga ni za tions (Blau 1971, p. 306; Blau and Schoenherr 1971). How ever, vari a tions in spans and in the length of hi er ar chi cal chains to some extent cancel each other out. Thus while the elegant mathematics no lon ger strictly ap ply, the ar gu ment re mains valid that any or ga ni za tion that approximates a pyramid must show a negative geo met ric curve of lev els with size. More over, any or ga ni za tional structure that seeks unity of com mand must ap prox i mate to a pyr a mid because all man ag ers must re port up to one top man ager. Hence the rate of in crease in struc tural dif fer en ti a tion de clines with in creas ing size be cause the py ram i dal shape of the hi er ar chy means that many more employees must be added to generate suc ces sive in creases in lev els as size grows. In that sense, size growth is pushing against the inherent bar rier of a pyr a mid so that its ef fects on lev els must be come weaker as it in creases. The bar rier co mes from the ge om e try of a pyr a mid, so that suc ces sively higher lev els taper off with fewer per sons at each level. The question arises as to whether other curvilinear relationships be tween size and other aspects of struc tural dif fer en ti a tion, such as the hor izontal, can be ex plained sim i larly by some in her ent lim i ta tion. One as pect of hor i zon tal struc tural dif fer en ti a tion is the di vi sion of labor, which conceptually is the narrowness of the scope of each job. Imag ine a pro duc tion pro cess of as sem bling a pen. If the firm is min i mally small, there is only one worker, who makes the whole pen. If the firm is larger, so that there are two work ers in pen pro duc tion, then each could spe cial ize on some of the op er a tions of as sem bling a whole pen. Thus now each worker does one half of the whole task, that is, spe cial iza tion has dou bled. If the firm is larger again, so that there are three work ers in pen pro duc tion, then spe cial iza tion in creases so that each worker does one third of the whole task. If there are four work ers, each does one fourth. For five work ers, each does one fifth. As size inc reases linearly (from 1 to 5 work ers), spe cial iza tion (as in dexed by the pro por tion of the whole task be ing per formed by each worker) in creases in the se ries: 1, .50, .33, .25, and .20, re spec tively. Hence each ad di tional worker in creases the spe cial iza tion in dex by .5, .17, .08 and .05, re spec tively. Thus, spe cial iza tion in creases at a less than pro por tion ate rate to the increase in size. Specialization increases as size in creases but at a de clin ing rate with re spect to size—a neg a tive geo met ric re lation ship. As dif fer en ti a tion in creases, that is to say, the work of the

Cau sal ity/Con tin gency in Bu reau cracy The ory

109

orga ni za tion be comes split up among its mem bers, the pro cess of di vision means in herently that a nega tive geometric relationship with size is in volved. Thus as specialization increases with size increase, it must do so at a declining rate with respect to size. This is in her ent in split ting up a sin gle en tity into more and more pieces. Thus for an or ganization whose task is constant, such as making the same type of products or ser vices, add ing em ploy ees must lead to spe cial iza tion that is pro gres sively smaller in its in cre ments. By im pli ca tion, any ba sis of struc tural dif fer en ti a tion must increase with size, but with successive in cre ments be ing less than the ear lier ones be cause the same set of tasks is be ing split up pro gres sively more finely. Thus the num ber of oc cupa tional ti tles, the num ber of di vi sions, the number of sec tions per d epartment, and the degree of spe cialization by func tion should all increase with size at a de clin ing rate with re spect to size, the em pir i cally observed relationships (Blau and Schoenherr 1971; Child 1973a). In this way the re la tion ships be tween size and hor i zon tal struc tural differ en ti a tion can be ex plained. The curvilinear rela tion ship between size and other mea sures of bu reau cratic struc ture can be de rived from this curvilinear re la tion ship between size and structural differentiation. The relationship be tween struc tural dif fer en ti a tion and for mal iza tion is pos i tive and lin ear, as is seen from the regression of overall docu men ta tion on overall role spe cial iza tion and qual i fi ca tions (Child 1973a, p. 182, Ta ble 9). Thus a unit increase in specialization produces a unit increase in for mal ization. Given that size has a curvilinear ef fect on spe cial iza tion, then in creasing size would flow through specialization to produce the same curvilinear re lationship of size with formalization. In this way the negative geo met ric re la tion ship be tween size and for mal iza tion is explained. The spe cial ized ex perts in ad min is tra tion, which are re corded in the Aston scales of func tional and role spe cial iza tion, to gether with the level of their qualifications (Pugh et al. 1968), pro duce new man agement sys tems, procedures, rules, regulations, and accompanying paperwork (Child 1973a). Each new type of specialist adds its new systems. But because the rate of addition of new types of spe cialists de clines as size in creases, the rate of in crease in formalization also de clines as size increases. Thus the curvilinear relationship between size and formalization can be explained, in addition to that between

110

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

size and struc tural dif fer en ti a tion. Can the ver ti calaspect of struc tural dif fer en ti a tion and its re la tion ship with de cen traliza tion be ex plained in a sim i lar way? As seen above, the pyramidal shape of the or ga ni za tional hi er ar chy means that there is a neg a tive geo met ric re la tion ship of size and ver tical structural differentiation (number of hierarchical levels). There fore, if the number of hi erarchical levels has a positive relationship with de cen tral iza tion, then the curvilinear ef fect of size on lev els flows through to produce a curvilinear relationship between size and de cen tral iza tion. The ex pla na tion is that if lev els cause de cen tral iza tion because of lack of information among top management, then the declining rate at which lev els are added with size in creases, means that decentralization will also increase at a declining rate with res pect to size. Thus the neg a tive geo met ric re la tion ships be tween size and many aspects of bureaucratic structure can be explained by size necessarily having such relationships with specialization and hierarchical levels. These struc tural vari ables, in turn, have lin ear ef fects with for mal iza tion and decentralization, respectively, so that size has negative geo metric relationships with formalization and de centralization. This is based on a sim ple kind of de ter min ism of bu reau cratic struc ture by size that in her ently must have de clin ing mar ginal ef fects on the ver ti cal and hor i zon tal struc tural dif fer en ti a tion that, in turn, af fect de cen traliza tion and for mal iza tion.

Functionalist Explanations The foregoing explains why the relationships between size and the bu reaucratic structural variables are curvilinear, but it overlooks a deeper and more troubling issue. Arguing that size causes structural differentiation because of inherent limitations in differentiation can ex plain the form of the re la tion ships, but it is not a func tion al ist ex planation because it does not explain structure by its consequences for performance. Fur ther, it is not a contingency type of functionalist the ory because the explanation omits reference to a fit that affects performance. We will now discuss func tionalist explanations of the re la tion ship of size and bu reau cratic struc ture, dis tin guish ing be tween uni ver sal ist ic and con tin gency forms.

Cau sal ity/Con tin gency in Bu reau cracy The ory

111

Functionalism ex plains that struc tures are adopted be cause they are ben e fi cial for the or ga ni za tion. This may oc cur through or ga ni za tional man ag ers guid ing the or ga ni za tion to adopt a more ef fec tive struc ture and thereby acting on be half of the organization. However, another possibility is that or ga ni za tional change may be driven by ac tors pur suing their own self-interests. As noted above, Child (1973a) argues that spe cial iza tion causes for mal iza tion be cause this ac cords with the occupational ideology of ad ministrative experts and furthers their self-interest. This political explanation an swers the ques tions of how and why ac tion oc curs to in crease for mal iza tion. How ever, the po lit i cal ex pla na tion is not nec es sar ily in com pat i ble with func tion al ism, in that the politically mo ti vated in creases in for mal iza tion could be func tional in their con se quences for the or ga ni za tion (i.e., in creas ing its ef fec tiveness). In such cases, the new sys tems that the spe cial ists are push ing to cre ate are needed by the or ga ni za tion. This would also fur ther their career fu ture, by be ing seen to help their or ga ni za tion. Such an in terpretation appears to be empirically valid in that the increases in for mal iza tion that accom pany size in creases are pos itively related to perfor mance, con sis tent with it be ing func tional for the or ga ni za tion (Child 1975). Thus pol i tics helps ex plain how ac tion oc curs and yet the increases in formalization are functional. Hence it is pos si ble in this way to combine political explanations of action with functionalism. This gives an account of how ben eficial systems change is brought about by ac tors, that is, it ties the lev els of a sys tem and ac tion to gether (Giddens 1984; Par sons 1951; Silverman 1970). Here a causal re la tion ship (spe cial iza tion causes for mal iza tion) has been ren dered func tionalist by sug gest ing that pos i tive per for mance con se quences may flow from formalization. Other aspects of structure may be explained by functionalism if they have pos i tive per for mance con se quences. Functionalism holds that organizations adopt structures that raise their performance. This means that a consequence (higher per formance) be comes a cause. How is this ac com plished? This ques tion has occasioned weighty discussion (Isajew 1968; Silverman 1968). Three an swers may be dis tin guished. The first is to say that the or ga ni za tion anticipates the benefits of adopt ing the new struc ture and this ideal im age of the fu ture in the minds of de ci sion mak ers causes the adoption of the new struc ture (Etzioni 1968). The sec ond is to say es sen tially that managers lack foresight, but experience problems with the existing struc ture and re act by adopt ing a new struc ture that will re store

112

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

per formance (Child 1972b). The third is to say that or ga ni za tions lack knowl edge of better struc tures and so are culled by the en vi ron ment and replaced by organizations with su pe rior struc tures (Etzioni 1961; Hannan and Freeman 1989). We will dis cuss this pop u la tion ad ap tation in Chap ter 6 when con sid er ing pop u la tion-ecol ogy the ory (Hannan and Free man 1989). Thus for the mo ment we shall fo cus on an tic i pation or reaction to problems as the two mechanisms by which organizations adopt structures that are functional. We will consider size and bu reau cracy first as ex plained by uni ver sal ist ic, func tion al ist the o ries. Next we will con sider con tin gency func tion al ist the o ries that involve direct cau sation of structure by size. Then we will consider contingency func tion al ist the o ries that in volve cau sa tion of struc ture by size in di rectly, through mis fit and per for mance, as SARFIT holds. Uni ver sal ist ic Func tion al ist Ex pla na tion In the case of Blau (1972), there are grounds for doubting whether his theory of struc tural dif fer en ti a tion is a con tin gency the ory. He ar gues that organizations in crease structural differentiation because spe cial iza tion leads to greater ef fi ciency, in that ho mo ge neous tasks are “eas ier” (Blau 1972, p. 14) and re duce costs while in creas ing ex per tise: “The di vi sion of la bor con se quently has a dou ble ad van tage. It makes it pos si ble to fill many po si tions with less trained per son nel, which fa cil itates re cruit ment and achieves econ o mies, and to fill the most dif fi cult jobs with more highly trained ex perts, which im proves the qual ity of per for mance” (Blau 1972, p. 15). Thus his the ory is uni ver sal ist ic, in that max i mum ef fi ciency would result from maximum specialization. However, the highest level of specialization that an organization can attain is set by its size, that is, the num ber of em ploy ees avail able for spe cial iza tion. Thus spe cial ization causes efficiency, with size acting as a constraint on how much specialization is pos si ble for a given or ga ni za tion: “the de gree of fea sible dif fer en ti a tion in an or ga ni za tion is lim ited by its size . . . dif fer en tia tion’s in stru men tal con tri bu tions cre ate the pres sures that pro mote it to the degree size permits” (Blau 1972, p. 16). This is a universalistic the ory with a sit u a tional con straint: size. It is not a con tin gency the ory be cause there is no no tion of fit. In par tic u lar, there is no con cept that an or ga ni za tion could have too much spe cial iza tion for its size and so be in mis fit. Sim i larly, an or ga ni za tion that is too lit tle spe cial ized for

Cau sal ity/Con tin gency in Bu reau cracy The ory

113

its size is not in mis fit, but rather is los ing some of its po ten tial per formance be cause it has not spe cial ized as much as it could given its size. Thus the op ti mal level of spe cial iza tion for an or ga ni za tion is not one that fits its size, but rather the max i mum it can at tain given its size constraint. In con trast, in con tin gency the ory the or ga ni za tion does not strive to attain the maximum possi ble level of a struc tural vari able, only to attain its op ti mal level, that is, the level that fits the con tin gency. Thus, in bureaucracy the ory, the organization attains its highest per formance by adopting the level of bureaucratic structure, for example, specialization, that fits its size (Child 1975), rather than by adopting the max i mum level of the struc tural vari able. The way ac tion cre ates struc tural dif fer en ti a tion in Blau’s the ory is consistent with it be ing a uni ver sal ist ic the ory. Man agers try to spe cialize as much as possible because they believe that specialization pro duces ef fi ciency (Blau 1972, pp. 13, 16, 22). Thus the man ag ers are following the universalistic theory. Given that universalistic the ories are simpler and there fore eas ier to fol low than con tin gency the o ries, which in volve com plex ity and no tions of fit, it is fea si ble to hold that managers follow universalistic the ories. There is empirical ev idence that the fit be tween struc ture and con tin gency may arise more be cause man ag ers fol low uni ver sal ist ic than con tin gency the ory be liefs (Priem 1994). While this ev i dence pertains to or ganic the ory rather than bureau cracy the ory, it cau tions that bu reau cracy may also arise be cause of universalistic beliefs of man ag ers. It sug gests the need to in quire em pir i cally into the be liefs of man ag ers re gard ing bu reau cratic structure to see whether they are uni ver sal ist ic or con tin gent. Uni ver sal ist ic beliefs of the classical man age ment type would support Blau’s the ory of how or ga ni za tional struc tures change over time. (How ever, uni versalistic managerial beliefs do not, of course, prove that universalistic theories about the objective relationship between structure and per for mance are cor rect and con tin gency the o ries in cor rect, which is why continued re search at the ob jec tive level is also needed.) The func tion al ism in Blau (1972) is there fore of a uni ver sal ist ic kind: Or ga ni za tions spe cial ize as much as pos si ble for their given size. Therefore, when an organization grows in size, the managers in crease specialization up to the level al lowed by its new size. This pro duces the causal con nec tion be tween size and spe cial iza tion. Given that spe cialization is enhanced by an in crease in any of the hor i zon tal as pects of

114

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

struc tural dif fer en ti a tion, the im plied causal model can be stated more gen er ally as size causes hor i zon tal as pects of struc tural dif fer en ti a tion that cause ef fi ciency. The ex plicit causal model of Blau (1970, p. 218, Figure 6) is that size causes the structural differentiation that causes ad min is tra tive in ten sity, with the reduc tion in ad min is tra tive in ten sity be ing the main ev i dence of func tion al ity in Blau’s re search (Blau 1970, 1972; Blau and Schoenherr 1971). The implicit and explicit causal models are of main effects between size and struc ture, and struc ture and functional out comes. There is no con tin gency ef fect of struc ture on functional outcomes that depends upon a fit to size. Neither are there feed back ef fects of mis fit on lower per for mance caus ing ad ap tations to at tain new fits, as in the con tin gency the ory of or ga ni za tional change. It seems unlikely, however, that the causal con nec tion be tween size and hierarchical levels is a universalistic main effect. If it were, that would mean that man age ment be lieves that per for mance is max i mized by max i miz ing hi er ar chi cal lev els, which runs coun ter to the pre vail ing so cial an tip a thy to ward tall hi er ar chies as be ing ineffectual and waste ful. It would also mean that up per-level man ag ers were seek ing to add as many sub or di nate lev els as pos si ble, given the size con straint. Mul tiplying high status subordinates by adding new intermediary levels could be seen as aid ing the self-in ter est of up per man ag ers, by boost ing their status and pay, consistent with a po litical explanation of their be hav ior (Par kin son 1957). How ever, this is con tra dicted by the ob serva tion that the span of con trol tends to in crease as size in creases (Blau 1970, 1972), so that man age ment is add ing new lev els more par si mo niously as size and lev els in crease, ar gu ing against po lit i cal self-in ter est as a pre dom i nant causal pro cess. Thus for the ver ti cal as pect of struc tural dif fer en ti a tion (i.e., the num ber of hi er ar chi cal lev els), it seems un likely that this is caused by size in a uni ver sal ist ic fash ion. By ex ten sion, this places a caution against accepting universalistic explanations of the horizontal as pects of struc tural dif fer en ti a tion. Con tin gency Func tion al ist Ex pla na tion of Di rect Cau sa tion Another variant of functionalism, however, is consistent with the con tin gency the ory that pos i tive per for mance out comes flow from the fit of structure to the contingencies. This contingency functionalist explanation would say that man ag ers know the fits and so at the same

Cau sal ity/Con tin gency in Bu reau cracy The ory

115

time as in creas ing size they also in crease the struc tural vari ables to at tain new fits, thereby an tic i pat ing, and so avoid ing, the per for mance loss that would have flowed from mis fit. Thus there is no feed back ef fect of mis fit through low per for mance. In stead, in this model there is a di rect, if not in stan ta neous ef fect of size on struc ture. Thus this function al ist model is com pat i ble with the causal model that size di rectly causes structure (as in the model in the first sec tion of this chap ter). Thus di rect ef fects of size on the vari ables of bu reau cratic struc ture can be con sis tent with con tin gency fit the ory, if the ad ap ta tion is through an tic i pa tion, rather than feed back from lost per for mance due to mis fit. While the an tic i pa tion of fit is a log ically possi ble type of con tingency functionalism, and an appealing image of proactive manage ment, it has difficulties. Management would need knowledge of the fits, which in volves know ing con tin gency the ory and also the scores on both the con tin gency and the struc tural vari ables that con sti tute the fit for their or ga ni za tion. Yet the struc tural vari ables are ab stract quan tities, such as spe cial iza tion by func tion (Pfeffer 1997), that are as sessed by ar cane mea sure ment scales (Pugh et al. 1968). More over, the fits are con tained in math e mat i cal for mu las that are re corded in the es o teric research literature (e.g., Child 1975). Thus it is unlikely that the fits between size and the bu reau cratic struc tural vari ables would be iden tified by the man age ment for their or ga ni za tion. Hence it is more fea sible that the organization moves into mis fit and that the man age ment then rec og nizes the prob lem and takes cor rec tive ac tion. Also, the low performance resulting from misfit provides the impetus to manage ment to give pri or ity to tak ing cor rec tive ac tion by adopt ing the new structure that fits the contingencies. For these reasons, anticipatory change is expected to be a less frequent occurrence than reactive change, which will now be dis cussed. Con tin gency Func tion al ist Ex pla na tion of In di rect Cau sa tion As seen, we can consider bu reaucratic structure to consist of four com po nents: hor i zon tal dif fer en ti a tion, formalization, ver ti cal dif ferentiation, and de cen tral iza tion. To be classed as a con tin gency the ory, there must be fits be tween each of these com po nents of struc ture and size that pos i tively af fect per for mance. The ef fect of mis fit on per formance feeds back to alter the structural vari ables. Thus the or ga ni zation might in crease in size, thereby cre at ing a de gree of mis fit with the

116

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

existing levels of the structural variables, which causes lower per formance. The man ag ers then seek to avoid this lost per for mance by in creas ing the struc tural vari ables to a new level that fits the new size, thereby regaining performance. This is the structural adaptation to regain fit (SARFIT) model. This functionalist explanation relies on causal connections between size and misfit, misfit and performance, and the feedback effect of per formance on structure and thence of struc ture on fit. How ever, it does not en tail a di rect causal ef fect of size on any of the struc tural vari ables, nor of any struc tural vari able on the other structural variables. Thus the contingency fit theory requires causal con nec tions with fit, but not be tween the con tin gency and the struc tural vari ables that com pose fit. The contingency fit model would, of course, mean that size pos itively affects structure, but indirectly through the feedback effect of mis fit on per for mance. Sim i larly, there would be a cor re la tion be tween size and struc ture, be cause in creases in size even tu ally cause in creases in struc ture, but not be cause size di rectly causes struc ture. By im pli ca tion, the the o ret i cal ra tio nales for a di rect causal ef fect of size on structure, or of one struc tural vari able on an other, would not ap ply be cause there are no such causal re la tion ships. Size does not in crease spe cializa tion at a decreasing rate with respect to size, because of in herent limitations in dividing up the same task into smaller and smaller pieces. Spe cial iza tion does not raise for mal iza tion be cause of ex perts utilizing their knowledge to cre ate new management systems. The correlations among structural variables, such as the components of bureaucracy, would arise be cause each of them sep a rately needs to fit with size. In this way, con tin gency fit the ory ex plains why these as soci ations ex ist with out hold ing that the struc tural vari ables cause each other. Figure 4.2 shows the causal relationships be tween size and bu reaucratic struc ture and among com po nents of bu reau cratic struc ture. For con ve nience only two com po nents of bu reau cratic struc ture are shown: horizontal structural differentiation and formalization. The causal connection between size and each structural variable is in direct through mis fit and per for mance as con tin gency the ory holds. Looking first at horizontal structural differentiation, size has a negative effect on fit, be cause in creas ing size moves the or ga ni za tion into mis fit with its ex ist ing level of hor i zon tal struc tural dif fer en ti a tion. This mis fit re duces per for mance. There fore the low per for mance that re sults from

Cau sal ity/Con tin gency in Bu reau cracy The ory

117

Figure 4.2. A Con tin gency The ory of Struc tural Change Caused by Feed back From Performance

the mis fit feeds back to cause an in crease in hor i zon tal struc tural dif ferentiation, which cre ates a new fit. Thus the ef fect of size on hor izon tal struc tural dif fer en ti a tion is indirect, through misfit and per formance. Size af fects fit neg atively, while fit af fects per formance pos itively, and performance negatively affects hor izontal structural dif fer en ti a tion. Al ge braically the re sult of this mul ti pli ca tion of a neg a tive sign by a pos i tive sign and then a neg a tive sign is a pos i tive sign. Hence size has a pos i tive cor re la tion with hor i zon tal struc tural dif ferentiation, which could be misinterpreted as a di rect pos i tive ef fect on horizontal struc tural dif fer en ti a tion. How ever, such an in ter pre ta tion is fallacious; the correlation is spurious because the ef fect of size on horizontal structural differentiation is indirect. While it is true to say that size causes horizontal struc tural dif fer en ti a tion, it is not because of some prop erty of size that di rectly causes an in crease in hor i zon tal

118

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

structural differentiation. Instead, it is because the size increase has cre ated a mis fit that ad versely af fects per for mance, which causes hor izon tal struc tural dif fer en ti a tion to in crease in or der to re gain fit and performance. A sim i lar pro cess oc curs for formalization. Again, change in for malization in response to size increase occurs because of misfit and the feed back from the re duced per for mance. Thus the ef fect of size on for malization is in di rect. The cor re la tion be tween size and for mal iza tion is spu ri ous. Size causes for mal iza tion, but through mis fit and per formance rather than from a property of size alone. Similar re marks would ap ply to ver ti cal struc tural dif fer en ti a tion and de cen tral iza tion. Thus the causal relationships between size and each component of bureaucratic structure are wholly indirect, mediated by fit and its effects on per for mance. Given that size (in di rectly) causes both hor i zon tal struc tural dif ferentiation and for mal iza tion in Fig ure 4.2, size in creases si mul ta neously with both these structural variables, so that these struc tural vari ables are pos i tively cor re lated. This might be in ter preted as a causal con nection be tween them, such as hor i zon tal struc tural dif fer en ti a tion caus ing formalization. Yet this would be fallacious on the logic of Figure 4.2, be cause there is no direct causal con nec tion be tween hor i zon tal structural dif fer en ti a tion and for mal iza tion (though an in di rect one). Thus, from the viewpoint of the contingency t heory of structural change as driven by fit and the feed back of per for mance, there is no re la tion ship between hor izontal structural differentiation and formalization. Sim ilar re marks ap ply to ver ti cal struc tural dif fer en ti a tion and de cen tralization. Thus there is no causal con nec tion among the four com po nents of bu reau cratic struc ture (hor i zon tal struc tural dif fer en ti a tion, ver ti cal struc tural dif fer en ti a tion, for mal iza tion, and de cen tral iza tion). Thus the SARFIT vari ant of con tin gency the ory holds that the ef fect of size on bu reau cratic struc ture is in di rect and there is no di rect cau sation among the components of bureaucratic structure. This gives a very different interpretation to bureaucracy the ory from the tra ditional one. Thus the causal models of Blau (1970), Child (1973a), and Mar sden, Cook, and Kalleberg (1994) are re ally very dif fer ent from those re quired by con tin gency the ory prop erly ap pre ci ated. The kind of sim ple, di rect ef fects be tween size and com po nents of bureaucratic struc ture that their causal models contain are to be distinguished from the contingency fit functionalism that is in te gral to con tin gency the ory.

Cau sal ity/Con tin gency in Bu reau cracy The ory

119

A Contingency Theory Causal Model of Size and Bureaucracy A con clu sion may be made to the fore go ing dis cus sion by of fer ing the fol low ing model that is based upon re ac tive con tin gency func tion al ism but that al lows some role for an tic i pa tion and that in cludes other el ements in a synthesis. Size affects struc tural dif fer en ti a tion in the man ner of con tin gency fit the ory, on both the hor i zon tal and ver ti cal as pects of struc tural differ en ti a tion. On the hor i zon tal as pect, in creas ing size adds em ploy ees, who ini tially stand idle or du pli cate and in ter fere with the task ex e cu tion of other em ploy ees, un til the work is re or ga nized so that old and new employees each have a specialized role, so that specialization in creases. Un til spe cial iza tion is in creased to fit the new size, the dis or gani za tion low ers per for mance. Thus size leads to mis fit and then to the in crease in spe cial iza tion that re stores fit and per for mance. Thus size causes in creases in spe cial iza tion, that is, hor i zon tal struc tural dif ferentiation, in di rectly through mis fit and lower per for mance. Similarly, on the vertical aspect, increasing size widens the span of control of a manager to the point where it becomes too wide and so perfor mance suf fers, which means that the span no lon ger fits the size. To avoid fur ther per for mance loss, a new level in the hi er ar chy is created, which re stores fit and per for mance. Thus size causes in creases in hi er ar chi cal lev els, that is, ver ti cal struc tural dif fer en ti a tion, in di rectly through mis fit and lower per for mance. Thus size causes both the hor i zontal and vertical aspects of structural differentiation through the feed back of the con tin gency effects of misfit on per f ormance. The ef fect of size on structural dif ferentiation, however, is not linear, but curvilinear. Struc tural dif fer en ti a tion in creases with re spect to size at a decreasing rate with re spect to size be cause of the in her ent lim i ta tions of spe cial iz ing work and add ing lev els to a py ram i dal hi er ar chy (as explained above). In some organizations, management may avoid misfit by ad justing the level of struc tural dif fer en ti a tion im me di ately af ter in creas ing size. As examples, specialization would be increased immediately n ew em ploy ees are re cruited, or a new in ter me di ary hi er ar chi cal level would be cre ated im me di ately the span of con trol of a man ager ex ceeds its op timal. In such cases there is no feed back ef fect of per for mance and so the ef fect of size on struc ture is not in di rect but di rect. The rea son why the

120

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

struc ture is changed is be cause of the con tin gency fit, so that the di rect ef fect of size on struc ture is ac com pa nied by the con tin gency fit ef fect on per for mance and is con sis tent with it. Specialization leads to formalization be cause the spe cial ists seek to have new sys tems adopted. This re flects their pro fes sional train ing and ideology and their political self-interest. However, while many new systems are likely to be ad vo cated by var i ous spe cial ists in the or ga ni zation, only some are adopted. New sys tems are most likely to find ac ceptance where they of fer to re solve prob lems, and such prob lems will tend to ex ist where the or ga ni za tion has low per for mance be cause of be ing in mis fit. Thus spe cial iza tion in creases for mal iza tion of those as pects of the organization that are in misfit because size increase has out stripped the level of for mal iza tion. The in crease in for mal iza tion is due to the new ad min is tra tive spe cial ists, cre ated by the re cent size in creases, who pro mote their man age ment sys tems. Thus, for mal iza tion is in creasing through ad ap ta tion to re gain fit and per for mance, as SARFIT holds, so that the causal ef fect of spe cial iza tion on for mal iza tion is ac com panied by the feedback ef fect of mis fit and lower per for mance. Vertical structural dif ferentiation, that is, the number of levels in the hierarchy, affects decentralization in a sim i lar man ner to the way horizontal structural dif fer en ti a tion (spe cial iza tion) af fects for maliza tion. An in crease in the num ber of lev els in a hi er ar chy im pedes the flow of in for ma tion to the up per lev els and so ren ders cen tral de ci sion mak ing less ef fec tive while also slow ing the re sponse of the or ga ni za tion to changes “down on the fir ing line.” There fore the ex ist ing de gree of cen tral iza tion be comes a mis fit to the new size be cause of the num ber of hi er ar chi cal lev els that size in duces, with ad verse con se quences for performance. This causes management to in crease the degree of decentralization and thereby re store the fit of de cen tral iza tion to size and re gain per for mance. Thus the ef fect of hi er ar chi cal lev els on de cen tral iza tion de pends upon mis fit and per for mance and so, is indirect. The management of some organizations may avoid misfit by in creasing lev els and de cen tral iza tion con cur rently. Thus, im me di ately a new in ter me di ary hi er ar chi cal level is cre ated, it is del e gated au thor ity to make some de ci sions. In such a scenario there is a direct effect of lev els on de cen tral iza tion, be cause of an tic i pa tion of fit and its ben e fits for performance. Thus, overall, there can be causal con nec tions be tween each of size and spe cial iza tion, spe cial iza tion and for mal iza tion, size and hi er archy,

Cau sal ity/Con tin gency in Bu reau cracy The ory

121

and hi er ar chy and de cen tral iza tion. This would be con sis tent with the causal models that arise from empirical research (Blau 1970; Child 1973a; Marsden, Cook, and Kalleberg 1994). Yet each of the causal con nections between these pairs of variables could be indirect, in volving con tin gency mis fit and the feed back from low per for mance, as SARFIT holds, rather than direct effects. In some organizations, on some occasions, management might correctly anticipate the need to adopt a new struc ture to fit the new level of the con tin gency. How ever, such proactivity may well be rare be cause it re quires knowl edge of fits that are com plex and sub tle, which may be lack ing, and also the will to change, which may also be lacking in the absence of per formance problems. Thus there are grounds for holding, theoretically, that the causal connections among size and the bureaucratic structural vari ables are pri mar ily in di rect ef fects that oc cur be cause of mis fit and low performance. It is possible that in some of these linkages uni versalistic beliefs might play a role, in the man ner of Blau (1972). For in stance, man agement might believe that specialization in creases performance and so will im me di ately in crease spe cial iza tion as em ploy ees are added, which in creases the spe cial iza tion level up to that which fits the new size. In this way, each time size increases, the or ga ni za tion moves into fit and gains beneficial con se quences for its per for mance. Thus the ob jec tive re la tion ship is be tween fit and per for mance, but fit is achieved be cause man age ment is guided by a sub jec tive be lief of the uni ver sal ist ic type. The re sult is that size di rectly causes spe cial iza tion. Thus there may be some role for uni ver sal ist ic be liefs even if uni ver sal ist ic func tion al ism is in valid ob jec tively. The role of uni ver sal ist ic be liefs and di rect cau sation by size may be lim ited to spe cial iza tion and for mal iza tion, as they are most con sis tent with the Tayloristic sci en tific man age ment prin ciples that may un der gird such man age ment be liefs. Over all, the re la tion ship among the size and bu reau cratic struc ture vari ables is held to be mainly ex pli ca ble by re ac tive con tin gency function al ism. A sec ond ary mech a nism is antic i pa tory con tin gency functionalism whereby causation occurs di rectly among size and the bu reau cratic struc tural vari ables. A third rank ing mech a nism is an tic ipa tory uni ver sal ist ic be liefs whereby cau sa tion oc curs di rectly among size and the bureaucratic structure variables. However, all three of these mech a nisms are functionalist, in volv ing ben e fi cial out comes for the or ga ni za tion. This is the view that emerges from seek ing to

122

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

ex plain the relationships among size and bu reaucratic structural variables by con tin gency the ory, with its func tion al ist em pha sis on the con se quences for per for mance and the cru cial roles of fit and mis fit. It takes us be yond the sim pler causal mod els of size and bu reau cracy in a di rec tion that en tails more com plex ity but thereby at tains more the o ret i cal co her ence within the con tin gency par a digm. This model is to be taken as a the o ret i cal state ment, so that the task for future em pirical research is to inquire into its validity, relative to other, per haps sim pler mod els. Do the re la tion ships be tween size and bureaucratic structure, and among the components of bureaucratic struc ture, arise be cause of con tin gency mis fit and per for mance feed back, as contingency theory in its SARFIT variant holds? Or do they arise because man ag ers cor rectly and ex pe di tiously an tic i pate the need to fit struc ture to new size lev els? Or do they, rather, arise be cause man ag ers cleave to er ro ne ous uni ver sal ist ic be liefs that nev er the less cre ate con tingency fits? These are some of the ques tions for fu ture em pir i cal re search to emerge from the the o ret i cal dis cus sion herein.

Conclusions In terms of a causal model of the relationship between size and the com po nents of bu reau cratic struc ture, the fol low ing can be pos tu lated, building on previous theoretical and empirical research. Size causes struc tural dif fer en ti a tion that causes for mal iza tion and cen tral iza tion, with formalization also being a cause of centralization. The effect of size on structural differ en ti a tion is a negative g eo met ric func tion, so that size also shows this same curvilinear re la tion ship with for mal iza tion and cen tral iza tion. The neg a tive geo met ric func tion of struc tural dif fer en ti a tion on size means that structural differentiation in creases with re spect to size at a de creas ing rate to size. Such curvilinearity may re flect in her ent phys i cal bar riers, in that spe cial iza tion be comes progres sively finer as it in creases with size, and the py ram i dal shape of a hierarchy means that pro gres sively more em ploy ees are re quired to build each hi er ar chi cal level. In contradistinction, functionalism stresses that structures are adopted be cause of their ben e fi cial out comes. Ac tion may be driven in part by po lit i cal self-in ter est of or ga ni za tional mem bers, but nev er theless still be functionalist in outcomes. Universalistic functionalism

Cau sal ity/Con tin gency in Bu reau cracy The ory

123

provides a simple model whereby size directly causes bureaucratic structure; however, this is considered to lack generality as a causal mechanism. Contingency theory holds that positive or ganizational per for mance re sults from the fit of struc ture to size. Where man age ment correctly anticipates the new fit, then a size increase could directly cause an in crease in bu reau cratic struc ture. How ever, the com plex ity and sub tlety of con tin gency fits makes this less fea si ble. More over, the im pe tus to struc tural change is im pelled more strongly by low per for mance. Therefore structural change in response to increasing size is held to be primarily indirect through misfit and low performance as SARFIT holds, that is, reactive con tingency functionalism. Thus the feed back ef fect of lost per for mance from mis fit could bring about the ob served as so ci a tions among size and the bu reau cratic struc tural variables, without size having a direct effect on structure. Anticipatory contingency functionalism may play a more restricted sec ond ary role to reactive contingency functionalism. A smaller role again may be played by uni ver sal ist ic be liefs that un wit tingly im pel struc tural change into fit. Hope fully, the the o ret i cal model ar tic u lated here will lead to a more in ci sive anal y sis of the role of con tin gency the ory fit caused re lation ships of size and bu reau cracy in fu ture em pir i cal re search, through in quiry into the various mech a nisms dis cussed herein.

THEtro Con CON versies TIN GENCY in Con tin THEORY gency The OFory ORResearch GA NI ZA TIONS

5

Controversies in Contingency Theory Research

W

ithin the con tin gency the ory lit er a ture there have been a num ber of con tro ver sial is sues. There have been de bates over the rel a tive importance of rival contingency factors such as technology and size. Also there is crit i cism of the de ter min ism in con tin gency the ory, and moves to re place it with stra te gic choice. Again the in fer ence that con tingency causes struc ture has been chal lenged by the ri val in ter pre ta tions that the contingency vari ables are mere correlates rather than being causes, or that struc ture causes the con tin gency. These in tel lec tual moves of fer an es cape from the idea that con tin gen cies make some struc ture, such as bu reau cracy, in ev i ta ble for an or ga ni za tion and open the door to al ter na tive struc tures, such as the or ganic. Again, some schol ars have sought to re place the multivariate con tin gency ap proach by con fig u rations based on con sis tency the ory. Con tin gency the ory has been used to ex plain power dis tri bu tions within and be tween or ga ni za tions, but these ex pla na tions have been crit i cized in var i ous ways. These six con troversial is sues of tech nol ogy, stra te gic choice, cau sal ity, re verse cau sal ity, configurations, and power will be considered in this chapter. 125

126

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

The crit i cisms of the orig i nal vari ant of con tin gency the ory and attempts to re vise its com po nent parts have led to deeper ex am i na tion of these is sues. This deeper ex am i na tion has tended to sup port the orig inal contingency the ory and to disconfirm the revisionist views. In particular there is support for the idea that contingency determines organizational struc ture. Thus much of this chap ter is an ar gu ment for contingency theory in its original form. More specifically, the discus sion of sev eral of these is sues pro vides sup port for bu reau cracy the ory.

Technology Versus Size De spite the rec on cil i a tion be tween the or ganic and bu reau cracy theories that we of fered in Chap ter 3, there is still a tension be tween these the o ries that is played out partly as an ar gu ment about the roles of the task and size contingencies. This has led to a long-running debate about the relative importance for organizational structure of task, in terms of tech nol ogy, and size, that con tin ues down to the pres ent. The de bate be tween tech nol ogy and size has fu eled an ex ten sive dis cus sion that has raised many is sues of the ory and method. The re sult ing lit er ature is com plex. Al most any state ment made about this lit er a ture re quires sev eral qual i fi ca tions and ex cep tions. How ever, we shall try to of fer an over view of some of its main points. The de bate con sid ered here is about whether tech nol ogy de ter mines overall organizational structure, relative to the explanation provided by size. The vari ables be ing ex plained are over all as pects of or ga ni zational struc ture, such as the num ber of lev els in the hi er ar chy, the spans of con trol at the top and bottom of the hierarchy, the percentages of administrative sup port staff, and the mech a nis tic ver sus or ganic structures. Thus the interest is in the effect of technology on the overall shape of the organizational structure and management system as a whole. This macro-level focus is to be distinguished from the microlevel is sue of whether in di vid u als’ roles (i.e., jobs) are af fected by characteristics of the task being performed, such as uncertainty. There is considerable ev i dence that greater task un cer tainty leads to less structured roles (Gerwin 1979b). Again, task in ter de pen dence be tween jobs affects role structures such as formation of teams and autonomous workgroups (Trist and Bamforth 1951; Wall et al. 1990). Such ef fects are part of the way that task af fects the mi cro-struc ture, and in Chap ter 3

Con tro versies in Con tin gency The ory Research

127

we sug gested that we can rec on cile bu reau cracy and or ganic the o ries by accepting that task shapes the micro-structure and size shapes the macro-structure. However, the issue being dis cussed here is whether technology shapes the macro-structure. Thus the debate is about whether tech nol ogy has grander ef fects than the mod est one at trib uted to it in the reconciliation in Chap ter 3. In outline, pro ponents of or ganic the ory have sought to show that technology leads away from mechanistic toward organic structures (Wood ward 1965). In con trast, pro po nents of bu reau cracy the ory have sought to show that technology has only weak effects rel ative to size (Child and Mansfield 1972; Hickson, Pugh, and Pheysey 1969). An implication is that the reduction in bureaucracy flowing from technology would be swamped by large increases in size, or nullified by more moderate in creases in size. Again, other bu reau cracy the o rists have argued that tech nol ogy ac tu ally leads to more bu reau cracy, so that advances in technology reinforce the ef fects of size in creases, rather than di min ish ing bu reau cracy (Blau, et al.1976). Wood ward (1965) found em pir i cally that tech nol ogy af fected or ga nizational struc ture and that size was not as so ci ated with or ga ni za tional structure. The Aston Study carefully sought to replicate Woodward’s study. Its au thors mea sured her op er a tional tech nol ogy con struct with a scale that they called production con tinuity (Hickson, Pugh, and Pheysey 1969). They mea sured many as pects of or ga ni za tional structure that differed from Woodward’s, but also had several the same as hers, for example, the span of con trol of the first-line su per vi sor. The Aston Study found few of the re la tion ships that Wood ward had found between her technology measure and her structural variables (Hickson, Pugh, and Pheysey 1969). The subsequent study by the Aston Group, the National Study, found even fewer of her re la tion ships (Child and Mansfield 1972). More over, the Aston and Na tional Studies found lit tle re la tion ship be tween Woodwardian tech nol ogy and their many other (non-Woodwardian) structural measures (Child and Mansfield 1972; Hickson, Pugh, and Pheysey 1969). Further, these studies measured other aspects of production technology, different from Wood ward, and these also had few, mainly weak re la tion ships to organizational structure (Child and Mansfield 1972; Hickson, Pugh, and Pheysey 1969). In con trast, size had re la tion ships with many differ ent as pects of or ga ni za tional struc ture, of ten strong and al most al ways stronger than technology. Therefore in multivariate analyses,

128

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

size rather than technology emerged as the underlying contingency fac tor (Child and Mansfield 1972; Hickson, Pugh, and Pheysey 1969). The Aston Group challenged Woodward, and their different findings cannot be ex plained away by meth od ol ogy. The cross-sectional na ture of the Aston Study was seized upon by some com men ta tors who ad vanced other causal in ter pre ta tions. These included that technology, as measured by a construct other than Wood ward’s, caused size and also struc tur ing, which in turn also was a cause of size—thereby turning the Aston interpretation on its head (Aldrich 1972a, 1972b; Heise 1972; Hilton 1972). Pugh and Hickson (1972) pointed out that in the ab sence of lon gi tu di nal data, causal in terpretations are haz ard ous. How ever, even with cross-sec tional data, rea son ing can be used to clar ify cau sal ity. If tech nol ogy is the cause of struc ture, then or ga ni za tions that have the same tech nol ogy would be ex pected to have the same struc ture. Yet Donaldson and Warner (1974) found that in la bor un ions, or ga ni za tions that have lit tle dif fer ence in technology, there were nevertheless wide variations in structure that were strongly correlated with size. This is consistent with size rather than tech nol ogy be ing the cause of struc ture. Many stud ies have con tin ued to in ves ti gate the rel a tive roles of technol ogy and size, us ing var i ous def i ni tions of tech nol ogy: Grinyer and Yasai-Ardekani (1981); Khandwalla (1974); Lincoln, Hanada, and McBride (1986); Marsh and Mannari (1976, 1980, 1981), Reimann (1977, 1980); and Singh (1986). But de bat ing the rel a tive strengths of technology and size assumes that there is an effect of technology on struc ture and that it is of the Wood ward (1965) kind, that is, pro mot ing the or ganic struc ture. It may there fore be fruit ful to fo cus on just the tech nol ogy ef fect on struc ture, to ver ify that it ex ists and to as cer tain whether its na ture is that which Wood ward (1965) ar gued. Returning to the Woodward mea sure of technology, Zwerman (1970) re ported rep li cat ing Wood ward’s find ings that the fit be tween technology and structure led to higher performance. However, a secon dary anal y sis of the study by Donaldson (1976) showed an absence of such relationships. Moreover, Donaldson (1976) also showed that none of the relationships be tween tech nol ogy and structure found by Woodward (1965) replicated consistently across the subsequent stud ies that had used the same tech nol ogy and struc tural vari ables as Wood ward (Child and Mansfield 1972; Hickson, Pugh, and Pheysey 1969; Zwerman 1970). He concluded that there was no valid

Con tro versies in Con tin gency The ory Research

129

gen er al iza tion con nect ing Woodwardian tech nol ogy and or ga ni za tional struc ture. How ever, the method em ployed, re quir ing each find ing to be sig nif i cant, has now been su per seded by meta-anal y sis. We shall therefore turn to the results of meta-analyses to form a more definitive con clu sion about the technology-structure relationship. However, at the time, the cri tique of Wood ward was seen as damaging (Eilon 1977), so that the im me di ately sub se quent con tri bu tions had to deal with this highly neg a tive eval u a tion. Reimann and Inzerilli (1979) countered on behalf of technology. They argued that there was a level-of-analysis problem in that some stud ies that found lit tle re la tion ship be tween tech nol ogy and struc ture failed to mea sure both tech nol ogy and struc ture at the same level, that is, the level of the over all or ga ni za tional sys tem. Reimann (1980) also ar gued that at the sys tems level, that is, the over all or ga ni za tion, where much work is ad min is tra tive, struc ture was af fected by the tech nol ogy at that level, that is, office automation through computerization and in for ma tion tech nol ogy. Thus we need to con sider whether the as pect of tech nol ogy that is the con tin gency for or ga ni za tional struc ture, and that thereby con sti tutes an im per a tive, is ac tu ally com put er iza tion. Whisler, Meyer, Baum, and Sorensen (1967) ar gued that com put eriza tion fa cil i tated the move ment of in for ma tion around the or ga ni zation, hor i zon tally and ver ti cally, that is, from de part ment to de part ment and between bottom and top. This was held to lead to a re duc tion in structural differentiation so that computerization would reduce the number of de partments and levels in the hierarchy, leading to a re centralization of de ci sion mak ing. In an em pir i cal study of com puter ization in the insurance in dustry, Whisler et al. (1967) found evidence sup port ing their the ory. How ever, com put er iza tion re duced the numbers of em ploy ees and this alone could pro duce the ob served struc tural changes. Given that size causes structural dif fer en ti a tion and de cen tral iza tion, a size re duc tion would de crease struc tural dif feren ti a tion and also lead to recentralization. Thus organizational structural change caused by com put er iza tion may flow through the in ter me di ary variable of size. Hence size would be the key to structure and there would be no in de pend ent ef fect of com put er iza tion re vers ing bu reauc ratization in the ab sence of size change. Indeed, contrary to Whisler et al. (1967), Blau and Schoenherr (1971) and also Blau et al. (1976) ar gue that com put er iza tion has effects similar to those of size, namely, that computerization in creases

130

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

struc tural dif fer en ti a tion, both hor i zon tally and ver ti cally. It also leads to more decentralization of de ci sion mak ing in that top management knows that lower-level em ploy ees are be ing forced to fol low stan dard procedures that are now programmed into the computers and are sub ject to computer sur veillance. Thus com puterization pro vides another indirect control over lower-level employees that enables top management to relinquish its direct control to a degree, thereby in creas ing del e ga tion. Thus computerization is really an other as pect of bureaucratization. Moreover, it feeds back to raise the lev els of other as pects of bu reauc ra ti za tion (more struc tur ing and de cen tral iza tion). Blau et al. (1976) see office automation as part of bureaucratization that re in forces the ef fect of size in push ing or ga ni za tional struc ture to be more bu reau cratic. Com put er iza tion can be con sid ered to be part of the wider trend toward formal ra tionality that Weber (1968) saw as char ac ter iz ing mod ern so ci ety and that pro duces bu reau cracy. Blau’s theory receives sup port from a meta-an alytic review of tech nology that shows that computerization is pos itively re lated to or ganizational struc ture (Caufield 1989). Spe cifically, computerization is pos i tively re lated to struc tur ing vari ables, though not to de cen traliza tion, show ing that Blau’s the ory mostly gen er al izes across the studies. Thus focusing on of fice au to ma tion re veals an ef fect of tech nol ogy on or ga ni za tional struc ture, but it is weak and sec ond ary to size, and pro motes bu reauc ra ti za tion. The meta-an a lytic re views also re veal sim i lar ef fects for op er a tions technology, whether measured by the Woodward or other measures. Operations technology has linear relationships with organizational structural variables (Caufield 1989), rather than the curvilinear re lations claimed by Woodward (1965) and others (including Blau et al. 1976). The relationships are positive between operations technology and or ga ni za tional struc ture. More spe cif i cally, op er a tions tech nol ogy is associated with the structuring variables, but the correlations are weak (around +.3). Operations technology has similar relationships with organizational structure to those between computerization and organizational structure. It does not matter whether operations or office tech nol o gies are ex am ined, the tech nol ogy ef fects are the same. The tech nol ogy-struc ture re la tion ships (in clud ing those with op er a tions technology) are unaffected by the size of the organizations studied, thus casting doubt on the idea that technology has more impact on

Con tro versies in Con tin gency The ory Research

131

macro-struc ture in smaller or ga ni za tions than in larger ones (Hickson, Pugh, and Pheysey 1969). The meta-analysis by Miller, Glick, Wang, and Huber (1991) also found that the relationship between technology and structure was robust and un af fected by some po ten tial mod er a tors, in clud ing sev eral of ten cited in the lit er a ture, such as dif fer ent def i ni tions of tech nol ogy and organizational size, though some moderators were found. How ever, their method is not a complete meta-analysis in that it does not ex am ine how much of the vari a tion in find ings be tween stud ies is due to ar ti facts and there fore is spu ri ous vari a tion. Caufield ex am ined the is sue of ar ti facts, pro duc ing ev i dence of the ro bust ness of the gen eral re la tion ship be tween tech nol ogy and struc ture. Much of the vari a tion, 70 percent, in findings from study to study is due to sampling error, that is, to small sample size. Range restriction explained another 17 per cent, so that ar ti facts ex plain 87 per cent of vari a tion across stud ies (Caufield 1989, p. 150). Thus any role for moderator variables that condi tions the tech nol ogy-struc ture re la tion ship is, at best, lim ited. Ad vances in tech nol ogy, whether op er a tional or ad min is tra tive, lead to more bu reau cracy, re in forc ing the ef fect of growth in size. Thus, in our view, the tech nol ogy-ver sus-size de bate may be re solved by say ing that size is the more im por tant con tin gency of or ga ni za tional struc ture than tech nol ogy, though both have the same type of ef fects. Tech nology, like size, in creases the bu reauc ra ti za tion of the macro-struc ture, rather than re duc ing it as or ganic the ory claims. This marks a ma jor restric tion in the va lid ity of or ganic the ory, so that its do main be comes limited to the micro-structure. It supports the reconciliation be tween bu reau cracy and or ganic the o ries pro posed in Chap ter 3.

Determinism Versus Choice As we have seen, struc tural con tin gency the ory holds that change in the con tin gency leads to change in or ga ni za tional struc ture, be cause of the need to regain fit and performance. Therefore con tingency the ory is deterministic (Astley and Van de Ven 1983). The organization re sponds to changes in the con tin gen cies and thereby to the en vi ron ment that in turn shapes those contingencies. Dis quiet has been ex pressed about this determinism, especially where contingency theory argues

132

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

that given cer tain con tin gen cies, the or ga ni za tion has no choice but to adopt cen tral ized or bu reau cratic struc tures that lack par tic i pa tory or democratic features (Schreyögg 1980). There fore those who favor participatory, or ganic struc tures have been mo ti vated to ar gue against contingency determinism or to show that contingencies have only weakly con strain ing power so that choice of or ganic struc tures is pos sible re gard less of the con tin gen cies. Some scholars have argued strenuously against the idea that the organization is determined by its situation and have instead as serted that man ag ers have free choice and are thereby to be held mor ally ac count able (Bour geois 1984; Whit ting ton 1989). Con tin gency the ory ap pears to some crit ics to be a man a ge ri ally con ve nient ide ol ogy that justifies as inev i ta ble or ga ni za tional char ac ter is tics that are not really in ev i ta ble, be cause they are not re ally re quired for or ga ni za tional ef fec tive ness, and that in jure the in ter ests of em ploy ees (Schreyögg 1980). Thus con tin gency the ory is op posed by free choice. An in ter me di ate po si tion within this de bate is that pro pounded by Child (1972b), which he terms “stra te gic choice” (though he is talk ing about choice re gard ing struc ture). This takes the con tin gency the ory of organizational struc ture but shows that some degree of choice can nev ertheless enter in at several stages in the process (Child 1972b). Organizations can have dominant economic positions and slack re sources, so avoiding the need to adapt. Adaptation can be of con tingency to struc ture, thereby avoid ing an im per a tive to change struc ture. The effect of contingency is quite limited, and managerial de cisions intervene be tween the ob jec tive sit u a tion and the struc ture cho sen. All of these and other means in crease the role of choice (Child 1972b; see also Child 1997). In these ways, perception, belief, political in terests, and power are added to the con tin gency the ory model so that human ac tion is en tered into a model that oth er wise deals in im per sonal vari ables such the en vi ron ment, size, struc ture, and per for mance. The con clu sion of Child (1972b) is that by em brac ing stra te gic choice, or ga ni za tional the ory can avoid the pes si mism and fa tal ism that in heres in ideas that bureaucracy or other managerially favored structures are somehow in evitable. Thus stra te gic choice is held to open the door to choos ing struc tures that in clude more dem o cratic or par tic i pa tory fea tures. The full stra te gic choice model is com plex and will not be re peated here. Donaldson (1985) of fers a crit i cal ex am i na tion of its logic that ar gues that it simply sub sti tutes one set of de ter mi nants for an other,

Con tro versies in Con tin gency The ory Research

133

thereby negating choice. Moreover, he ar gues that more eman ci pated choices about or ga ni za tional struc ture come not from pur su ing re search into choice but rather from de vel op ing more valid mod els of or ga ni zational de sign that can better in form de ci sion mak ing. Thus Donaldson (1985) ac cepts the idea of choice but ar gues that reach ing the ob jec tives of mak ing wider choices in volves re search ing the out comes of choice options rather than in ves ti gat ing how choice op er ates. Sub se quently, how ever, new ev i dence led Donaldson (1996b, pp. 14-40) to ques tion how much choice re ally ex ists over struc ture and to ar gue that con tin gen cies largely de ter mine struc ture. Thus the orig i nal the sis of con tin gency the ory, that there is an im per a tive on or ga ni za tions to adopt the structure that is dictated by their contingencies to avoid performance loss, is truer than the strategic choice thesis of a large degree of choice over struc ture. Donaldson (1996b) goes through the ev i dence and shows that many of the prop o si tions in stra te gic choice the ory are not sup ported. He ar gues that the em pir i cal ev i dence to date shows the va lid ity of the de ter min is tic con tin gency the ory of or ga ni zational struc ture. For ex am ple, the av er age cor re la tion be tween size and functional specialization, when corrected for measurement error in functional specialization, is +.82 (Donaldson 1986, pp. 86, 90; also, 1996a, p. 31). This correlation is so high that there is little scope for other fac tors to in ter vene and cre ate choice. More over, as we saw in the previous chapter, functional specialization causes other aspects of structure, such as standardization and doc u men ta tion (Child 1973a, p. 182 Ta ble 9), so that these struc tural vari ables are also highly de ter mined by the size con tin gency, in di rectly through the ef fect of size on functional specialization. As another ex ample, whereas traditional con tingency the ory holds that organizations in misfit change their structure to regain fit (e.g., Chan dler 1962; Wood ward 1965), Child (1972b) ar gues that they could in stead change their con tin gency to re gain fit, thereby avoid ing hav ing to fit their structure to the contingency. In this way structure is not determined by the contingency and instead man agers have a choice. How ever, em pir i cal re search shows that in less than 5 per cent of cases or ga ni za tions re gain fit through al ter ing the con tin gency of di ver si fica tion to fit the func tional struc ture (Donaldson 1987). Even in these few cases the rea son for chang ing the con tin gen cies might not be re gaining fit, but rather strategic change forced by very low performance. While some dediversifying of or ganizations has occurred, em pirical

134

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

studies do not conclude that they are attempts to regain fit (Davis, Diekmann, and Tinsley 1994; Hoskisson and Hitt 1994). Thus the idea of a route to re gain fit through con tin gency change is a the o ret i cal notion that is seldom used in reality, so that organizations adapt their struc tures to their con tin gen cies, as con tin gency the ory holds. Stra te gic choice the ory draws on the im age of the large or ga ni za tion that is so dominant that its wealth allows it to absorb the costs of structural mis fit and also to con trol, rather than be de ter mined by, its environment (Child 1972b). While dom i nant mar ket share and other factors can to a degree off set the depressing effect on organizational performance of struc tural mis fit, the for mer are not so much more im por tant than the lat ter as to ren der the mis fit in ef fec tual for or ga ni zational change. Ham il ton and Shergill (1993, p. 79 Ta ble 8.3) in ves ti gate the effect of in dustry concentration, a proxy of the degree of oligo polization , which in dexes mar ket share. They show that this affects 27 percent of growth in corporate financial per formance, but that struc tural mis fit af fects it al most as much, by 23 per cent. For profit, indus try con cen tra tion af fects 28 per cent, while struc tural mis fit af fects 16 per cent. Thus, while mar ket power af fects per for mance more than structural misfit, struc tural mis fit is not triv ial rel a tive to the market power. Hence struc tural mis fit will sig nif i cantly de grade per for mance and in crease the prob a bil ity of ad ap ta tion. Many of the stud ies of or ganizations adapting to their changing strategy contingency are of the larg est cor po ra tions in their coun tries, such as the largest 100 or Fortune 500, and so would possess dominant market positions (e.g., Chandler 1962; Channon 1973; Dyas and Thanheiser 1976; Fligstein 1985; Ham il ton and Shergill 1992, 1993; Rumelt 1974; Suzuki 1980). Despite their large size, dominant market position, and wealth, even these or ga ni za tions have, sooner or later, bowed to the dic tate to ad just their struc ture to their strat egy con tin gency, con sis tent with the ob serva tion that the per for mance loss from struc tural mis fit is sub stan tial. Large or ga ni za tions can in flu ence their en vi ron ments through lobby ing gov ern ments, ad ver tis ing, and other means (Perrow 1986; Pfeffer and Salancik 1978). Nev er the less, this in flu ence of the or ga ni za tion on its en vi ron ment can co ex ist with the en vi ron ment in flu enc ing the or ganization and causing changes to certain internal as pects, such as its or ganizational structure (Burns and Stalker 1961; Thompson 1967). Moreover, the capacity of even large corporations to “en gi neer their environments” to make them more con du cive can eas ily be over stated.

Con tro versies in Con tin gency The ory Research

135

Hirsch (1975) shows that highly regulated environments (pharma ceuticals) al low firms to make higher prof its than lowly reg ulated en viron ments (pho no graph re cords). How ever, the phar ma ceutical industry is highly reg u lated be cause of pub lic con cerns about possible adverse ef fects of drugs, whereas much less pub lic pres sure occurs about phono graph re cords. Firms can not, at will, change their en vi ron ments so as to be highly reg u lated and prof it able. Gen eral Motors (GM) is sometimes seen as ex em pli fy ing how large corporations con trol their en viron ments. Not with stand ing some such ca pac ity, GMsuf fered declining market share in its core market in the 1980s and profit decline to the point where its CEO was dis missed, so GM and its man age ment were not able to con trol key fea tures of its en vi ron ment and their fates. Thus even GM has been forced to adapt its struc ture to changes in its stra tegic contingencies (Chan dler 1962). A number of studies have supported stra tegic choice theory by showing the ef fects on or ga ni za tional form (e.g., struc ture) of char acter is tics of their man ag ers, such as per son al ity or func tional back ground (At-Twaijiri and Montanari 1987; Finkelstein and Hambrick 1996; Fligstein 1985, 1991; Miller and Droge 1986; Miller, Droge, and Toulouse 1988; Miller and Toulouse 1986; Montanari 1979). How ever, the effect of these managerial char ac ter is tic variables tends to be secondary rel a tive to the con tin gency vari ables such as or ga ni za tional di ver si fi ca tion or size (Donaldson 1996b, pp. 41-57). Moreover, some of these man a ge rial char ac ter is tics are de ter mined by con tin gen cies, so that they are means by which con tin gen cies in di rectly af fect or ga ni zational form. For instance, the functional background of the CEO af fects the strat egy and struc ture of a firm (Fligstein 1985, 1991). How ever, provid ing a CEO is a man i fes ta tion of a func tional de partment be ing power ful rel a tive to other de part ments, and de part mental power is affected by the con tin gency of the crit i cal en vi ron men tal chal lenge facing the com pany (Hambrick 1981; Hickson, Hinings, Lee, Schneck, and Pennings 1971; Hinings, Hickson et al. 1974; Fligstein 1990a; Pfeffer and Salancik 1978). Thus, while some managerial char ac ter is tics re search pro vides ev i dence of some de gree of stra te gic choice, other re search is re veal ing the de ter min ist ef fects on or ga ni za tions of contingencies. Thus the best ev i dence to date disconfirms stra te gic choice and argues that it is mostly false, so that the ex tent of choice over struc ture is, at most, lim ited (see Donaldson 1996b, pp. 14-57 for fur ther ev i dence

136

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

and argument). Contingencies determine structure. Organizations change their structures to fit the existing level of their contingency fac tors, such as size or diversification, in order to avoid performance loss from misfit. The adaptation of structure to contingency often has a time lag, which in some cases may be for years (Dyas and Thanheiser 1976). Whittington (1989) argues that such lags constitute strategic choice. However, such time lags oc cur be cause the cri sis of low per for mance that trig gers adap tive struc tural change is de layed due to the in flu ence of other causes of per for mance. When the joint event of chronic mis fit and the other causes of per for mance de press it low enough, then the struc ture will change, so mov ing the or ga ni za tion into fit, ful fill ing the contingency imperative (Donaldson 1987). In this way, time lags in struc tural ad ap ta tion re flect the de ter min ist ef fect of per for mance rather than re sult ing from free choice or pro vid ing oc ca sions for its ex er cise. In sum, despite the attractiveness of the stra tegic choice theory, it fails to withstand scrutiny. Organizational struc tures are de ter mined to a high de gree by con tin gen cies, as shown by very high cor re la tions with con tin gen cies such as size. Or ga ni za tions al most in vari ably move into fit through adaptation of their structure to the contingency, so that the contingency is an imperative to which structure must be molded. While or ga ni za tions may af fect as pects of their en vi ron ment, nevertheless, the environment and contingencies de termine their struc ture. This holds also for large or ga ni za tions that en joy dom i nant market positions, because of the substantial performance they lose from be ing in mis fit. Or ga ni za tional slack can lead to de lays in structural adaptation, rather than its avoidance. Subjective factors affect structure, but they are minor rel ative to the contingencies. Some are de ter mined by the con tin gen cies and so are part of the way that con tingen cies de ter mine struc ture. Man agers choose new struc tures, and in that way add value, but choose the struc ture that fits the con tin gen cies, so that the struc tures are situationally de ter mined (Donaldson 1996b, pp. 50-52). While the stra te gic choice idea is ap peal ing, it has turned out to lack validity in many of its arguments. The strategic choice theory has forced a closer examination of many of the ideas that compose the origi nal contingency theory. However, the more those ideas are inspected by empirical in quiry, the more they turn out to be valid. Latterly, Whit ting ton has been in volved in an em pir i cal anal y sis of the

Con tro versies in Con tin gency The ory Research

137

adop tion of di vi sional struc ture by Eu ro pean firms, and this leads him to re verse his prior view that or ga ni za tions are gov erned by free choice and instead to accept that they are determined by diversification, as positivist theories hold (Donaldson 1996b). As he and his colleagues con clude: “. . . Donaldson’s (1996) ‘posi tiv ist’ pro ject for or ga ni za tion theory has again found international support, and in a normative context much less favourable than that of early divisionalization” (Whittington, Mayer, and Curto 1999, p. 546). Thus closer empirical study has led Whit ting ton to re verse much of his ear lier op po si tion to contingency determinism (Whit ting ton 1989).

Size as Cause of Bureaucratic Structure As we have seen, size is a prom inent con tin gency in con tin gency the ory re search. There are many studies show ing cross-sectional correla tions be tween size and var i ous as pects of or ga ni za tional structure (e.g., Blau 1972; Pugh and Hinings 1976). The theory of Blau (1970) categorically states that size causes structural differentiation and other struc tural vari ables. The Aston Group was less ex plicit, but size is al most in vari ably treated as an in de pend ent vari able in anal y ses of organizational structure in Aston program studies (Pugh and Hickson 1976; Pugh and Hinings 1976). Moreover, Child (1973a), in his theoretical interpretation of the Aston Group work, posited that size caused structure. Subsequent researchers have tended to follow this lead of treat ing size as a cause, so that cor re la tions be tween size and struc ture are in ter preted as size caus ing struc ture. How ever, this convention has been challenged as lacking theoretical justification (Kimberly 1976). Fur ther, path anal y ses have been made of the Aston data in which size is caused by struc ture (see Heise 1972; Hilton, 1972). A diachronic anal y sis (i.e., com par i son across time) in which changes in size are compared with subsequent changes in structure failed to sup port size as be ing a cause of struc ture, thereby lead ing to the ar gu ment that Blau’s general relationships are static scale phenomena and not dy nam ics caused by size (Cul len, An der son, and Baker 1986). However, other diachronic analyses show that size causes bu reaucratic struc ture. Meyer (1972) an a lyzed diachronic em pir i cal data and found that size caused struc ture and that al most none of the pos i tive, cross-sec tional cor re la tion be tween size and struc ture arose from any

138

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

re verse causal ef fect of struc ture on size, thus sup port ing Blau’s (1970) the ory that size causes struc tural dif fer en ti a tion. Marsh and Mannari (1989, p. 88, Fig ure 1) an a lyzed lon gi tu di nal data from their study of Japanese manufacturing firms and found that size caused structural complexity (horizontal struc tural dif fer en ti a tion mul ti plied by ver tical structural differentiation); moreover, the numerical estimate was only somewhat weaker than the cross-sectional association be tween size and structural complexity. Similarly, size caused administrative intensity, though the lon gi tu di nal es ti mate was about half that of the cross-sectional (for a discussion of this study, see Donaldson 1996b, pp. 98-104). Hence the Marsh and Mannari (1989) study supports Blau’s causal the ory that size causes struc tural dif fer en ti a tion and admin is tra tive in ten sity, rather than size be ing a mere cor re late of structure. Overall, the results from studies, while having some conflicting find ings, lend sup port to the view that size causes struc ture rather than to there just be ing a static as so ci a tion be tween scale and struc ture.

Reverse Causality As we have stressed, con tin gency the ory sees the con nec tion be tween con tin gency and struc ture as be ing be cause con tin gency causes structure. This implies determinism of structure by contingency, which leads to conclusions about the inevitability of bureaucracy that are resisted by some schol ars, as we saw in the sec tions above. An other way to es cape the im passe is to ar gue that as so ci a tions be tween con tin gency and struc ture arise be cause struc ture causes con tin gency. Thus, ob served as so ci a tions be tween con tin gency and struc ture lose their im pli ca tion of the de ter min ism of struc ture by con tin gency. In for mal terms, the argument is that causality is retained but inverted. Therefore con tingency de ter min ism has to an swer the re verse cau sal ity ar gu ment. The is sue has also been raised of whether con tin gency causes structure and also struc ture causes con tin gency, that is, re cip ro cal cau sal ity. Contingency theory would not be dam aged, if, in ad di tion to con tingency causing structure, structure causes con tin gency. How ever, it is damaged, if, instead of contingency caus ing struc ture, as con tin gency the ory sup poses, struc ture causes con tin gency, that is, re verse cau sality. In that case, the whole im ag ery of or ga ni za tions adapt ing to their contingency situation would be false.

Con tro versies in Con tin gency The ory Research

139

The pos si bil ity that the cau sal ity be tween con tin gency and struc ture is re ally the re verse, that is, struc ture caus ing the con tin gency, con tains a deeper problem for contingency theory. It raises the possibility that the con tin gency vari able is not re ally a con tin gency, thus call ing into ques tion the fun da men tal ba sis of con tin gency the ory. As we have seen, the strategic diversification contingency causes adoption of a di visional structure because that structure fits diversification and so its adoption restores per for mance, that is, the ex pla na tion is through an un der ly ing con tin gency fit. How ever, any re verse cau sal ity of struc ture caus ing strat egy may not be be cause of any un der ly ing con tin gency fit. For in stance, the ar gu ment that divisionalization causes di ver si fi ca tion because di visional structures have more general man agers and are thereby more likely to see the mer its of di ver si fi ca tion, is a uni ver sal ist ic theory that does not involve any contingency fit idea (Rumelt, 1974; Scott, 1971). In such a case, proving re verse causality, that is, that divisionalization causes diversification rather than diversification causes divisionalization, would disprove not only the contingency the ory of change, but also the fit idea, which is at the core of the par adigm. Hence the is sues raised by re verse cau sal ity can go be yond mere di rec tion of cau sal ity. Chan dler (1962) and oth ers (e.g., Channon 1973; Rumelt 1974) ar gue that strat egy causes divisionalization. How ever, the ar gu ment has been made that structure leads to strategy (Hall and Saias 1980), and this may be taken as chal leng ing the con tin gency the ory idea that strat egy leads to structure. As we shall show, the structure-leads-to-strategy arguments do not suc cess fully chal lenge the idea that strat egy leads to structure. Amburgey and Dacin (1994) find both that strat egy causes struc ture and that structure causes strategy, though the effect of strategy on struc ture is stron ger than vice versa. This dem on strates reciprocal cau sa tion be tween the strat egy con tin gency and struc ture, and shows that reverse causality coexists with strategy, causing structure rather than replacing it. In the Chandlerian the sis that “strat egy leads to struc ture” (Chan dler 1962), the operational meaning is principally that diversification causes divisionalization. Studies made across time (i.e., lon gi tu di nal or diachronic studies) show that diversification precedes and causes division alization. This holds both for qual i ta tive, in ten sive case his to ries (Chandler 1962) and also for quantitative stud ies of large num bers of

140

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

big corporations (Channon 1973, 1978; Dyas and Thanheiser 1976; Fligstein 1985; Rumelt 1974; Suzuki 1980). As Pfeffer (1997, p. 161) comments of these stud ies: “the ev i dence is that firms di ver si fied first and then changed their struc ture, not the other way around, as would be ex pected if struc ture caused strat egy (e.g., Rumelt, 1974; Channon, 1973).” More spe cif i cally, these stud ies show that ini tial di ver si fi ca tion causes divisionalization and that divisionalization does not cause the initial diversification (Channon 1973, 1978; Dyas and Thanheiser 1976; Fligstein 1985; Rumelt 1974; Suzuki 1980). There is also the is sue of whether, once a firm has adopted the di visional struc ture, that di vi sional struc ture in creases subsequent di ver sifi ca tion (e.g., Rumelt 1974)—with Donaldson (1982b) and Ham il ton and Shergill (1992, p. 104, Ta ble 3) find ing no such ef fect. Even if this causal re la tion ship is proved in fu ture re search, it does not contradict the pre vi ous one, that ini tial di ver si fi ca tion causes di vi sional struc ture be cause each re fers to dif fer ent causal re la tion ships. The ini tial move from be ing un di ver si fied to be ing di ver si fied causes the adop tion of the di vi sional struc ture, which, in turn, then might sub se quently raise the level of di ver si fi ca tion. Only if divisionalization were shown to pre cede and cause the ini tial di ver si fi ca tion would this re verse the cau sa tion of initial diver si fi ca tion caus ing divisionalization and thereby challenge the con tin gency the ory. In the Chandlerian the sis that “strat egy leads to struc ture,” strategy refers to achieved strat egy, such as the level of di ver si fi ca tion that the firm has at tained. Thus, “strat egy leads to struc ture,” means con cretely that firms that have achieved di ver si fi ca tion adopt a di vi sional struc ture. How ever, if, in stead, strat egy is in ter preted as stra te gic in ten tion, then presumably creating a corporate planning de part ment leads to more strategic planning, and in that sense structure can be said to cause strategy (Hall and Saias 1980). Thus one might say that “structure causes strat egy” and that this is the re verse of “strat egy causes struc ture.” However, the terms strategy and structure are here being used in a highly abstract sense to mean potentially any aspect of strategy or structure. The contingency theory idea that initial diversification causes divisionalization is only chal lenged by re verse cau sa tion be tween diversification and initial divisionalization. It is not challenged by a reverse cau sa tion be tween other mean ings of strat egy or struc ture. In sum, there are analy ses across time that show that strate g y causes structure. Attempts to date to nullify this by arguing that

Con tro versies in Con tin gency The ory Research

141

instead structure l eads to strategy are not valid, because the evidence points to reciprocal causality so that re verse causality coexists with strat egy caus ing struc ture. Regarding size and bureaucratic structure, it could be argued that for mal or ga ni za tion pre cedes size, in that man age ment cre ates a new organizational department and then recruits its per son nel. How ever, this spec u la tive causal pro cess would not pro duce some of the re search re sults that have found the cor re la tions be tween size and bu reau cratic structure. Much of those studies measure struc ture by the struc ture ac tually in ex is tence, rather than by a struc ture that ex ists purely formally and not in re al ity. Thus, for ex am ple, the Aston Group spe cialization is the num ber of dis tinct func tionally specialized roles that ex ist and to which peo ple are al lo cated, not of fi cially de fined roles that are not being per formed be cause they are un filled (Pugh et al. 1968). There fore the cor re la tions be tween size and these struc tural vari ables are due to size caus ing bu reau cratic struc ture, not the re verse. Overall, di ver si fi ca tion causes divisionalization and size causes bu reaucratic struc tur ing, and neither re lationship can be rejected on the grounds that the only ev i dence is cor re la tions that are re ally produced by re verse cau sal ity.

Configurationalism Versus Cartesianism Configurationalism pres ents two prob lems to con tin gency the ory. It rejects con tingency and it rejects multivariatism. First, it rejects the idea that organizational characteristics need to fit contingencies to produce high per for mance. In stead it ar gues that or ga ni za tional char ac teristics need to be con sis tent with each other to pro duce high per for mance. Sec ond, whereas con tin gency the ory is multivariate and sees many fits, one for each level of the con tin gency vari able, configurationalism tends to see only a few configurations. The Aston Group stud ies and other con tin gency the ory re search use what has been termed a Car te sian ap proach (af ter Des cartes; Donaldson 1996b). Cartesianism means that a multidimensional framework composed of mul ti ple vari ables is used, con sist ing of sev eral di men sions inde pend ent of each other, with each di men sion be ing a con tin uum of fine gra da tions that run from low to high (e.g., rang ing in lev els from 1 to 10). An organization could be at the lowest end of a dimension

142

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

(e.g., 1) such as spe cial iza tion, and an other or ga ni za tion could be at the high est end (e.g., 10), but other or ga ni za tions could each be at dif fer ent in ter me di ary points along the di men sion (e.g., 3, 5, 6, 8, and 9). An or ganization has a quan ti ta tive score on each of these dimensions that de fines its po si tion in the mul ti di men sional space. Thus, if there are three in depend ent di men sions, an or ga ni za tion might score 3 on the first di men sion, 7 on the sec ond di men sion, and 5 on the third di men sion. In contrast to Cartesianism, some organizational re search ers ar gue for the use of configurations (also known as types, gestalts, or arche types; Meyer, Tsui, and Hinings 1993). Configurationalism enjoys a considerable following, as seen in a special issue of the Academy of Man age ment Jour nal (Meyer, Tsui, and Hinings 1993). There is a growing literature that analyzes organizations in terms of configurations (Doty and Glick 1994; Doty, Glick, and Huber 1993; Ketchen et al. 1993; Ketchen et al. 1997). Configurationalism em ploys some el e ments of the con tin gency ap proach, such as its con cepts or find ings, for ex am ple, ma chine bureau cracy (Miller 1986; Mintzberg 1979), but uses them in ways different from con tin gency the ory re search. Configurationalism as serts that every or ga ni za tion falls into one of a few dif fer ent configurations, defined as a particular combination of char ac ter is tics or scores on vari ables. This has the ap peal of par si mony in that a model com posed of a few types is held to ex plain much of the organizational world. For example, Mintzberg (1979) discusses five configurations. The reason why there are so few configurations, Mintzberg (1979) argues, is that each one is based upon a sin gle one of the modes of coordination given by Thomp son (1967). In ter me di ary positions be tween the con fig u ra tions are seen as in ter nally in con sistent and therefore in effective and not viable. Configurationalism is based on con sis tency the ory, which holds that or ga ni za tional char acteristics need to fit each other for high performance to result (Child 1977, 1984; Khandwalla 1973); these fits be come the con fig u ra tions in con fig u ra tion the ory. From Mintzberg (1979), Miller (1986) de rives four strat egy and structure con figurations of business firms: simple structure, innovating adhocracy, machine bureaucracy, and divisionalized conglomerate. According to (Miller 1986), each con fig u ra tion is an in ter nally con sistent bundle of characteristics that constitutes a viable and effective type. For in stance, Miller (1986) ar gues that there is a sim ple struc ture organizational configuration, which is low on size and bureaucratic

Con tro versies in Con tin gency The ory Research

143

struc tur ing, and that there is also ma chine bu reau cracy or ga ni za tional configuration, which is high on size and bureaucratic structuring. There are, however, no organizational configurations that are inter mediary in size and bu reau cratic struc tur ing. There fore an or ga ni zation ven tur ing into an in ter me di ary po si tion be comes vul ner a ble and unstable and either goes back to its previous con fig u ra tion or moves rapidly to an other con fig u ra tion. Most or ga ni za tions re main sta ble in their existing con fig u ra tion, and when they do move this often takes the form of a quan tum jump from one con fig u ra tion to an other (jumping over the in ter me di ary, unviable po si tions; Miller 1986). Thus, because configurationalism holds that there are few fits, that is, con fig u ra tions, widely sep a rated in con cep tual space, this leads to its pos tu late that organizational change consists of in frequent, large movements from one fit to an other, that is, quan tum jumps. Some ver sions of con fig u ra tion the ory hold that the con fig u ra tions need not fit any contingencies, thereby contradicting contingency theory, while other versions of configuration theory hold that the configurations need to fit con tin gen cies for high per for mance to re sult (see Doty, Glick, and Huber 1993). This lat ter ver sion is more com pat ible with contingency the ory and with contingency de terminism, in that the configuration needs to fit the sit u a tion. The main dif fer ence is that con tin gency the ory holds that there are many fits and these form continua, for example, the line of fits be tween size and spe cial iza tion (Child 1975). In con trast, fits of con fig u ra tions to their con tin gen cies can take the form of a limited number of ideal fits, separated in the concept space formed by the con tin gency and or ga ni za tional char acteristic (Doty, Glick, and Huber 1993). Thus configurationalism con tains a vari ant that denies that there are con tin gen cies and also a vari ant that holds that there are contingencies of the con fig u ra tions. The noncontingency vari ant of configurationalism lends sup port to stra te gic choice. Be cause there are no con tin gen cies, the con fig u ra tions need not fit the sit u a tion (see Doty, Glick, and Huber 1993). There fore an organization can choose any one of a num ber of con fig u ra tions and still have high per for mance. Thus there is more than one way to achieve high performance, which is termed equifinality. Equifinality allows stra te gic choice, so that con tin gency de ter min ism is re jected. An other vari ant of con fig u ra tions holds that there are hy brids, so that many more than a few con fig u ra tions are vi a ble fits that lead to high per for mance (Doty, Glick, and Huber 1993). This hy brid con fig u rations

144

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

approach also has contingent and noncontingent variations (Doty, Glick, and Huber 1993). The con tin gent vari ant of hy brid con fig u rations pos tu lates that there are quite a num ber of con fig u ra tions, each of which fits a par tic u lar value of the con tin gency vari able. Clearly, this is closer to contingency theory. The more hybrids there are, and the nearer each is to one of the oth ers, the more com pat i ble with con tin gency theory is the concept of contingent hybrid con figurations. However, if there are nu mer ous hy brids (whether con tin gent or noncontingent), then configuration alism loses the par si mony and el e gance that is part of its appeal. The most prominent versions of configur ationalism tend to ar gue that there are few con fig u ra tions, and so we shall give pri mary at ten tion to the few con fig u ra tions pos tu late. This is appropriate because it is such variants of configurationalism, espe cially where noncontingent, that are the most different from, and so pose the stron ger chal lenge to, con tin gency the ory. In these ways configurationalism, especially in its noncontingent, nonhybrid form, takes elements of the pre ceding contingency theory lit er a ture and stands them on their head. Configurationalism is a ma jor form of re search us ing con tin gency vari ables and yet is an ti thet i cal to the contingency the ory from which those variables arose. This anti thesis exists regarding the number of configurations, whether they need to fit con tin gen cies, equifinality, choice, sepa ra tion in space, infre quency of move ment, or mag ni tude of jumps. We will now crit i cally ex am ine configurationalism and ar gue that it is less sound than the orig i nal con tin gency theory. Cri tique of Configurationalism There are many prob lems with configurationalism, as we shall now see, so that it is a coun ter pro duc tive move ment in or ganizational theory re search. The prob lems range from lack of the o ret i cal co her ence to lack of em pir i cal va lid ity. Miller (1990) sub se quently went on to write about con fig u ra tions in a way that contradicted his basic argument that configurations are bun dles of char ac ter is tics that to gether lead to high per for mance. He asserted in stead that configurations, while initially leading to high performance, were subject to an “Icarus Paradox” in that they tend eventually to become pathological, leading to poor performance, because the con fig u ra tions be come so pure as to be come ex treme and

Con tro versies in Con tin gency The ory Research

145

unbalanced. This argues against the whole idea of configurations as being func tional com bi na tions that are vi a ble po si tions for an or ga niza tion long term. Thompson (1967) argues that effective organizations can, and should, have sev eral co or di na tion de vices within them. There fore, the argument (Mintzberg 1979) that each viable configuration needs to have just one coordination mechanism contradicts the theory of Thomp son (1967) and the sup port ing em pir i cal re search (Gerwin and Christoffel 1974; Van de Ven, Delbecq, and Koenig 1976; Van de Ven and Ferry 1980). Hence the rationale for defining the configurations and ex plain ing why there are so few of them is un sound. The empirical claims of configuration the ory are disconfirmed by re search, some of it con ducted be fore the con fig u ra tion ap proach was de vel oped. For in stance, most or ga ni za tions are at in ter me di ary points along the continua that make up the dimensions of or ganizational context and structure, such as size and bureaucratic structure. Most organizations are not con fined to the end points of these di men sions that define the configurations, such as the small structure and the machine bu reau cracy. The reader can eas ily ver ify this for him self or her self by look ing at the scat ter-plots of data in pub lished stud ies (e.g., Blau 1970; Blau and Schoenherr 1971; Marsden, Cook, and Kalleberg 1994). The num ber of vi a ble fits be tween struc ture and con tin gency is very many more than the five held to be the con fig u ra tions. For ex am ple, the fits between bureaucratic structure and size are so many that they form a line stretching widely across the range of these variables (see Child 1975, p. 21, Fig ure 1). The fit line is a con tin uum of points of fit. There is a struc ture that fits each size level. The fits are ad ja cent to each other, so that an or ga ni za tion can pro ceed along the line from one fit to another, thereby growing incrementally from being small and unbureaucratic to large and bureaucratic. Sim i larly, or ga ni za tional change is far more com mon than the in frequent changes that con fig u ra tion the ory states, and it sel dom in cludes quantum jumps, such as between the small structure and machine bureaucracy con fig u ra tions. On the con trary, or ga ni za tions fre quently make changes, al most all of which are in cre men tal, so that they grad u ally move along the dimensions such as size and bureaucracy. For in stance, Inkson, Pugh, and Hickson (1970) stud ied the amount of change in four teen or ga ni za tions over a four-year pe riod. The quan tum jump theory would lead to the expectation that most organizations would

146

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

not change and a few would change a lot, so that the dis tri bu tion would be bimodal (zero and high), with no organizations scoring at inter mediary lev els. How ever, the per cent age change scores on the struc tur ing of ac tiv i ties vari able are: -6, 0, 3, 5, 6, 10, 14, 14, 16, 17, 21, 50, 60, and 81 (Inkson, Pugh, and Hickson 1970, p. 322). This is not a bi modal dis tri bution with organizations clustering at zero and a single, very high value—that is, re main ing sta tion ary or mak ing a quan tum jump. Instead the dis tri bution is scat tered over the range be tween high est and low est val ues. Only one or ga ni za tion scored zero (i.e., was sta tion ary), and eleven organizations increased, but by less than the maximum. Most or ga ni za tions made only in cre men tal changes. This study fails to sup port the quan tum change no tion and in stead sup ports the con tin gency the ory idea of in cre men tal change. Sim i lar re sults come from a study of the amount of change in 16 or ga ni za tions by Dewar and Hage (1978). For in stance, in the first of two three-year pe ri ods, the per centage change in struc tural com plex ity was: 0, 0, 0, 0, 0, 0, 0, 11, 14, 17, 17, 20, 26, 30, 59, and 66. Hence, half the organizations are distributed at in ter me di ary val ues be tween zero and the max i mum, so that most of them are changing incrementally (for further details, see Donaldson 1996b, pp. 120-121). The data from these two empirical stud ies re fute the quantum jump idea and instead support the contingency theory idea that if an or ga ni za tion makes a large change in its struc ture over its life, this comes about gradually, as a result of a se ries of in cremental changes. Study Fails to Con firm Con fig u ra tions Ketchen et al. (1997) conducted a meta-analytic review of stud ies examining the relationship between or ganizational configurations and performance, and concluded that configurations had a positive effect. However, some of the studies included are of strategic groups (Cool and Dierickx 1993; Fiegenbaum and Thomas 1990), which es sen tially partition firms in an industry by their performance, thereby creating a tautologous relationship between types and performance, rather than in ves ti gat ing whether or ga ni za tional struc tural con fig u rations are re lated to per for mance. Thus the re sults of this meta-anal y sis are not proof of con fig u ra tions. There fore we should fo cus in quiry on a specific study (Doty, Glick, and Huber 1993) from their review that does ex am ine struc tural con fig u ra tions and per for mance.

Con tro versies in Con tin gency The ory Research

147

Doty, Glick and Huber (1993) test Mintzberg’s (1979) five con fig u rations of context and structure and find no relationship between the con fig u ra tions and per for mance. They also find that most or ga ni za tions do not conform to one or other of the five con fig u ra tions. Thus their find ings chal lenge the whole typology of Mintzberg (1979) and those typologies de rived from his typology (i.e., Miller 1986). Doty, Glick, and Huber (1993) also test the strat egy con fig u ra tions of Miles and Snow (1978). They find these con fig u ra tions re late to perfor mance, sup port ing configurationalism. More over, they find no support for these con fig u ra tions as need ing to fit con tin gen cies, thus challenging contingency theory. However, there are problems with their analysis. Their data do not re ally sup port the no tion that there are few con fig u ra tions nor that the con fig u ra tions that ex ist re late pos itively to perfor mance, as will now be shown. Doty, Glick, and Huber (1993) find two sorts of con fig u ra tions that relate to per formance: ideal types and hybrid types. (We will follow Doty, Glick, and Huber in referring to their “ideal types” without a hyphen, while having used the hyphenated “ideal-types” to refer to those pro pounded by Weber, 1968). The ideal types mean, in the study by Doty, Glick, and Huber, that there are only three (equally ef fec tive) types that an or ga ni za tion may chose among, thus operationalizing the idea of few configurations. The hybrid fits mean that there are many more fits to chose among (because each hybrid is a combination of types and there are many such combinations). Regarding the central claim of configurationalism, that there are few con fig u ra tions, the ev idence from this study is rather dis concerting. While ideal (i.e., three) fits and hybrid (i.e., many) fits explained 24 per cent of the vari ance in per for mance, the ideal fits alone ex plained only 1 per cent and the hy brid fits explained 7 percent (Doty, Glick, and Huber 1993, p. 1236). Thus hy brid fits ex plained seven times as much as ideal fits, which on their own explain only a triv ial amount of per for mance. (The bulk of the vari ance ex plained, 16 per cent, may be due to some kind of in ter action of ideal with hy brid fits.) Thus the cen tral claim of con fig u ra tions, that there are few viable types be cause only a few combinations of variables pro duce high per for mance, is not sup ported by the data in this study. Turn ing to the find ings that sup port the re la tion of ideal and hy brid types to per for mance, the data are again con trary. As Doty, Glick, and Huber (1993) state, neither the ideal nor the hybrid types cor relate

148

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

positively and significantly with any of the five measures of per formance. The ev i dence sup port ing the re la tion ship be tween the ideal and hy brid fits and perfor mance co mes from ca non i cal anal y ses that aggregate across vari ables (Doty, Glick, and Huber 1993). However, the relationships in cluded differ in sign, so that their aggregation is prob lem atic. Of the ten cor re la tions be tween the ideal and hy brid fits and the five measures of effectiveness, eight are positive and two are negative. Therefore aggregating positive and negative correlations would appear to be meaningless, because the positive correlations support the theory but the negative correlations refute it. Moreover, one of the neg a tive cor re la tions is the larg est, and the only sig nif i cant correlation, so that this counterfinding should be given more weight than the more numerous but nonsignificant findings that support the the ory. There fore aggregating to gether pos i tive and neg a tive cor re lations boosts the as so ci a tion be tween fits and per for mance, but could mis lead, in that the stron gest source of as so ci a tion could come from a correlation that ne gates the the ory that fit and per for mance are pos i tively re lated. Dis re garding sign, the av er age of the ten correlations is +.078; allowing for sign, however, their average correlation drops to only +.017, which is al most nil. Thus the ca non i cal anal y ses that sup port the re la tion of ideal and hy brid fits to per for mance may be mean ingless. The more mean ing ful re sults are the ten cor re la tions be tween the ideal and hybrid fits and per for mance and, as noted al ready, none of these correlations is significantly positive. The study by Doty, Glick, and Huber (1993) should be coded as fail ing to sup port the idea that configurations lead to high performance. By implication, their study should not be seen as supporting the idea that there are multiple equally effective configurations, because, in fact, no configuration leads to higher lev els of ef fec tive ness. In sum, even for the Miles and Snow typology, the study by Doty, Glick, and Huber (1993) fails to support the idea of a few types, sup porting in stead the existence of many hybrids. However, neither the ideal types nor the hybrid types re late to per for mance. Thus the study should be coded as not supporting con fig u ra tions. More over, there is no sup port for the idea of equifinality, that is, sev eral, equally highly effective types, among which choice can be made with out los ing per formance, be cause no type relates positively to performance. If

Con tro versies in Con tin gency The ory Research

149

config urationalism is not re ally sup ported by their data, do their findings sup port con tin gency theory? Doty, Glick, and Huber (1993) find no evidence that contingency formulations of their types are pos i tively re lated to per for mance, and so contingency the ory is not supported. However, Doty, Glick, and Huber (1993) them selves warn against in ter pret ing their study as ev idence against the con straints implied by con tin gency the ory. As they cau tion, there may still be con straint op er at ing within their sam ple in that only certain combinations of structure and contingency may be viable and so the organizations have already moved into alignment. This is a point made by Van de Ven and Drazin (1985) that, in their terms, natural se lec tion brings struc ture into cor re la tion with con tin gency and so elim i nates de vi a tions awayfrom fit, mak ing unobservable the as so ci a tion be tween mis fit and lower per for mance. Thus lack of empirical support for contingency, in the Doty, Glick, and Huber (1993) study (and other stud ies) does not mean such con straint is not op er at ing and that con tin gency the ory is not valid. More over, while the study pro vides no sup port for con tin gency theory, this may re flect the way that con tin gency fit is operationalized (Doty, Glick, and Huber 1993). Fol low ing con fig u ra tional logic, con tin gency fit is operationalized as be ing the de gree of fit of the struc ture to the ideal type clos est to the con tin gency level of the or ga ni za tion. This is a procedure that only par tially cap tures the de vi a tion from the fit line of structure on contingency, which is the correct test of contingency theory. An organization that lies on the fit line and thereby is in fit could nev er the less be at a dis tance from what is claimed to be its ideal type, so that that or gani za tion would be scored as be ing in mis fit. The problem arises be cause contin gency theory sees ev ery point along the fit line as a fit, whereas con fig u ra tion the ory, in its con tin gent form as used in this study, sees few points on the fit line as fits (i.e., as the configurations). To il lus trate the point, if the fit line where de fined as required struc ture equals the con tin gency, then there are fits at each point of struc ture equals 1, contin gency equals 1; at 2, 2; at 3, 3; at 4, 4; at 5, 5; and so on. But con tin gent configurationalism would hold that there were few ideal types, such as at 1, 1 and at 4, 4, so that an or ga ni za tion in fit at 3, 3 would be coded as be ing a mis fit with its near est ideal type (4, 4). Hence the mis fits could con tain some fits, so con found ing

150

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

the em pir i cal re la tion ship be tween fit and per for mance. There fore failure to support contingency theory may reflect a failure to operation alize fit cor rectly. Miller (1986) ob serves that many the o ries use sim i lar types and sees in this a ver i fi ca tion of configurationalism. How ever, the sim i lar ity of types re flects that the o rists have tended to draw from the same pool of ideas. Miller (1986) also justifies configurations by reference to the typologies of or ga ni za tional the o rists such as Burns and Stalker (1961) with their mech a nis tic and or ganic types of struc tures. How ever, Burns and Stalker (1961) cau tion that or ga ni za tions are dis trib uted along the con tinua formed by these two ex treme types, with some or ga ni za tions oscillating be tween the two types and many or ga ni za tions in cor po rating both types, ideas that all contradict configurationalism: the two forms of sys tem [mech a nis tic and or ganic] rep re sent a po lar ity, not a di chot omy; there are, as we have tried to show, in ter me di ate stages be tween the ex trem i ties em pir i cally known to us. Also, the re la tion of one form to an other is elas tic, so that a con cern os cil lat ing be tween rel a tive sta bil ity and relative change may also os cil late be tween the two forms. A concern may (and fre quently does) op er ate with a man age ment sys tem which in cludes both types. (p. 122)

Subsequent or ganizational the o rists work within this tra di tion of viable structures as distributed along a continuum be tween extreme, po lar types. For in stance, Gresov (1990, p. 506, em pha sis as in orig i nal) writes of his own work: The view em braced here is that unit de sign pat terns can be con cep tu al ized as ranging on a spectrum from the mechanistic (tight) to the organic (loose), in which the “pure” forms ap proach the end points. The spec trum ap proach has been used re cently in a num ber of stud ies (e.g., Al ex an der and Randolph 1985; Koberg and Ungson 1987; Tushman 1979).

Thus the po lar types of mech a nis tic and or ganic are part of a con tin uum, or spectrum, thereby contradicting the notion that they are con figurations. Cri tique can also be made of sev eral of the con fig u ra tions of fered in the popu lar config u ra tional the ories, such as the “divisionalized con glomerate” configuration (Miller 1986, p. 115). Research into

Con tro versies in Con tin gency The ory Research

151

divisionalized corporations reveals many types other than the con glomerate, such as the re lated prod uct and ver ti cally in te grated types, each of which re quires their own vari ants of the di vi sional struc ture to fit them (Hill, Hitt, and Hoskisson 1992; Lorsch and Al len 1973). Sim i lar critiques can be made of other con fig u ra tions, such as the in no vating adhocracy and ma chine bu reau cracy con fig u ra tions (see Donaldson 1996b, pp. 114-117). How ever, the way for ward is not to amend one or more con fig u ra tions nor to in crease their num ber, but rather to aban don configuration alism in fa vor of Cartesianism. This en tails us ing the kind of mod els of con tin gency, struc ture, and fits as con tinua as pre sented in this book. The con fu sion in duced by configurationalism is, how ever, to be distin guished from an other use of con fig u ra tions. Be cause or ga ni za tional types can ex ist in the mind, as men tal mod els, they can guide peo ple and be used to eval u ate ac tual or ga ni za tions. The fact that the types may re main ide als that are never fully at tained in prac tice need not di min ish their ori ent ing role. More over, peo ple may only see a few types within their minds. Some configurations are used in this way as archetypes that provide a framework for human thinking about organizations, which, in turn, guide actions to try to change organizations to make them more like the ar che types, that is, more like the ide als. For in stance, Coo per, Hinings, Green wood, and Brown (1996) iden tify two con trast ing archetypes in professional service firms: a traditional professional part ner ship model and a busi ness man age ment model. They chart how mem bers of some pro fes sional ser vice firms have sought to move their firms away from the first ar che type and to ward the sec ond ar che type, along var i ous tracks and with vary ing out comes, in part be cause some other mem bers seek to re tain el e ments of the first ar che type. The or ganizations just move some way along the imaginary continuum from one ar che type to an other, so that the ar che types re main ex treme states that are not at tained in re al ity, that is, the ar che types are ide al iza tions. Real organizations are distributed along the continuum between the archetypes. This use of organizational types is valid because they are ab stract ide al iza tions, which al lows there to be few of them. The archetypes part of the configurations school is legitimate and scientifically fruitful. It is to be distinguished from the doctrine that there are few or ga ni za tional types or con fig u ra tions in the world. We may term the ar che types ascognitiveconfigurations (i.e., men tal mod els) to dis tin guish them from the no tion that re al ity is com posed of a few

152

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

types, which we may term ex is ten tial con fig u ra tions. As we have seen, existential config u ra tions fail to ex plain the or ga ni za tional world. In con trast, cog ni tive con fig u ra tions help to ex plain or ga ni za tional de sign, organizationalchange, and mem ber re ac tions to or ga ni za tional de signs. How ever, ex is ten tial configurationalism would seem to be more com mon in or ga ni za tional the ory re search than cog ni tiveconfigurationalism (con sider, for ex am ple, the ar ti cles in the spe cial issue on con fig u rations in the Acad emy of Man age ment Jour nal; Meyer, Tsui, and Hinings 1993). The influence of cognitive configurationalism seems to be stron gest in the spe cial ist topic of pro fes sional ser vice firms, so that it is confined. Therefore, at present, the net effect of configurationalism would seem to be more neg a tive than pos i tive. In sum mary, ex is ten tial configurationalism has at trac tions of par simony, but this is beguiling. Real organizations are widely distributed along both the structural and the contingency di men sions. There are many more fits than the few con fig u ra tions pos tulated. Fits form con tinua along which or ga ni za tions can move incrementally. Or ganiza tions do not change through oc ca sional quan tum jumps, but frequently and incrementally. Configurationalism, in its most de vel oped vari ant, holds that con fig u ra tions are un bounded by con tin gen cies so that there are many struc tures that an or ga ni za tion can adopt and still have high performance, thereby constituting equifinality and choice. However, these ideas remain as assertions unproven em pirically and con tra dicted by the em pir i cally proven fits of struc ture to con tin gency (e.g., Child 1975). We must therefore re ject ex is ten tial configurationalism: the notion that fits are localized to being at discrete points in space with no vi a ble in ter me di ary fits be tween them (for fur ther crit ical dis cus sion, see Donaldson 1996b, pp. 108-129). In stead, we should use the Cartesian ap proach that contingencies and structures form a multidimensional framework within which fits are continuous. Thus the fits be tween a struc tural and a con tin gency di men sion are a line in two-di men sional space, so that fit is a con tin uum. Be cause fit forms a con tin uum, each fit has an ad ja cent fit so that an or ga ni za tion can migrate be tween fits, form ing a path way. This ex plains how or ga ni za tions can change over time, incrementally, from be ing small, struc tur ally simple, and local to being, in extreme cases, large, structurally com plex, and mul ti na tional.

Con tro versies in Con tin gency The ory Research

153

Contingencies of Organizational Power A branch of struc tural con tingency the ory is concerned with the power as pect of or ga ni za tional struc ture. It can be sub sumed un der resource dependence theory (Pfeffer and Salancik 1978), which holds that or ga ni za tions are de pend ent upon the en vi ron ment for re sources and seek var i ous means to at tain these re sources while try ing to main tain their au ton omy from the ex ter nal pro vid ers. One such means is the internal organizational structure. The external resource dependence becomes a con tin gency that shapes the in ter nal or ga ni za tional struc ture in two regards: the power of departments and the structure of the board of di rec tors. Re source de pend ence the ory spec i fies many other means for the or ga ni za tion to gain re sources and in flu ence its en vi ronment, but here we will fo cus just on in ter nal or ga ni za tional struc ture in keep ing with the theme of this book (for a crit i cal dis cus sion of re source dependence theory in general, see Donaldson 1995b, pp. 129-163). If re source de pend ence the ory is valid, then re source de pend ence be comes an other con tin gency of or ga ni za tional struc ture, ad di tional to the task and size con tin gen cies that con sti tute the main struc tural con tin gency theories. We will first discuss the strategic contingencies theory of intra-organizational power and then discuss co-optation theory. The Stra te gic Con tin gencies The ory of Intraorganizational Power The de pend ence of the or ga ni za tion upon ex ter nal re sources af fects which de part ment has the most power, thereby in flu enc ing an as pect of the power dis tri bu tion within the or ga ni za tion. This the ory is re ferred to as the stra te gic con tin gen cies the ory of intraor ga ni za tional power. The term contingency re fers here narrowly to some environmental chal lenge with which the or ga ni za tion must cope (Hickson et al. 1971). For ex am ple, in a study of brew er ies in west ern Can ada, gov ern men tal regulations precluded the use of many competitive tac tics such as market ing, and there fore low-cost pro duc tion was the crit i cal, en vi ron mentally de fined, chal lenge that de ter mined or ga ni za tional per for mance (Hinings et al. 1974). There seems usu ally to be only one chal lenge so we will use the sin gu lar, con tin gency, rather than contingencies. The

154

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

word contingency is used in this the ory to re fer not to the broad ar ray of contingency varia bles in contingency theory (i.e., size, technology, etc.), but rather to just one fac tor, the crit i cal chal lenge. This use of the word contingency dif fers from most of the con tin gency lit er a ture, but is consistent with the use of that word in Thompson (1967). While the theory of Thomp son is clearly of the contingency type, specifying modes of organization as appropriate to various situ a tions, he does not present it in con tin gency terms. Instead, Thomp son (1967, p. 24) uses the term “con tin gency” more nar rowly to re fer to some thing that var ies and is be yond the con trol ofthe or ga ni za tion, so that the or ga niza tion must make vary ing re sponses. Tech nol ogy and the other sit u a tional fac tors that later would be called con tin gen cies are not dis cussed in these terms by him. Nev er the less, Thomp son put forth the stra te gic contingency theory of or ganizational power, and the Thompsonian use of the word “con tin gency” is seen sub se quently in the for mal statements of that the ory. Hence, in this theory we find the id iosyncratic use of the word “contingency” to refer narrowly to only one con tingency, the critical challenge facing the organization, rather than the broad range of contingencies dis cussed in con tin gency theory. Further, while the term strategic contin gen cies is used, the word strategy here is re lated to the idea of stra te gic be hav iors, such as work groups acting to maintain their power and status (Crozier 1964). “Strategy” in the strategic contingency theory of intraorganizational power does not em brace other mean ings of strategy found in the con tingency research, such as re fer ring to di ver si fi ca tion or ver ti cal in tegration (Chan dler 1962; Rumelt 1974) or de fender ver sus pros pec tor (Miles and Snow 1978). For these rea sons, we be lieve that re fer ring to the stra te gic con tin gency the ory is un help ful. We pre fer to clas sify the con tin gency re ferred to in the stra te gic con tin gency the ory as the crit ical chal lenge. This is con sis tent with it be ing only one type of con tin gency, rather than be ing called the con tin gency (or the con tin gen cies), which is mis lead ing given that there are other con tin gen cies in con tingency the ory (e.g., size and tech nol ogy). The stra te gic con tin gency the ory of intraorganizational power holds that or ga ni za tions are most ef fec tive where the de part ment that deals with the most critical environmental challenge of the organization is the most pow er ful de part ment in the or ga ni za tion. This idea fea tures in early contingency the ory research (Lawrence and Lorsch, 1967, pp. 142, 143; Perrow 1970). A for mal model was spec i fied by Hickson

Con tro versies in Con tin gency The ory Research

155

et al. (1971), in which a de part ment has more power than its peer departments (i.e., those at the same level in the hierarchy) to the de gree that it deals with the critical challenge, does so successfully, and i s nonsubstitutable. This model was empirically validated by subsequent re search (Hambrick 1981; Hinings et al. 1974; Kenny and Wil son 1984). How ever, Hambrick (1981) showed that which chal lenge is de fined as most crit i cal for an or ga ni za tion is in flu enced by the strat egy adopted by that or ga ni za tion. In this way there is an el e ment of stra te gic choice by man age ment, so that com plete de ter min ism of the crit i cal chal lenge by the en vi ron ment is not a valid view. Thus it may not be wholly true to clas sify the crit i cal chal lenge asan en vi ron men tal chal lenge. This undermines, to a degree, the argument of resource dependence the ory that the organization is externally controlled (Pfeffer and Salancik 1978). Thus this contingency of power should be termed the critical chal lenge, in rec og ni tion that it could be ei ther the en vi ron ment or the man age ment that de ter mines which chal lenge is crit i cal. Astley and Zajac (1990) pro vide an em pir i cal cri tique of the the ory that power re sides with the de part ment that con trols the crit i cal challenge. In a study, they test the al ter na tive ex pla na tion that the de partment that is most cen tral to the work flow is the most pow er ful. Their find ing is that cen tral ity, not the crit i cal chal lenge, ex plains de part men tal power. Thus the contingency of intraorganizational power could be centrality, not critical challenge. This possibility places a caution against the va lid ity of the crit i cal chal lenge con tin gency the ory and deserves at ten tion in fu ture re search into intra-or ga ni za tional power. Co-optation Co-optation the ory states that the or ga ni za tion co-opts mem bers of pow er ful out side or ga ni za tions so that they work to fur ther the in terests of the organization that has co-opted them, thereby modifying the power relationships be tween or ga ni za tions (Pfeffer and Salancik 1978). An im por tant in stance is the ap point ment to the board of di rectors of an or ga ni za tion, of per sons who hold po si tions in or ga ni za tions that have power over that or ga ni za tion, so that those ex ter nal or ga ni zations then as sist the organization by pro vid ing the re sources it needs. Pfeffer (1972) argues a contingency ver sion of this the ory, con sis tent with resource dependence theory, by stating that or ga ni za tional perfor mance will be greater, the more that a com pany in cludes in its board

156

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

di rec tors from or ga ni za tions that pos sess the re sources that the com pany needs. He supports this empirically with a study that finds that the pro por tion of out side di rec tors is higher for com pa nies that have higher leverage (i.e., are dependent on more debt financing) and are more regu lated (i.e., are more subject to governmental control). In par tic u lar, the higher the proportion of directors who are attorneys, the more the company re lies on debt and is regulated (Pfeffer 1972). The more the company de vi ates from the propor tion of out sid ers ex pected from the con tin gen cies of debt and reg u la tion, the worse is its financial p erformance (controlling for industry; Pfeffer 1972). This supports the contingency co-optation theory that boards need to fit these contingencies for the company to perform at a higher level. However, whether the attorneys and other outside di rec tors are from or ga ni za tions that could pro vide such sup port to the or ga ni za tion and whether their be ing ap pointed to the board has led them to use their influence to ben e fit the com pany (i.e., to be co-opted) are un demon strat ed (Mintzberg 1983, p. 87). Thus there is scope for more em pir i cal research to con firm that the pro cesses oc cur ring are those pos tu lated by co-optation the ory. An other study of boards of directors shows that outside directors have a neg a tive ef fect on the fi nan cial per for mance of the firm, thereby sug gest ing that such ap point ments may not al ways suc ceed in co-optation (Muth and Donaldson 1998). This negative re lationship only holds where there is high network connections, meaning that the board directors are highly linked to other firms (Muth and Donaldson 1998). Thus there may be a proc ess of dysco-optation whereby di rec tors use their board seats to ex tract wealth from the fo cal firm for the ben e fit of other economic organizations in which they have an interest. Hence board directorships could reduce, rather than increase, the resources avail able to an or ga ni za tion. Thus not all ap point ments of out sid ers to boards are co-optation, and the re verse pos si bil ity should be con sid ered in fu ture stud ies of power and boards. It could be that some di rec tor ships func tion as co-optation and oth ers as dysco-optation. Which of these interorganizational processes occur could be determined by some con tin gency so that it would be ex pli ca ble by a new the ory of the contingency type. Thus con tin gency the o ries have been used to ex plain power in and between or ganizations. For both the critical challenge contingency the ory of intraorganizational power and the con tin gency co-optation

Con tro versies in Con tin gency The ory Research

157

theory of interorganizational power, there is evidence supporting them, but also either counterevidence or meth odological criticisms that ren der them some what con tro ver sial. It is hoped the dis pu ta tion around both the o ries will stim u late more re search to clar ify the va lid ity of these con tin gency the o ries of or ga ni za tional power.

Conclusions Whereas pi o neer ing con tin gency the ory treated tech nol ogy as a ma jor contingency fac tor, sub sequent bureaucracy theory research argued against the tech nol ogy the sis and in fa vor of size as the more im por tant con tin gency the ory of or ga ni za tional struc ture. In re sponse, some schol ars have sought to re as sert the tech nol ogy the sis. How ever, tech nol ogy is a less im por tant con tin gency than size. Also, it has ef fects different from those claimed in the pioneering technology thesis. Instead of mak ing struc tures more or ganic, it makes struc tures more bu reau cratic. This holds for in f or ma tion tech nol ogy as well as for operations tech nol ogy. Thus the ef fect of tech nol ogy par al lels that of size. Advances in tech nol ogy re in force the ef fects of size in creases, mak ing the or ga ni za tion even more bu reau cratic than it would be if size alone increased. On the issue of the technology contingency, bureaucracy the ory is supported rather than or ganic the ory. Against the determinism of structure by contingency pos ited by contin gency the ory in its orig i nal ver sion, some schol ars have as serted instead that there is a substantial zone of choice. However, a closer inspection has revealed dif ficulties with many of the arguments for stra te gic choice and con sid er able de ter min ism of struc ture by con tingen cies. Con tin gencies such as size cor re late very highly with piv otal aspects of struc ture. Or ga ni za tions in mis fit have to adopt the structure that fits their contingency. This is due to the performance loss from mis fit, which is sub stan tial even for large cor po ra tions with market power. Sub jec tive fac tors play a mi nor role in struc tural cau sa tion and some of them are paths by which contingencies in di rectly de termine struc ture. There is very lit tle ev i dence of any sub stan tial amount of stra te gic choice that al lows or ga ni za tions to re ject bu reau cracy and in stead adopt the or ganic struc ture. Critics some times con tend that con tin gency re search just con sists of correlations be tween con tin gen cies and struc ture, which could mean

158

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

that the con tin gen cies are just cor re lates of struc ture or that struc ture causes the con tin gen cies. How ever, there is ev i dence that size is a cause of bu reaucratic struc ture. Moreover, diachronic data and causal anal y ses show that the strategy contingency is a cause of structure. Some attempts to establish reverse causality have not used the same concepts of strategy and structure and hence failed to reverse the causality meant by “strategy leads to structure.” Others have shown that re verse cau sal ity co ex ists with, rather than re places, the causal effect of strat egy on struc ture. Thus re verse cau sal ity fails to con sti tute a refutation of the contingency theory proposition that contingency causes structure. Hence arguments for reverse causality may lead toward a richer model, but they do not damage contingency theory. The de ter min ism of struc ture by con tin gen cies as stated in bu reau cracy the ory is in tact. In stead of study ing or ga ni za tions in terms of con tinua of con tin gencies and struc tures with many fits, some schol ars have as serted few fits, or con fig u ra tions. How ever, there are prob lems with configuration alism. The con ten tion that there are few fits be cause each or ga ni za tion must use only one co or di na tion mode con tra dicts the the ory of Thomp son (1967), so that the con fig u ra tion ar gu ment lacks a co her ent the o ret i cal base. There is no com pel ling ev i dence that there are only a few fits, nor that they re late pos i tively to or ga ni za tional per for mance. Sim i larly, the contention that organizational change consists of quantum jumps, with most or ga ni za tions mak ing no change and a few making a large change, is con tra dicted by ev i dence that shows that most or ganizations make changes that are in cre men tal in mag ni tude. A more co her ent and more em pir i cally valid view is that con tained in con tin gency the ory re search. The con tin gency the ory view is that there are many more fits than the few configurations. These fits are arranged adjacent to each other, providing mi gration routes for organizations to change incrementally, so that or ga ni za tions may change grad u ally. Cartesian ism, which treats contingencies, structure, and fits as continua, re mains sounder. Thus mod ern con tin gency the o ries, whether bu reau cracy or or ganic, should con tinue with the Car te sian con tin gency ap proach. Thus, over all, the crit i cisms of bu reau cracy the ory, whether ex plicit or im plicit, that are pro pounded by tech nol ogy, stra te gic choice, re verse causality, and configurationalism fail to damage it. There are sound rea sons for cleav ing to contin gency the ory in its orig inal form: a Car tesian frame work, in which con tin gency causes struc ture de ter min-

Con tro versies in Con tin gency The ory Research

159

istically. The main vari a tion from the orig i nal ver sion of con tin gency the ory is to hold that size, rather than technology, is the more major contingency of structure. Re gard ing power, struc tural con tin gency the o ries pro vide an ex pla nation for its dis tri bu tion in and among or ga ni za tions. While sup port has been garnered for the crit i cal chal lenge the ory of intraor ga ni za tional power, later work has challenged its va lid ity. Equally, the con tin gency vari ant of co-optation the ory re ceives some em pir i cal sup port, but has also been chal lenged by later re search. Thus the con tin gency the ory of power is a can di date for fu ture re search to clar ify the ex tent of its va lidity. The outcome of such research might lead in some cases to new forms of con tin gency the ory.

THE CON Chal lengesTIN From GENCY OtherTHEORY The oriesOF OR GA NI ZA TIONS

6

Challenges From Other Theories

C

ontingency theory has been chal lenged by other the o ries, such as institutional the ory. In this chapter we will consider several challenges from other organizational theories. It will be seen that the con tin gency the ory ex pla na tions con tinue to en joy va lid ity. Com peting explanations from other theories are not as valid as sometimes be lieved. Thus the present chapter will present re buttals on be half of contingency the ory. While pre sent ing a rea soned case for con tin u ing to en ter tain con tin gency the ory, some of the points re main in con ten tion, thereby constituting is sues for fu ture re search. Much of the challenge to contingency the ory is provoked by the func tion al ism and ra tio nal ism of con tin gency the ory. As we saw above (in Chap ter 1), con tin gency the ory ex plains or ga ni za tional change as a functionalist process of ad ap ta tion. The organization adopts a new struc ture that better fits its new level of the con tin gency vari able. For ex am ple, the or ga ni za tion adopts a di vi sional struc ture to fit its di versi fied strat egy. Or again, the or ga ni za tion in creases its size and then inc reases the level of its bu reau cratic struc ture to bring it into fit with 161

162

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

size. In such ways, the or ga ni za tion re places dys func tional with functional (i.e., ef fec tive) struc tures. This is ra tio nal for the or ga ni za tion in that it fulfills its goal to be effective. Given that the organizational change co mes about through the de ci sions made by its man ag ers, they are act ing in a pro-or ga ni za tional fash ion. Such ra tio nal ism and func tion al ism, how ever, are quite at odds with other or ga ni za tional the o ries.Institutional the ory em pha sizes rit ual more than rationality and processes of con formity, such as those to gain legitimacy with pow er ful out side bod ies, even with out gain to in ter nal organizational effectiveness (DiMaggio and Powell 1983; Meyer and Rowan 1977; Meyer and Scott 1983; Scott 1995). Population-ecology em pha sizes or ga ni za tional in er tia, so that an or ga ni za tion may fail to adapt and be come dys func tional, leav ing change to oc cur at the level of the population through death and birth of organizations because of lack of rational organizational ac tion (Hannan and Freeman 1977, 1984, 1989). Political explanations hold that or ga ni za tional change is ra tio nal for in di vid u als as they pur sue their self-in ter est, but of ten costly to the organization, for example, where managers “empire-build,” so that many changes are organizationally irrational and dysfunctional (Child 1973b; Freeman and Hannan 1975; Parkinson 1957; Reimann 1979). Sim i larly, in or ga ni za tional eco nom ics, Wil liam son (1970, 1985), in hisM-form the ory, sees this struc ture as adopted by or ga ni za tions in an at tempt to con strain the dam age from man a ge rial self-in ter est, so that it is an or ga ni za tion ally ra tio nal re sponse to or ga ni za tion ally ir ratio nal, po lit i cal man a ge rial be hav ior. Fur ther, the psy cho log i cal the ory of threat-rigidity sees or ga ni za tional change as gov erned by irrational and dys func tional pro cesses (Staw, Sandelands, and Dutton 1981). Thus, arising from the chal lenges of these other the o ries, we dis cuss in this chap ter the ra tio nal ist, func tion al ist con tin gency the ory, compared with the institutional, population-ecology, political, M-form, and threat-rigid ity theories. These different views have prompted fur ther work on the phe nom ena of strat egy and struc ture, so that we extend our discussion of this con tingency relationship to critically assess the validity of these alternative views. Also, the challenges to contingency the ory have in volved anal y ses of the re la tion ship be tween size and bureaucratic struc ture so, again, we re visit these phe nom ena to see that the challenges have not succeeded in overturning con tingency the ory.

Chal lenges From Other The ories

163

Divisionalization Is Not an Institutional Effect Institutional theory holds that organizational change is less rational than contingency theory states and is not oriented toward adopting struc tures of su pe rior ef fec tive ness, rather their adop tion is spurred by institutional isomorphism (DiMaggio and Powell 1983; Meyer and Rowan 1977; Meyer and Scott 1983; Powell and DiMaggio 1991; Scott 1995; Scott and Meyer 1994; Zucker 1977, 1987). The explanation of organizational change is change by coercive, mimetic, or normative pro cesses that are ori ented to ward gain ing le git i macy through con formity to in sti tu tion ally ap proved or ga ni za tional forms (DiMaggio and Powell 1983). Above, in pre sent ing ev i dence for con tin gency, we used stud ies of divisionalization; how ever, there have been three stud ies of divisionalization that are of ten in ter preted as sup port ing the in sti tutional the ory and con tra dict ing the ra tio nal, func tional view of con tingency theory (Scott 1995, pp. 119-120): Rumelt (1974), Fligstein (1985), and Ar mour and Teece (1978). We will briefly re view each and show that, at the least, a contingency theory interpretation is equally plausible, so that none of the stud ies pro vide decisive evi dence for insti tu tional the ory against con tin gency the ory. Rumelt (1974) in ter prets much divisionalization in the 1960sas due to fol low ing a fash ion rather than be ing ra tio nal and func tional in ef fectiveness. However, his analysis lacks an examination of whether the firms stud ied were mov ing into a fit that raised their fi nan cial per formance when they divisionalized. A secondary analysis of his data shows that all the firms that divisionalized in the six ties moved into fit (Donaldson 1995b, p. 90). Thus the firms stud ied by Rumelt should be coded as making rational, sociologically functional changes, rather than fol low ing fash ion. Fligstein (1985) pres ents ev i dence that firms are in flu enced in their divisionalization by whether other firms have adopted the divisional form. Yet this could be be cause the fo cal firm with its func tional structure is in misfit and suf fering poor competitiveness relative to its divisionalized com pet i tors, which, as we have seen, are in fit. There fore the influence of in dustry peers could be competitive pressure that makes divisionalization economically ra tio nal. Actually this in ter preta tion is more con sis tent with the data, be cause it im plies that, for any number above zero, the more industry peers that divisionalized, the

164

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

greater the prob a bil ity of the fo cal firm divisionalizing—which is what Fligstein (1985) shows. In contrast, a mimetic interpretation would hold only once the divisional structure was in the majority (which is con sis tent with the idea in in sti tu tional the ory that mi metic pro cesses oc cur only for late adop tions, when the new or ga ni za tional form would be being used by most organizations [Tolbert and Zucker 1983]). There fore proof of a mi metic pro cess would re quire show ing that only in those in dus tries where more than half the firms had divisionalized was more divisionalization of in dus try peers in creas ing the prob a bil ity of divisionalization of the fo cal firm—which Fligstein (1985) does not examine. Again, in sti tu tional the ory em pha sizes that the adop tion of or ga niza tional forms is not for rea sons of ef fec tive ness, so that it is rit ual, not ra tio nal. While Fligstein (1985) gives this in sti tu tional the ory ex pla nation, for the pur port edly mi metic ef fects in his data to be de ci sive there would need to be evidence that firms adopting divisional structures were not thus mov ing into fit and so rais ing their per for mance. How ever, Fligstein does not classify his firms into fit or mis fit, so that his anal y sis can not re ject func tion al ist in ter pre ta tion of divisionalization. Scott (1995, p. 120) makes the same crit i cism of Fligstein (1985): “the [cases of mimetically caused divisionalization] are treated as non rational. How ever, with out in for ma tion on the ef fect of these de ci sions on per for mance mea sures, it is not pos si ble to make this de ter mi na tion.” As wehave just seen, an ex am i na tion of divisionalization among large U.S. firms (Donaldson 1987) instead of a set similar to those studied by Fligstein, shows that overwhelmingly the adoption of the divisional form was a move into fit and there fore ben e fi cial for their ef fec tive ness and so so cio log i cally func tional. Hence the the o ret i cal mean ing of his study is equiv o cal and should not be coded as proof that in sti tu tional theory is a more valid explanation of divisionalization than con tingency the ory. Fu ture re search may dem on strate, un equiv o cally, mi metic pro cesses, but if firms that divisionalize are fol low ing firms whose division alization has brought them into fit and im proved their fi nan cial per for mance, then the outcome could be economically rational and functional for the fo cal firm. By show ing that firms are in flu enced in their adop tion of structures by other firms, such as industrial peers, one of the ideas of in sti tu tional the ory could be drawn upon—though call ing this pro cess mim icry would be in ac cu rate be cause it is pe jo ra tive. In this way,

Chal lenges From Other The ories

165

se lected elements of institutional the ory could become com ple mentary to con tin gency the ory, il lu mi nat ing the so cial and psy cho log i cal processes through which firms see what new struc ture they need to adopt to fit their new con tin gen cies. How ever, this rap proche ment would require abandoning the institutional theory assertion that organiza tional form is dictated by ritual rather than rationality (i.e., effectiveness in an instru men tal sense). The study of divisionalization by Armour and Teece (1978) found that divisionalization raised per for mance in an ini tial pe riod, but not in later years. In sti tu tional the ory in ter prets this as mean ing that early adoption of the divisional struc ture was economically rational, but that later adoptions were conforming to the norm and so mimetic. However, Ar mour and Teece (1978) do not examine whether their firms were in fit or mis fit and so could not show that divisionalizing in the ini tial pe riod was move ment into fit with strat egy or other con tin gen cies, whereas for late adopt ers it was less eco nom i cally ra tio nal by being less likely to be a movement into fit. Thus there is no decisive evidence to elim i nate a con tin gency ex pla na tion that these firms were divisionalizing to move into fit after hav ing diversified, in both the initial and later pe ri ods. Moreover, the positive performance benefits of divisionalization found by Ar mour and Teece (1978) in the ini tial pe riod and the ab sence of per for mance ben e fits of divisionalization in the later period could be com pat i ble with con tin gency the ory. Again, the is sue re volves around fit and misfit. In Donaldson (1987), misfit overwhelmingly occurs empirically when firms have diversified and retained the functional struc ture, whereas divisionalized firms are over whelm ingly in fit (because they have di ver si fied). It is pos si ble that in the ini tial pe riod of the Armour and Teece (1978) study the functional firms were a mixture, some re main ing un di ver si fied and so in fit and some hav ing di ver si fied and so in mis fit, so that the av er age func tion ally struc tured firm was in partial misfit. In contrast, the divisionalized firms would pre dom inantly be in fit. Therefore the divisionalized firms would perform higher than the func tional firms—as was found em pir i cally (Ar mour and Teece 1978). In the later period, however, sufficient time might have elapsed for the firms that had ear lier di ver si fied into mis fit to have divisionalized and so moved into fit. Therefore there would be fewer functionally struc tured, di versified firms left in misfit, leaving the functional firms pre dom i nantly in fit. Thus the com par i son in the later

166

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

pe riod would be be tween func tion ally struc tured firms that on av er age are in fit, with divisionally struc tured firms that also on av er age are in fit. Hence, in the later pe riod, there would be no sig nif i cant dif fer ence in per for mance be tween func tional and di vi sional firms—the re sult found empirically (Ar mour and Teece 1978). Whether this in ter pre ta tion of Ar mour and Teece is valid must await a fu ture, sec ond ary anal y sis that ex am ines fit and mis fit. For the pres ent, their study can not be seen as decisive support for institutional theory against contingency theory, be cause it is open to mul ti ple in ter pre ta tions. Palmer, Jennings, and Zhou (1993, p. 120) found that large cor po rations were more likely to divisionalize if their CEO had been educated at a ma jor grad u ate school. They in ter pret this find ing in in sti tu tional theory terms as being nor ma tive isomorphism. How ever, an al ter native in ter pre ta tion of the find ing is that CEOs with busi ness ed u ca tion would have more knowl edge of the di vi sional struc ture and so be more in clined to rec og nize the sit u a tions in which their cor po ra tions needed to adopt it. Hence the find ing can be in ter preted as show ing the role of knowl edge among top man ag ers in in duc ing ra tio nal or ga ni za tional adaptation. Mahajan, Sharma, and Bettis (1988) in ves ti gated whether the adoption of the multidivisional struc ture among firms con formed to an im i ta tion pro cess and con cluded that it did not. Thomp son (1983) found that the diffusion of the multidivisional structure among firms was affected by diversification and size, that is, by contingency variables, thus sup port ing con tin gency the ory. A way that normative isomorphism could influence firms to adopt the multidivisional struc ture is through man age ment con sul tants. Yet a sur vey of or ga ni za tions that had un der gone struc tural change (pre dominantly divisionalization) found that in only 11 per cent of cases did their top management said that management consultants were the reason for the change (Hill and Pickering 1986, p. 35). This was much less than some other reasons for structural change: 50 percent for strategic change and 21 percent for performance decline (Hill and Pickering 1986, p. 35). These findings sup port the con tin gency the ory idea that struc tural change is driven by change in the stra te gic con tin gency and the resulting performance decline from the ensuing misfit; they give lit tle sup port to the nor ma tive isomorphism ex pla na tion of division alization from in sti tu tional the ory.

Chal lenges From Other The ories

167

Whittington, Mayer, and Curto (1999) stud ied large cor po ra tions in France, Germany, and the United Kingdom over the period 1950 to 1993 and conclude that the institutionalist explanations of division alization are false and the Chandlerian func tion al ist ex pla na tion sound: . . . the di vi sional form emerges as more than the prod uct of Amer i can postwar po lit i cal and eco nomic dom i na tion. The di vi sional was not just foisted on Eu rope at a time of rel a tive weak ness (Djelic, 1998); it has been steadily ab sorbed into the Eu ro pean main stream long af ter the wan ing of Amer i can he ge mony since the 1970s. . . . In dig e nous forms of or ga ni za tion—such the hold ing or the French cen tral ized func tional or ga ni za tion—have gen er ally withered, regardless of chang ing fash ions. Institutionalist skepticism has been de feated by the ro bust qual i ties of the Chandlerian multidivisional. Against institutionalist ar gu ments for the lo cal con struc tion of ef fi cient forms of or ga niz ing (Whit ley, 1994) or the fluc tu a tions of busi ness fash ions (Fligstein, 1990), the multidivisional appears to have enduring and generalizable ad van tages over its ma jor ri vals. (p. 546)

Thus we con clude that at tempts by in sti tu tional the o rists to pro mote an in ter pre ta tion of divisionalization that is contrary to, and would re place, con tin gency the ory are not at pres ent de ci sive. As ser tions that institutional effects are a cause of divisionalization founder, mainly because stud ies fail to re fute the ex pla na tion of divisionalization as a move into contingency fit (Armour and Teece 1978; Fligstein 1985; Rumelt 1974). There fore there are grounds for main tain ing the view of divisionalization as be ing ra tio nal and so cio log i cally func tional as shown empirically by contingency fit research (Donaldson 1987; Hamilton and Shergill 1992, 1993). The conflicting in terpretations seen here prompt the need for fu ture re search that would re solve these ri val in terpretations and shows which is most empirically valid, or shows a more valid the o ret i cal model of divisionalization that blends in sti tu tional and contingency theories. Any study that is seeking to refute con tin gency the ory must show that firms chang ing their struc tures are not thereby mov ing into fit with their con tin gen cies and so in creas ing their per for mance ef fec tive ness. This involves em pirically measuring fit and performance. To falsify any theory, it is necessary to use its concepts and mea sures in or der to put them to the test. How ever, the empirical ev idence to date from studies that examine fit and per formance supports rather than refutes contingency theory, sug gesting

168

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

that future studies will not in fact refute the contingency the ory of divisionalization. At the present time, it is reasonable to hold that divisionalization em pir i cally sup ports the con tin gency the ory of or gani za tional change (in its SARFIT form), rather than sup port ing in stitutional the ory. (For further cri tique of institutional theory, see Donaldson 1995b, pp. 79-128.)

Not Population-Ecology Effect An other ma jor con tem po rary school of or ga ni za tional the ory that ri vals con tin gency the ory is pop u la tion-ecol ogy (Hannan and Free man 1977, 1984, 1989). This crit i cizes con tin gency the ory for as sum ing that adaptation co mes about by an or ga ni za tion chang ing it self (Hannan and Free man 1977). In stead, population-ecology pro poses that ad ap tation of ten co mes about by changes in the mem ber ship of the pop u lation, so that mis fit ting or ga ni za tions die out and are re placed by new, better-fit ting or ga ni za tions (Hannan and Free man 1977). In par tic u lar, pop u la tion-ecol ogy holds that large or ga ni za tions are in er tial so that pop u la tion ad ap ta tion is par tic u larly strong in cre at ing change among them (Hannan and Freeman 1984). If the or ga ni za tional changes ex plained by contingency theory were really due to population ad ap tation and thus better explained by population-ecology theory, this would challenge contingency theory. We will show that the changes among large firms in their or ga ni za tional strat e gies and ad ap ta tions in their struc tures of the kind dis cussed above are mainly caused by ad aptation of ongoing organizations, as contingency theory holds, rather than by pop u la tion change. Whereas popu la tion-ecol ogy theory holds that large or ga ni za tions are in er tial (i.e., un chang ing), large cor po ra tions have changed their strategies and structures in adap tive ways, which invalidates population-ecology theory. Hannan and Freeman (1989, p. 79) list strategy among the core characteristics of an organization that are strongly sub ject to in er tia and so re sis tant to change by an or ga ni za tion. Again, Hannan and Freeman (1989, p. 81) include hierarchical cen tral iza tion to be sub ject to their the ory and this is em pir i cally linked to divisionalization (Chenhall 1979), so that divisionalization should also be sub ject to their the ory. Thus strat egy and struc ture are phe nomena to which population-ecology theory applies and, consistent with

Chal lenges From Other The ories

169

this, Hannan and Free man (1989) cri tique Chan dler’s (1962) ac count of strat egy and struc ture. Yet studies that follow the same large corporation over time show that the av er age large cor po ra tion in creased its level of stra te gic di versification and consequently altered its structure from functional to divisional (Channon 1973, 1978; Dyas and Thanheiser 1976; Fligstein 1985; Rumelt 1974; Suzuki 1980). Most of the large cor po ra tions studied by Rumelt (1974) diversified and divisionalized so that most are chang ing in these re gards rather than be ing in er tial. As we have seen, the di vi sional struc ture fits di ver si fi ca tion (Donaldson 1987; Ham il ton and Shergill 1992, 1993) and so adopt ing it af ter di ver si fy ing is adap tive. While some re search finds that di ver si fi ca tion is coun ter pro duc tive (e.g., Rumelt 1974),other re search shows di ver si fi ca tion to ben e fit or ga ni zational per for mance (Fligstein and Brantley 1992). More over, division alization moderates the effect of diversification on performance, so that diversification is adaptive when supported by divisionalization (Hill, Hitt, and Hoskisson 1992). Thus large cor po ra tions have changed their strat e gies and struc tures and this con sti tutes adap tive change by ongoing or ganizations, which con tradicts the population-ecology theory that large or ga ni za tions are in er tial and do not adapt. A population-ecologist might reply that while the large organizations that sur vived had adapted, other large cor po ra tions may not have adapted and so not sur vived, and may have been re placed in the pop u la tion by newly founded or ga ni za tions that were better adapted. How ever, there is lit tle change in mem ber ship of the pop u la tion of large corporations. Chandler (1977, p. 371) found that of the 278 largest industrials in 1917, only 14 had disbanded fifty years later. Similarly, Fligstein (1990b, p. 65) shows that among the largest 100 U.S. corporations over a sixty-year pe riod, on av er age 98 per cent re main at the end of each year. With so little change in the population of large corporations, there is little scope for adaptive organizational change through changes in pop u la tion mem ber ship. Or ga ni za tional disbandings is not a ma jor mech a nism for the elim i na tion of maladapted large corporations because few of them disband. Equally, organizational foundings is not a major mechanism for adding well-adapted large corporations to the pop u la tion be cause few of them are founded. (Of those corporations that enter the population of large corporations, some will have done so by grow ing larger, so they are not foundings, thus keeping the number of foundings smaller than 2 percent per

170

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

annum.) Fligstein (1985, p. 388, Ta ble 4) finds no sig nif i cant ef fect of comers and leavers to the top 100 U.S. corporations as causes of divisionalization over a fifty-year pe riod. There fore ex am i na tion of large cor po ra tions re futes the pop u la tion-ecol ogy claim that large or ga ni zations are in er tial and that ad ap ta tion co mes from pop u la tion change. More specifically, the changes in strategy and structure among large corporations that have been used in this book to empirically sup port con tin gency the ory are not re ally due to causal pro cesses of the kind asserted by pop u la tion-ecol ogy the ory (for more de tails, see Donaldson 1995b, pp. 73-74). In sum, population-ecology theory holds large organizations to be in er tial so that ad ap ta tion oc curs by pop u la tion change, whereas contin gency the ory sees on go ing or ga ni za tions as adapt ing, in clud ing large organizations. Divisionalization is adaptive when following divers ification, so that divisionalization by large corporations is adaptation by them and helps ren der di ver si fi ca tion adap tive also. Hence or ga niza tional ad ap ta tion oc curs even for large cor po ra tions, and has oc curred in many of them. Little of the ad aptation among large corporations could be at trib uted to pop u la tion change be cause there is lit tle pop u lation change. Thus stra te gic and struc tural change among large cor po ra tions fails to sup port pop u la tion-ecol ogy the ory and in stead sup ports contingency theory (for further critiques of population-ecology, see Donaldson 1995b, pp. 42-78, and Young 1988).

Rational Not Political Explanation The structural contingency the ory of organizations is a rational explan ation in that the or ga ni za tion adopts a new struc ture to at tain fit and performance. However, some scholars have asserted instead that or ga ni za tions are in part shaped po lit i cally, that is, by the self-in ter est of their mem bers, es pe cially the mem bers with most power, their man agers (Child 1972b). In this way, while in di vid ual man ag ers act ra tio nally, organizations are not guided by an organizational rationality. Some writers hold that managers build empires, so that as organizations grow, the ra tio of man ag ers and ad min is tra tive staff to to tal em ploy ees rises disproportionately, which unnecessarily in creases costs so that or ga ni za tional ef fec tive ness suf fers (Child 1973b; Free man and Hannan 1975; Par kin son 1957; Reimann 1979). Wil liam son (1970, 1985) ar gues

Chal lenges From Other The ories

171

that in or der to curb such man a ge rial pref er ences for ex pense, or ga nizations resort to the M-form, multidivisional structure as they grow larger, which re stores dis ci pline over man ag ers by hold ing them ac countable for profitability to the head of fice. We will ex plore each of these two ideas: that man ag ers and ad min is tra tive staff grow more than proportionately to or ga ni za tional size and that the M-form is adopted in re sponse to size growth to curb such op por tun ism. We shall see grounds for doubting both that organizations are plagued by managerial em pire build ing and that the M-form is adopted to curb such man a ge rial op por tun ism. This pro vides rea sons for re ject ing the po lit i cal ex pla nation of organizations and for reaffirming the rational, functionalist view of con tin gency the ory. Man agers and Staff Not Growing Dis pro por tion ately The po lit i cal view of organizations is seen starkly in the writ ings of the pop u lar sat i rist North cote Par kin son (1957), who as serted his fa mous law, that man ag ers mul ti ply sub or di nates not ri vals. In this way, managers are said to build empires of ad ministrators to gratify their own power and sta tus, at cost to the or ga ni za tion and its ef fec tive ness (Par kin son 1957). (His in flu en tial sat ires have them selves been sub ject to cri tique [Donaldson 1996b].) The hy poth e sis from Par kin son is that the ratio of managers and administrators to workers increases with size, but, as seen earlier (in Chap ter 3), the work of Blau (1970, 1972) and oth ers refutes this, show ing in stead that the ra tio tends to de crease with size. Empirical sci entific studies by Child (1973b) and Reimann (1979) into the re la tion ship be tween size and ad min is tra tive in ten sity (ra tio of man ag ers and ad min is tra tors to to tal em ploy ment) con clude in support of Parkinson, but this has been challenged (Donaldson 1996b). Child (1973b) in ter prets his data as be ing ev i dence of com plex cau sal ity con gruent with po lit i cal pro cesses. How ever, a more mean ing ful andconsistent in ter pre ta tion is that ad min is tra tive in ten sity is driven al most w holly by size (see Donaldson 1996b, pp. 69-70) and does not in crease with size, thus contradicting Parkinson. Similar remarks apply to the study by Reimann (1979; see Donaldson 1996b, pp. 76-80). A meta-an a lytic re view of stud ies of size and the pro por tion of man ag ers to to tal em ploy ees (Donaldson and Caufield 1989) shows that of sev en teen stud ies (to tal ing 2,218 organizations), the average cor relation was negative: –.45

172

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

(Donaldson 1996b, p. 105, Table 5.1). This supports Blau’s (1970) theory of in creas ing econ o mies of ad min is tra tion as size in creases. These types of find ings, how ever, were thrown into doubt by an influ en tial ar ti cle by Free man and Kronenfeld (1973). They ar gued against cor re lat ing size with a vari able whose de nom i na tor was also size, such as ad min is tra tive (or man a ge rial) in ten sity, as sert ing that this cre ated a problem of definitional dependency such that the observed neg ative correlations were spurious. However, the article by Freeman and Kronenfeld (1973) contains seve ral damaging er rors (see Donaldson 1996b, pp. 88-93). In par tic u lar, they claim that curves, of the sort used by Blau to prove his the ory, show ing ad min is tra tive in ten sity de creasing with size, can be produced spu riously by transforming random data into the ra tio vari able of ad min is tra tive in ten sity (Freeman and Kronenfeld 1973, pp. 110-112). In their data, how ever, ad min is tra tive intensity a nd size are not independent and, instead, ad ministrative int en sity de creases with size. There fore Free man and Kronenfeld fail to demonstrate their claim that such findings can be artifactually pro duced just by transforming one variable into a ratio (administrative intensity). What they ac tu ally show is that in data where the num ber of ad min is tra tors grows less than pro por tion ately to the num ber of workers, the ra tio of ad min is tra tors to work ers de creases as the num ber of work ers in creases (see Donaldson 1996b, p. 92). Thus there is no in herent artifactual flaw in tro duced by us ing ra tio vari ables, such as ad minis tra tors to work ers or man ag ers to to tal em ploy ees. This conclusion is supported by analyses by MacMillan and Daft (1979, 1984), who show that such supposedly flawed ratio variables and the sup pos edly sounder meth ods rec om mended by Free man and Kronenfeld (1973) yield sim i lar re sults. Thus the find ings us ing ra tio vari ables by Blau, and in the above meta-analysis, are not flawed and pro vide meth od olog i cally re li able find ings. They are em pir i cally valid tests that sup port the the ory of Blau (1970) that ad min is tra tive in tensity de clines with size. In con trast, there is very lit tle rea son to see Parkinson’s theory of managerial empire building as supported by empirical so cial sci ence (for fur ther ar gu ments, see Donaldson 1996b, pp. 58-107). The evidence shows increasing econ o mies of scale in administration as size increases, sup port ing the ra tio nal ex pla na tion of or ga ni za tional struc tures as grow ing in so cio log i cally func tional ways,

Chal lenges From Other The ories

173

rather than the political explanation of man agers serving their per sonal in ter ests at cost to the or ga ni za tion. The case for the political ex planation of or ga ni za tions against the ra tional explanation co mes also from some studies of organizational decline. These studies contrasted or ga ni za tions that were growing in size with those that were declining in size (Freeman and Hannan 1975; Marsh and Mannari 1989). Freeman and Hannan (1975) found that changes in the ad min is tra tive ra tio dif fered be tween grow ers and de cliners. This led them to crit i cize tra di tional con tin gency the ory re search as incorrectly presuming that growth and decline are symmetrical. Spe cifically, ad min is tra tive in ten sity tended to shrink dur ing growth in total em ploy ees but to in crease dur ing de cline, be cause ad min is tra tors were shed dis pro por tion ately less dur ing de cline. The in fer ence made was that man a ge rial in ter ests came into play pro tect ing man ag ers’ jobs and those of their staff. This was pre sented as ev i dence that or ga ni zations are shaped politically, by the interests of power-holders, rather than rationally and functionally (Free man and Hannan 1975). Thus, sup pos edly, func tion al ist the ory, such as Blau’s of econ o mies of scale in ad min is tra tion, held only in growth but not de cline. The theory of econ omies of scale in administration holds that administrative in ten sity de creases with size (Blau 1970). There fore admin is tra tive in ten sity de creases for grow ing or ga ni za tions and in creases for declining organizations. The theory implies that the processes of growth and de cline will be asym met ri cal, not sym met ri cal. Hence the finding of asymmetry between g rowth and decline is consistent with Blau’s the ory, not a ref u ta tion of it. Thus sev eral of the em pir i cal findings pre sented as part of a cri tique of the the ory ac tu ally con firm it. For in stance, in school sys tems, when the num ber of teach ers (i.e., operating per son nel) de creases, fewer sup port staff are cut than are added when teachers in crease by the same number (Freeman and Hannan 1975, pp. 222, 227). This means that ad min is tra tive in ten sity is decreasing for grow ers and in creas ing for de clin ers, as Blau’s (1970) the ory holds (see also discussion of Montanari and Adelman 1987, and Marsh and Mannari 1989, in Donaldson 1996b, pp. 97-104). Thus, overall, the functionalist and organizationally rational explanation of organiza tional struc ture is valid, and the po lit i cal ex pla na tion is not proven by stud ies that com pare growth and de cline. (For a crit i cal dis cus sion of

174

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

other pro po nents of the po lit i cal view of or ga ni za tions such as Roy (1990) and Smith (1978), see Donaldson 1996b, pp. 58-84). Divisionalization Not Caused by Opportunism As noted earlier, Chandler (1962) explained the change from the functional to the divisional structure by the contingency of strategy. How ever, Wil liam son (1970, 1985) has ex plained it by the con tin gency of size in re sponse to man a ge rial op por tun ism. The af fin ity be tween the the ory of man a ge rial op por tun ism of Wil liam son, and other in flu en tial organizational economics views, such as agency theory (Jensen and Meckling 1976), has helped give the op por tun ism the ory cred i bil ity, as has the stat ure of Wil liam son from his trans ac tion cost the ory (Wil liam son 1975, 1985). Ear lier we have ar gued that divisionalization is caused by the strat egy con tin gency, whereas Wil liam son is ar gu ing that divisionalization is caused by the size con tin gency. Thus we need also to ex am ine his the ory to as sess whether the strat egy con tin gency is the ma jor contin gency of divisionalization, that we have main tained thus far. Williamson’s (1970, 1985) struc ture, the M-form, is a multidivisional structure in which the general manager in charge of each division is held accountable for its profitability and is closely monitored and sanctioned by the head office managers and staff. This structural model is con sis tent with the con cept of divisionalization we have been discussing, with Wil liam son choos ing to em pha size the con trol aspects that are ger mane to his the ory. This af fin ity be tween divisionalization and the M-form stems from the fact that both de rive from the work of Chan dler (1962). Williamson (1970) argues that in large, functionally structured corporations, the man ag ers in charge of de part ments are only weakly con trolled by top man age ment and so are able to act op por tu nisti cally. Such mid dle man ag ers act to the det ri ment of cor po rate ob jec tives by fur ther ing their own self-in ter est through em pire build ing and the like. Divisionalization im poses a dis ci pline that curbs such waste by holding di vi sional heads ac count able for the prof it abil ity of their di vi sions. This is furthered by creating an elite staff at the corporate office that audits the di vi sions. For this rea son cor po ra tions change from a func tional to a divisional structure as they increase in size, according to Williamson (1970). The di vi sional struc ture con fers other ad van tages, such as superior re source allocation and strate gic decision mak ing,

Chal lenges From Other The ories

175

but these benefits might be expected to flow from divisionalization with out size in crease. Thus the ar gu ment con nect ing divisionalization with a size con tin gency seems made most firmly by the op por tun ism argument. Cross-sec tional stud ies have ex am ined whether divisionalization is related more to size or strategy, with mixed results (Chenhall 1979; Donaldson 1982a; Grinyer and Yasai-Ardekani 1981; Grinyer, YasaiArdekani, and Al-Bazzaz 1980; Khandwalla 1977). Moreover, studies examining changes over time support strategy as causing division alization (Donaldson 1987; Hamilton and Shergill 1992, 1993). Also, Fligstein (1985, p. 388, Ta ble 4) finds that strat egy is a cause of multi divisional struc ture in all five time pe ri ods, whereas size is a cause in only two out of the five pe ri ods, thereby sup port ing strat egy as a more major cause than size (however, size is in dexed by as sets that may not capture fully the effects of size on divisionalization). Palmer et al. (1987, p. 37, Table 3) find that size has no effect on adoption of the multidivisional struc ture once di ver sity is in cluded in the model. Sim ilarly, Palmer, Jennings, and Zhou (1993, p. 118, Ta ble 3) find that size has no ef fect on adop tion of the multidivisional struc ture whereas di ver sity has. Again, Mahoney (1992, p. 60, Ta ble 5) found that di ver si fication but not size causes adoption of the multidivisional structure. These find ings lend sup port to Chan dler’s the ory that division alization arises because of the problems of coordinating complex, diversified cor po ra tions ef fec tively, rather than to Wil liam son’s the ory of man a gerial opportunism. More over, other the o ret i cal ex pla na tions are avail able as to why size might cause divisionalization: lim its to the econ o mies of scale of plants (Khandwalla 1977) and limits to the cognitive ability of man ag ers to cope with the complexity of large size (Jaques 1976). Clearly these explanations for an effect of size on divisionalization do not involve divisionalization being in troduced to stem greater opportunism that results from size. Also, whether divisionalization does reduce empire building, that is, the addition of un necessary ad ministrative staff, is open to question. Grinyer and Yasai-Ardekani (1981, p. 478, Table 2) find em pir i cally that the ra tio of ad min is tra tive to pro duc tion work ers is pos i tively cor re lated with divisionalization, rather than the neg a tive correlation that would be expected from Williamson’s (1970) theory that the divisional structure disciplines managers and curbs empire building. There is scope for fu ture research to test these vari ous

176

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

com peting explanations. Nevertheless, as we have seen from the empirical evidence, it cannot be maintained that divisionalization is caused by size and not by di ver si fi ca tion, Thus the ex pla na tion of di visional struc ture by strat egy that we have in voked repeatedly is not in validated by the ar gu ment that divisionalization is shaped po lit i cally.

Not Threat-Rigidity The contingency the ory model of the organization and its managers mak ing ra tio nal ad ap ta tions dif fers also from the threat-ri gid ity theory (Staw, Sandelands, and Dutton 1981; for a dis cus sion, see Ocasio 1995). This holds that threat leads to ri gid ity, so that or ga ni za tions and their man ag ers make ste reo typed re sponses that fail to re solve the prob lem and tend to worsen the sit u a tion. While the threat-ri gid ity the ory may apply to organizations facing a threat such as a liquidity crisis (e.g., neg a tive cash flows and im pend ing bank ruptcy), this is a far worse level of per for mance than that seen as trig ger ing or ga ni za tional change in con tin gency the ory. The low per for mance that is be low the satisficing level (e.g., “15 per cent re turn on cap i tal”) is con sid er ably higher than teetering on bankruptcy (i.e., a neg a tive re turn on capital). Thus low per for mance en gen der ing cri sis and ra tio nal ad ap ta tion is dis tin guishable from abys mal per for mance that may lead to feel ings of threat and irrational, coun ter pro duc tive decisions. Studies show that low performance caused firms to adopt new strategies, such as refocusing by re duc ing di ver si fi ca tion or to be come more strat egically rational in other ways (Cibin and Grant 1996; Donaldson 1994; Grant 1993; Johnson 1987; Mintzberg and Waters 1982; Smith, Child, and Rowlinson 1990). As we have stressed, the em pir i cal stud ies of struc tural change in large cor po ra tions show clearly that their re sponse to low per for mance is ra tio nal ad ap ta tion, that is, to adopt a new struc ture that fit ted their strat egy (Donaldson 1987). For ex am ple, when the fi nan cial cri sis hit Du Pont, it adopted a di vi sional struc ture that fit ted its di ver si fied op er a tions (Chan dler 1962). Moreover, such divisionalization con sti tutes de cen tral iza tion, which is the op po site of the cen tral iza tion that is as cribed to or ga ni za tions un dergo ing threat, in threat-ri gid ity the ory, and pro vides a struc turalmeans

Chal lenges From Other The ories

177

to ri gid ity (Cameron, Kim, and Whetten, 1987; Cameron, Sutton, and Whetten, 1988; Cascio 1993; Khandwalla 1977; Whetten 1980). Thus the response by or ga ni za tions and their man ag ers to cri ses of low per for mance is rational, adap tive struc tural change, as held by con tin gency the ory, rather than ir ra tio nal, self-de feat ing ac tions, as held by threatri gid ity the ory. Whether the level of per for mance acts as a mod er a tor between ra tional ad aptation and irrational threat-rigidity should be re garded as a hy poth e sis for fu ture re search. In par tic u lar, the lev els of performance that activate the rational and the irrational responses should be iden ti fied em pir i cally. There are, of course, other views of organizational change. Meyer and Zucker (1989) write about “permanently failing organizations.” They are re fer ring to or ga ni za tions whose per for mance is so low that they would be ex pected to have dis banded, yet that per sist. The ex pla na tion of Meyer and Zucker (1989) is by pol i tics, in that vested in ter ests intervene to prevent the or ga ni za tion from tak ing the ra tio nal course of dis band ing it self be cause it is not at tain ing ad e quate per for mance. While the organization has failed against the stan dards set by usu ally dominant stakeholders (such as owners), other stakeholders with an in ter est in the con tin u ing ex is tence of the or ga ni za tion are able to en sure its con ti nu ity (Meyer and Zucker 1989). The per sis tence of an or ga ni za tion de spite di sas trous per for mance is a dif fer ent phe nom e non from that be ing dis cussed here, the ad ap ta tion of struc ture to con tin gency. Struc tural ad ap ta tion in con tin gency the ory can be thought of as be ing achieved by the dom inant control lers, the manag ers playing their tradi tional role of man ag ing the or ga ni za tion, and as oc cur ring when the per for mance of their or ga ni za tion drops be low the level ac cept able to themselves or to company directors or owners. This acceptable or satisficing level is higher than the fail ing level, for ex am ple, bank ruptcy, discussed by Meyer and Zucker (1989). We have seen that organiza tions may per sist in struc tural maladaptation while their per for mance is above the satisficing level, but this is to be distinguished from the phenomenon of permanent failure, which involves a lower level of per formance and pertains to the existence of the organization. Thus con tin gency the ory and per ma nently fail ing or ga ni za tions the o ries are dif fer ent and may com ple ment each other, rather than con tra dict ing or refuting each other.

178

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

Conclusions The or ga ni za tional ra tio nal ity and func tion al ity pos tu lated by con tingency the ory has come un der chal lenge, and irrationalist and po lit i cal explanations have been of fered in its place. How ever, these explan ations have been re but ted in this chap ter. In sti tu tional the ory has tried to show that new or ga ni za tional structures are adopted out of conformity, but the evidence is that these structures are instrumentally effective, supporting the contingency theory in ter pre ta tion. In par tic u lar, the in sti tu tional the ory con ten tion that divisionalization is a mere fash ion can not be sus tained, given that divisionalization has been shown to be a move into fit and therefore beneficial for or ga ni za tional per for mance. Population-ecology holds that large organizations are inertial, but large cor po ra tions change their strat e gies and struc tures in adap tive ways. The changes in strategy and structure discussed in this book are not pro duced by changes in the pop u la tion of large firms, as populationecology would sug gest, but rather by changes in on go ing or ga ni za tions, as con tin gency the ory states. The organizational poli tics view stresses that manag ers further their in ter ests at costs to the or ga ni za tion, such as by em pire build ing. Yet empirical stud ies of changes in ad min is tra tive in ten sity as or ga niza tions grow fail to find em pire build ing and in stead find econ o mies of scale in ad min is tra tion, that is, changes that are or ga ni za tion ally ra tio nal and functional. Studies of decline find asymmetries with growth but these are com pat i ble with econ o mies of scale and so are not ev i dence of managerial empire building. While organizational economics holds that the M-form divisional structure is adopted to curb managerial em pire build ing, other ex pla na tions are avail able that see the adop tion of the M-form as a re sponse not to pol i tics but rather to co or di nat ing complexity or to cog ni tive lim i ta tions. Threat-rigidity theory holds that poor per for mance trig gers ir rational and dysfunctional man agerial behaviors. How ever, the struc tural changes in con tin gency the ory, such as divisionalizing in re sponse to di ver si fy ing, are adap tive, not ir ra tio nal. The per for mance level that trig gers struc tural ad ap ta tion, while low, may not be as low as the poor level that may cause threat-rigidity. Thus both the contingency and threat-ri gid ity the o ries may hold, be ing ac ti vated by dif fer ent lev els of organizational per formance: contingency theory where performance

Chal lenges From Other The ories

179

drops below the satisficing level and threat-rigidity theory where performance drops far ther to a cat a strophic level. In a sim i lar vein, the theory that some organizations can be permanently failing may be complementary to con tin gency the ory, rather than re fut ing it. Thus overall, the contingency theory of organizations meets the challenge of the or ganizational theories of in stitutional, organiza tional economics, organizational politics, population-ecology, and threat-rigidity. It remains valid despite attempts to offer alternative explanations de rived from these the o ries.

THE Fit Con CON cept TIN and GENCY Anal yTHEORY sis OF OR GA NI ZA TIONS

7

Fit Concept and Analysis

F

it is the concept at the heart of contingency the ory. Con tingency analysts strive to iden tify what con sti tutes fit. They seek to show the effect of fit on performance. This has been an increasing fo cus of contingency theory research. Because fit is so central to contingency the ory and re search we need to dis cuss it in de tail. We need to ask what ex actly we mean by fit. We also need to ask how we em pir i cally es tab lish the ef fect of fit on per for mance in terms of re search de sign and em pir ical anal y sis. This and the next chapter are intended to be read together. In con trast to pre vi ous chap ters, they have a tech ni cal fla vor by con sid er ing is sues of mod el ing and meth od ol ogy. Readers seek ing a dis cus sion of contingency theory in general terms may prefer to omit them, or to come back to them after reading the other chapters. However, those read ers who want to per form con tin gency re search will find that these chap ters con tain ad vice about how to con duct as pects of such re search. Other read ers who wish to eval u ate the re search find ings in the lit erature will also find guid ance in these chap ters. We will be gin with questions about the con cept of fit and then move to meth od olog i cal issues of re search de sign and em pir i cal anal y sis. 181

182

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

Philosophical Issues in Contingency Theory Research Con tin gency The ory Is Not a Tau tol ogy Contingency theory is sometimes criticized for being a tautology, that is, for be ing true by def i ni tion and there fore be ing cir cu lar rea soning. In logic, a tautology is an an a lytic state ment mas quer ad ing as a synthetic state ment; this means a def i ni tion pre sented as if it as serts something empirical about the world, whereas the statement de fines the world to be that way (Ayer 1936). Be cause a tau tol ogy is true by defi nition it can never be fal si fied, so a tau tol ogy is not falsifiable (Bachrach 1989). The tautology crit icism of contingency theory centers on the con cept of fit. Con tin gency the ory holds that fit pro duces high per for mance. Why does fit produce the highest performance? Because, by definition, fit is the combination of con tingency and struc ture that produces high performance. At this most abstract level the fitperformance relationship is true by definition. Therefore the relationship be tween fit and per for mance is a tau tol ogy. Yet it is a mis take to dis miss con tin gency the ory for be ing a tau tol ogy, be cause sci en tific the o ries can con tain such tau tol o gies in some of their parts, while being logically and empirically valid scientific theories. Many of the most general scientific laws are true by definition. For example, Dar win ian evo lu tion ary the ory states: the sur vival of the fit test. Why are they the fit test? The an swer is: “Be cause they sur vive.” Thus the most ba sic idea or high-level gen eral law is a tau tol ogy. How ever, Dar win ian the ory over all es capes be ing tau tol o gous by giv ing the abstract the o ret i cal state ment more specific con tent. Survival properties are iden ti fied for each crea ture, for ex am ple, the long necks of giraffes help them survive because they can eat the leaves they need for food from the tops of trees. This more spe cific prop o si tion is em pir ically testable and falsifiable. It is a lower-level proposition derived from the higher-level the ory idea that the fit test crea tures sur vive. Being a tau tol ogy, the high-level gen eral law is not falsifiable, but the lower-level, derived propositions are, because they are not tautologies. Thus the theory is tested empirically by testing its lower-level propositions rather than its high-level gen eral law. Thus Dar win ian the ory needs to be as sessed as an over all the o ret i cal struc ture com posed of b asic laws and spe cific prop o si tions to see that the the ory is not a tautol ogy and is falsifiable.

Fit Con cept and Anal y sis

183

The same is true for the contingency theory of organizations. The most ab stract gen eral state ment of con tin gency the ory about fit is a tautology. However, contingency theory, like Darwinian theory, moves beyond tautology by spe cific propositions that give content to the abstract idea of fit. This is done when con tin gency the ory states what or ga ni za tional struc tures fit which con tin gency and why. For ex am ple, a di vi sional struc ture fits a di ver si fied firm, be cause a di ver si fied firm’s greater com plex ity of prod ucts and mar kets (rel a tive to an un di ver sified firm) needs more in for ma tion-pro cess ing, which is better done in autonomous divisions. Thus under the contingency the ory umbrella nests a series of specific theories that connect a contingency with a struc ture. Each com bi na tion of the vary ing lev els of the con tin gency and the struc ture is stip u lated to be ei ther a fit or a mis fit, and a rea son is given. It is these lower-level, more spe cific con tin gency prop o si tions that are nontautologous and falsifiable. They ren der con tin gency theory, as an overall theoretical structure, as not being a tautology and as falsifiable. Fruit ful ness of Re search Pro gram Not Fal si fi ca tion Al though con tin gency the ory is falsifiable, should we be attempting to falsify it? To be sound, a theory should meet the criteria of being potentially falsifiable. How ever, ad her ence to this log i cal cri te rion does not mean that con tin gency the ory re search should be seek ing to fal sify the theory. In stead, con tin gency the ory re search should be seek ing to de velop con tin gency the ory and to re veal as fully as pos si ble its po tential. Thus there is a dis tinc tion to be drawn be tween falsifiability as a log i cal cri te rion and the act of fal si fy ing a the ory. Popper (1945) holds that the o ries should be falsifiable and that sci en tific work takes the form of an en er getic at tempt to fal sify ex ist ing theories to open the door to new, superior theories—albeit avoiding pre ma ture and er ro ne ous dis missal of the o ries. The doc trine of fal sification was created to demarcate be tween science and metaphysics (Pop per 1959). The doc trine of falsifiability has in flu enced so cial and or ga ni za tional sci en tists, lead ing some to the erroneous view that sci en tific prac tice not aimed at fal si fi ca tion is im proper and un sci en tific. How ever, phi los o phers of sci ence sub se quent to Pop per have stressed that, his tor i cally, fruit ful sci ence often d e vel ops de spite it be ing fal sified. Some early falsifications of a scientific theory can be erroneous

184

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

be cause its va lid ity can be seen only af ter the cre ation of an aux il iary sci ence con sist ing of im proved meth ods and com ple men tary sci en tific theory, which is needed to reveal the truth of the basic theory (Feyerabend 1975). The de vel op ment of the aux il iary sci ence can take a long time. For in stance, early tests of the Co per ni can the ory of the so lar sys tem found it to be false (Feyerabend 1975). If as tron o mers had adhered to the doctrine of falsification it would have been dismissed, wrongly. A proper test of the Co per ni can the ory re quired the de vel opment of an aux il iary sci ence that in volved tools such as ac cu rate telescopes, and com ple men tary the ory such as New ton’s me chan ics. It took 150 years be fore valid sci en tific ev i dence con firm ing the Co per ni can the ory accumulated and be came con clu sive (Feyerabend 1975). Thus the mod ern phi los o phy of sci ence stresses that fruit ful sci ence con sists of pro grams of re search within a par a digm whereby pro po nents pur sue the basic theories and practices, acting as advocates for their theory (Feyerabend 1975; Kuhn 1970; Lakatos 1974). This in volves de vel opment of the aux il iary sci ences that are re quired to show the truth of the the ory. For con tin gency the ory—as for any or ga ni za tional the ory—the same ap plies. Contingency the ory will only be re vealed in its true light by care fully re solv ing tech ni cal prob lems and de vel op ing com ple men tary theory. This en tails a lengthy pro gram of re search over many years, to be pur sued by ad her ents of the par a digm who act as its ad vo cates. Only af ter lengthy pros e cu tion will the full power of con tin gency the ory be able to be assessed. Only then can the true value of contingency the ory be known rel a tive to other or ga ni za tional the o ries (whose ad her ents will also pur sue their pro grams). Mer ci fully, it should not take 150 years to get an accurate as sessment of contingency theory! In topics such as organizational struc ture, where it has been pur sued now for forty years (Wood ward 1958), there is al ready con sid er able ev i dence es tab lish ing its va lid ity, as we have seen. But even in the case of struc tural con tin gency theory there is need for further de velopment to reveal fully its true validity. This chapter is about pursuing the contingency theory research pro gram, in clud ing by build ing on ex ist ing work to further de velop the re quired body of aux il iary meth od ol o gies. This is not to say, how ever, that struc tural con tin gency the ory could not, or should never, be falsified in principle. If, even af ter pro longed at tempts to de velop struc tural con tin gency the ory, em pir i cal re search failed to sup port its ma jor tenet (fit of struc ture to con tin gency causes

Fit Con cept and Anal y sis

185

high per for mance), then the the ory would have been shown to be false. How ever, the ac cu mu lat ing ev i dence is much more pos i tive, as we shall see in this and the following chapter, so that future fal sification of struc tural con tin gency the ory is un likely.

The Relationship Between Fit and Performance As we have seen in pre vi ous chap ters, con tin gency the ory holds that if the or ga ni za tional struc ture fits the contingency, then higher per formance results. Conversely, if the organizational structure mis fits the con tin gency, then lower per for mance re sults. There fore or ga ni za tions move to ward fit to gain better per for mance. This ex plains or ga ni za tional change by con tin gency the ory. It also ex plains why con tin gen cies and structures are associated em pir i cally. The idea that un der lies all these explanations is that there is a fit between struc tures and con tin gency that positively affects performance. Some combinations of the con tingency and or ga ni za tional struc ture are better than oth ers for per formance. For each level of the con tin gency vari able, there is a level of the or ga ni za tional struc tural vari able that pro duces the high est perfor mance and thereby con sti tutes the fit. Thus fit is cen tral to con tin gency theory because it explains variations in organizational performance, organizational change, and associations between contingencies and structures. There has been considerable discussion in organizational stud ies about the def i ni tion of per for mance. Or ga ni za tions can be con sid ered to be seek ing their goals, so that per for mance is the de gree to which an or ga ni za tion at tains its goals, which are set by those in au thor ity over the or ga ni za tion (Par sons 1961). Alter na tively, an or ga ni za tion can be con ceived of as a sys tem, so that its per for mance is equated to var i ous as pects of sys tems func tion ing (Yuchtman and Sea shore 1967). Al ter natively again, organizations can be considered to be composed of numerous individuals and external constituencies that constitute stake holders so that each in di vid ual or stake holder group has dis tinct cri te ria and or ga ni za tional per for mance is there fore mul ti di men sional (Pfeffer and Salancik 1978; Pickle and Friedlander 1967). Each of these views of per for mance could be the ba sis on which ap pro pri ate op er a tional measures of per formance are selected in empirically studying the relationship be tween fit and per for mance.

186

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

Meyer and Gupta (1994) point out that measures of organizational per for mance may be mul ti ple and uncorrelated with each other. They ar gue that there is a pro cess whereby per for mance mea sures lose vari ation across or ga ni za tions over time, so that they are replaced by new measures, uncorrelated with the old, which restore vari a tion. To the degree that this is true, it potentially adds complexity to the per formance measurement issue. However, within a period, performance mea sures re main vi a ble so that stud ies can use them, and also the re sults of one study can be com pared with those of an other study. Across time pe ri ods, per for mance mea sures might dif fer, though con cep tu ally they might still all be considered to be measuring per formance at a more abstract level and there fore be com pa ra ble. For in stance, if or ga ni zational goals change, the new performance measures are still assessing the degree of goal attainment, so can be compared to the de gree of attainment of dif fer ent goals at an other pe riod. Given the importance of fit-performance, it has been investigated em pir i cally in con tin gency the ory re search on or ga ni za tional struc ture from the earliest studies onward (e.g., Child 1975; Donaldson 1987; Drazin and Van de Ven 1985; Hamilton and Shergill 1992, 1993; Hill, Hitt, and Hoskisson 1992; Jennings and Seaman 1994; Khandwalla 1973; Law rence and Lorsch 1967; Powell 1992; Wood ward 1965). What exactly is meant by fit operationally in contingency the ory research? There are two main op er a tional con cepts in the lit er a ture: con gru ence and in ter ac tion (Pennings 1987). Fit as Con gru ence The op er a tional con cept of fit as con gru ence holds that fit is a com bination of the levels of the contingency and structure that produce higher per for mance. Other com bi na tions are in con gru ent so that the level of the struc ture does not fit that re quired by the level of the con tin gency and hence lower per for mance re sults (i.e., lower per for mance than in fit). Pfeffer (1997, p. 158) re fers to this as pect of struc tural con tin gency the ory as the “con so nance hy poth e sis,” mean ing “‘that those or ga ni za tions that have struc tures that more closely match’ or fit ‘the re quire ments of the con text’ will be ‘more ef fec tive than those that do not’ (Pfeffer 1982, p. 148).” Sim i larly, Pennings (1987, p. 225) re fers to the match ing con cept of fit, which he de fines as: “. . . a value on a structural di men sion for each level of an en vi ron men tal di men sion which

Fit Con cept and Anal y sis

187

will max i mize ef fec tive ness.” As we have dis cussed, in con tin gency theory the con tin gency in cludes the en vi ron ment, but is a wider con cept that ex tends to “con text” more broadly, thereby en com pass ing intra organizational vari ables, such as size. Congruence(orconsonance) is seen in the ex em plar of fit-per for mance relationships—between technology and organizational structure in Wood ward’s (1965) pi o neer ing con tin gency study. The struc tural variable of the span of con trol of the first-line su per vi sor was as so ci ated with tech nol ogy in a curvilinear re la tion ship, so that it rose and then fell as tech nol ogy ad vanced. For unit and small batch the span was low, for mass production it was high, and for process production it was again low. Those firms that were at or about the mean span of con trol for their tech nol ogy cat e gory per formed higher than the firms whose spans of control deviated from the means (being ei ther lower or higher). This held for each of the three tech nol ogy cat e go ries. Thus, in this study, fit is having the mean structural value for the level of the tech nol ogy con tin gency, and mis fit is de vi a tion from the mean. This is the general idea across all studies that operationalize fit as congruence. There is some line of fit, which in Woodward (1965) is curvilinear but is usually linear. This is the line that con nects all the points of fit. For each level of the con tin gency vari able there is a level of the or ga ni za tional struc tural vari able that is the fit (i.e., yields the high est performance). Deviation from this fit line constitutes misfit and so pro duces lower per for mance. This kind of logic is seen in many con tingency anal y ses of or ga ni za tional struc ture. For ex am ple, Child (1975, p. 21, Fig ure 1) found that there was a fit whereby larger or ga ni za tions were more bureaucratically struc tured. Or ga ni za tions that lay on this fit line had higher per for mance. Those firms that lay off this fit line had lower per for mance. De vi a tion from the fit line is mis fit. This idea can be taken far ther by con cep tu al iz ing de grees of mis fit. The far ther the or ga ni za tion is away from fit, the greater is its mis fit and the lower is its re sult ing per for mance ex pected to be. The dis tance from the line at which the or ga ni za tion lies be comes its de gree of mis fit (see Drazin and Van de Ven 1985, p. 520, Figure 2[a]). The fit line may be represented empirically by the regression line of the organizational struc ture on the con tin gency vari able, among the higher-per form ing or ga ni za tions in the sam ple. The ac tual po si tion of an or ga ni za tion in mis fit is the re sid ual, that is, the dif fer ence be tween the ac tual value and that ex pected from the re gres sion. In other words, the de gree of mis fit

188

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

of an organization is the amount of discrepancy between actual and ideal (i.e., the fit) scores. This allows quantification of the de gree of mis fit. It ar tic u lates the idea of fit as con gru ence be tween struc ture and contingency, and of misfit as incongruence be tween them. Fry and Smith (1987) sug gest that fit or con gru ence has many pos sible re la tion ships with per for mance. They hold that the tra di tional view of the re la tion ship be tween con gru ence and per for mance as pos i tive is only one pos si bil ity. They state that con gru ence could have a neg a tive or curvilinear relationship with per for mance. How ever, if more con gru ence leads to less per for mance, then it is not congru ence or fit, be cause by definition congruence or fit leads to higher performance. Thus the relationship between congruence or fit and performance is always pos i tive. Again, Fry and Smith (1987) state that con gru ence can have pos i tive ef fects on some per for mance out comes but neg a tive ef fects on oth ers. But if a com bi na tion of or ga ni za tional struc tures and sit u a tional fac tors has a neg a tive ef fect on per for mance, then that com bi na tion is not a con gruence or fit, instead it is an incongruence or misfit. Thus while the com bi na tion may be a con gru ence or fit for some per for mance outcomes, it is an in con gru ence or mis fit for other con gru ence out comes. It would be clearer to say that a com bi na tion is con gru ent for those perfor mance out comes on which its ef fect is pos i tive and in con gru ent on those performance outcomes on which its effect is negative. Thus in spec i fy ing whether a com bi na tion is con gru ent or in con gru ent, there is a need to spec ify which per for mance out comes are pos i tively af fected. There may be some other per for mance out comes on which the ef fect is neg a tive. In this way we pre serve the mean ing of a con gru ent or fit state as be ing one that al ways raises per for mance. Fur ther, Fry and Smith (1987) hold that con gru ence can have pos itive ef fects in the short run but neg a tive in the long run. By def i ni tion, how ever, a con gru ence or fit al ways pro duces pos i tive out comes. Once again, it would be clearer to say that the com bi na tion that is con gru ent for the short run is incongruent for the long run and vice versa. This discussion brings out that congruence of fit is always em bedded in a causal the ory. That the ory needs to be made ex plicit when say ing that one combination of contingency and organizational structure is congruent or fitting. Thus a con gru ence or fit is with re spect to some particular out come over some time pe riod. For an other per for mance outcome or time period that combination may be incongruent. The

Fit Con cept and Anal y sis

189

discussion here is of log i cal pos si bil i ties, in the spirit of Fry and Smith (1987). Whether a con gru ence that raised per for mance would re duce some other as pect of per for mance, or whether con gru ence dif fers for short- and long-run per for mances, would need to be es tab lished em pir i cally. The log i cal pos si bil ity of such dif fer ences does not guar an tee that they ac tu ally ex ist. Fit Is Not In ter ac tion There is a sec ond op er a tional def i ni tion of the fit con cept in the literature (Pennings 1987). This is fit as an interaction between the contingency and the or ga ni za tional struc tural vari able. Spe cifically, fit is measured by a multiplicative in teraction term, that is, the con tingency variable multiplied by the organizational structure variable (Schoonhoven 1981). As Pennings (1987, p. 225) states: “it is as sumed that ef fec tive ness is high when high lev els of both en vi ron men tal . . . and struc tural . . . di men sions are pres ent, but that it is low when ei ther di men sion is low or ab sent.” How ever, the rea son that such an in ter action term is con sid ered to be an operationalization of the fit con cept is unclear. Fit combines the contingency and organizational structure vari ables. In ter ac tion terms are cus tom arily used for com bi na tions of variables in multiple regression. Theref ore the interaction term has been applied to test the fit concept in multiple regression analysis. How ever, this pre sumes that a fit is a multi pli ca tive in ter ac tion (i.e., the con tin gency and struc ture vari ables are mul ti plied to gether to give an interaction term), and this is not necessarily so. Cer tainly some ar gu ment is required to justify the assumption that fit is a multiplicative interaction. Conversely, there are very many problems in operation alizing fit as an in ter ac tion term, and these will be dis cussed at var i ous points in this chap ter. Fit is not a multi pli ca tive in ter ac tion in the way that the fit concept has been used in the seminal contingency theory studies. As seen above, in the seminal stud ies fit is congruence. For Wood ward (1965) fit is con gru ence: High per for mance oc curs when the structural variable matches the technological contingency. Sim ilarly, for Child (1975), fit is congruence: High performance occurs when the struc tural vari able matches the size con tin gency. Moreover, a multiplicative interaction term does not capture the relationship between con gruence (fit) and performance. Consider a model in which the struc tural vari able, S, is shown on the ver ti cal axis

190

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

and the contingency variable, C, is shown on the horizontal axis (see Figure 7.1). The structural variable, S, needs to fit the contingency variable, C, for high per for mance to re sult. Sup pose, for sim plic ity, that the fit is that S needs to equal C. There fore the fit line is at 45 de grees to both the vertical and horizontal axes. When C is 2, S must be also of value 2 for it to be in fit. When C is 2, if S is 3, S is too big to fit so it is actually a mis fit, so lower performance would result. However, the multi pli ca tive in ter ac tion term would say that C of 2 with S of 3 equals per for mance of 6 (= 2 x 3) and so is better than C of 2 with S of 2 which equals per for mance of 4 (= 2 x 2). Thus the multi pli ca tive in ter ac tion attributes higher per for mance to the mis fit than to the fit. Clearly, the multiplicative term fails to cal i brate the ef fect of mis fit on per for mance. The es sence of a good operationalization of fit is that or ga ni za tions in mis fit must score lower than those in fit. The multi pli ca tive interaction term does not do this and so is not a mea sure of fit. Another issue is that the multiplicative in teraction term assumes that the fit is a straight line. How ever, some of the fits are not lin ear, but are curvilinear. For example, Woodward (1965) produced ev idence that some of the fits be tween tech nol ogy and struc ture were curvi linear. As we have seen, the fit of span of con trol of the fore man to tech nol ogy was that the span first increased but then decreased as technology advanced, so that it was an in verted U-shaped curve (Wood ward 1965). A multi pli ca tive term as sumes that the fit of span to tech nol ogy would be a higher span at the more ad vanced tech nol ogy than at the me dium tech nol ogy. The multi pli ca tive term would there fore award a high score to a firm that had a large span and had the most ad vanced tech nol ogy, even though that firm was in mis fit. Thus the multi pli ca tive interaction term fails to operationalize curvilinear fits. Van de Ven and Drazin (1985) state that misfit is appro pri ately measured by a deviation score and advocate the Euclidean dis tance for mula. However, they dis cuss this un der the head ing of in ter ac tion con cepts of fit. This choice of lan guage seems un help ful, given that, as Van de Ven and Drazin (1985) ar gue, in ter ac tion terms fail to cap ture the de vi a tion-from-fit con cept. T here fore it seems pref er a ble to use the word congruence rather than interaction to con cep tu al ize fit. Thus the multi pli ca tive in ter ac tion term is not a cor rect operationalization of the fit con struct. The un der ly ing model of the re la tion ship among the con tin gency, the or ga ni za tional struc ture, and per for mance in the multiplicative in ter ac tion term is not that pos tu lated by the fit

Fit Con cept and Anal y sis

191

Figure 7.1. Per for mance From Fit and Mis fit: Fit as a Multi pli ca tive In ter ac tion of Con tin gency and Struc ture

idea. An in ter ac tion term tells us some thing that may be of in ter est but con cep tu ally it does not have the mean ing of fit that has ex isted in con tingency theory re search. This meaning of fit is as con gru ence and so re search ers wish ing to mea sure the de gree of fit should use mea sures of how close the or ga ni za tion is to the fit line. However, a study by Powell (1992, pp. 127-128) finds sim i lar em pir ical re sults from us ing the two dif fer ent operationalizations of fit: con gru ence and in ter ac tion. Thus the choice be tween the two meth ods

192

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

may sometimes not affect the results. Nevertheless, the two methods are not the same conceptually and so there can be no guarantee that they will al ways pro duce sim i lar re sults, so re search wish ing to study fit in the sense that has been meant in con tin gency the ory re search should use the con gru ence mea sures, not the in ter ac tion term. The Fit Line as Iso-Per for mance In an im por tant ar ti cle that has helped de fine think ing about fit in con tin gency the ory re search, Van de Ven and Drazin (1985) state, inter alia, that the fit line is a line of iso-performance. Iso-performance means equal per for mance, so that each point on the line causes per for mance equal to ev ery other point. This is seen for ex am ple in Wood ward’s (1965) study of the fit of span of control to technology. Those firms lying at fits were the high est per form ers in the whole study even though there were three different fits, one for each technology category. For instance, Firm A would be in fit with its mass production technology and Firm B in fit with its pro cess pro duc tion tech nol ogy. De spite be ing in dif fer ent fits, the two firms have equal per for mance rat ings, that is, both are in the highest per forming sub group of all the firms in the study. Thus per for mance was in creased by be ing on the fit line, but there is no in crease in per formance from moving along the fit line. Iso-performance means that the fit to low levels of the contingency vari able pro duces the same per for mance as the fit to high lev els of the contingency variable. Hence the iso-per for mance con cept ar tic u lates some thing very ba sic about con tin gency the ory re search: All fits are equally good. For each value of the con tin gency vari ables there is a value of the or ga ni za tional structure that is a fit that produces the highest performance for that value of the con tin gency. This, as we have seen, is the fun da men tal fit, or con gru ence, idea of con tin gency the ory. But the em pir i cal re search from the pi o neer ing study of Wood ward (1965) on ward also con tains the idea that the high performance of one fit is the same as the high performance of ev ery other fit be tween that or ga ni za tional struc tural and that contingency variable. Thus the high performance for one value of the or ga ni za tional struc tural vari able that is a fit is the same as that produced for any other value of that organizational structural vari able that is a fit. Again, the high per for mance of a fit for one value of

Fit Con cept and Anal y sis

193

the con tin gency vari able is the same as that pro duced by the fits for any other value of that con tin gency vari able. The im pli ca tion is that each point of fit on the line of fit of bu reaucratic structuring to size produces equal performance to every other point. Sim i larly, each point of fit on the line of organicness to task un cer tainty pro duces equal per for mance to ev ery other point. How ever, the performance ef fects of fit be tween one as pect of struc ture and its con tin gency can dif fer from that of an other as pect of struc ture and its con tin gency. For ex am ple, the ef fect on per for mance of the fit be tween size and bu reau cratic struc tur ing might be dif fer ent from the ef fect on performance of the fit between task uncertainty and organicness. Whether these two ef fects are sim i lar or dif fer ent is a mat ter for em pir ical re search. Iso-per for mance means that fit ef fects on per for mance are the same within one struc tural as pect and its con tin gency rather than be tweenstruc tural as pects and their con tin gen cies. An other prob lem with fit as a multi pli ca tive in ter ac tion term is that it is in com pat i ble with fit as iso-per for mance. Re verting to our ear lier ex am ple (Fig ure 7.1), if the fit line is de fined as S equals C, then when S and C are both 2 then there is a fit, and when S and C are both 3 then there is an other fit. Ac cord ing to iso-per for mance, both these fits would pro duce the same per for mance. How ever, us ing a multiplicative in ter ac tion term, S mul ti plied by C means that the first fit is 4 (= 2 x 2) and the second fit is 9 (= 3 x 3); there fore the per for mance of the sec ond fit would be much greater than that of the first fit. Thus multi pli ca tive in ter ac tion terms hold that per for mance is in creas ing as the or ga ni zation goes along the fit line, which is not a line of iso-per for mance. This is an other rea son to avoid the multi pli ca tive in ter ac tion term in seek ing to operationalize the tra di tional con tin gency the ory re search con cept of fit and show ing its per for mance ef fect. Prob lems With Equifinality Some schol ars ar gue that there is “equifinality,” by which they mean sev eral dif fer ent ways to ob tain the same out come (Gresov and Drazin 1997). Doty and Glick (1994), writing as configurationalists, use “equifinality” to mean that each con fig u ra tion or type is equally ef fective. This of course is contrary to contingency logic, which holds that the type that fits its sit u a tion is more ef fec tive than the other types that

194

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

are mis fits. We have al ready of fered a crit i cal ob ser va tion about equifinality in configurationalism (in Chap ter 5). In discussions of contingency the ory, equifinality may be used to mean that there are sev eral dif fer ent fits that are all equally ef fec tive in the same sit u a tion. Thus, for a given level of the con tin gency fac tor, an or ga ni za tion could have any one of sev eral widely dif fer ing struc tures and still be in fit and have high per for mance. This goes against the con tingency logic that for a particular level of the contingency variable there is one level of the or ga ni za tional struc ture vari able that pro duces the high est per for mance and thereby is the fit (Pennings 1987, p. 226). Where different or ganizational structures have been shown to be equally ef fec tive in the same in dus try, the firms can be seen to dif fer in their con tin gency so that each of the dif fer ent struc tures is the fit for its sit u a tion. For ex am ple, Child (1977, 1984) has ar gued that, in the same in dus try, firms that had func tional struc tures had equally high per for mance as firms with di vi sional struc tures. How ever, Donaldson (1985, pp. 148-151) has pointed out dif fer ences be tween the firms and ar gued that these were dif fer ences in con tin gen cies that ex plain the dif fer ent struc tures so that each firm had to adopt a par tic u lar struc ture to fit its contingencies, so that there was no equifinality. Equifinality could also be taken to mean that the fit is not a line but, in stead, a broad band, so that for a given level of the con tin gency vari able, the fit is a range of val ues of the or ga ni za tional struc ture. How ever, the ex is tence of such broad bands of fit has yet to be shown em pir i cally. Empiricalanalyses show fit to be a line (e.g., Child 1975, p. 21, Fig ure 1). Thus the equifinality no tion is not sup ported to date. Rather than equifinality being a range of fits for each level of the contin gency, some au thors use equifinality to mean that the fit to each level of the con tin gency pro duces the same (or sim i lar) per for mance. In this mean ing of equifinality it be comes the same as iso-per for mance, so that showing iso-perf ormance establishes equifinality. However, equi finality car ries con no ta tions of choice, in that if there is more than one way to achieve high per for mance, then an or ga ni za tion can choose between them. The fact that there are nu mer ous po si tions that pro duce high per for mance, one at each level of the con tin gency, does not mean that an or ga ni za tion can freely choose among them. An or ga ni za tion is con fined to the fit to its level of the con tin gency vari able, or to lev els of the contingency vari able that it can feasibly attain. For example, an or ganization has a cer tain size, so it is re stricted to the fit to its level of

Fit Con cept and Anal y sis

195

the size con tin gency, or to fits to ad ja cent sizes that it could feasibly at tain. Therefore to equate equifinality with iso-performance is potentially confusing be cause equifinality tends to be used to mean that there is choice, whereas iso-per for mance can ex ist with out choice. Jennings and Seaman (1994) found two combinations of strategy and struc ture that pro duced high per for mance in their em pir i cal study: de fender strat egy with mech a nis tic struc ture and pros pec tor strat egy with organic structure. They int erpret these findings as support for equifinality, in that the per for mance of the two groups in fit was sim i lar even though each had different combinations of strat egy and struc ture. However, this in ter pre ta tion as sumes that each or gani za tion was free to choose be tween the prospector and de fender strategies. The pos si bil ity of any sit u a tional con straint was not ex am ined, such as by in ves ti gat ing whether the or ga ni za tions fol low ing one strat egy differed in their sit u a tion from those fol low ing the other strat egy. In so far as strat e gies such as be ing a low-cost de fender hang on con tin gen cies such as size that give scale ad van tages (Por ter 1980), then these strat egies are not free choices for an or ga ni za tion. Thus where a study dem on strates iso-per for mance, such as that by Jennings and Sea man (1994), we should be cautious about accepting that there is also equifinality un til the lack of de ter mi nants has been dem on strated. Man a ge rial De ci sion Not Se lec tion Fit Van de Ven and Drazin (1985) dis tin guish three con cepts of fit: se lection, interaction, and systems. They argue that confusion and prob lems in the con tin gency the ory lit er a ture can be re solved by fol low ing their schema. Van de Ven and Drazin dis cuss or ga ni za tional struc tures but say that they are ap pli ca ble to other or ga ni za tional char ac ter is tics. There is much to be welcomed in the concepts of Van de Ven and Drazin, but we shall argue that their three concepts of selection, interaction, and sys tems would be better reconceptualized as man a gerial de ci sion, con gru ence, and multifit, re spec tively. We have dis cussed interaction fit above, ar gu ing that it is better termed congruence , and will discuss systems fit in the next section. Here we will discuss the concept of se lec tion fit. Van de Ven and Drazin (1985) use the word selection to de scribe the pro cess whereby or ga ni za tion struc ture co mes to be cor re lated with the contingencies. Selection consists of natural selection through which

196

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

misfitted organizations are culled and managerial selection through which man ag ers make the de ci sions about struc ture. Drazin and Van de Ven (1985) sup port their con cep tual arguments with an empirical anal y sis. They show that there are cor re la tions be tween sev eral struc ture variables and the contingency, which they see as supporting natural selection. Further, they show that the correlations are greater for aspects of structure and process under man age ment con trol, thereby sup port ing the con cept of man age ment selection. Nat u ral se lec tion would explain fit among firms by say ing that the mis fit ted firms failed to sur vive. How ever, the units of anal y sis within the em pir i cal study of Drazin and Van de Ven (1985) are units within a public-sector organization, so that a misfitted unit would not go bankrupt and disband, hence mis fit and low per for mance would not nec es sar ily lead to lack of sur vival. There fore the cor re la tions be tween struc ture and con tin gency in their em pir i cal anal y sis would not seem to be brought about by natu ral selection, though their conceptual point may ap ply to free-stand ing or ga ni za tions, such as firms. Thus for some organizations, none of the correlations between their structure and contingency variables are caused by natural selection and all are brought about by managerial selection. However, where referring to man a ge rial de ci sions over struc ture it might be pref er a ble not to use the word selection, which, be cause of the in flu e nce of pop u la tion-ecol ogy, has come to have strong connotations of natural se lec tion in mod ern or ga ni za tional the ory. It may be better to use the term man a ge rial de cision to distinguish it from natu ral selection. This helps better recognize that al most all or ga ni za tions are shaped to some de gree by the decisions their managers make, without natural se lection having nec es sar ily played a part.

Com bining Con tin gencies and Fits The issue arises of how to com bine more than one con tin gency to deter mine the fit of an or ga ni za tional struc ture. The is sue also arises of how to combine multiple fits. In our view all of these combinations should be seen as fol low ing a sim ple pro cess of ad di tion (or sub trac tion).

Fit Con cept and Anal y sis

197

Mul ti ple Con tin gencies There may be more than one contingency for an organizational struc tural vari able. The ef fect of one con tin gency fac tor is added to the ef fect of the sec ond fac tor to de ter mine the fit. For ex am ple, large size requires a high de gree of bu reau cracy, and if the task is rou tine, then this also re quires bu reau cracy, so that the fit is even more bu reau cratic than it would be for just large size or rou tine on their own. Some times con tin gen cies may make op pos ing re quire ments for the or ga ni za tional struc tures needed to fit each of them. In such a sit u a tion, the contingencies have “con flict ing im pli ca tions” (Child 1972b, p. 16) for struc ture, and they are termed “con tra dic tory con tin gency fac tors” by Mintzberg (1979, p. 474). It is some times felt that the contradictory im pli ca tions of each con tin gency leads to a zone of choice (Child 1972b, p.16). Yet the logic of contingency theory is that if one contingency specifies a high level of the organizational structure as the fit and a second contingency specifies a low level of the same organizational struc ture as the fit, then the fit is a me dium level of the or ga ni za tional structure. Where contingencies make conflicting prescriptions, the lower level of a struc tural vari able that one pre scribes is sub tracted from the higher level prescribed by the other. This means mathematically that the ef fect of each con tin gency in de ter min ing the fit of an or ga ni zational struc ture that is sub ject to more than one con tin gency is ad di tive. Hence the fit will be sin gu lar and so no zone of choice ex ists. Con sider the case where an or ga ni za tion is large and yet in an un stable en vi ron ment that man dates the or ga ni za tion to in no vate. The large size re quires a bu reau cratic struc ture while the in no va tion re quires an or ganic struc ture, thereby cre at ing an ap par ent con flict. How ever, the large size is dealt with by a bu reau cratic macro-struc ture, such that the over all or ga ni za tion has many hi er ar chi cal lev els, many de part ments, many administrative specialists, and many administrative rules and pro ce dures. The re quire ment for in no va tion is dealt with by adopt ing or ganic el e ments in some parts of the mi cro-struc ture, that is, in those parts that in no vate, for ex am ple in R&D and in pro ject teams, leav ing other parts quite mech a nis tic, for ex am ple, the pro duc tion de part ment (Lawrence and Lorsch 1967). Thus the op pos ing ten den cies can co ex ist be cause some parts of the or ga ni za tion re flect the logic of large sizeand

198

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

bureaucracy while other parts reflect the logic of innovation and organic ness (as shown in Chap ters 2 and 3). More spe cif i cally, large size re quires high for mal iza tion, while in no va tion re quires low for mal ization, so that the pre scrip tions of these two con tin gen cies for for mal ization are in con flict. The over all level of for mal iza tion in the or ga ni zation re flects the ef fects of both large size, which is push ing to raise it, and in no va tion that is push ing to re duce it. The ad min is tra tive macrostruc ture has high for mal iza tion, while within this, de part ments that deal in innovation (e.g., R&D) have low for mal iza tion. The re sult ing, overall level of formalization is less than it would be for a large but noninnovative or ga ni za tion and higher than it would be for a small, innovative organization. Hence the lower level of formalization pre scribed by in no va tion is sub tracted from the higher level of for mal ization pre scribed by large size. Thus the level of for mal iza tion re flects the ad di tive ef fects of in no va tion and size. It is a sin gle value rather than a zone of sev eral fits that cre ate choice. Gresov (1989) found an empirical effect of the conflicting con tingen cies idea. He ex am ined the fit of struc ture to both the task un cer tainty and horizontal de pendence contingencies simultaneously. He found that fit and per for mance were re lated where both con tin gen cies re quired similar structures, but that where their requirements conflicted, the relationship broke down and units had lower per for mance, sup port ing the ar gu ment that con flict ing con tin gen cies lead to mis fit. Con flicting con tin gen cies may in crease the prob a bil ity of er ro ne ous man age ment choice and hence of mis fit, rather than sig ni fy ing equi finality and a range of equally ef fec tive struc tural choices. There is scope for rep li ca tion of these results and further empirical examination of the conflicting contingencies is sue.

Com bining Mul ti ple Fits As we have seen, within the contingency the ory of organizational struc ture there are a num ber of dif fer ent fits be tween var i ous as pects of organizational structure and their contingencies. The question arises of how to com bine more than one fit to as sess their to tal ef fect on or ga ni za tional per for mance. Some an a lysts have sug gested that the fits are additive, in that the first fit is added to the second fit to yield the over all ef fect on or ga ni za tional per for mance (Randolph and Dess 1984,

Fit Con cept and Anal y sis

199

Exhibit 2, p. 123). In our view, this additive model of the effects of multi ple fits is sound. Van de Ven and Drazin (1985), how ever, ar gue that the over all ef fect of mul ti ple fits is not the sum of their in di vid ual ef fects and so is not additive. They ad vance a model of the com bi na tion of mul ti ple fits as be ing a sys tems fit. In a sys tems fit, the ef fect of mul ti ple fits on or ga niza tional per for mance is not just the sum of the ef fect of each fit on perfor mance. In stead, there is some ho lis tic prop erty that is not cap tured by an atomistic anal y sis of each fit sep a rately that then just com bines them to gether (Drazin and Van de Ven 1985). Thus, they ar gue, the effect of mul ti ple fits can not be cal cu lated by sim ply add ing up the effect of each fit on per for mance. Drazin and Van de Ven (1985) ex am ine sep a rately the pairwise fit of each struc ture and pro cess vari able to the con tin gency vari able and its relationship with per for mance (mea sured by ef fi ciency and sat is fac tion). They show that each fit attains only very limited support—only four out of twenty-two are sig nif i cant and one of these has a sign con trary to the the ory. They then show that the sys tems fitis sig nif i cantly cor related with performance (both efficiency and satisfaction) in the way theoretically expected. While this is evidence that multiple fits are stron ger than pairwise fits, this is no proof of sys temic prop er ties of a ho lis tic kind whereby the whole is more than the sum of its parts. The Eu clid ean dis tance for mula used by Drazin and Van de Ven to cal cu late sys tems fit sums the ef fects on per for mance of each of the pairwise fits. It is an ad di tive model. Thus their sys tems fit mea sure is noth ing more than the sum of its parts. There fore the greater strength of the sys tems fit than each of the pairwise fits is caused by the sum ma tion of many small ef fects. Fur ther more, pairwise fits suf fer un re li abil ity of measurement that reduces their observed effect on performance. Summing across the pairwise fits produces an ad ditive in dex that will be less unreliable and so pro duce a truer cor re la tion. Hence the su pe ri or ity of the multiple fit over the pairwise fits is compatible with it being the sum of the parts and is not ev i dence of ho lis tic prop er ties. Drazin and Van de Ven (1985) also ar gue, in part, by an il lus tra tion. This is a cu ri ous il lus tra tion be cause in it the ef fect of the mul ti ple fits on over all or ga ni za tional per for mance is cal cu lated by sim ply add ing up the ef fects of each fit on per for mance. Thus in the il lus tra tion the whole is simply the sum of its parts. It therefore illustrates the point against which the ar gu ment is be ing made. In their il lus tra tion (Drazin

200

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

and Van de Ven 1985), an in di vid ual fit is of ten not as so ci ated with to tal per for mance of the or ga ni za tion. How ever, this is be cause most of the fits are neg a tively cor re lated with each other so that most of them have zero cor re la tion with to tal per for mance. Hence each fit af fects per formance, but most of the fits have no cor re la tion with to tal per for mance be cause of the sup press ing ef fect of the other fits. Thus the lack of as so ci a tion be tween many of the in di vid ual fits and to tal per for mance in this il lus tra tion is no more than con founded cor re la tion. Hence, the seemingly par adoxical result is explained through consid er ing the correlations between each of the variables, that is, reductionism and not holism. Thus we can see that the sys tems fit model re ally boils down to the ad di tive model. We sug gest that re search ers in con tin gency the ory use ad di tive mod els to an a lyze the ef fects of mul ti ple fits on per for mance. This ad di tive model ap plies both to com bin ing mul ti ple fits of an or ga nizationalstruc tural vari able to var i ous con tin gen cies and to com bin ing mul ti ple fits of var i ous or ga ni za tional struc tural vari ables to a con tingency. It also applies to more encompassing analyses that combine multiple fits of various organizational struc tural vari ables to var i ous con tin gen cies. Thus re search ers should use the con cept of mul ti ple fits, that is, multifits.

The Identification of the Fit Line So far we have talked about the fit line, but how do we know where it is? Con cep tually it is the line of points of fit, that is, the se ries of val ues of the or ga ni za tional struc tural vari able each of which con sti tutes the fit for a par tic u lar value of the con tin gency. But how do we iden tify what is the value of the or ga ni za tional struc tural vari able that fits each value of the con tin gency? Two ways of an swer ing this ques tion may be dis tin guished: the the o ret i cal approach and the em pir i cal ap proach. The the o ret i cal approachis that for a par tic u lar or ga ni za tionalstruc ture a con tin gency fac tor is iden ti fied through the o ret i cal anal y sis. The spe cific val ues of the or ga ni za tional struc ture that fits each different value of the con tin gency are then iden ti fied by think ing through the logic of the ar gu ment. A ma trix of all the pos si ble com bi na tions ofthe val ues of the con tin gency and or ga ni za tional struc ture is con structed. Then each com bi na tion is iden ti fied as ei ther a fit or a mis fit. One can then in spect

Fit Con cept and Anal y sis

201

the pat tern of cells in the ma trix that are fits and see whether it forms a straight line or a curve or some other non lin ear pat tern. Hav ing gen er ated a the o ret i cal model on an a priori ba sis, we can test its va lid ity em pir i cally. This in volves ex am in ing whether the com bina tions designated as be ing fits actually perform higher than those designated as being misfits. The first step may be to calculate the performances of all the cells that are fits and com pare this sub to tal with the per for mances of all the cells that are mis fits (e.g., Donaldson 1987; Hamilton and Shergill 1992). This ag gre gate anal y sis has the disad vantage that some cells la beled as fits may ac tu ally be mis fits and vice versa; the ac tual per for mance of the in di vid ual cell is not known, so it may be ob scured by the ag gre ga tion. How ever, it has the ad van tages of larger numbers, whereas the number of cases in the individual cells may be too small to be reliable. The aggregate method allows a quick test of the overall model and if positive encourages fur ther re search, whereas a neg a tive re sult is an im por tant sig nal that may lead to a re assess ment of the fit model or, in deed, of the the ory. The sec ond step is to cal cu late the per for mances of each cell and as cer tain that each of those designated as fit actually has a performance higher than those desig nated as a misfit (and vice versa for the mis fit cells). A neg a tive re sult for a cell may in di cate that it is ac tu ally a mis fit rather than a fit (or vice versa) and lead to a reassessment of the theory for that particular combination of con tin gency and or ga ni za tion struc ture. Con versely, in stead of start ing with a the ory, the em pir i cal ap proach starts with the data and then seeks to find a pat tern, which is then in terpreted theoretically. The analysis examines var ious com binations of the lev els of the con tin gency and struc tures to see which pro duce highest performance. The high-performing combinations are the fits and the low-per form ing are the mis fits. A the ory would then be de vel oped of why cer tain com bi na tions are fits that are ef fec tive and oth ers mis fits that are in ef fec tive. In a com pletely em pir i cal ap proach the con tin gency fac tor it self may be discovered empirically. Initial attempts fail to find an association be tween an or ga ni za tional struc ture and or ga ni za tional per for mance. No main ef fect hav ing been found, the search is then for mod er a tors of the relationship between the structure and performance. A frequent move is then to look for vari ables that are as so ci ated with the struc ture. Such an as so ci a tion would be ex pected if the vari ables were mod er a tors of the effect of the structure on performance, that is, contingencies.

202

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

The rea son for the as so ci a tion is that or ga ni za tions would tend to move into fit to gain the resulting higher performance. Therefore variables found to be as so ci ated with the struc ture are then sub ject to an anal y sis to see whether they are in fact contingencies, by seeing whether these vari ables mod er ate the re la tion ship be tween the or ga ni za tional structure and per for mance. What we have just de scribed is an em pir i cally based pro ce dure (i.e., an emer gent or data-driven pro ce dure). Some would cas ti gate it as “a fish ing trip,” but there is some role for it. Con tin gency the ory re search on or ga ni za tional struc ture has quite of ten gone from ini tially seek ing as so ci a tions be tween con tin gency and struc ture to then ex am in ing the effects of combinations of contingency and structural variables on performance. This is seen in the re search on tech nol ogy and or ga ni zational struc ture (Wood ward 1965) and on size and bu reau cracy (Child 1975; Pugh et al. 1969). The two dif fer ent ap proaches for iden ti fy ing the fit line, the the orydriven and em pir i cal-driven, are ex tremes. Less ex treme than ei ther of these two research procedures might be the use of some mixture of theory-driven and em pir i cally driven ap proaches.

Validation of the Fit Model Once a model of what constitutes fits and misfits between structure and contingency has been created, it has to be validated empirically. This raises two methodological is sues: causal in fer ence and mea surement of the re la tion ship be tween fit and per for mance. Contingency the ory holds that the fit of struc ture to the con tin gency pos i tively af fects per for mance. But how can we be sure that fit is a cause of per for mance? Given that a cause pre cedes an ef fect, fit needs to be ass essed at one point in time and per for mance at a subsequent time point. There is no theory that spec i fies the time lag, but a two-year pe riod be tween fit and per for mance pro duces a pos i tive ef fect of fit on performance (Donaldson 1987; Hoffman, Carter, and Cullen 1994). Thus we can have more con fi dence in mak ing the causal in fer ence that fit is a cause of per for mance—and that per for mance is an ef fect of fit— where fit is mea sured at a time prior to the point at which per for mance is mea sured. Thus a diachronic re search de sign is pref er a ble. How ever,

Fit Con cept and Anal y sis

203

some research designs are cross-sectional (i.e., synchronic), so that they study fit and performance at the same time. Indeed, being more ex act, while some such stud ies col lect the struc tural and per for mance data at the same time, the performance is for a time prior to that of struc ture (Dal ton, Daily, John son, and Ellstrand 1999, p. 680), such as an av er age of several years of profitability or sales growth prior to the date for which struc ture is mea sured (e.g., Child 1974). In such cases, strictly speaking, an observed association is evidence f or an effect of per for mance on fit, so that per for mance is a cause of fit. Such stud ies are conven tion ally in ter preted, how ever, as in ves ti ga tions of the ef fects of fit on performance. Nevertheless, causal inference that fit affects performance is more certain where fit pre cedes per for mance, so that re search de signs should in cor po rate this fea ture. (The is sue of whether per for mance af fects fit will be dis cussed more fully be low.) Other causes of performance may confound the effect of fit on performance and these need to be controlled in the research design. Because in con tin gency the ory the de pend ent vari able is per for mance, which is ob vi ously af fected by many vari ables other than con tin gency fit, there is am ple scope for such con found ing vari ables. The com bined effects of the causes of performance other than fit may be great and may be greater than the fit, so rendering confounding more feasible. Other causes of per for mance only be come con founds, how ever, if they are cor re lated with fit. Some times such a cor re la tion arises be cause the confound and the con tin gency fit vari ables are causally con nected or have a systematic association, other times a correlation will exist by chance in a sam ple of or ga ni za tions. The appropriate procedure is to seek to control for such possi bly con found ing ef fects of other causes of per for mance. This may be achieved by hold ing con stant other fac tors by study ing only or ga ni za tions that are in the same cat e gory, for ex am ple, the same in dus try (Dess, Ire land, and Hitt 1990). Alternatively, it may be achieved by using a control group of sim i lar or ga ni za tions that dif fer only in their con tin gency fits. However, the more usual approach is to measure each possible con found ing vari able and en ter it into a multivariate sta tis ti cal anal y sis that as sesses the im pact of fit on per for mance, con trol ling for the con founds. Such other causes of or ga ni za tional per for mance that may be con trolled sta tis ti cally in anal y ses of fit in clude in dus try con cen tra tion, firm size, risk, and le ver age (Ham il ton and Shergill 1993).

204

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

Turning to the measure ment of the rela tion ship between fit and per for mance, there are three main ways this is done sta tis ti cally in the lit er a ture: sub group anal y sis, re gres sion anal y sis, and de vi a tion anal y sis. Sub group Anal y sis In subgroup analysis the sample of organizations is broken into a number of groups, and their performances are compared. Typically there are two sub groups: the fits and the mis fits. As noted ear lier, it is an ad van tage that the num ber of cases in the sub groups can be rea son ably large be cause the sample is be ing bro ken into only two subsamples. The comparison be tween the sub groups may be of var i ous sta tis tics: means, re gres sion co ef fi cients, and cor re la tions. Means Per haps the sim plest form of subgroup anal y sis is to cre ate two sub groups—fits and misfits—and then show that the mean per formance of the fits is su pe rior to that of the mis fits. An ad van tage of this procedure is that the performances of organizations are aggregated within each sub group, and this tends to pro duce a more re li able per for mance score than for orga ni za tions in di vid u ally. Regression If there is fit be tween an or ga ni za tional struc tural vari able and a contin gency that leads to higher per for mance, then or ga ni za tions will tend to move into fit, leading to an as so ci a tion be tween the organizational struc ture and the con tin gency. This as so ci a tion pro duces a cor re la tion be tween the or ga ni za tional struc ture and the con tin gency. This in turn leads to a slope in a re gres sion anal y sis of the or ga ni za tional struc ture on the contingency. For example, size and structuring are positively cor re lated and struc tur ing has a pos i tive slope in a re gres sion on size (Child 1975). This slope may be used as the empirical es ti mate of the line of fit. How ever, a pref er a ble ap proach is to take a subsample of the best-performing organizations and use their regression slope as the estimate of the fit line. Be ing the high per form ers, more of them should be in fit than for the sample as a whole, which would contain more

Fit Con cept and Anal y sis

205

organizations that are in mis fit, thereby low er ing the per for mance of the whole sample. Sub group re gres sion takes the re gres sion line of the high-per form ing subsample, the fits, and then compares it with the re gres sion line for the low-per form ing subsample, the mis fits. The hy poth e sis is that these two re gres sion lines will be sig nif i cantly dif fer ent. The re gres sion lines are hy poth e sized to dif fer be cause the mis fits lie away from the fit line. The re gres sion line is there fore ex pected to dif fer in ei ther slope co ef ficient, or con stant term, or both. For ex am ple, Child (1975, p. 21, Fig ure 1) found that the re gres sion line of the low-per form ing firms had a different slope and constant (i.e., intercept) than the high-performing firms in re gres sions of spe cial iza tion on size. This shows that the low-per form ing firms, the mis fits, lie away from the fit line. Spe cifically, the misfits are less specialized than the fits at large size and are more spe cial ized than the fits at small size (Child 1975, p. 21, Fig ure 1). This illustrates how subgroup regression analysis can be used to show a relationship be tween fit and per for mance. The two regression lines of the fits and the misfits, however, cross over each other at a lowish size point (Child 1975, p. 21, Fig ure 1). This might seem to mean that at the point of intersection fit and mis fit are the same, so that the structure both fits and misfits that level of the contingency variable. Clearly that would be impossible, or would contradict the ideas of fit and mis fit. How ever, there is re ally no cross over, the ap pear ance of such is an artifact of the re gres sion tech nique. The regression line creates a continuum connecting disparate data points. If the mis fits to low size are above the re gres sion line of the fits and the mis fits to high size are be low that line, then the re gres sion line of the misfits will intersect that of the fits. The portion of the misfit regression line that ap pears to cross over the fit line is just an imag i nary line achieved by con nect ing up the mis fits on ei ther side of the fit line. In such a sce nario there is no ac tual cross ing of a line of fits and mis fits; there is no struc tural level that is both a fit and a mis fit. There are rea sons for believing the conjecture about the position of the misfits, as to where they are above and be low the fit line, so that this in ter pre ta tion gains credence. Child (1975, p. 21, Fig ure 1) found that the re gres sion line of structure on the size contingency for the high-performing subgroup was steeper than that of the low-performing subgroup. Moreover, the low-performing subgroup had a higher positive intercept than the

206

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

high-performing group. The result was that the two lines intersected to ward the lower level of the con tin gency vari able. Yet it is rea son able to hold that positive in tercepts are false. Certainly their theoretical meaningfulness is prob lem atic. A pos i tive in ter cept means that when size is zero, the organization has a structure. Thus despite having no employees, the organization nevertheless has specialized functions that are oc cu pied by peo ple! The func tional spe cial iza tion scale mea sures the extent to which each specialization is performed by at least one full-time per son (Pugh et al. 1968), so it would have to score zero if there were no em ploy ees in the or ga ni za tion. Thus the fit line of structure on size is theoretically mean ing ful only if struc ture is zero when size is zero, that is, the fit line goes through the origin (0, 0). (A case could be made that struc ture should be come pos i tive only when there are a num ber of em ploy ees, so that, tech ni cally speak ing, the in ter cept of the re gres sion of struc ture on size is neg a tive. How ever, for sim plicity we will as sume an in ter cept of zero, that is, the fit line passes through the or i gin. Even if the in ter cept is actually neg a tive, the fol low ing ar gument holds and so the as sump tion be ing made is con ser va tive.) If the true fit line goes through the or i gin, then, at lower lev els of the con tin gency, sub stan tial mis fits would have to lie above the true fit line due to the re stricted space be low the line. Sim i larly, at higher lev els of the con tin gency, sub stan tial mis fits would have to lie be low the true fit line due to the restricted space for misfits above it. Therefore, at the extremes of the contingencies, misfits would tend to be toward the mid dle of the struc tural vari able. Re gres sions are very sen si tive to such out li ers, there fore the slope of the mis fits would be shal lower than the slope of the fits and also the intercept of the mis fits would tend to be higher than that of fits. Thus, given that the misfits have a regression line with a shal lower slope and a higher in ter cept than the true fit line, the low-performing subgroup would have a shallower slope and a higher in ter cept than the high-per form ing sub group. There fore their two regression lines would intersect, yielding the observed finding (Child 1975, p. 21, Fig ure 1). While the ob served differ ences in the regres sion slopes between the low- and high-performing subgroups cap ture the pattern of the dif fer ences between the misfits and the fits, they under state those dif fer ences. The low-per form ing sub group would tend to be con tam inated to some de gree by fits, be cause the low-per form ing sub group is those or ga ni za tions that are only be low av er age per for mance and not

Fit Con cept and Anal y sis

207

just the lowest per form ers. There fore the true mis fit line would be of even shallower slope and more positive intercept than the observed regression line of the low-performing organizations. Similarly, the high-performing subgroup would tend to be contaminated to some de gree by mis fits, be cause the high-per form ing sub group is those or gani za tions that are only above av er age per for mance. There fore the true fit line would be of even steeper slope and less positive in tercept than the ob served re gres sion line of the high-per form ing or ga ni za tions. Given that, as we have seen, positive intercepts of the fit line make little the o ret i cal sense, the fit line should go through the origin, so its slope should be steeper. Therefore the observed positive intercept of the high-per form ing sub group may be artifactual. The slope of the fit line could be steeper than that found in the re gres sion anal y sis. Researchers should be aware of this possibility when mak ing their inter pretations of re gres sion re sults. The em pir i cal re gres sion re sults should not dis suade us from the o ret i cally mean ing ful fit lines, such as those that go through the or i gin. Sim i larly, if true fit lines are steeper than the regres sion lines of high-per form ing sub groups, then it be comes fea si ble to hold that they may have a slope of 1 as is pos tu lated by some the o ret ical ap proaches to spec i fy ing the fit line (Al ex an der and Randolph 1985; Keller 1994). In sum mary, pos tu lat ing that the fit line goes through the or i gin is more the o ret i cally mean ing ful and ex plains the pat tern of the mis fits, in clud ing their ap par ent cross-over of the fits. Also, if the fit line is steeper than the high-performance re gression and the mis fit line is shal lower than the low-per for mance re gres sion, then the com par i son of these two sub groups un der states the mag ni tude of the dif fer ence be tween the fits and mis fits in slopes (and also in ter cepts). Thus the dif fer ence in slopes (or in ter cepts) might be found in some stud ies to be small and “not sig nif i cant,” even where the fits dif fer sub stan tially from the mis fits, lead ing to the er ro ne ous con clu sion of no re la tion ship be tween fit and per for mance. An other po ten tial weak ness of the subgroup regression analysis is that the mis fits could all lie away from the fit line but yet there could be no dif fer ence in the re gres sion lines. Logically, mis fits can be above or below the fit line. If the misfits are both above and below the fit line, then their re gres sion line could be iden ti cal to the fit line, that is, have the same slope and con stant co ef fi cient. In such a situation, however, the cor re la tion co ef fi cient of the mis fit group would be lower than that of the fits, because the misfits are scattered farther away from the fit

208

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

line. One might be able to con firm this in ter pre ta tion by show ing that the misfits can be par ti tioned into two groups: one with a re gres sion line above the fit line and one be low. The cor re la tion of these two new re gres sions would be greater than that of the rather fuzzy line formed from all the mis fits scat tered on ei ther side of the fit line. Thus if there is a fit line in the data, then sub group re gres sions should be able to show that it differs from the misfits by comparing high-performing with low-per form ing or ga ni za tions, though some times sup ple men tal correla tional anal y sis may be re quired to help in the ver i fi ca tion.

Correlations Correlations may also be used as the main tool in a sub group anal ysis, by split ting the sam ple into low- ver sus high-per form ing or ga nizations and cor re lat ing the con tin gency and struc tural vari ables within each subgroup (e.g., Khandwalla 1973, p. 490, Table 2). Again, the lower-performing subgroup is expected to have a lower correlation between the contingency and the organizational structure than the high-performing subgroup, because the mis fits are scattered around the fit line. There is also the point that higher correlations in di cate a higher re gres sion slope (be cause cor re la tion co ef fi cients and re gres sion slopes are definitionally con nected). Thus higher cor re la tions pro vide in di rect ev i dence of dif fer ences in slopes and so the anal y sis be comes a subtle form of the sub group re gres sion anal y sis dis cussed above. The disadvantage of sub group correlational anal y sis, how ever, is that if all the misfits lay on one side of the fit line, then they could have the same or an even higher correlation than the fits. In such a case, how ever, a supplementary re gres sion anal y sis would show that the mis fits had a different slope or constant than the fits. Thus correlational analysis may need to be sup ple mented by re gres sion anal y sis. An al ter na tive correlational method is to split the sam ple into subgroups of low versus high lev els of the con tin gency vari able and then correlate structure and performance within each subgroup. Fit is supported by demonstrating that there are sta tis ti cally sig nif i cant differences between the sub groups, produced by the correlation being pos i tive for one sub group and neg a tive for the other subgroup (Argote 1982, p. 430, Table 2).

Fit Con cept and Anal y sis

209

Re gres sion Anal y sis In re gres sion anal y sis, the ef fect of fit on per for mance is as sessed by re gress ing per for mance on fit. Fit may be operationalized as a con gru ence term, so that the main effect of fit on per formance is assessed. Alternatively, fit may be operationalized as an in ter ac tion term, so that a mod er a tor ef fect is as sessed. Ei ther way, the ef fect of other causes of performance may be controlled by including them in the regression along side fit (e.g., Ham il ton and Shergill 1992 p. 106, Ta ble 5). Main Ef fect Re gres sion Anal y sis In main effect regression anal ysis, fit is operationalized as a single vari able tak ing two or more val ues, such as fit or mis fit. There may be more than one combination of the lev els of the contingency and structural variables that constitute a fit, for example, a functional struc ture fits very low di ver sity and a di vi sional struc ture fits di ver si fied strat e gies (Ham il ton and Shergill 1992, p. 99, Fig ure 2). Sim i larly, there may be more than one com bi na tion of the lev els of the con tin gency and struc tural vari ables that con sti tute a mis fit, for ex am ple, a func tional struc ture mis fits di ver si fied strat e gies and a di vi sional struc ture mis fits very low di ver sity (Donaldson 1987, p. 8, Fig ure 4). An advantage of this ap proach (rel a tive to mod er ated re gres sion dis cussed be low) is that the fit term operationalizes the idea of fit as congruency. Each of the combinations of the lev els of the con tin gency and struc tural vari ables that con sti tutes a fit is a state of con gru ence be tween con tin gency and struc ture, that is, a match of a level of struc tural vari able to a level of the con tin gency vari able. Sim i larly, each of the com bina tions of lev els of the con tin gency and struc tural vari ables that constitutes a mis fit is a state of incongruence between contingency and structure, that is, a mis match of a level of struc tural vari able to a level of the con tin gency variable. The effect of fit on performance is empirically assessed by testing whether the fit term shows a positive effect, that is, whether there is a main ef fect of fit on per for mance. Moderated Re gres sion Anal y sis In mod er ated re gres sion anal y sis, fit is rep re sented by a multi plicative in ter ac tion term, that is, the con tin gency mul ti plied by the

210

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

or ga nization structure (e.g., size times structuring). Orga ni za tional per for mance is re gressed onto this in ter ac tion term to see the mag ni tude of the slope coefficient. A slope significantly different from zero is taken to indicate an effect of fit on performance (Hill, Hitt, and Hoskisson 1992; Venkatraman 1989). Alternatively, the amount of perfor mance vari ance ex plained by the in ter ac tion term is com pared with that ex plained by the main ef fects of its two con stit u ent vari ables, the contingency and the structure. If there is a significant in crease from the in ter ac tion, then that is taken as ev i dence for an ef fect of fit on per for mance (Argote 1982, p. 429, Ta ble 1). The prob lem is that, as discussed above, a multiplicative interaction term does not re flect the con cept of fit as con gru ence and so it is not an operationalization of fit as that concept has been meant in con tin gency the ory re search. Therefore researchers wishing to test for the traditional type of fit should avoid moderated regression analysis and instead use one of the other tech niques, such as main ef fect re gres sion or sub group anal y sis.

De vi a tion Anal y sis In de vi a tion anal y sis the de gree of mis fit of an or ga ni za tion is measured by its dis tance from the fit line. This deviation could be from a theoretically given fit line. For instance, Alexander and Randolph (1985) and Keller (1994) define fit as a line in which the level of the structural variable equals that of the contingency vari able, that is, it passes through the or i gin and has a slope of 45 de grees to the hor i zon tal (see Fig ure 7.2). The con tin gency and struc tural vari ables each range from 1 to 5 in lev els. A mech a nis tic struc ture (i.e., struc ture level 1) fits a rou tine level of the tech nol ogy con tin gency (i.e., tech nol ogy level 1). In contrast, an or ganic structure (i.e., structure level 5) fits a non routine tech nol ogy (i.e., tech nol ogy level 5). For ev ery in crease of one level in tech nol ogy, the struc ture needs to in crease by one level in or der for it to fit. A one-unit deviation from the fit line, that is, the mildest degree of mis fit, re duces per for mance by one unit to –1. The great est mis fit pos si ble is a de vi a tion of four units, which re duces per for mance by four units to –4. Thus the greater the deviation, the greater the misfit, and so the lower the performance. The level of performance pro duced by each point of fit or mis fit i n the two-di men sional sp ace is given a pri ori. In con trast, where the fit line is de fined em pir i cally

Fit Con cept and Anal y sis

211

Fig ure 7.2. Mis fit and Per for mance

by regression, then the misfit is a re sidual from that regression (Pennings 1987). A vari ant of de vi a tion anal y sis is to mea sure the de gree of mis fit by the Eu clid ean dis tance for mula. Van de Ven and Drazin (1985, pp. 350-351) de fine the Eu clid ean dis tance as the dis tance be tween the ac tual score of an or ga ni za tion on a struc tural di men sion and the ideal score that the or ga ni za tion would re quire to be in fit with its level of the con tin gency vari able. This dis tance is then squared for each struc tural di mension and then added to the squared deviations scores of all the other struc tural di men sions for that or ga ni za tion. Finally, the square root is taken. Thus the Euclidean distance is the sum of all the deviations

212

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

across all the structural dimensions. The Euclidean distance for an organization mea sures its de gree of mis fit. This Eu clid ean dis tance can be correlated with performance to test the hypothesis that misfit is negatively cor re lated with per for mance. The deviation score has the advantage of providing a mea sure of misfit for each or ga ni za tion, which can be cor re lated with the per formance of each or ga ni za tion to yield a di rect es ti mate of the strength of the relationship be tween fit and per for mance. How ever, the de vi a tion score is a dif fer ence score, be tween the ideal level of the or ga ni za tional struc ture that is re quired to fit the con tin gency and the ac tual level of that structure. Therefore it is prone to low reliability, as will be dis cussed below (in Chapter 8). Thus the correlation be tween fit and performance tends to un der state the true ef fect of fit on per for mance. Moreover, the fits are to a few ideal profiles, rather than to a line of many fits (Drazin and Van de Ven 1985, pp. 532-534). There fore many points on a fit line would lie away from the few, ideal pro file fits and so be treated as misfits, thereby contaminating the analysis of fit and performance (as dis cussed for con fig u ra tions in Chap ter 5). Overall, the discussion of methodological issues has suggested the advantages and disadvantages of various approaches. By using more than one ap proach, the an a lyst may avoid the pit falls in her ent in any one ap proach and reach a sounder con clu sion.

Conclusions Fit is cen tral to con tin gency the ory and so is re ceiv ing in creas ing study in em pir i cal re search. The idea that fit raises per for mance moves be yond be ing an empty tau tol ogy by be com ing more spe cific about which or ganizational struc tures fit which con tin gen cies. Fit is the congruence be tween the or ga ni za tional structure and its contingency. The line of fit may be straight or curved. The empirical tests for fit in volve show ing that de vi a tions from the fit line cause lower per for mance. Fit as tra di tion ally un der stood in con tin gency the ory is not cap tured by a multi pli ca tive in ter ac tion term. Fit is a line of iso-performance in contingency the ory re search, in that each point on the line pro duces the same per for mance as any other point. Re search fails to sup port the no tion of equifinality. Equifinality would be ev i dence for choice only if struc tures yield ing the same

Fit Con cept and Anal y sis

213

perfor mance ex ist at the same level of the con tin gency. If they ex ist for different lev els of the con tin gency, then the or ga ni za tion may be constrained by its level on that con tin gency so that there is no choice. Whereas it has been sug gested pre vi ously that there are three con cepts of fit: se lec tion, in ter ac tion, and sys tems, these would be better reconcept ualized as man a ge rial de ci sion, con gru ence, and multifit, re spec tively. Mul ti ple fits can be com bined by add ing them to gether to pro duce the to tal ef fect on per for mance of mul ti ple fits. Con flicting con tin gen cies can be han dled within this ad di tive model by sub tract ing the struc tural pre scrip tions of one contingency from an other to yield a net ef fect. The pro cess whereby fit is iden ti fied may be ini tially driven by ei ther the ory or data. Ei ther way, fit needs a the o ret i cal ra tio nale and em pir i cal validation. Causal in fer ence is aided by us ing a re search de sign in which fit tem po rally pre cedes per for mance, and other causes of per for mance are controlled. There are several techniques for empirically dem onstrat ing fit. These in clude sub group anal y ses, re gres sion anal y ses, and deviation anal y sis. The char ac ter is tics of each have been dis cussed.

THE Fit AfCON fectsTIN Per GENCY for mance THEORY OF OR GA NI ZA TIONS

8

Fit Affects Performance

C

o ntingency theory states that fit affects performance. There is over whelm ing ev i dence that this is true. In this chap ter we will review the em pir i cal stud ies and see how they sup port the con tin gency idea that fit pos i tively af fects per for mance. This re view will be followed by a discus sion of meth od olog i cal issues that influ ence the ob served strength of the re la tion ship between fit and per formance. This reveals cer tain pitfalls in re search ing the fit-per formance re lationship, which means that the strength of the re la tion ship may not al ways be shown fully by em pir i cal stud ies. We will close by ex tract ing eight les sons that may help us surmount these po ten tial prob lems in future research, so that it dem on strates the true strength of the ef fect of fit on or ga ni za tional performance. While we will dis cuss these is sues in the con text of struc tural contin gency re search, many of them ap ply gener ally to any con tin gency theory.

215

216

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

Studies of Contingency Fit and Performance Dalton, Todor, Spendolini, Fielding, and Porter (1980) provided a criti cal re view of the re search con nect ing struc ture and per for mance. They commented that despite the importance of performance there was a pau city of re search, weak nesses in meth ods, and in con sis ten cies in find ings, both re gard ing the main ef fects of struc ture and the con tingent ef fects of struc ture. They called for more and better re search. Two early stud ies of the re la tion ship be tween con tin gency fit and perfor mance failed to find a pos i tive re la tion ship (Mohr 1971; Pennings 1975). This led to the va lid ity of con tin gency the ory be ing ques tioned. For in stance, Pfeffer (1997, p. 161) states: “This [con so nance, i.e., fit- performance] hypothesis has received sup port in some studies (e.g., Wood ward, 1965) but not in oth ers (Mohr, 1971; Pennings, 1975).” He goes on to state that some support f or the hypothesis comes also from Schoonhoven (1981) (Pfeffer 1997, p. 162). How ever, Pfeffer (1997) con cludes about struc tural con tin gency the ory that the Schoonhoven (1981) study “marked vir tu ally the end of em pir i cal re search on or ga ni za tional struc ture.” Thus, ac cord ing to Pfeffer (1997) not only is the balance of findings from studies of fit and per for mance not pos i tive, but no further work has addressed the fit-performance relationship. This de fi ciency has oc curred de spite the ac knowl edge ment by Pfeffer (1997, p. 162) that “empirical sup port for the con so nance hy poth e sis has been in con sis tent. But that could con ceiv ably be rem e died by more careful studies and measures.” However, subsequent to Mohr (1971) and Pennings (1975), there have been other studies that have found a positive re lationship between fit and per formance and so sup port con tin gency the ory (Al ex an der and Randolph 1985; Argote 1982;Dewar and Werbel 1979; Donaldson 1987; Drazin and Van de Ven 1985; Gresov 1989, 1990; Hamilton and Shergill 1993; Hill, Hitt, and Hoskisson 1992; Hoskisson 1987; Jennings and Sea man 1994; Keller 1994; Kraft, Puia, and Hage 1995; Nohria and Ghoshal 1997; Pennings 1987; Powell 1992). We have dis cussed some of these stud ies al ready and will now ex am ine the other stud ies to draw out their find ings and their im pli cations for re search into the re la tion ship be tween fit and per for mance. Taken to gether, the stud ies of the re la tion ship of fit and per for mance, overall, sup port the con tin gency the ory that fit pos itively affects per for mance. More stud ies find for the pos i tive re la tion ship than find against it. More over, some of the stud ies find ing for the relationship

Fit Af fects Per for mance

217

ind icate that some of the ear lier stud ies may have failed to find the re la tion ship be cause of meth od olog i cal lim i ta tions. Mohr (1971) studied the fit between unpredictability of jobs and participativeness in local health departments. There was almost no association between fit and effectiveness, so contingency theory was not sup ported. Uni ver sal ist ic the ory was sup ported in that ef fec tive ness was higher for high par tic i pa tion—sup port ing hu man re la tions the ory. Pennings (1975) stud ied the fit of struc tural to en vi ron men tal vari ables in bro ker age of fices. He con cluded that there was lit tle sup port for such effects, and the contingency theory had to be questioned. He found sup port for main ef fects of struc ture, par tic u larly power, on ef fec tiveness, again lend ing sup port to uni ver sal ism. De spite these two studies failing to support contingency the ory, however, other studies have supported con tin gency the ory. A sub se quent study by Pennings (1987) of branches of a large commercial bank found support for contingency theory. He shows that misfit is correlated with lower performance, particularly financial performance. Ideal pro files of struc tural (com mu ni ca tions and power) vari ables for low, me dium, and high lev els of five en vi ron men tal contingency vari ables (as pects of com pe ti tion and cus tom ers) were iden tified from high-per form ing branches. The de vi a tions from these ideas were then computed by the Euclidean distance formula for the remaining branches. For the three fi nan cial mea sures of per for mance, all cor re la tions were in the the o ret i cally expected direction and eight out of fif teen were sig nif i cant. In par tic u lar, the ra tio of in ter est in come to de pos its was sig nif i cant in four out of five anal y ses with mis fits and the av er age cor re la tion can be cal cu lated as –.31 (Pennings 1987, p. 231, Ta ble 1). Ca non i cal anal y ses show that the re la tion ship be tween en vironmental contingency and structural variables is stronger for the high-per form ing than for the low-per form ing branches, which is con sis tent with their struc tures fit ting the con tin gen cies (Pennings 1987). The de gree of cor re la tion be tween mis fit and per for mance is en courag ing, especially the interest to deposit average correlation of –.31. It should be noted, more over, that these cor re la tions un der es ti mate the true correlation, because the highly fitting branches are not used in correlating misfit with performance, thus restricting the range of misfit. (This pro ce dure is to avoid, per haps un nec es sar ily, the seem ing tau tol ogy of mea sur ing mis fit of the same or ga ni za tions that were used to de fine fit.) A weak ness of the study (Pennings 1987) is that it does

218

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

not report the main effects of environment and structure on per formance, and so the extent to which these constituent variables of fit work to create a spurious degree of misfit-performance relationship (as will be dis cussed be low) is un known. Mul ti ple struc tural vari ables are ap par ently summed to gether to yield the mis fit with each en vi ronmen tal con tin gency, in the man ner of Drazin and Van de Ven (1985). However, the misfits with each environmental contingency are not summed to gether to yield an as sess ment of the to tal ef fect of mul ti ple mis fit on per for mance. Over all, the find ings of Pennings (1987) tend to con firm the an a lytic ap proach of Drazin and Van de Ven (1985), upon which Pennings drew. The study shows that it can be ex tended to dif ferent sorts of or ga ni za tions (bank branches), to dif fer ent struc tural variables, and to multiple contingencies that differ from the single one used by Drazin and Van de Ven (1985). Fur ther Studies of Task Con tin gency While Pennings (1987) stud ied the con tin gency of the environment, oth ers have stud ied the con tin gency of the task it self and in that way exam ined a con tin gency sim i lar to the job un pre dict abil ity that Mohr (1971) stud ied. Un like Mohr, these sub se quent stud ies found pos i tive evidence supporting contingency theory. There have been four such stud ies of the task con tin gency (Al ex an der and Randolph 1985; Argote 1982; Gresov 1990; Keller 1994). Argote (1982) studied the fit of organizational co or di na tion structures to in put un cer tainty in hos pi tal emer gency units. She found that fit was positively cor re lated with ef fec tive ness as mea sured by pa tient care. These re sults held us ing both mod er ated re gres sion and sub group anal y ses (i.e., cor re la tions be tween struc ture and ef fec tive ness for low and high un cer tainty). Thus these two meth ods yielded sim i lar re sults. It may be noted in pass ing that in put un cer tainty was not sig nif i cantly cor re lated with most of the struc tural vari ables in this study (Argote 1982). Hence a preliminary anal ysis look ing for contingencies by seeking variables correlated with structure would not have detected that in put un cer tainty was a con tin gency of most of the struc tural variables. There fore, while con tin gency re la tion ships will of ten be re vealed by associations between contingency and organizational structure, such associations are not always present. Their absence should not

Fit Af fects Per for mance

219

fore stall trivariate anal y ses of con tin gency, struc ture, and out comes to test di rectly for con tin gency re la tion ships. Sim i lar to Argote (1982), Al ex an der and Randolph (1985) found, in a study of nurs ing, that the fit be tween struc ture and tech nol ogy (i.e., task) affected the quality of patient care. This effect of fit on patient care held af ter con trol ling for the ef fects of both struc ture and tech nology on pa tient care, that is, for the con stit u ent vari ables of fit. How ever, this fit effect holds only for the fit of horizontal participation (struc ture) to vari abil ity (tech nol ogy). The fit of for mal iza tion (struc ture) to un cer tainty (tech nol ogy) was sig nif i cantly cor re lated with care qual ity, but the sign was op po site to that the o ret i cally ex pected (Al ex an der and Randolph 1985). Whereas con tin gency the ory holds that un cer tainty needs less formal iza tion, in this set ting it required more. Thus the gen eral the ory con nect ing un cer tainty and for mal iza tion may be ar gued to have an exception in nursing aimed at producing qual ity care, as Al ex an der and Randolph (1985, p. 856) sug gest. Al ex an der and Randolph (1985) used, as their mea sure of mis fit, de vi a tion from the fit line de fined as struc ture score equals the tech nol ogy score. Thus the fit line is given a pri ori. The theory is that an in crease in tech nol ogy re quires the same in crease in struc ture for the or ga ni za tional unit to be in fit. This con trasts with the empirical derivation of the fit line from data that are used in other contingency the ory re search (e.g., Pennings 1987). The positive findings at tained by Alexander and Randolph for the fit of horizontal par tic i pa tion to vari abil ity con firm that their the oretical fit line does ap prox i mate the true fit line for these vari ables, be cause devi a tions from it are cor re lated with lower per for mance. The same ap proach to mod el ing fit was used by Keller (1994), who also examined the effects of fit on performance in or ganizational subunits. How ever, in this study, the sub units were pro ject groups from four in dus trial re search and de vel op ment or ga ni za tions, com posed of highly qualified professionals, mostly male. The fit of project group struc ture to tech nol ogy was found to af fect pro ject group per for mance. Spe cifically, more in for ma tion-pro cess ing struc tures fit ted more nonroutine task technology and led to higher-quality project outcomes. Keller (1994) also found a main ef fect of struc ture (infor ma tion process ing) on per for mance, though this was weaker than fit, thereby confirming the im por tance of con tin gency. More over, the effect of fit on performance is not a spurious effect of the constituent variables (struc ture and tech nol ogy) be cause both were con trolled (as discussed

220

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

below). Again, the positive findings confirm the fit line used, even though it was de rived sim ply and on an a priori theoretical ba sis. The pos i tive ef fect of fit on per for mance held not only for pro ject quality measured concurrently with fit, but also one year later (which was stronger than concurrently). This means that fit affected subsequent performance, thereby pro vid ing ev i dence that fit is a cause of per formance. This is an im portant finding in that most stud ies of the re lationship between fit and performance have been cross-sectional and thereby less def i nite about cau sal ity. Fo cusing on the causal ef fect of fit on per for mance should re duce any neg a tive ef fect of per for mance on fit (dis cussed be low), so this may be a rea son why this re la tion ship was stron ger than for con cur rent per for mance. In Keller (1994), how ever, a pos tu lated fit of in for ma tion pro cess ing to unanalyzable task tech nol ogy had a weak and nonsignificant ef fect on pro ject qual ity. The smaller vari a tion of this fit (rel a tive to that with nonroutineness [Keller 1994, p. 173]) may have ar ti fi cially re duced the cor re la tion be low its true size. Fur ther, nei ther fit had a sig nif i cant effect on the ability of teams to keep their projects on budget and sched ule (Keller 1994, p. 173). Both the Al ex an der and Randolph (1985) and Keller (1994) stud ies—like Argote (1982)—found pos i tive ef fects of fit on qual ity. It may be that qual ity mea sures the en hanced abil ity of groups to solve prob lems posed by more un cer tain and vari able tasks. As the task becomes more uncertain and vari able, so it re quires more prob lem solv ing, and this need is met by more or ganic struc tures (i.e., more participation and more information processing). Thus quality may be one of the mea sures that is more suit able for test ing the ben e fits of fit ting organicness to task. In con trast, mea sures that tap ad min istrative efficiency such as budget attainment may not tap the benefits from this fit. Fut ure research should be sensitive to whether the measures of per for mance used are those that are rel e vant for the fit under study. It is worth noting that, in both studies (Al exander and Randolph 1985; Keller 1994), the vari ables called tech nol ogy are re ally as pects of the task rather than hard ware, so that the pos i tive find ings support the task contingency rather than technology. Gresov (1990) found that the ef fi ciency of or ga ni za tional work units was re lated to the fit of their struc tures to the task un cer tainty con tin gency and also to the horizontal dependence contingency. High task un cer tainty and high hor i zon tal de pend ence each re quired an or ganic rather than a mech a nis tic struc ture. In sum, the re la tion ship of the fit

Fit Af fects Per for mance

221

of struc ture to the task con tin gency with per for mance is sup ported in four empirical studies (Argote 1982; Alexander and Randolph 1985; Gresov 1990; Keller 1994). Fit to the Strat egy Con tin gency There have been studies also of the fit between strategy and struc ture. These are to be dis tin guished from stud ies of the main ef fects, that is, whether ei ther strat egy or struc ture each on its own af fects per formance, that is, universalistic rather than contingency effects (Cable and Dirrheimer 1983; Ca ble and Yasuki 1985; Har ris 1983; Hill 1985a; Hill and Pickering 1986; Hoskisson and Galbraith 1985; Teece 1981). Rumelt (1974) in ves ti gated the ef fects of strat egy and struc ture on perfor mance (see also Channon 1978), but did not an a lyze the ef fects of fit between strat egy and struc ture on per for mance. Sub se quent schol ars developed operational measures of fit and analyzed their effects on per for mance, find ing nil (Grinyer, Yasai-Ardekani, and Al-Bazzaz 1980) or pos i tive ef fects (Donaldson 1987; Ham il ton and Shergill 1992, 1993; Hoskisson 1987; Jennings and Sea man 1994). Ham il ton and Shergill (1992, 1993) con ducted a study of the larg est corporations in New Zealand. They found that the fit of struc ture to diversification strat egy led to higher fi nan cial per for mance. This held for both of their indices of fi nancial per formance: growth (in sales, divi dends, as sets, and earn ings per share) and prof it abil ity (re turn on as sets and re turn on eq uity). They also com pared the vari ance ex plained by fit to that ex plained by the ef fect of strat egy on its own (i.e., its main effect) and to organizational size. The results were that there was an effect of fit that held con trol ling for a main ef fect of strat egy, so that its strategy constituent did not spuriously cause the fit effect. Moreover, the effect of fit on profitability was much larger than size, a variable often found to affect profitability. Other re search has tested for ef fects of strat egy-struc ture fit on per for mance but failed to find it (e.g., Grinyer, Yasai-Ardekani, and Al-Bazzaz 1980). In or der to reveal the ef fect of strat egy-struc ture fit on performance, the or ganization needs to be in that state long enough to be definitely in fit rather than misfit and for its effects to cumulate and so be visible. In order to tap the cumulative effect it is necessary to con fine the study to or ga ni za tions that have been in one state, ei ther fit or mis fit, for a lengthy period, such as ten years. Both the

222

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

stud ies that use mul ti ple (10) year pe ri ods of fit ver sus mis fit find effects of fit on per for mance(Donaldson 1987; Ham il ton and Shergill 1992, 1993). In con trast, the study that fails to find such an ef fect (e.g., Grinyer, Yasai-Ardekani, and Al Bazzaz 1980), examines fit at only one point in time, which may mean the or ga ni za tion has not been in fit long enough to have pro duced a de tect able effect of fit. There fore, the ef fect of be ing in fit for ten years should be stronger than being in fit for a short time. In other words, the more “treat ment,” the greater the ef fect, and so the more chance of the ef fect be ing de tected. More over, to iso late the true ef fect for or ga ni za tions it is nec es sary to ensure that an organization remains in fit for the entire period over which fit is mea sured. An or ga ni za tion that is in fit at time pe riod 1 and then in fit at time pe riod 2 (say 10 years later) may not have re mained in that state dur ing the in ter ven ing years if it di ver si fies dur ing that pe riod. The dy namic pat tern is that an or ga ni za tion in fit that di ver si fies moves into mis fit and then even tu ally changes its struc ture and moves into fit. Therefore an organization in fit at times 1 (i.e., having a functional struc ture) and 2 (i.e., hav ing a di vi sional struc ture), that di ver si fied between those times, can have spent much of the in ter ven ing time in misfit. There fore to clas sify the or ga ni za tion as in fit is in cor rect in that for most of the time pe riod it was in mis fit and this con tam i nates the findings. The analysis of Hamilton and Shergill (1993) is of corporations that changed neither their strategy nor their structure over the study pe riod, in or der to con trol for the ef fects of such changes. This is sim i lar to Donaldson (1987), who ex cludes cor po ra tions that di ver si fied dur ing the study pe riod. Hence stud ies of ef fects of fit on per for mance would do well to confine the organizations to those that remain in the one state for at least sev eral years. For or ga ni za tions mea sured at two points in time, those or ga ni za tions chang ing strat e gies should be ex cluded to be cer tain that the or ga ni za tions re main in the same state. Donaldson (1987) stud ied the ef fect of be ing in strat egy-struc ture-fit (for 10 years) on per for mance in the en su ing years. Thus the re search de sign was not cross-sec tional but rather diachronic, add ing con fi dence that performance was an effect of fit. Hamilton and Shergill (1993) used a cross-sec tional de sign, but, be cause fit at the start of the pe riod precedes growth (in sales, dividends, as sets, and earnings per share) dur ing that pe riod, we can make the causal in fer ence that fit was a cause of the growth di men sion of per for mance. Thus their study also pro vides ev i dence that fit is a cause of per for mance.

Fit Af fects Per for mance

223

Hoskisson (1987) examines the performance of firms before and af ter divisionalizing (i.e., adop tion of the M-form struc ture) and in this way as cer tains the causal im pact of divisionalization on per for mance. Con trols were ap plied for an nual gross na tional prod uct growth, as set growth, com pany trend residuals, early ver sus late adopt ers, in dus try, and sales (Hoskisson 1987, pp. 634-635). Hoskisson (1987) finds that the ef fect of divisionalization on per for mance is mod er ated by strat egy, con sis tent with the con tin gency idea of fit. Firms with the high est level of di ver si fi ca tion (i.e., un re lated diversifiers) had per for mance in creases resulting from adoption of the divisional struc ture (Hoskisson 1987, p. 639, Fig ure 1). This is ex pli ca ble in con tin gency the ory be cause the divisional structure is a fit with unrelated diversification, so that, by divisionalizing, these firms are mov ing into fit and there fore in creas ing their per for mance. In con trast, ver ti cally in te grated firms had a per formance decline resulting from adoption of the divisional structure (Hoskisson 1987, p. 639, Figure 1). This is explicable in contingency the ory as the di vi sional struc ture not be ing a fit with ver ti cal in te gra tion, be cause the in ter de pen dence among the “di vi sions” pre cludes their operating au tonomously as well as assessment as independent profit cen ters, so that the fully fledged divisional struc ture, that is, the M-form, is dysfunctional (Lorsch and Allen 1973). Firms with the medium level of di ver si fi ca tion (i.e., related diversifiers) had milder per for mance de creases re sult ing from adop tion of the di vi sional struc ture (Hoskisson 1987, p. 639, Fig ure 1). This is ex pli ca ble in con tin gency the ory as the di vi sional struc ture not nec es sar ily be ing a fit with re lated di ver si fi cation. Where in no va tion is sought, the greater flex i bil ity and cus tomer re spon sive ness makes the divi sional struc ture a fit, but where cost re duction is the priority, as in a price-com petitive market, then a functional struc ture is the fit for a related diversifier (Donaldson 1979, 1985). If more of the re lated diversifiers faced cost rather than in no va tion as their critical challenge, then these results of mild decrease in performance would be expected. The inclusion of this addi tional con tingency factor in subsequent studies would clarify these results. Thus the find ings of Hoskisson (1987) are con sis tent with con tin gency the ory and il lus trate a re search de sign based on com par ing per for mance be fore and af ter the adop tion of a par tic u lar struc ture. Jennings and Seaman (1994) illustrate the way in which contin gency the ory can il luminate the fits be tween in novation strat egy and structure in strate gic man agement. They studied savings and

224

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

loans organizations. Some of these had a pros pec tor strat egy and high ad ap ta tion, in that they were ag gres sively pur su ing the new op por tu nities made possible by deregulation. Some had a defender strategy of conducting traditional business and low adaptation. Ef fective imple mentation of these two con trast ing strat e gies was pos tu lated to in volve adopting a fitting organizational structure: organic for prospectors and mechanistic for defenders. Among the prospectors, the group of or ga ni za tions with the most or ganic struc tures had the high est per formance (Jennings and Sea man 1994, p. 469). Sim i larly, among the de fenders, the group of or ganizations with the most mecha nis tic structures had the high est per for mance. This shows the im por tance of fit ting struc ture to strat egy in order to at tain high per formance. Specifically, among prospectors, the most organic group had prof itability over 100 times greater than the least organic group. Similarly, among defenders, the most mechanistic group had profitability over thirty times greater than the least mechanistic group (Jennings and Sea man 1994, p. 469). These re sults in di cate that the best fit for the pros pec tor strat egy was the highest level of organic structure, possessed by the most organic group of firms (Jennings and Sea man 1994). Of the two other groups of pros pec tor firms, both had less or ganic struc tures, and the less or ganic their struc tures, the lower their per for mance. Sim i larly, the best fit for the defender strategy was the highest level of mechanistic structure, possessed by the most mechanistic group of firms. Of the two other groups of de fend ers, both had less mech a nis tic struc tures, and the less mechanistic, the lower their per for mance. Hence the best per form ing groups of both pros pec tors and de fend ers were at the ex tremes of the or ganic-mech a nis tic con tinua. This is con sis tent with the idea that there was a fit line running di agonally from prospector-high organic to defender-low or ganic (i.e., high mech a nis tic). The two high-performing groups were on the fit line (or closer to it than the other groups). The fur ther an or ga ni za tional group was from this fit line, the lower was its per for mance. Thus or ga ni za tions whose struc tures were at an in ter mediate level between organic and mechanistic had lower performance be cause they were not in fit with their strat egy. In this study the broad na ture of fit of struc ture to strat egy was given from a pri ori theory, but the level of struc ture that is more op ti mal was dis cov ered em pir i cally. A main limitation of the study was that fit and performance were

Fit Af fects Per for mance

225

apparently mea sured for the same time pe riod, so that causal in fer ence is less cer tain. Steer and Cable (1978) investigated the effect of organizational struc ture on per for mance. Their anal y sis was in the con tin gency spirit in that they used op ti mal form to cap ture the fit of di vi sional struc ture to the firm’s “op er at ing en vi ron ment” (Steer and Ca ble 1978, p. 17), but un for tu nately the na ture of this fit is un clear (see also Thomp son 1981). Other Studies of Fit and Per for mance Kraft, Puia, and Hage (1995) reanalyzed the data from Child’s National Study. Un like Child (1975), who used sub group re gres sions to iden tify fit and mis fit, they mea sured mis fits by de vi a tions from a fit line of an a priori the o ret i cal sort (the required structure equals the contingency) and also by residuals from a regression analysis. They found that the deviation method was su pe rior in re veal ing fit ef fects, which sug gests that the the o ret i cally de fined fit line was closer to true fit than the regression line. They also examined for main effects and found as many as their fit ef fects. Kraft et al. an a lyzed ser vice or ga ni zations separately from manufacturing or ganizations and found more relationships among ser vice or ga ni za tions. They also used a wider set of performance measures than the financial measures used by Child (1975). Their re sults sug gest that while fit may lead to higher fi nan cial per for mance, it may be at a so cial cost such as in dus trial un rest, that is, work stop pages. This study il lus trates how con tin gency the ory re search can dif fer en ti ate among the out comes of fit for dif fer ent stake holders of the organization. The results of fit can be positive for one set of stake holders but neg a tive for oth ers, so that a fit for one group may be a mis fit for an other. Schlevogt and Donaldson (1999) ex am ined the re la tion ship be tween fit and per for mance of firms in China. Fit was of three struc tural vari ables (centralization, formalization, and integration) to two con tingency vari ables (size and un cer tainty), ex em pli fy ing a multistructure, multicontingency approach to fit, that is, multifit. In an analysis in which eleven other en vi ron men tal and or ga ni za tional causes of per formance were controlled, structural fit was shown to positively affect performance. Moreover, fit was the third strongest out of the twelve

226

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

causes of per for mance, thus sug gest ing the con tin u ing im por tance of fit for both re search ers and man ag ers. Over all, while the two stud ies by Mohr (1971) and Pennings (1975) fail to sup port con tin gency the ory, other stud ies sup port con tin gency the ory by find ing a pos i tive re la tion ship be tween fit and per for mance (Alexander and Randolph 1985; Argote 1982; Child 1975; Donaldson 1987; Drazin and Van de Ven 1985; Gresov 1990; Ham il ton and Shergill 1993; Hill, Hitt, and Hoskisson 1992; Hoskisson 1987; Jennings and Sea man 1994; Keller 1994; Khandwalla 1973; Kraft, Puia, and Hage 1995; Pennings 1987; Powell 1992; Schlevogt and Donaldson 1999). More stud ies support the positive relationship of fit and performance than fail to support it. More over, the later and more so phis ti cated re search pro vides more sup port for con tin gency the ory than the ear lier re search. This shows that the contingency ap proach is sound. It should re ceive in creas ing at ten tion to at tain more ac cu rate def i ni tions of fit and more ex act es ti ma tions of the ef fect of fit on per for mance, to re veal the full po ten tial of con tin gency the ory.

The Empirical Estimation of the Effect of Fit on Performance In the pre vi ous chap ter we dis cussed how to iden tify the fit line and val i date its ex is tence by show ing that fits dif fer from mis fits in their effect on organizational performance. But we may also want to know how much fit af fects per for mance. The es ti ma tion of the strength of fit on performance is prone to certain prob lems such that the strength is of ten un der es ti mated in em pir i cal stud ies. More over, if the fit-per for mance relationship is un der stated suf fi ciently, such as if it fails sta tis ti cal signif i cance tests, then the er ro ne ous con clu sion might be drawn that fit has no effect on performance and thus that the fit concept is invalid. Knowledge of the pitfalls in estimating the strength of fit on per formance may en able us to avoid them. Some times they can not be avoided in an em pir i cal study, but be ing aware of these pit falls as sists us to make a better in ter pre ta tion of the study. We may ex plore these is sues us ing three methodological categories that affect the strength of empirical re sults: un re li abil ity, range restriction, and con founds. These gen eral methodological considerations have spe cific significance in the topic of fit and per for mance.

Fit Af fects Per for mance

227

Unreliability If two vari ables have a true cor re la tion of, say, .6, this will be found only if both vari ables are mea sured with com plete re li abil ity. Any unreli abil ity low ers the ob served cor re la tion be low the true correlation, because its ef fect is to cre ate ran dom er ror (Nunnally 1978). Thus any un re li abil ity in the mea sure ment of ei ther fit or per for mance leads the empirical es ti mate of the ef fect of fit on per for mance to un der state the true mag ni tude of that ef fect. If per for mance is mea sured by the ac count ing fig ures of a firm, say profit, then there will be some un re li abil ity due to er rors of mea surement. If firm per for mance is mea sured by re turn on as sets (ROA), that is, the ra tio of profit to as sets, then this is li a ble to be even more un re li able because of the well-known unreliability in the measurement of assets. Mea suring firm per for mance by ROA is quite pop u lar in stud ies of fit-performance of firms and so these studies are underestimates (e.g., Hamilton and Shergill 1992). Again, changes in performance, such as growth mea sures, are prone to be of lesser re li abil ity be cause their nu mer a tor (e.g., profit now less profit 2 years ear lier) is a dif fer ence score and, as such, subject to mea sure ment er ror, as will be dis cussed be low. Thus what ever the operationalization of per for mance, any un reliability in its measurement renders the observed fit-per formance relationship con ser va tive. How ever, larger prob lems of un re li abil ity oc cur re gard ing fit. Fit is a combination or con gru ence of two vari ables, the con tin gency and the or ga ni za tion struc ture. There fore any un re li abil ity in the mea sure ment of ei ther of these will lower the ob served fit-per for mance re la tion ship be low its true strength. How ever, mis fit is a dif fer ence be tween the level of the or ga ni za tion struc ture that is re quired to fit the con tin gency and the actual level of the organizational struc ture. This as pect is readily ap par ent by re fer ring to mea sures of mis fit as dif fer ences (Powell 1992) or re sid u als from the fit line (Van de Ven and Drazin 1985). In psycho metrics, such dif fer ence scores are seen as being much less re li able than the vari ables of which they are com posed (Johns 1981; Venkatraman 1989). The ra tio of mea sure ment er ror to true score is much higher for dif fer ence scores than for the vari ables from which the dif fer ence is calcu lated. (Nev er the less, Al ex an der and Randolph 1985 and Keller 1994 argue that their the o ret i cally derived mis fit scores are re li able despite be ing dif fer ence scores.)

228

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

To put it sim ply, if vari able X is mea sured on a ten-point scale (from 1 to 10) and an or ga ni za tion scores five, the er ror of mea sure ment may be one point, so that the true score could be four or six, that is, an er ror of mea sure ment of 20 per cent. How ever, the level of X re quired to fit the con tin gency value might be seven, so that the mis fit is two, but with a one-point er ror of mea sure ment, the er ror for that mis fit is 50 per cent. Thus, in this ex am ple, the mis fit, be ing a dif fer ence score, has a ra tio of mea sure ment er ror to true score that is two-and-a-half times that of vari able X, from which it is con structed. Hence ev ery time we es ti mate the fit-per for mance re la tion ship, this problem of un derestimation due to unreliability potentially occurs. The prob lem is pres ent, for in stance, when ever cor re la tions or re gres sions are used that in volve the mis fit or fit of each in di vid ual or ga ni za tion be ing mea sured and then en tered into the anal y sis sep a rately. It is partic u larly a prob lem where we mea sure the mis fit of each or ga ni za tion and then cor re late this with the per for mance of each or ga ni za tion, because both the mis fit and the per for mance mea sure may be low on re liabil ity. There fore all such find ings must be re garded as un der es ti mates and in ter preted ac cord ingly. For ex am ple, the cor re la tion be tween the Euclidean distance and performance of organizational subunits by Drazin and Van de Ven (1985) will un der es ti mate the true re la tion ship considerably. A so lu tion that re duces this prob lem is to use ag gre ga tion. The av erage score on a vari able of a sam ple is more re li able than the score of an in di vid ual or ga ni za tion in that sam ple. The er rors of mea sure ment are ran dom and there fore com bin ing across in di vid u als to form a sample av er age can cels out much of the er rors. This is achieved in the pro cedure dis cussed ear lier of form ing two sub groups, the fits and the misfits, and then cal cu lat ing the mean per for mance of each. Be cause of the aggregation into the two subgroups, the mean performances of each sub group are more re li able than those of their mem bers. By com par ing the dif fer ences in the means of the fit and mis fit sub groups, a fairly re liable es ti mate of the ben e fits of fit may be ob tained. Such an es ti mate, how ever, is of the form: “fit raises per for mance by 20 per cent rel a tive to mis fit.” This is in for ma tive, but it is not an es ti mate of the fit-per for mance re la tion ship of the kind that says: “fit and per for mance are cor re lated .4.” Where state ments of that kind are sought, correlational or regression techniques can be used, but the analyst should be aware that they are li a ble to yield un der es ti mates. How ever,

Fit Af fects Per for mance

229

the cor re la tions can be cor rected for the un re li abil ity of their two variables by applying a formula (Hunter, Schmidt, and Jackson 1982). If the re li abil ity of ei ther fit or per for mance (pref er a bly both) can be calcu lated, then the cor rec tion for mula can be ap plied. The cor rec tion of ob served cor re la tions be tween fit and per for mance would tend to raise their value, lead ing to a better ap pre ci a tion of the strength of the ef fect of fit on per for mance. Range Re stric tion An ob served em pir i cal re la tion ship can also un der es ti mate the true re la tion ship be cause of re stric tion in range. If two vari ables X and Y are cor re lated .6 in the real world, an em pir i cal study of a sam ple of or ga niza tions will only find that cor re la tion if the vari a tion on X in the sam ple is as great as that in the world, and similarly for Y. If variable X ranges in value from 1 to 10 in the world, but only from 3 to 8 in the sample, then its range is re stricted. In or der to re veal the true mag ni tude of a cor re la tion, a study must first at tain the true vari a tion on both the variables being correlated. Yet we often read published studies whose sam ple is con fined to one in dus try or re gion, which may re strict the range on ei ther X or Y, so that when they find that, con trary to their hypothesis, there is no significant correlation be tween X and Y, this could be a false con clu sion. The gen eral re la tion ship con nect ing X and Y may hold for or ga ni za tions in that in dus try or re gion, but there is not enough vari a tion therein to re veal it. Again, a pro ject may ex am ine the re la tion ship be tween size and bu reau cratic struc tur ing in a set of small or ga ni za tions (e.g., Miller, Drogue, and Toulouse 1988), which con siderably restricts the range in size, so that the resulting estimate of the size-struc tur ing re la tion ship will be sub stan tially at ten u ated, lead ing to an underestimate. In the fit-per for mance re la tion ship, the prob lem of range re stric tion occurs in several forms. It oc curs in the sub group anal y ses dis cussed earlier. These dichotomize the sample into fits and misfits. Therefore some of the organizations in the fit sub group are truly in fit whereas others classed as fits are some way off fit. Similarly, the misfits sub group con tains some or ga ni za tions that are com plete mis fits and some that are just moderate misfits. Therefore, in comparing the mean performances of the two sub groups we are re ally com par ing the per formances of the average fitness of the fit subgroup with the average

230

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

misfitness of the mis fit sub group. The sam ple may have the full range by con tain ing an or ga ni za tion that is com pletely in fit and one that is com pletely in mis fit. Nev er the less, the vari a tion be tween the two subgroups is less than full be cause the av er ages of the two sub groups are to ward the mid dle of the con tin uum that runs from fit to mis fit. Thus the com par i son is anal o gous to com par ing the first and third quartiles, not the bot tom and top of the scale. Hence if the fit group has a mean profit of $15 mil lion and the mis fit group has a mean per for mance of $10 mil lion profit, the ben e fit of fit seems to be $5 mil lion profit or 50 percent. But the difference in per for mance would be greater be tween the or ga ni za tion that is com pletely in fit com pared with the or ga ni zation that is com pletely in mis fit, which is the more ex act mea sure of the ef fect of fit on per for mance. Sim i lar re marks ap ply to re gres sion and correlational meth ods that compare sub groups. As dis cussed above (in Chap ter 7), the com par i son of re gres sion lines is not of complete fit versus com plete misfit but only of a group in fit rel a tive to one in mis fit. There fore the dif fer ences in slope co ef fi cients, con stant terms, and cor re la tions un der state the true ef fect of mis fit so that no sig nif i cant dif fer ence may be ob served, which leads to a false con clu sion of no ef fect of fit on per for mance. This could oc cur even if, in the sam ple, the or ga ni za tions that are in com plete mis fit are very dif fer ent in their po si tion in the two-di men sional contingency-structure space from the organizations that are in com pletefit. There fore de spite the re li abil ity ad van tage of sub group anal yses that were discussed above, they suffer a disadvantage of range re stric tion that leads to an un der es ti mate of the strength of the ef fect of fit on performance. This should be borne in mind when interpreting such anal y ses. There is less range re stric tion where the anal y sis of the re la tion ship be tween fit and per for mance is con ducted at the level of the in di vid ual case. For in stance, the fit of each or ga ni za tion is mea sured and re lated to the per for mance of that or ga ni za tion, with the cases in the sam ple be ing used to com pute the cor re la tion be tween fit and per for mance. If an or ga ni za tion is com pletely in fit or com pletely in mis fit, then this value is en tered di rectly into the anal y sis with out the re duc tion in extreme val ues that oc curs where data are ag gre gated (e.g., sub group av er ages are calculated). Thus anal yses such as cor relations or re gres sions preserve the full range of fit that is in the sam ple. This helps them to

Fit Af fects Per for mance

231

reveal the true correlation. Hence they suffer less from range restric tion than sub group anal y ses. Yet even correlation and regression analyses can suffer from some range re stric tion. To re ally as sess the true strength of the ef fect of fit on per for mance, we need to know the dif fer ent per for mances of cases in com plete fit and com plete mis fit. Yet cases in the real world may not be as ex treme as ei ther com plete fit or mis fit. For this rea son even the best sam ple would have less range than po ten tial vari a tions in fit. Em pir i cal research can study only variations that actually occur. High misfit would be very dam ag ing to or ga ni za tional per for mance, so that or gani za tions in mis fit tend to change their struc ture and move into fit, as the SARFIT model states. Moreover, as March (1999) argues, firms tend to im i tate more suc cess ful com pet i tors, thereby elim i nat ing vari ation in the independent variables (here contingency and structure) that cause per for mance and also per for mance it self. Al ter na tively, misfitted organizations having low performance are consequently less likely to survive; in particular, a firm may become bankrupt and dis band. Thus, while above (in Chap ter 6) we ex pressed res er va tions about eco log i cal se lec tion, any eco log i cal se lec tion against mis fit that oc curs will re duce the ex tremes of mis fit oc cur ring in the pop u la tion (Hannan and Free man 1989; Van de Ven and Drazin 1985). For these rea sons of ad ap ta tion and se lec tion, few ac tual or ga ni za tions may be in high misfit at the time of study, de spite some be ing in high mis fit at other times and mis fit be ing very dam ag ing for per for mance. The em pir i cal studies can only assess the relationship of fit and performance as revealed by vari a tions in fit at the time of the study. If these vari a tions are only mod er ate rel a tive to po ten tial vari a tion, then t he fit-per for mance re la tionship may be only moderate or weak. This might lead analysts, or some one read ing their re sults, to state that fit only weakly af fected or ga ni za tional per for mance. Per haps the ob served cor re la tion is so weak that it is not even sta tis ti cally sig nif i cant, lead ing to the con clu sion that fit has no effect on performance; this in turn would tend to be con strued as a ref u ta tion of the idea that there was such a fit. How ever, the true ef fect of fit on per for mance might be strong so that or ga ni za tions avoid high misfit, thereby restricting the range of fit and leading to misleading em pir i cal re sults. The strength of a con tin gency the ory fit may be assessed by comparing it to some other cause of organizational performance. If the

232

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

correlation of fit and performance is weaker than the correlation between that other cause and per for mance, then the an a lyst may con clude that the con tin gency the ory fit has a weaker ef fect than that other cause. The in fer ence may be drawn that con tin gency the ory fit is un important and can be virtually ignored by management, who should rather con cen trate on us ing the other cause to boost the per for mance of their or ga ni za tion. Yet the contingency the ory fit could ac tu ally be more im por tant than that other cause, so much so that or ga ni za tional management has ad justed its or ga ni za tional struc ture so that there is little misfit. In contrast, they did not work on the other cause yet because that is a weaker de ter mi nant of per for mance, but there fore it has more ac tual vari a tion and shows up as ap par ently a stron ger cause in anal y ses. There is thus a par a dox that or ga ni za tional struc tures that strongly affect or ga ni za tional per for mance may be re stricted in their range, because of adaptation or se lection, and possibly more so than other or ga ni za tional char ac ter is tics that have weaker ef fects on per formance. Liv ing or ga ni za tions are not un think ing ex per i ments to al low so cial sci en tific ob serv ers to make easy in fer ences, but rather are dy namic, adaptive sys tems in which managers seek to avoid loss and increase gain. The propensity of adaptation to eliminate variation in factors affecting per for mance should al ways be borne in mind when in ter preting empirical re sults in or ga ni za tional sci ence. Van de Ven and Drazin (1985) argue that, rather than adopting structures that fit the contingencies, managers may follow universal istic rules that im pose a stan dard struc ture re gard less of the sit u a tion. Again, lo cal man ag ers may be more likely to tai lor their struc ture to its sit u a tion, in con tin gency the ory fash ion, but up per-level man ag ers are more likely to impose a standard structure in universalistic fashion. Where managerial decisions impose structural uniformity there may be fewer fits than would occur if structures were allowed to vary to match their contingencies. Once again, this could reduce the cor re lation between fit and performance, leading to an underestimate of its true strength. Con forming in this way to some gen eral rule, de spite it being counterproductive, is consistent with institutional theory. Therefore, isomorphic ef fects, by re duc ing vari ance of ei ther con tin gency or struc ture, could lead to range re stric tion and hence at ten u a tion of the cor re la tion be tween con tin gency and struc ture. Thus SARFIT, population-ecology, and in stitutional the ories all point to ways in which range restriction could occur (through adaptation, selection,

Fit Af fects Per for mance

233

and con for mity, re spec tively), re duc in g the ob served fit-per for mance relationship. Confounds As is widely un der stood in so cial sci ence, the as sess ment of the true ef fect of any cause can be up set by con founds. Such con found ing vari ables can ob scure the im pact of the cause, or they can lead to spu ri ous effects that appear to be due to a factor but are really due to the con found. Such highly gen eral con sid er ations ap ply also in es ti mat ing the ef fect of fit on per for mance, as dis cussed above. How ever, there are reasons why con founds may be par tic u larly trou ble some when study ing the fit-perf or mance re la tion ship. Con founds of the ef fect of fit on perfor mance are more than just the usual idea that “other things are not equal.” Instead there are liable to be systematic causal forces arising from within con tin gency the ory it self. The vari ables en dog e nous to fit and per for mance can pro duce con founds to the re la tion ship of fit on per for mance, which are ad di tional to the more usual sit u a tion in so cial sci ence of con founds by ex tra ne ous vari ables. There are two main system atic forces that cre ate con founds of the re la tion ship be tween fit and performance. These are the negative effect of per for mance on fit and the spurious ef fects in tro duced through as so ci a tions of the con stit uents of fit with performance. Neg a tive Ef fects of Per for mance on Fit There is reason to be lieve that there is a negative effect of per formance on fit. As we have stressed, contingency the ory re search shows that or ga ni za tions in mis fit tend to adapt into fit only when the over all per for mance of the or ga ni za tion be comes low so that a cri sis of poor performance occurs (Chandler 1962; Donaldson 1987). High per for mance tends to keep an or ga ni za tion that is in mis fit in that state. In contrast, low performance of an organization in misfit tends to trig ger adaptive change so that it moves into fit. Thus the theory of performance-driven organizational change holds that there is a neg ative ef fect of per for mance on fit (as will be dis cussed fur ther in the next chap ter). More over, high per for mance tends to cause an organization in fit to ex pand, for ex am ple, by us ing slack re sources to in crease in size

234

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

or di ver si fi ca tion, so that it moves into mis fit. Hence a neg a tive ef fect of per for mance on fit arises. Thus a more fully spec i fied con tin gency the ory model would be that fit af fects per for mance pos i tively, but also that per for mance af fects fit negatively. Thus a cross-sec tional cor re la tion between fit and per formance is the net re sult of these two op pos ing causal forces. A neg a tive ef fect tends to can cel out the pos i tive ef fect so that the ob served cor rela tion be tween fit and per for mance could be zero. A stron ger neg a tive than positive ef fect would make the ob served cor re la tion between fit and per for mance be come neg a tive. The ob served cor re la tion reg is ters whichever cause is stronger in a particular sample. In this way, the results of stud ies us ing cross-sec tional meth ods to as sess the ef fect of fit on per for mance could be mis lead ing. They could lead to the er ro neous con clu sion that there was no pos i tive ef fect of fit on per for mance, when it was just be ing masked by a neg a tive ef fect of per for mance on fit. Even if this mask ing was in suf fi cient to cause the correlation of fit and per for mance to be come neg a tive or zero, it would de press the magnitude of the positive correlation so that the positive effect of fit on performance was un der es ti mated. In fact, it is not unusual for fit at one point in time to be related to performance averaged over several prior years, so that fit is actually after per for mance (e.g., Child 1974). This means that the cor re la tion between fit and performance is ac tu ally reg is ter ing quite a lot of the effect of performance on fit. If that effect is negative, as held by the theory of per formance-driven or ganizational change, then it will be mask ing the ef fect of fit on per for mance, so that any ob served pos i tive correlation is more likely to be a sub stan tial un der es ti mate. Thus the timing of the measurement of the fit and performance variables can amplify the masking of the positive fit-per for mance re la tion ship by the negative-effect per for mance-fit re la tion ship. There is a be nign im pli ca tion of per for mance hav ing an ef fect on fit, which can help make more cor rect causal in fer ences from cross-sec tional data. Some em pir i cal stud ies of the ef fects of fit on per for mance mea sure fit and performance at the same point in time, so that they are cross-sec tional rather than diachronic. Thus they do not show un am biguously that fit causes performance, be cause a positive association be tween fit and per for mance could log i cally be due to a pos i tive ef fect of per formance on fit. However, as we have just seen, the theory of performance-driven organizational cha nge holds that the effect of

Fit Af fects Per for mance

235

performance on fit is neg a tive. There fore, ac cord ing to this un der ly ing theoretical model, any pos i tive as so ci a tion be tween fit and per for mance can not be caused by an ef fect of per for mance on fit (be cause that would be neg a tive). Hence, pos i tive as so ci a tions be tween fit and per for mance found in cross-sectional studies are not spurious. Thus, to the degree that it is true that per for mance neg a tively ef fects fit, this adds con fi dence to interpreting ob served pos i tive cor re la tions be tween fit and per formance as re flect ing a pos i tive ef fect of fit on per for mance. This rea soning applies also to empirical studies where performance is actually mea sured at a time prior to fit, which will am plify the neg a tive ef fect of per for mance on fit, so that any ob served pos i tive cor re la tion sig ni fies a stron ger un der ly ing pos i tive ef fect of fit on per for mance. In order to control for this systematic con found it is necessary to con trol for the neg a tive ef fect of prior per for mance on fit. This con trol can be achieved in an em pir i cal study by mea sur ing fit be fore per formance, which is lagged sev eral years af ter fit, while also con trol ling for ex ist ing (i.e., prior) per for mance. This helps to cap ture the true ef fect of fit on performance by controlling the negative feedback effect of performance on fit. Another method is to study only organizations with slim profit mar gins, such as in com pet i tive in dus tries, that would suf fer low per for mance and adapt promptly when they came into misfit. In this case, the values of the other causes of performance are not cor re lated with fit in such a way as to mask the ef fect of fit on per for mance. Such or ga ni za tions lack the slack that would al low them to be in mis fit with high per for mance, the con di tion that pro duces the neg a tive association between performance and fit that con founds the pos i tive ef fect of fit on per for mance. Notwithstanding the control on prior performance, there is still a need to con trol di rectly for other causes of or ga ni za tional per for mance when estimating the effect of fit on perf ormance. As we have argued, con trol ling for prior per for mance turns off the con found ing, neg a tive feed back ef fect from per for mance to fit, yield ing a truer es ti mate of the effect of fit on per for mance. This con trol for prior per for mance also con trols for those ef fects of other causes on or ga ni za tional per for mance. However, prior performance is performance several years earlier, which re flects the ef fects of the other causes of per for mance at an even ear lier time. These other causes of or ga ni za tional per for mance could, however, af fect organizational per formance within the past several years. If any of these causes were themselves cor re lated with fit, then

236

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

they would con found the re la tion ship be tween fit and per for mance. In or der to con trol for these pos si ble con founds, it is nec es sary to con trol for these other causes of performance as well as to control for prior performance. This con trol can be achieved by the meth ods men tioned above (e.g., in clu sion of con found ing vari ables in sta tis ti cal anal y ses). The the ory of per for mance-driven or ga ni za tional change is an ex am ple of an aux il iary sci ence. It is a body of the ory that of fers ex pla na tions of its own phenomenon, that is, how per for mance drives or ga ni za tional change. Yet it also sheds light on the phenomenon con sid ered in this chapter, that is, the empirical study of how fit affects performance. Thus, in the con text of the study of the ef fect of fit on per for mance, the the ory of per for mance plays an aux il iary role (Feyerabend 1975). Constituents of Fit Contingency the ory is pe cu liarly prone to an other type of con found. As we are aware, misfit is a dis crep ancy score be tween the level of the organizational structure re quired by the contingency and the actual level of the organizational structure. The contingency and structural vari ables are the con stit u ents of mis fit. There is a defi ni tional con nection between the discrepancy and its constituents that can lead to a correlation be tween them (see Co hen and Co hen 1983; Cronbach and Furby 1970; Ed wards and Parry 1993; Johns 1981; Wall and Payne 1973; Werts and Linn 1970). Hence dis crep ancy scores tend to have a built-in correlation with each of their constituent vari ables. Where these con stituents hap pen to be, in turn, cor re lated with the de pend ent vari able, an ap par ent ef fect of the dis crep ancy score on the de pend ent vari able can be spu rious. Similarly, the constituent variables can have a re lation ship with per for mance that is the op po site of that be tween fit and performance. Then the constituent variables mask the effect of fit on performance, mak ing its ef fect seem smaller than it is, or nil, or even neg a tive. In these cases, the ap par ent lack of a sig nif i cant pos i tive re lation ship be tween fit and per for mance may be false, due to its con stit uent variables. The rea son why dis crep ancy scores tend to have a built-in cor re la tion with their constituent variables is because of range restriction. Con sider again the sim ple model where the fit of the or ga ni za tional struc ture to its contingency is a line in which, mathematically, structure

Fit Af fects Per for mance

237

equals con tin gency. Struc ture is the ver ti cal axis and con tin gency is the hor i zon tal axis, and the fit line is 45 de grees to the hor i zon tal axis. If all mis fits are be low the fit line, then large mis fit is pos si ble only where con tin gency is large; where con tin gency is small the mis fit must also be small. Thus the level of the con tin gency vari able de ter mines the max imum value of mis fit, and so mis fit is pos i tively cor re lated with con tingency. By similar reasoning, large val ues of mis fit are pos si ble only if structure is small, so misfit is negatively correlated with structure. These relationships follow from the way that the space of possible misfits has been defined as a triangular area below the fit line. Thus these cor re la tions be tween mis fit and its con stit u ents are artifactual in that they arise be cause fit and its con stit u ents are not vari ables that are in de pend ent of each other, tak ing any value what ever the other; rather their possible val ues are linked by def i ni tion. Be cause of such artifactual cor re la tions be tween mis fit and its constituents, if either constituent has a relationship with performance, then it could lead to a spu ri ous re la tion ship be tween mis fit and per formance. For in stance, given that con tin gency is artifactually pos i tively cor re lated with mis fit, if con tin gency also has a neg a tive ef fect on performance, then there will be a spu ri ous neg a tive effect of misfit and per for mance. Thus an ob served neg a tive ef fect of mis fit on per for mance could be wholly due to the neg a tive ef fect of con tin gency on per for mance. In that case, in fer ring that mis fit causes lower per for mance would be er ro ne ous be cause the real ef fect on per for mance is by the con tin gency. The true ef fect is uni ver sal ist ic, in that higher val ues of the con tin gency variable cause less performance, not contingency misfit. Similarly, given that struc ture is artifactually neg a tively cor re lated with mis fit, if structure also has a positive effect on performance, then this would produce a spurious, negative effect of mis fit on performance. Both con tin gency and struc ture could each con trib ute to a spu ri ous, neg a tive ef fect of mis fit on per for mance, thereby mak ing it more likely to oc cur. Fur ther more, if con tin gency has a pos i tive ef fect on per for mance and structure has a negative effect, then both con tin gency and structure would produce spu rious positive effects of misfit on performance. These could mask any true negative effect of misfit on performance, leading to erroneous inferences, such as that misfit had a non significantly negative effect on performance, or nil effect, or even a positive effect. In order to avoid these confounds, it is desirable to

238

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

control both the con tin gency and or ga ni za tional struc ture vari ables by including them as con trol vari ables in anal y ses that test for an ef fect of fit on performance. The artifactual cor re la tions be tween con tin gency and struc ture will not always arise and depend upon the distribution of or ganizations around the fit line. If all the mis fits oc curred above the fit line, then this also creates artifactual or in-built correlations with contingency and structure that can confound the fit-performance relationship. How ever, these artifactual correlations are of reverse signs to those where the mis fits are all be low the fit line. Thus, if mis fits are scat tered equally above and be low the fit line, then they off set each other and there is no artifactual correlation between fit and contingency and structure. In such a sit u a tion, even if ei ther con tin gency or struc ture has an ef fect on performance, the problem described of spurious effects of misfit on performance does not arise. An inspection of the scatter of misfits around the fit line should in form the anal y sis. Hence whether the con stituents of fit spuriously affect the fit-performance relationship is determined by the scat ter of data around the fit line and whether ei ther of the constituents is cor re lated with per for mance. The safe course is rou tinely to in clude both con tin gency and or ga ni za tional struc ture as con trols in a multivariate anal y sis when test ing for the ef fect of fit on performance. Given that either con tingency or or ganizational structure can be definitionally cor re lated with mis fit, there is a pos si bil ity that any claimed ef fect of fit on per for mance could re ally be due to a uni ver sal ist ic ef fect of con tin gency or struc ture on per for mance. Con trolling for con tin gency and struc ture in a multivariate anal y sis of the ef fect of fit has the ad van tage of show ing that the ef fect of fit on per for mance is a con tin gency re la tion ship, and is not re ally due to a uni ver sal ist ic re la tion ship of the con tin gency or struc ture on per for mance. While we have dis cussed the prob lem of causal ef fects of con stit u ents of fit on per for mance con found ing the ef fects of mis fit on per for mance, the same holds for other as so ci a tions be tween mis fit and per for mance (e.g., cor re la tions). If there is a cor re la tion be tween per for mance and one of the con stit u ents of mis fit, then, given an artifactual cor re la tion be tween that con stit u ent and mis fit, this can cre ate a spu ri ous cor re lation between misfit and performance. Such a cor relation between a con stit u ent and per for mance can arise for rea sons other than an ef fect of the constituent on per formance (as discussed so far): an effect of

Fit Af fects Per for mance

239

per for mance on the con stit u ent, or just by an as so ci a tion be tween the constituent and performance. Thus the potentiality for confounding con stit u ents of mis fit is rather broader in the case of correlational analy sis. Again, how ever, correlational anal y sis may be pro tected from this con found by in clud ing both the con stit u ents as con trol vari ables. The study by Keller (1994) pro vides an ex am ple of how fit mea sures are correlated with their constituent variables. Fit of information processing (i.e., struc ture) to nonroutineness (i.e., the tech nol ogy con tin gency) is pos i tively cor re lated with each of its con stit u ent vari ables: information pro cess ing (.57) and nonroutineness (.52). In turn, both information processing and nonroutineness are positively correlated with performance (project quality), thereby introducing a degree of spu ri ous cor re la tion be tween fit and per for mance. This is dealt with by con trol ling for both con stit u ent vari ables in the anal y sis of the ef fect of fit on performance, by including both in the multivariate analysis (Keller 1994). In sum, con founds are li a ble to oc cur quite sys tem at i cally in anal y ses of the effects of fit on performance. Confounds can be controlled by entering potential confounding variables in the statistical analysis. Such po ten tial con found ing vari ables in clude prior per for mance and also the con stit u ent vari ables of fit. An al ter nate pro ce dure to con trol ling for prior performance in the analysis is to sam ple or ga ni za tions with lit tle slack (such as in com pet i tive in dus tries). Ad di tional con trols for the other causes of per for mance should also be made. Over all, the dis cus sion has brought out a num ber of meth od olog i cal prob lems. It is pos si ble that em pir i cal re sults will not al ways be good es ti mates of the un der ly ing re la tion ships. In par tic u lar, the mag ni tude of the ef fect of fit on per for mance is li a ble to be un der es ti mated in empir i cal stud ies. It is hoped that the an a lyst will make better in ter pre tations in the light of the considerations dis cussed here.

Lessons for Future Research Be cause of the cen tral ity of fit to con tin gency the ory, em pir i cal val i dation that fit leads to higher per for mance is im por tant. As we have seen, how ever, stud ies of the ef fect of fit on per for mance are prone to a number of meth od olog i cal prob lems that can ob scure the true ef fect, leading to un der es ti mates or to spu ri ously nil or neg a tive find ings. Below

240

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

we ab stract eight les sons from the pre ced ing dis cus sion. Drawing upon these les sons can re duce these meth od olog i cal prob lems and im prove the ac cu racy of es ti mates of the ef fect of fit on per for mance. Les son 1: Use mul ti ple fits As seen, there are nu mer ous fits be tween var i ous con tin gen cies and organization structural variables. Therefore fit is more completely captured if the de gree of mis fit of each is mea sured to yield the to tal misfit. If only one aspect of fit is studied, then fit is in a sense underspecified and this will makes its effect understated. Moreover, mul ti ple fits pro vide a multi-item mea sure of fit that is thereby more re li able than a sin gle item. The un re li abil ity of fit mea sures is a prob lem, and the con struc tion of multi-item mea sures helps im prove re li abil ity. Les son 2: Use re li able mea sures of per for mance Unreliability of the per for mance vari able also low ers the cor re la tion be tween mis fit and per for mance. Per for mance can be mea sured more re li ably by avoid ing vari ables such as as sets that are prone to ar bi trary val u a tions and id io syn cratic changes. Changes in per for mance, such as changes in profitability, also suffer from unreliability because their numerator is a change score (e.g., profit this year mi nus profit last year) and so would have lower reliability than the constituent per for mance vari ables (e.g., profit). Thus mea sures should be used that are re li able, such as sales and profit or the ra tio of profit to sales. More over, multiitem scales improve reliability and this could be done by combining performance measures. This is best done after establishing that they form a com mon fac tor in fac tor anal y sis so that the scale has con struct validity. Les son 3: In con trast ing the ef fects on per for mance of fits and mis fits, use or ga ni za tions that have re mained as ei ther fits or mis fits for several years This is in or der for the ef fects on per for mance of fit to cu mu late into sub stan tial vari a tions in per for mance be tween the fits and the mis fits, which will be clearly visible dif fer ences and show the true strength of the effect.

Fit Af fects Per for mance

241

Les son 4: Use a diachronic re search de sign in which fit pre cedes per for mance This adds confidence to the causal inference that fit is a cause and that per for mance is its ef fect. Les son 5: Con trol for prior per for mance In the anal y sis of the ef fect of fit on sub se quent per for mance, in clude per for mance as a con trol vari able. More specifically, use the per for mance for the same time as fit as a con trol vari able. This switches off the neg a tive feed back ef fect from per for mance to fit, thereby pre vent ing a ma jor po ten tial con found. Les son 6: Use the con stit u ent vari ables of fit, that is, the con tin gency and or ga ni za tional struc tural vari ables, as con trol vari ables in the anal y sis of the ef fect of fit on per for mance This is to rule out any con found ing of the re la tion ship be tween fit and per for mance that is ac tu ally due to the pos si ble defi ni tional con nection of fit and its constituent vari ables. This con trol also dem on strates that the ef fect of fit on per for mance is in de pend ent of any main ef fect on per for mance of ei ther the con tin gency or the or ga ni za tional struc tural vari ables, in clud ing uni ver sal ist ic ef fects. Les son 7: Con trol for other causes of per for mance This is the stan dard meth od olog i cal pro ce dure to en sure that cau sality at trib uted to fit in its re la tion ship with per for mance is valid and not spurious. It also helps to control for any negative effects of per formance on fit not con trolled for by con trol ling for prior per for mance. Les son 8: Cor rect ob served cor re la tions for un re li abil ity of mea sure ment Be cause mis fit is a dif fer ence score, it is much less re li able than the organizational struc ture and con tin gency vari ables that con sti tute it. The for mula for cor rec tion of un re li abil ity should be used to ob tain a more ac cu rate es ti mate of the true pos i tive cor re la tion be tween fit and

242

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

performance (Hunter, Schmidt, and Jackson 1982). However, if con founds are not controlled, then spu ri ous neg a tive cor re la tions may be ob tained and cor rect ing them for un re li abil ity will sim ply in flate the negative correlations. This shows the im por tance of con trol ling confounds before ap ply ing the for mu las for cor rec tion of un re li abil ity. The best es ti mates of the fit-per for mance re la tion ship will be ob tained by ap ply ing all the les sons in the same re search pro ject. They should be ap plied in the fol low ing or der: • measure fit comprehensively and over sev eral years, and fit and sub sequent performance as re li ably as pos si ble

• then con trol in the sta tis ti cal anal y sis of the ef fect of fit on per for mance for prior per for mance, fit con stit u ents, and other causes of per for mance

• then cor rect for un re li abil ity

Applying these lessons will tend to increase the magnitude of the resulting re la tion ship be tween fit and per for mance, better re veal ing its true strength.

Conclusions Overall, empirical studies show that fit positively affects performance, thereby supporting the central idea of contingency theory. Whereas two early stud ies of the re la tion ship be tween fit and per for mance failed to find it, many sub se quent stud ies have done so. The dis cus sion of in di vid ual stud ies has brought out some of the meth od olog i cal im provements that al low the ef fect of fit on per for mance to be seen. The empirical estimation of the effect of fit on performance is af fected by un re li abil ity, range re stric tion, and con founds. These stan dard methodological considerations have particular meanings here. They tend to make the em pir i cal es ti mates un der state the strength of the effect of fit on per for mance. Fit is prone to un re li abil ity be cause it is a difference score. Similarly, performance may be measured unreliably. The un re li abil ity of both fit and performance makes the correlation between them lower than their true re la tion ship. Again, range will be re stricted where or ga ni za tions are ag gre gated, such as into sub groups of fits and mis fits. Also, range becomes fur ther restricted through

Fit Af fects Per for mance

243

or ganizational ad ap ta tion, se lec tion, and con for mity. These sources of range restriction lower the correlation between fit and performance. Confounds of the fit-performance relationship arise in a systematic way, in that per for mance feeds back to mask it. Fur ther, con founds can also arise through the defi ni tional con nec tion be tween fit and its con stit u ent vari ables, which may mask the fit-per for mance re la tion ship or cre ate it spu ri ously. To the ex tent that past (and some fu ture) re search will be af flicted with some of these prob lems, this needs to be borne in mind in in ter pret ing re sults. Eight lessons have been iden ti fied here for avoid ing meth od olog i cal problems. Their use in future contingency research should help to ascertain truer estimates of the effects of fit on performance. These methodological improvements constitute the development of auxil iary tech niques so that the full ex plan a tory power of con tin gency the ory can be re vealed as its re search pro gram pro gresses.

THE CONtin Neo-Con TIN gency GENCY Theory THEORY OF OR GA NI ZA TIONS

9

Neo-Contingency Theory

S

o far we have re mained within the tra di tional frame work of con tingency the ory re search. How ever, there are cer tain prob lems with the tra di tional view. In this chap ter we iden tify these prob lems and sug gest how to re solve each of them. This leads to a re vised form of con tin gency the ory that may be termed neo-con tin gency the ory. Contingency theory is traditionally concerned with organizational performance and in that broad sense is consistent with economics. How ever, con tin gency the ory has mostly not drawn much on eco nomics and tends to re main iso lated from it. Some at tempts have been made to import elements of economics into organizational theory more gen erally, such as agency theory (Jensen and Meckling 1976) and transaction costs economics (Wil liam son 1975). How ever, these have proven con tro ver sial (Bar ney 1990; Donaldson 1990a, 1990b; Ghoshal and Moran 1996; Moran and Ghoshal 1996; Wil liam son 1996). Moreover, they deploy only a narrow range of economic con cepts. In our reformulation of con tin gency the ory we shall draw on eco nom ics and the related dis ci pline of fi nance, at nu mer ous points, to gain con ceptual in sights. In so do ing, we will uti lize as pects of eco nom ics dif fer ent from those used to date in or ga ni za tional the ory. We hope that in this 245

246

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

way we may break some new ground and encourage an infusion of ideas from eco nom ics and fi nance into or ga ni za tional the ory.

Problems in Traditional Contingency Theory In the pres ent book and else where (Donaldson 1985, 1995a, 1996a) we have defended contingency theory from criticism made against it. However, there are three problems that exist within contingency theory, and these have not been widely can vassed to date. As we have seen, traditional con tin gency the ory holds that an or ganization in misfit will change its structure to move into fit with its contingency (e.g., Burns and Stalker 1961). How ever, the or ga ni za tion is in mis fit be cause of a prior change in the level of a con tin gency vari able, such as task un cer tainty, so that the or ga ni za tion moved out of fit. The ques tion arises as to why the or ga ni za tion changed its con tin gency. This is not ex plained by con tin gency the ory. Such a move seems problem atic be cause it cre ates mis fit that low ers or ga ni za tional per for mance. Thus there is a need to ex tend con tin gency the ory to ex plain changes in the contingency. Con tin gency the ory holds that an or ga ni za tion in mis fit will typ i cally, even tu ally move into fit. The ques tion arises as to how man ag ers know what the fit is. If man age ment is un sure about what struc tures fit their contingencies, as is quite plausible, then they cannot move decisively into fit. Thus we again need to ex tend con tin gency the ory, to pro vide a more realistic ac count of how change in the organizational structure occurs so that the or ga ni za tion moves to ward fit. As we saw in Chap ter 7, con tin gency the ory holds that the fit line is a line of iso-per for mance so that each fit pro duces as high per for mance as any other fit (Van de Ven and Drazin 1985). The ques tion arises as to why or ga ni za tions move from one fit to an other if there is no per for mance gain from so do ing. We need to make re vi sions to the con cept of the line of fit be ing a line of iso-per for mance. These are the three problems with traditional con tingency theory that this chap ter will ad dress. Their res o lu tion will lead to the re for mula tion of con tin gency the ory into neo-con tin gency the ory. While re taining some of the build ing blocks of con tin gency the ory, we will replace some of the traditional assumptions in order to produce what is, we hope, an im proved the ory.

Neo-Con tin gency The ory

247

Disequilibrium and Ongoing Change Traditional contingency the ory is a theory of equilibrium in that it de scribes or ga ni za tional change as a pro cess of re gain ing equi lib rium (Burns and Stalker 1961). An or ga ni za tion in mis fit moves into fit, but is then as sumed to stay there. There is no the ory of why the or ga ni za tion moves out of fit into mis fit in the first place. The move into mis fit by chang ing the con tin gency, such as task un cer tainty or tech nol ogy or size, is just a given. Contingency theory provides a motivation for an or ga ni za tion in mis fit to move into fit: to gain the higher per for mance that fit pro duces (Burns and Stalker 1961; Wood ward 1965). How ever, when the organization moves out into misfit it consequently loses performance, so that there is no incentive for it to move out of fit. There fore fit is an equi lib rium be cause once at tained, the ten dency of the organization would be to stay there. The equilibrium aspect of contingency theory is consistent with it being a functionalist type of the ory. It is con sis tent also with con tin gency the ory be ing a cy ber netic type of the ory (Hage 1974) that is fo cused on def i cit re duc tion, that is, re duc ing mis fit. Other types of the o ries in or ga ni za tional be hav ior explain how def i cits are cre ated; for in stance, goal-set ting the ory (Locke and Latham 1990; Wood and Locke 1990) ex plains that peo ple set new performance targets that exceed their present performance, thereby cre at ing a “gap” or def i cit, which is anal o gous to a mis fit. Such def i citcre at ing the o ries at trib ute to the in di vid ual or or ga ni za tion the qual ity of being proactive, thereby creating a tension. In contrast, deficit re duction theories, such as structural contingency theory, attribute to the individual or or ga ni za tion the qual ity of be ing re ac tive, so re moving an existing irritant. Given that in contingency theory the deficit, mis fit, is re moved by re gain ing fit, the ques tion be comes how mis fit is attained, that is, how and why an organization moves into misfit, thereby cre at ing a def i cit. Why would an or ga ni za tion move from the com fort of fit to the dis com fort of mis fit? The or ga ni za tion moves out of fit be cause of a change in the level of one or more of its con tin gen cies, such as an in crease in size or di ver si fication. What would cause an increase in a contingency? Increases in many con tin gen cies re quire in creases in re sources, for ex am ple, ad ditional financial re sources to hire new employees, who constitute an increase in size. An increase in the degree of technological ad vance (Wood ward 1965) re quires sur plus re sources to pay for the new cap i tal

248

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

equip ment and em ployee train ing. Di ver si fi ca tion in volves ac quir ing a new busi ness or in ter nally gen er at ing new prod ucts or ser vices (Rumelt 1974), so that diversification re quires surplus resources to acquire firms or to pay in ter nal de vel op ment costs (Chan dler 1962). In creases in the innovation rate, with its ac com pa ny ing in crease in task un certainty, re quires sur plus re sources to pay for re search and de vel op ment, in clud ing for the hir ing of pro fes sion ally qual i fied tech nol o gists. Thus in creases in the con tin gen cies of size, tech nol ogy, di ver si fi ca tion, in nova tion, and task un cer tainty all re quire sur plus re sources to fund them. Sur plus re sources and wealth are often gen er ated by the or ga ni za tion per form ing at a high level, such as by a firm pro duc ing high prof its. An or ga ni za tion is more likely to per form highly if it is in fit. There fore, an or ga ni za tion in fit that ex pe ri ences high per for mance may in con se quence in crease its con tin gen cies. More spe cif i cally, the greater the per for mance of an or ga ni za tion in fit, the higher the rate at which its con tin gen cies increase. As a result of in creases in the levels of its contingencies, an organization in fit moves into misfit. Thus high performance feeds back to cause an organization to move from fit into misfit. The high performance of such an organization is partly caused by its fit and partly by other causes (e.g., an up swing in the busi ness cy cle). Thus there is a nat u ral ten dency for or ga ni za tions in fit to move into mis fit. The the ory is no lon ger an equi lib rium the ory of chang ing and then stay ing put. The the ory is now one of dis equi lib rium. It pre dicts that or ga ni za tions in mis fit will move into fit and also that organiza tions in fit will move into mis fit. In this way it be comes sym met ri cal. It be comes sym met ri cal also in regard to the feed back ef fects of per formance. As we have seen, tra di tional con tin gency the ory holds that low per for mance feeds back to cause ad ap ta tion by an or ga ni za tion so that it moves from mis fit into fit (Chan dler 1962; Donaldson 1987). There is no consideration in traditional contingency theory of a feedback effect of high per for mance. In con trast, our re vised for mu la tion states that there is a feed back ef fect of high per for mance, caus ing in creases in the contingencies, such as size. The whole theory be comes more dynamic, for now change in one thing leads to change in the other, which feeds back to cause further change in the first thing, causing recurrent change. Specifically, by adapting its struc ture to its contingencies, the organization moves from mis fit into fit, which raises its per for mance, which feeds back to

Neo-Con tin gency The ory

249

in crease its con tin gen cies that cre ate mis fit again and so on. Thus the or ga ni za tion can cy cle re cur rently through these causal paths, mov ing be tween mis fit and fit. In so do ing the or ga ni za tion al ter nates in in crements of increase in the contingencies, such as size growth, and in increments of change in its struc tures, such as in creases in its bu reau cratic struc tur ing. By recurrently making incremental changes in its contingencies and struc tures, the or ga ni za tion can grad u ally change, grow ing from one form to an other. For ex am ple, the or ga ni za tion can grow from small to large, or un di ver si fied to highly di ver si fied, or lo cal to mul ti na tional. Thus the the ory can ex plain the in cre men tal change that many or ga ni za tions dis play (as seen in Chap ter 5; see also Donaldson 1996a, pp. 120-122). Further, some organizations will have gone through the cy cle more times than oth ers, which ac counts for the wide vari a tions in or ga ni za tional con tin gen cies (e.g., size and strat egy) and also struc tures, which are ob serv able in pop u la tions of or ga ni za tions (Donaldson 1996a; Hannan, Ranger-Moore, and Banaszak-Holl 1990). Thus the new the ory is more com plete, more sym met ri cal, and accounts for a change in a con tin gency and for the pos si bil ity of re peated changes in contingencies and struc tures that fun da men tally al ter the or ga ni za tion over time.

Performance-Driven Organizational Change We are now in a position to discuss more systematically the idea that per for mance drives or ga ni za tional change. As we have seen, tra di tionally con tin gency the ory held that a cri sis of poor per for mance was required to trig ger needed adap tive change (Chan dler 1962; Donaldson 1987). This is con sis tent with the the ory of Si mon (1976), that man a gerial decision making is boundedly rational, so that organizations do not maximize their performance, but satisfice. Managerial decision mak ing is prob lem solv ing that is ac ti vated when or ga ni za tional per formance falls be low the level deemed to be sat is fac tory (i.e., satisficing). Thus low per for mance of an or ga ni za tion whose struc ture mis fits its con tin gency leads to a change in the or ga ni za tional struc ture to bring it into fit. In ad di tion, we have just ar gued that high per for mance feeds back to increase the contingency variables. Hence low performance tends to cause an increase in the organizational structure, and high

250

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

per for mance tends to cause an in crease in the con tin gency. These are the central ideas of the theory of performance-driven change (see Donaldson 1999 for a fuller version). If organizational per formance drives organizational change, then this ar gues that to un der stand or ga ni za tional change we have to at tend to the causes of organizational per formance. As we have seen, misfit leads toward lower performance and fit leads toward higher per formance. However, organizational performance is affected by many causes other than fit. Some of these causes are in ter nal to the or ga ni zation, such as di ver si fi ca tion and debt. Other causes are ex ter nal to the organization, such as the business cycle and competition. The low performance that leads to adap tive change in the or ga ni za tional structure is caused by the ef fect of mis fit be ing re in forced by other causes depress ing or ga ni za tional per for mance. Sim i larly, the high per for mance that leads to an in crease in the con tin gen cies, such as growth in or ga nizational size, is caused by the effect of fit being reinforced by other causes rais ing or ga ni za tional per for mance. Thus to un der stand or gani za tional change we need to an a lyze the con junc tion be tween fit and the other causes of performance. Port fo lio The ory In or der to an a lyze the way mul ti ple causes in ter act to de ter mine the overall per formance of an organization, we can make use of the risk con cept and port fo lio the ory from fi nance. Or ga ni za tional risk is the vari a tion over time in the level of or ga ni zational per for mance. A cause of or ga ni za tional per for mance that var ies over time will thereby cause fluctuations in organizational per formance, that is, will cause organizational risk. For organizational change to oc cur, the causes of or ga ni za tional per for mance need to act to gether in such a way as to pro duce nontrivial or ga ni za tional risk. An or ga ni za tion with widely fluc tu at ing per for mance, when it is in mis fit, will ex pe ri ence the low per for mance needed to trig ger adap tive change in the or ga ni za tional struc ture. The or ga ni za tion will also ex pe ri ence the high per for mance, when it is in fit, that leads to rapid growth in the con tin gen cies (e.g., size). Thus to have re peated ep i sodes of growth and adaptation, the or ga ni za tion needs to have re cur rent swings of its per for mance from high to low and back again. Such re cur rent in cre ments in each of size and or ga ni za tional struc ture pro duce long-term growth

Neo-Con tin gency The ory

251

and suc cess of the or ga ni za tion. Thus, for an or ga ni za tion to de velop over time from small to large and suc cess ful, it needs to have fluc tu a tions in its performance. There fore the op ti mal or ga ni za tional risk is not zero, because that would forestall organizational ad aptation and growth. Port fo lio the orystates that the risk of a port fo lio is af fected by the risk of each el e ment and also by the cor re la tion among the el e ments. If two el e ments have high risk but are neg a tively cor re lated (or even are only weakly, pos i tively cor re lated), then they pro duce lower risk in the portfolio than in each of the elements (Brealey and Myers 1996). The up ward fluc tu a tion in one el e ment is off set by the down ward fluc tu a tion in the other, so that the fluc tu a tion in the over all port fo lio is less than in either element. Or ganizational portfolio theory applies the portfolio theory from finance to the organization, so that the organization is con sid ered to be a port fo lio com posed of the causes of over all or ga nizational performance. Each cause of organizational performance is a fac tor in the or ga ni za tional port fo lio. Each fac tor has a cer tain degree of risk (i.e., vari a tion over time) and cor re la tion with the other factors. Overall organizational per formance is thus determined by the risks of the portfolio factors and their correlations. If one cause of organizational per formance is negatively correlated with a second, then the fluc tu a tion in over all per for mance is re duced. Thus the portfo lio of causes of per for mance de ter mines the fluc tu a tions in or ga ni zational per for mance that are nec es sary for or ga ni za tional change. If the causes are high risk and are sub stan tially, pos i tively cor re lated, then the organization will have large fluctuations in its performance and will change frequently, recurrently adapting and growing to become suc cess ful over the long term. How ever, if the causes are low risk or neg a tively correlated, then the or ga ni za tion will have small fluc tu a tions in its performance and will not change, lead ing to stag na tion. These are the key ideas of organizational portfolio theory (Donaldson 1999). This the ory com ple ments con tin gency the ory by giv ing a more ar tic ulated treat ment of the way per for mance me di ates changes in struc ture and con tin gen cies. Organizational portfolio the ory identifies eight portfo lio factors that are major causes of organizational performance and thereby of organizational change. Four factors tend to promote or ganizational change: the busi ness cy cle, com pe ti tion, debt, and di vi sional risk. They work by either in creas ing or ga ni za tional risk (the business cycle and

252

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

divisional risk) or decreasing the organizational perf ormance level (competition and debt). Four fac tors tend to fore stall or ga ni za tional change: diversification, divisionalization, divestment, and directors. They work by either decreasing organizational risk (diversification, divisionalization, and directors) or increasing the organizational performance level (di vest ment). These eight factors in ter act with or ganizational maladaptation or adaptation to determine whether a maladapted organization makes adaptive change and whether an adapted or ga ni za tion grows at a high rate. Fac tors That Pro mote Or ga ni za tional Change Organizational change is promoted where the depressing effect of mis fit on or ga ni za tional per for mance is re in forced by other causes of per for mance also de press ing it. Some or ga ni za tional port fo lio fac tors, competition and debt, work directly on the level of performance, lowering it. Other portfolio factors, the business cycle and divisional risk, are causes of per for mance that fluc tu ate highly and so con trib ute to or ga ni za tional risk, thereby in duc ing episodes of low performance that pro mote change when the or ga ni za tion is in mis fit. The Busi ness Cy cle The busi ness cy cle is a suc ces sion of changes in the level of ac tiv ity in the econ omy, so that eco nomic ac tiv ity fluc tu ates up and down re currently over the years (Kuczynski 1986). It strongly af fects the sales and profitability of firms in the economy (Dotsey and King 1987). The upswing of the busi ness cy cle leads to growth in firms, so that a firm in fit will in crease its size so that its new size co mes to mis fit its ex ist ing organizational struc ture. While the mis fit works to re duce firm per for mance, the buoy ant con di tion of the econ omy more than off sets the mis fit so that the firm con tin ues to pros per. At this point, the ef fect on performance of the business cycle is negatively correlated with the effect on per for mance of the mis fit, thereby coun ter ing it and pro duc ing no or ga ni za tional adap tive change, de spite change be ing needed. Af ter the business cycle turns down, however, the depressive economic conditions reinforce the depressive effect of misfit on performance. When or ga ni za tion per for mance be comes low enough, adap tive or ganizational change occurs, and the organization adopts the required

Neo-Con tin gency The ory

253

new struc ture (Chan dler 1962). At this point, the ef fect of the busi ness cy cle is pos i tively cor re lated with the ef fect of mis fit, thereby re in forc ing it and pro duc ing the needed adap tive or ga ni za tional change. Thus the business cy cle promotes organizational growth on its upswing and adaptive change in organizational structure on its downswing. It moves the or ga ni za tion into mis fit on the up swing and into fit on the down swing. Hence the fluc tu a tions in the econ omy drive fluc tu a tions in or ga ni za tional per for mance that in turn drives or ga ni za tional change. Some in dus tries have deeper busi ness cy cles than oth ers and so would be pre dicted to have more or ga ni za tional changes in their firms (for a fuller set of the o ret i cal prop o si tions about the busi ness cy cle and or gani za tional change, see Donaldson 1999). Competition Competition also pro motes adap tive change in or ga ni za tional structures. The greater the competition, the lower the profit mar gin of the firm (Samuelson 1980) and the more likely that mis fit will force a cri sis of poor per for mance and trig ger adap tive change. Com pe ti tion is affected by the num ber of other or ga ni za tions in that in dus try, which in turn is affected by globalization (Dunning, Kogut, and Blomström 1990). How ever, com pe ti tion is also af fected by the qual ity of the competitors. O ther things being equal, a firm in misfit will suffer loss of sales to competitors if those competitors are themselves in fit, be cause they are better or ga nized, but not if the com pet i tors are them selves in misfit. For example, a firm in mis fit, hav ing di ver si fied but re tained a functional struc ture, that has com pet i tors that are in fit, will suf fer low performance and so re gain fit by division alizing. Whereas if its com pet i tors are in mis fit, also hav ing di ver si fied yet retaining a func tional struc ture, the firm would not suf fer low per for mance and so would not divisionalize. This introduces a relativity perspective into analyses of the effect of fit on performance and organizational change that has been lack ing from the con tin gency the ory literature. Debt Fi nan cial strat e gies also play a role in organizational change. Debt re duces the sur plus profit, thereby low er ing the per for mance level, so that a cri sis of poor per for mance has a higher prob a bil ity of oc cur ring.

254

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

Therefore an organization in misfit is more likely to make adaptive change the more highly it is indebted, because its performance has a higher chance of becoming low. Therefore debt must be added to the port fo lio fac tors of the busi ness cy cle and com pe ti tion as fac tors pro moting adap tive or ga ni za tional change. These three factors are at the cor po rate level, that is, the level of the over all or ga ni za tion. How ever, there is an other or ga ni za tional port fo lio fac tor that works to pro mote or ga ni za tional change, and it is at the di vi sional level: di vi sional risk. Divisional Risk The risk of a di vi sion is the fluc tu a tion in its per for mance; this af fects divisional change in a way that parallels how organizational per formance affects organizational change. Low performance of a division that has a misfit between its structures and its contingencies leads to adoption of the required structures that fit the contingencies. Con versely, high performance of a division in fit raises its rate of growth and increases its contingencies, leading the division into misfit. The higher the risk of a di vi sion, the more its per for mance fluc tu ates, cre ating adaptation, growth, and the long-term success of that division. Such pos i tive de vel op ments for a di vi sion con trib ute to the suc cess of the corporation as a whole. Therefore di vi sional risk helps cor po rate success. In sum, the busi ness cy cle, com pe ti tion, debt, and di vi sional risk all work to in crease or ga ni za tional change. Fac tors That Fore stall Or ga ni za tional Change There are, however, four organizational port folio factors that work to prevent or delay organizational change. These are diversification, divisionalization, di vest ment, and di rec tors. Di ver si fi ca tion and Divisionalization It is a fa mil iar idea in fi nance that di ver si fi ca tion re duces corporate risk (Salter and Weinhold 1979). The rea son is that di ver si fi ca tion spreads the corporation across dif fer ent in dus tries so that their dif fer ent cy cles off set each other. This leads to more sta ble prof its for the corporation than for its di vi sions. Di ver si fi ca tion leads to division alization (Rumelt

Neo-Con tin gency The ory

255

1974), and this enhances the dampening effect of diversification on corporate risk, because the divisional struc ture re duces risk (Buhner and Möller 1985; Hoskisson 1987). The au ton omy of di vi sions al lows their profits to fluc tu ate in de pend ently of each other and fully reflect the di ver gence of their in dus trial cy cles. More over, di vi sional heads are held accountable for the profitability of their divisions (Williamson 1970), and this leads them to manipulate their reported profits, smooth ing out fluc tu a tions in di vi sional prof its in order to re li ably attain budgeted targets. In portfolio theory terms, divisionalization in creases the neg a tive cor re la tions be tween the busi nesses of a di ver si fied corporation while also re duc ing their risk, so that cor po rate risk is reduced. Thus divisionalization interacts with diversification to lower the risk of the corporation (Donaldson 1999). The lower or ga ni za tional risk re duces the like li hood of cri ses of poor per for mance for the di ver si fied-divisionalized cor po ra tion. There fore when it en ters mis fit it tends to avoid mak ing needed adap tive changes, drift ing into me di o cre per for mance in the lon ger term. Or ga ni za tional theory has tra ditionally cel ebrated the advantages of the divisional struc ture, such as su pe rior stra te gic de ci sion mak ing, re source al lo cation, and discipline on divisions (Williamson 1970). Contingency theory, of course, holds that a divisional structure fits diversification and thereby leads to higher per for mance (Chan dler 1962). While these benefits in here in the divisional structure, organizational portfolio the ory points to the pos si bil ity of a lon ger-term dis ad van tage: the lack of ad aptation because of in sulation from crises of poor corporate performance. Divestment This tendency of the diversified, divisionalized corporation not to adapt may be re in forced by an other fac tor: di vest ment. A cor po ra tion that is ex pe ri enc ing a cri sis of poor per for mance—be cause it is in mis fit and so needs to make adaptive change, such as to its organizational struc ture—may in stead di vest some busi ness or busi nesses. The sale of these businesses will often generate large amounts of cash, raising profits and so pre vent ing per for mance from be com ing low, so that no needed changes are made to the re main ing or ga ni za tion. More over, a highly di ver si fied cor po ra tion with many in de pend ent di vi sions could divest them one by one, over an extended pe riod, thereby keep ing its

256

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

profitability satisfactory and so avoiding changes to its core. Thus divestment may be a tactic that is most available to the diversified, divisionalized cor po ra tion and re in forces the ten dency of such cor pora tions to avoid adap tive change and to con tinue in me di oc rity.

Directors Nonexecutive directors also may work to reduce the risk of the corporations on whose boards they sit as di rec tors or chair. Agency and transaction costs the o ries hold that nonexecutive di rec tors pro vide a needed control on managers on behalf of the shareholders, thereby rais ing cor po rate profit and share holder re turns (Jensen and Meckling 1976; Williamson 1985). Stewardship theory states the opposite: that free ing man ag ers from con trols in creases in trin sic mo ti va tions and so boards pre dom i nantly of ex ec u tive di rec tors raise cor po rate profit and shareholderreturns(Davis, Schoorman, and Donaldson 1997; Donaldson 1990a). Re search into the ef fect of nonexecutive di rec tors on cor po rate profit and shareholder returns has been mixed (e.g., Baysinger and Butler 1985; Donaldson and Davis 1991; Ezzamel and Watson 1993; Kesner 1987). Some reviews calculate the average effect to be prac tically nil (Boyd 1995). In stead of fo cus ing on the level of performance of a com pany as the out come of board struc ture, or ga ni za tional port fo lio the ory fo cuses on the fluc tu a tion in the per for mance of a com pany. Nonexecutive di rec tors are less likely to ap prove risky courses of ac tion for their com pany than are ex ec u tive di rec tors. Any de ci sion in volves bal anc ing costs and ben e fits. Se nior man ag ers for mu late proposals for ex pan sion that in volve risky ac tions such as in vest ment in a new, large-scale plant or in re search and de vel op ment. Their fa mil iarity with their com pany leads them to have con fi dence in their predictions about the future. When they sit on the board as executive directors, they carry this con vic tion with them and can in flu ence other directors through their ad vo cacy at the board meet ing. How ever, a nonexecutive di rec tor is li a ble to see such ex ec u tive pre dic tions as mere prom ises and per haps as self-jus ti fi ca tions. The nonexecutive di rec tor can feel more certainty about the costs of the proposed project and accord them greater weight than the supposed benefits. Therefore non executive directors lead to less risky strategies by corporations. Re search finds empirically that nonexecutive directors lead to less risky corporate

Neo-Con tin gency The ory

257

strat e gies (see Baysinger, Kosnik, and Turk 1991; Hill and Snell 1988), de spite this be ing con trary to the as ser tions of agency the ory. Thus or ga ni za tional port fo lio the ory states that nonexecutive di rectors reduce corporate risk and thereby forestall wide fluctuations in corporate profit (Daily and Dalton 1994a, 1994b). In this way non executive directors tend to prevent the episodes of low performance that trig ger needed adap tive or ga ni za tional changes. Thus, un wit tingly, nonexecutive directors become an other fac tor preventing long-term or ga ni za tional suc cess. In sum, the four port fo lio fac tors of di ver si fi ca tion, division alization, di vest ment, and di rec tors com bine to re duce or ga ni za tional risk and so fore stall adap tive or ga ni za tional change, lead ing to long-term me di ocre performance. Many large corporations are high on several, if not all, of these fac tors, which would tend to make their problems of ad aptation chronic. These four factors tending to prevent or ganizational change are, as we have seen, op posed by the four fac tors that pro mote it: the busi ness cy cle, com pe ti tion, debt, and di vi sional risk. The risk of an or ga ni za tion and thereby its pro pen sity to adapt and grow is af fected by its level on each of the eight or ga ni za tional port fo lio factors. An organization is more likely to make needed changes and grow if it is undiversified, functionally structured, in a competitive industry with a large busi ness cy cle, and it has mainly ex ec u tive di rec tors who take on debt and do not divest. An organization is less likely to make needed changes and grow if it is di ver si fied, divisionally structured, in industries with little competition and small business cycles, and has mainly nonexecutive directors who refuse to take on debt and who readily divest if per formance becomes low. Thus by specifying fac tors that affect organizational per formance and their interaction with misfit and fit, we can develop a systematic theory about or ga ni za tional change. This builds on the in sight of the pi o neers of contingency the ory that organizations needing to make adaptive changes often fail to do so until performance falls low enough that there is a cri sis (Chan dler 1962).

Quasi-Fit We have been talk ing about the or ga ni za tion mov ing from mis fit into fit. But how does the management know what fit is? The tra ditional

258

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

contingency the ory tends toward an implicit view that management knows what a fit for its or ga ni za tion would be. How ever, this view may be un re al is tic. We shall there fore first ex am ine the tra di tional view, then ex plain why it is un re al is tic, and then pro pose a more re al is tic view. The traditional contingency literature seems to take it as largely unproblematic that when an or ga ni za tion is in mis fit, its man age ment can see what would be a fit. Yet how can man age ment know what would fit its or ga ni za tion in its par tic u lar sit u a tion? Per haps, one might rea son, management has been educated in contingency theory. Contingency the ory holds that as size in creases, an or ga ni za tion should in crease its level of for mal iza tion. Yet few peo ple, apart from re search ers, know the re gres sion equa tions that spec ify the level of for mal iza tion that fits size (such as that in Child 1973b, 1975). More over, to ap ply that knowl edge pre scrip tively, man age ment would need to know the re quired score of its or ga ni za tion on the scale used in the re search to mea sure for mal ization (Pugh et al. 1968). Even if a man ager knew the quan ti ta tive level, vari ables such as for mal iza tion are ab stract and so not di rectly ac tionable (Pfeffer 1997). Thus the management of most organizations would not know ex actly which level of each or ga ni za tional struc ture it needed to adopt to move into fit with the con tin gen cies of its or ga ni zation, such as its size. Man age ment might more plau si bly know the di rec tion in which fit lay, and head in that direction. For example, they might realize that having grown, they need to increase the level of their formalization. But how could they know by how much they should in crease it? How do they avoid increasing the formalization too little or too much, given that they do not know the exact level that is re quired to fit the size of their or ga ni za tion? It might be ar gued that they use trial and er ror, that is, un der shoot ing and over shoot ing through nu mer ous it er a tions, until they even tu ally land on the tar get, that is, the level of for mal iza tion that fits their or ganizational size. Yet that would make organizational ad ap tation very troublesome and protracted. Moreover, it im plies numerous in crements and decrements in quick succession for many or ga ni za tions un til the or ga ni za tion con verges on the fit line through ex ten sive trial and er ror. Yet this seems im plau si ble. Thus we need another the o ret i cal ac count of or ga ni za tional ad ap ta tion. We suggest that organizational adaptation is a move into quasi-fit. For the rea sons al ready given, or ga ni za tions in mis fit will only rarely, and to a de gree with luck, move into full fit. They will much more

Neo-Con tin gency The ory

259

typ ically move to ward fit, but at tain only quasi-fit. That is to say, the organization will move closer to fit but not attain fit completely. The organization will adjust the level of its structure to narrow the gap between its ac tual level and that re quired to fit its con tin gen cies, with out completely eliminating the gap. Some support for the view that when organizations ad just their struc ture they of ten en ter only quasi-fit rather than full fit, is found in empirical studies of the divisionalization of large firms. There is a strong ten dency for firms that have di ver si fied to even tu ally adopt the divisional structure (Channon 1973; Dyas and Thanheiser 1976; Fligstein 1985; Mahoney 1992; Palmer et al. 1987; Palmer, Jennings, and Zhou 1993; Rumelt 1974; Suzuki 1980). Replacing their func tional or holding struc tures with a di vi sional struc ture is a move away from misfit between structure and diversification, toward fit, which in creases performance (Donaldson 1987; Hamilton and Shergill 1992, 1993). However, a closer examination within these divisional struc tures shows that many of them are incomplete, hav ing failed to adopt all the at trib utes that make a fully de vel oped di vi sional struc ture (Al len 1978; Hill 1985a, 1985b, 1988a; Hill and Pickering 1986; Steer and Ca ble 1978; Williamson and Bhargava 1972). For instance, some division alized firms fail to decentralize de ci sion mak ing (Hill 1985a, p. 744). Thus many divisionalized firms are only in quasi-fit to the level of their diversification contingency, rather than being in full fit. How ever, quasi-fit would still pro duce higher lev els of per for mance than mis fit, so that firms could be in quasi-fit and yet be mod er ately ef fec tive, suffi cient to have the ca pac ity for further growth. Moreover, given that in com plete divisionalization is so com mon among large firms, there is a chance that a firm with in complete divisionalization will face com pet i tors who are sim i larly, incom pletely divisionalized. There fore the firm may nev er the less be able to main tain mod er ately high per formance and growth de spite its be ing only in quasi-fit. By moving from misfit to quasi-fit the organization im proves its degree of fit and thereby improves its performance. This increase in per for mance is suf fi cient to feed back and com mence grow ing again. In turn, such in crease in the con tin gency in creases the mis fit and be gins to reduce performance, so that eventually a further in crease in the organizational structure is triggered. Hence the organization can go through the cy cle of in creas ing the con tin gency and struc ture de scribed in the pre vi ous sec tion, de spite not at tain ing full fit. The or ga ni za tion

260

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

needs only to at tain quasi-fit for the feed back ef fect of per for mance on the con tin gency to cause an other round of the cy cle of growth. Thus an organization needs only to enter quasi-fit recurrently to exp erience recurrent in cre men tal in creases in con tin gen cies and struc tures. The con cept of the or ga ni za tion mov ing into quasi-fit is more re al istic than that of the or ga ni za tion mov ing into full fit. The man age ment of an or ga ni za tion does not need to know ex actly what fit is, but only has to know the cor rect di rec tion. It can in fer the cor rect di rec tion by ex trap o lat ing from past de ci sions. For ex am ple, when the or ga ni za tion grew and had prob lems in the past, man age ment in creased the level of formalization in the organizational structure and this reduced the prob lems. Now that the or ga ni za tion has grown again and has sim i lar prob lems again, a sen si ble course of ac tion is to in crease again the level of for mal iza tion. This model of man a ge rial de ci sion mak ing is consis tent with bounded ra tio nal ity. Lack ing com plete knowl edge, man ag ers tend to act only when there is a problem and to use a so lu tion close at hand that is not per fect but that satisfices, that is, re duces the mag ni tude of the prob lem to an ac cept able level (Si mon 1976). If man ag ers were to move the or ga ni za tion into full fit, they would need to have the con tin gency fit model in their minds, but we have seen that this is im plau si ble. In con trast, if man ag ers be lieved in a uni versalistic the ory, they could follow it by simply increasing whichever struc tural vari able the uni ver sal ist ic the ory said should be max i mized. This would provide enough guidance for the managers so that they increased the level of their struc ture, so that they moved into quasi-fit. Priem (1994) pro vides ev i dence for or ganic the ory that close ness to fit is more strongly as so ci ated with CEO be liefs of a uni ver sal ist ic than of a contingency theory type. Hence managers just increase a structural vari able, which closes the gap some what, thereby mov ing the or ga ni zation from misfit closer to fit, without necessarily attaining full fit, so that the or ga ni za tion en ters only quasi-fit. By chance an or ga ni za tion mov ing to ward fit will move into full fit, but this will be the oc ca sional ex cep tion rather than the gen eral pat tern. Hence, an or ga ni za tion mov ing only into quasi-fit rather than into full fit is com pat i ble with organiza tional change be ing guided by uni ver sal ist ic man a ge rial beliefs. Fig ure 9.1 shows the con cept of quasi-fit di a gram mat i cally. There is a quasi-fit line at some dis tance away from the full-fit line. In deed there is a quasi-fit line both below and above the full-fit line. The quasi-fit line is the line that con nects up all the points of quasi-fit. For an

Neo-Con tin gency The ory

261

Fig ure 9.1. Lines of Quasi-Fit: Showing Growth and De cline Paths

or ga niza tion that is be low the full-fit line, if it is on the quasi-fit line at point A, then it is suf fi ciently in fit that its per for mance is high enough for it to have some sur plus re sources so that it grows. This in crease in or ga nizational size while retaining the existing level of organizational structure, for example, f ormalization, means that the organization moves from quasi-fit into mis fit. The low per for mance from this mis fit even tu ally causes the or ga ni za tion to in crease its level of for mal iza tion, so that it at tains quasi-fit again, this time at a new point, B. The re stored per formance even tu ally causes a fur ther in crease in or ga ni za tional size, which then leads to a fur ther in crease in for mal iza tion un til the or ga niza tion re gains quasi-fit again, this time at point C. Sim i larly, the or ga ni za tion sub se quently moves to point D. The grow ing or ga ni za tion follows a zig zag path such that it moves along the quasi-fit line (A, B, C, and D) that is be low the full-fit line. In this way the or ga ni za tion could grow over its life from very small with very low formalization, to very large with very high for mal iza tion.

262

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

If at any point in its life the organization decreases in size, that is, declines or downsizes, then initial size decreases will mean that it ap proaches closer to the full-fit line. It may even tu ally come to lie on the full-fit line at E. Fur ther size de crease brings the or ga ni za tion to F on the quasi-fit line above the full-fit line. Then a further size decrease again brings the or ga ni za tion into mis fit at G above that quasi-fit line. To attain quasi-fit, the organization must decrease its for malization suf fi ciently to move to H. If the or ga ni za tion con tin ues to de crease its size, it again moves into mis fit and must again re duce its for mal iza tion so that it at tains the quasi-fit line again at I. Thus once hav ing moved onto the quasi-fit line above the full-fit line, an organization may continuously decline by moving in a zigzag fash ion down along that quasi-fit line (F, H, and I). If at any time an or ga ni za tion that is on the quasi-fit line above the full-fit line in creases its size (i.e., re verses its decline), this ini tially moves it to ward fuller fit and hence higher per formance. Only when the organization grows further into misfit by crossing the full-fit line and then the quasi-fit line that is be low the fullfit line, will the or ga ni za tion en ter mis fit that is se vere enough to stop it grow ing.It will then in crease its for mal iza tion and will move into a new quasi-fit on the quasi-fit line that is be low the full-fit line. In terms of Figure 9.1, an organization at H that grows, will grow to (ap prox imately) B and then reattain quasi-fit at C. Hence an or ga ni za tion that is continuously grow ing will move up the quasi-fit line that is be low the full-fit line, while an organization that is continuously de clin ing will move down the quasi-fit line that is above the full-fit line. In summary, while contingency the ory states that organizations adapt and move into fit, they may not move into full fit. They may move only into quasi-fit, being that level of fit that is sufficient to raise performance enough that con tin gen cies in crease again (e.g., growth). This in turn means that the or ga ni za tion be gins to move to ward a new mis fit. In this way the or ga ni za tion can cy cle to ward and away from fit with out ever en ter ing full fit. Thus the or ga ni za tion may de velop over time, in both its contingency and structural variables, without ever moving along the line of full fits. There are alternate increments of increases in the or ga ni za tional struc tural vari able and in the contingency (e.g., size). A grow ing or ga ni za tion could move along a quasi-fit line that lies below the full-fit line. Conversely, a declining organization could move along a quasi-fit line that lies above the full-fit line.

Neo-Con tin gency The ory

263

Fit as Hetero-Performance As we have seen, tra di tional con tin gency the ory holds that fit is a line of iso-performance. This holds that the performance of each fit is the same as any other fit, so that the per for mance pro duced by a fit to low lev els of the con tin gency vari able is equal to that pro duced by fit to the high lev els of the con tin gency. The ex em plar is the pi o neer ing study by Woodward (1965) in which the fit for low-technology contingency (unit and small batch pro duc tion) pro duces the same level of per for mance as the fit for medium tech nol ogy (mass pro duc tion) and high technology (process production). If this is true, what is the incentive for a firm to ad vance its tech nol ogy from low to me dium to high? More gen er ally, what is the in cen tive for an or ga ni za tion to in crease the level of its contingency beyond the lowest level? The iso-performance concept is incompatible with the observed fact that organizations increase their lev els of the con tin gen cies. Oth er wise all or ga ni za tions would have low tech nol ogy and be small. The rea son for this the o ret i cal hi a tus lies in the fun da men tal na ture of structural contingency theory. As we noted at the outset, con tingency logic is that the performance or effectiveness of an organiza tional structure is moderated by some fac tor, which is termed the con tin gency. The fo cus is upon the or ga ni za tional struc ture as a cause of per for mance. The only in ter est in the con tin gency is as a mod er a tor of the causal effect of or ga ni za tional struc ture on per for mance. There is no concept that the contingency is itself a cause of organizational performance. Such a view is, strictly speaking, il logical. As we saw earlier (in Chapter 1), a contingency is a mod erator, and any moderator is itself also a cause of the de pend ent vari able. If W mod er ates the causal ef fect of X on Y, then this re la tion ship means also that X mod er ates the causal effect of W on Y. Thus both X and W have a causal effect on Y. It is a matter of con ve nience whether X or W is taken as the pri mary cause or as the mod er a tor. Hence, it fol lows that, if an or ga ni za tional struc ture has a pos i tive ef fect on or ga ni za tional per for mance that is con di tional upon a con tin gency, then that con tin gency also has an ef fect on per formance that is con di tional upon the struc ture. Both structure and con tin gency are causes of ef fec tive ness. They are in ter change able, and each could play the other’s role.

264

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

There is a kind of theoretical myopia in structural con tingency theory that treats vari ables, such as size or tech nol ogy, only as con tin gen cies and does not con sider them as hav ing a larger role. This my o pia is con sis tent with con tin gency the ory be ing a the ory of how or ga ni zational struc tures af fect per for mance. It is, how ever, in con sis tent with a more rounded the ory of or ga ni za tional per for mance. Such a broader view would have to con sider the causal con tri bu tion of these con tin gencies to or ga ni za tional per for mance. Where or ga ni za tions are gen er ally ob served to in crease their con tin gen cies, such con tin gen cies will of ten be causes of performance. This is the hypothesis from functionalism; whether a contingency does in fact raise performance is a matter for empirical val i da tion. An example of con tingency myo pia may be seen regarding the concept of strat egy. In struc tural con tin gency the ory, strat egy is treated as being a contingency of divisional structure (Hoskisson 1987). As such, the level of strategy determines whether a divisional structure produces high or low performance. A divisional structure produces high performance when the strategy of the firm is a high level of diversification, be cause a di vi sional struc ture fits a di ver si fied strat egy. Thus strat egy makes no con tri bu tion to per for mance and is merely a moderator of the ef fect of struc ture on per for mance, so that struc ture, not strategy, is the cause of performance. This is the view taken by approaching strat egy from the the o ret i cal lens of a the ory of or ga ni zational struc ture, struc tural con tin gency the ory. However, the o ries within the field of strat egy, nat u rally enough, view strategy as a cause of performance. Given that studies failed to find con sis tently the ex pected uni ver sal pos i tive ef fect of stra te gic di ver si fication on performance (e.g., Rumelt 1974), strategy the ory re search examines whether there is a moderator of the strategy-performance relationship that accounts for the variations in findings. Strategic management research shows that structure moderates the strategyperformance relationship. Hill, Hitt, and Hoskisson (1992) suggest that the ef fect of di ver si fi ca tion on per for mance is con tin gent upon the or ga ni za tional struc ture of the com pany. For each de gree of di ver si fication there is an or ga ni za tional struc ture that fits it, such that high effectiveness re sults if the com pany adopts the struc ture that fits its di ver si fi ca tion. Hill, Hitt, and Hoskisson (1992) ar gue that for a highly diversified com pany, producing prod ucts un related to each other, a highly de central ized struc ture is the fit. In con trast, for a medium

Neo-Con tin gency The ory

265

diversified com pany, pro duc ing prod ucts that are re lated to each other, a medium decentralized structure is the fit. They show that the fit of structure to the degree of diver si fi ca tion pos itively affects perfor mance, thus sup porting their contingency the ory of strat egy as contin gent upon struc ture. Hence di ver si fi ca tion is treated as the cause of per for mance, with struc ture be ing the mod er a tor. This il lus tratesthat a vari able considered by struc tural con tin gency the ory to be a contin gency can be con sid ered to make a causal con tri bu tion. Thus vari ables treated as con tin gen cies can also be con sid ered as causes of perfor mance. Whether a vari able is con sid ered to be a cause of per for mance or a contingency tends to be a mat ter of con ven tion in a re search lit er ature, re flect ing the the ory that dom i nates therein. More over, there is a the o ret i cal rea son for hold ing that what are custom arily treated as the con tin gency vari ables in struc tural con tin gency the ory have a causal im pact on or ga ni za tional per for mance. With out the con tin gency con trib ut ing to per for mance there is no rea son for the organization to increase its level on the contingency variable. For example, why should a firm make the capital in vestment required to move from batch to mass production without some benefit flowing from that in crease in tech nol ogy? If the con tin gency it self con trib utes to rais ing or ga ni za tion per formance, then the fit line is not one of iso-per for mance, but rather is one of het ero-per for mance. As a line of het ero-per for mance, the fit line is reconceptualized so that, instead of each point on it producing the same level of per for mance (i.e., iso-per for mance), each point along it enjoys higher perf ormance. Therefore the fits do not all produce the same per for mance; rather they pro duce dif fer ent per for mances. More specifically, the performances of the fits increase as the level of the con tin gency in creases. Thus the point on the fit line that cor re sponds to the lowest level of the contingency variable has the lowest per formance. The next point on the fit line, which corresponds to the next from low est level of the con tin gency vari able, has a higher level of perfor mance than the first point. Like wise the next point up from that point en joys higher per for mance again. In this way each suc ces sive point along the fit line, going from low to high values of the contingency, produces higher performance. A fit to a higher level of the con tingency pro duces a higher per for mance than a fit to a lower level of the contingency. Thus a hetero-performance fit line is a monotonically increasing function of performance. Hence performance rises as an

266

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

organization goes along the fit line from the low est to high est value of the contingency. The fit line be ing a line of het ero-per for mance pro vides an in cen tive for the or ga ni za tion and its man age ment to move along the fit line. As we have just seen, be ing in fit to the low est level of the con tin gency vari able pro duces less per for mance than be ing in fit to the next higher level of the contingency vari able, there fore the organization has an inc entive to move from the first fit to the second fit. Because each fit produces higher performance at each successive level of the con tingency vari able, there is in cen tive for the or ga ni za tion to move up along the fit line from low to high levels of the con tin gency. This in creased per for mance from in creas ing the level of the con tingency vari able ex plains why or ga ni za tions in crease their con tin gency level and do not sim ply re main in fit at the low est level of the con tin gency. It is there fore con sis tent with our re for mu la tion of con tin gency the ory as a the ory of dis equi lib rium. An or ga ni za tion in fit will tend to move into mis fit be cause it will in crease its level of the con tin gency vari able (e.g., size) in or der to gain the ben e fits of that higher level of the con tingency. Initially, the misfit will reduce per formance. But when the organization adopts a new value of the organizational structure and thereby moves into the new fit, the organization will experience the higher per for mance that mo ti vated its ear lier in crease in the con tin gency variable. Thus an organization will wish to move along the fit line, increasing its con tin gency and its struc ture, so that it moves from one fit to the next, because the or ga ni za tion in creases its performance at each step. Thus the hetero-performance concept of fit is compatible with the dis equi lib rium the ory of or ga ni za tion that an or ga ni za tion in fit does not re main there in def i nitely. More over, an or ga ni za tion in mis fit has an in cen tive to move into the new fit, that is, into the next fit on the line of hetero-performance, in or der to reap the ben e fit of the en hanced per for mance. The or ga ni zation has a disincentive to return to the old fit, because it is at a lower level on the hetero-performance line and therefore has lower per formance. Thus the hetero-performance con cept of fit is also con sis tent with the SARFIT model of struc tural ad ap ta tion to re gain fit. An org ani za tion in fit will tend to in crease its con tin gency, thereby moving into mis fit, and then in crease its struc ture in or der to en ter the new fit. The fit line be ing one of het ero-per for mance pro vides in cen tive both for the ini tial move away from fit and for the even tual move into

Neo-Con tin gency The ory

267

the new fit. The rep e ti tion of this cy cle of the or ga ni za tion mov ing into new mis fits and then new fits cre ates suc ces sive in cre men tal in creases in the contingency and the structure. Thus the organization moves over time along the fit line, grow ing on both the con tin gency and the struc tural vari ables. Hence the het ero-per for mance fit con cept pro posed here is compatible with the SARFIT model and its explanation of the dy nam ics of re cur rent, in cre men tal or ga ni za tional change. The model of fit and per for mance also needs to state mis fit and per formance by spec i fy ing the level of per for mance that in heres in be ing in mis fit at each lo ca tion off the fit line. The con cept of het ero-per for mance has implications for the per for mance of mis fits rel a tive to fits. Consider the per for mance pro duced by mild mis fit, that is, be ing just be low the hetero-performance fit line. The level of per for mance of a mild mis fit is that it must be lower not only than the new fit but also than the old fit (oth er wise it might re gain fit by mov ing back to the old fit). Given that, as we have seen, the old fit has lower per for mance than the new fit, the mild mis fit must be lower per for mance than the old fit as well as be ing lower still than the new fit. Thus a mild mis fit has not only lower per for mance than the ad ja cent fit at the same level of the con tin gency variable (i.e., the new fit), but also lower per for mance than the fit at the next lower level of the con tin gency vari able (i.e., the old fit). In this way the per for mances of mis fits rel a tive to fits can be spec i fied. The per for mances of mis fits rel a tive to other mis fits can also be speci fied by the het ero-per for mance fit con cept. Be cause het ero-per for mance fit is a line on which fits fur ther along the line pro duce higher per formances, points just off the line, that is, in mild mis fit, have dif fer ent performances. A point just below the low performance end of the fit line (i.e., cor re spond ing to the low est level of the con tin gency vari able) has lower per formance than the ad ja cent fit, which itself has low per formance, so that the mild misfit has very low performance. In contrast, a point just below the high performance end of the fit line (i.e., corresponding to the highest level of the contingency variable) has lower performance than its adjacent (new) fit, which itself has high per for mance, so that the mild mis fit has quite high per for mance. Thus mild mis fit has lower per for mance near the low than near the high end of the fit line. More gen er ally, the same de gree of mis fit that is the same distance away from the line (i.e., deviation) has higher per formance at higher lev els of the fit line (cor re spond ing to higher lev els of the con tingency) than at lower lev els of the fit line. Thus a cor ol lary of

268

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

the fit line being hetero-performance is that the misfits are also heteroperformance. In these ways, elements of the new misfit-performance model are also de riv able from the SARFIT the ory, spec i fy ing the per formances of mis fits rel a tive to fits and other mis fits. In sum mary, fit is a line of het ero-per for mance on which each point along the line (cor re spond ing to each higher value of the con tin gency vari able) is higher in per for mance than the one be fore it. Mis fit is also hetero-performance, so that the same de gree of mis fit (i.e., de vi a tion from the fit line) pro duces higher per for mance at higher lev els of the fit line. A mis fit ad ja cent to the fit line pro duces per for mance lower than the old fit and even lower than the new fit. In these ways an organiza tion has in cen tives to move out from fit into mis fit and then into a new fit, so mov ing along the fit line, point by point, thereby in creas ing in con tin gency and struc ture, as stated by the SARFIT model of or ga ni zational dy nam ics. The same holds for the quasi-fit line con cept.

Discussion Overall we have sought to make good several de ficiencies in con tingency the ory. We have striven to fill in gaps and make the the ory more coherent. In so do ing we have created a more dy namic pic ture of the organization as tend ing al ways to be chang ing in some way or other, if not in organizational structures then in contingencies. Change feeds upon change, rather than roll ing the or ga ni za tion into some equi librium cor ner po si tion from which it does not move. This has re quired at tend ing to the fuller con se quences of per for mance that feed back to af fect not only the or ga ni za tional struc ture but also the con tin gen cies. The cau sa tion be comes more sym met ri cal, cir cu lar, and com plete. The focus of the theory broadens from the organizational structures to include a more rounded ap pre ci a tion of the role of the con tin gen cies. These emerge from just be ing ad juncts to the or ga ni za tion struc ture, to being variables in their own right that make performance con tri butions, in con junc tion with struc ture, and are in turn molded by those per for mance con se quences. It is now a the ory of or ga ni za tional perfor mance and organizational change. It may be termed a theory of organizational ad ap ta tion. At nu mer ous points we have drawn on eco nom ics and fi nance. This is seen in the con cept that in creases in con tin gen cies, such as size and

Neo-Con tin gency The ory

269

di ver si fi ca tion, flow from sur plus re sources gen er ated by high per formance (Penrose 1959). It is seen in the use of the con cepts of risk and port fo lio from fi nance. It is fur ther seen in the na ture of sev eral of the portfolio factors. The business cy cle comes from macro-economics. Competition is, of course, cen tral to eco nom ics. Di ver si fi ca tion, debt, and di vest ment are tools of cor po rate fi nance. Finally, the fit line as a line of het ero-per for mance brings in the in cen tives that are re quired to im pel ac tion in eco nom ics. In these ways we hope to bring some new in sights into or ga ni za tional the ory and to en cour age a more pro ductive ex change with eco nom ics.

Conclusions Con tin gency the ory may be usefully developed in a num ber of di rections. The point of de par ture is to rec og nize dif fi cul ties in tra di tional contingency the ory and then to de vise so lu tions to them. Three such difficulties are rec og nized here, and a so lu tion to each is of fered. First, con tin gency the ory seems to be too static and only to dis cuss change as a movement from misfit into equilibrium. In particular it does not ex plain why the or ga ni za tion moved out of fit and into mis fit. We propose that organizations fre quently change their con tin gen cies when in fit be cause of the feed back from high per for mance. Thus an or ga ni za tion in fit has higher per for mance as a re sult and this (to gether with other causes rais ing per for mance) pro vides the sur plus re sources that cause increases in contingencies such as size and diversification, that is, hiring new employees or adding new product lines. These in creases in contingencies move an or ga ni za tion from fit into mis fit. The the ory is sym met ri cal in that both low and high val ues of per formance lead to change. Low performance leads from misfit to fit by change in the organizational structures, but high performance also leads from fit to mis fit by change in the con tin gen cies. Thus changes in both organizational structure and contingencies are affected by the feed back from per for mance. Hence there is a built-in ten dency toward disequilibrium that coun ters the ten dency to ward equi lib rium that is traditionally the fo cus of contingency theory. Misfit leads toward fit, but also fit leads toward misfit. Therefore there can be a cycle of perennially re new ing change rather than sta sis. In this way an or ga ni za tion can ex pe ri ence re peated in cre ments of change in con tin gen cies

270

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

and or ga nizational structure, so that it can grow large and internally elab o rate over time. Thus con tin gency the ory be comes more dy namic. Given that or ga ni za tional per for mance drives or ga ni za tional change, there are other causes of performance than just the fit and misfit discussed in con tin gency theory. For the low or high val ues of per for mance to oc cur that are re quired to cause or ga ni za tional change, mis fit or fit must be ac com pa nied by the right val ues of these other causes of performance. Mis fit needs to be accompanied by other causes of performance that are depressive in order to create low performance and hence ad ap ta tion. Fit needs to be ac com pa nied by other causes of per for mance that are buoy ant in or der to cre ate high per for mance and hence high growth rates. We need to an a lyze the in ter ac tion be tween fit and these other causes of per for mance to un der stand when change will and will not oc cur. Organizational portfolio the ory, in spired by portfolio the ory from fi nance, pro vides con cepts that help this anal y sis. Each cause pro duces a cer tain de gree of fluc tu a tion of or ga ni za tional per for mance and therefore con sti tutes a cer tain de gree of or ga ni za tional risk. The in ter ac tion of these causes with fit and with each other con sti tutes a port fo lio that affects the risk of the overall organization and thus its propensity to change. Port fo lio fac tors can re in force each other and in duce or ga ni za tional change or they can offset each other and prevent needed organizational change. Or ga ni za tional port fo lio the ory iden ti fies eight portfolio fac tors that affect performance in addition to fit. Four of these raise or ga ni za tional risk or lower performance and so promote or ga ni za tional change: the busi ness cy cle, com pe ti tion, debt, and di visional risk. Four of these re duce or ga ni za tional risk or raise per for mance and so forestall organizational change: diversification, divisional ization, divestment, and directors. The interaction of each of these organizational port fo lio fac tors with mis fit in flu ences whether needed adap tive or ga ni za tional changes oc cur or not. Sim i larly, their inter action with fit in flu ences the rate of or ga ni za tional growth. An im pli ca tion is that the di ver si fied, divisionalized cor po ra tion may fail to make needed adap tive changes in a timely man ner, es pe cially if it has di rec tors who are mainly nonexecutives who respond to performance problems by di vest ing. This pro vides a cau tion to the rather op ti mis tic ap pre ci a tion of the divisionalized cor po ra tion in the lit er a ture at pres ent. The second problem in traditional contingency the ory is that it is unclear how managers know what exact organizational structures fit

Neo-Con tin gency The ory

271

their contingencies. The suggestion is that, rather than having such knowl edge and mov ing their or ga ni za tion into full fit, it is more re al istic to say that or ga ni za tions move only into quasi-fit. Man agers know the di rec tion in which to move their or ga ni za tion but not the amount of movement required. Therefore organizational change often stops short of full fit. How ever, by re duc ing the de gree of mis fit, or ga ni zational per for mance in creases suf fi ciently to feed back and in crease the con tin gen cies. Thus the cy cle of in cre men tal growth in or ga ni za tional struc ture and con tin gency can oc cur by the or ga ni za tion mov ing into quasi-fit. The third problem in traditional con tin gency the ory is the con cept of the fit line as be ing one of iso-per for mance. This is a line of points all of which pro duce equal or ga ni za tional per for mance. This begs the ques tion of why an or ga ni za tion should move from one fit to another by chang ing its level of the con tin gency vari able, es pe cially when that entails in curring ad ditional expenditure. The suggestion here is that the fit line be reconceptualized as a line of hetero-performance, so that fits vary in their performance ef fects. In particular, the fits to higher lev els of the con tin gency vari able pro duce higher per for mances than the fits to lower lev els of the con tin gency vari able. Thus or ga ni zational performance increases as the organization goes along the fit line. There fore the or ga ni za tion has an in cen tive to move along the fit line by in vest ing in higher lev els of the con tin gen cies, such as size or technology. Over all, the mod i fi ca tions to con tin gency the ory pro posed here seek to make it more realistic and more dynamic. They also make it more internally coherent. In so doing, the theory broadens out from just explaining the effect of organizational structures on performance as be ing condi tioned by contingencies. The contin gen cies be come contri bu tors to or ga ni za tional per for mance in their own right.Changes in con tin gen cies are also ex plained by the feed back of performance. It be comes a more rounded the ory of orga ni za tional per formance and or ga ni za tional change. In so do ing, it forges closer connections be tween or ga ni za tional theory and the dis ci plines of economics and finance.

THE Fu ture CON Re search TIN GENCY Op porTHEORY tu nities OF OR GA NI ZA TIONS

10

Future Research Opportunities

I

n this chap ter we will con sider some pos si ble fu ture de vel op ments for con tin gency the ory re search. These include research questions arising from the need to prove that the relationships among size and bu reau cracy are caused by con tin gency the ory fits. They in clude also research questions that stem from the reformulation of contingency theory into neo-contingency theory, such as those pertaining to the ideas of disequilibrium, the organizational portfolio, and quasi-fit. Fur ther, the neo-con tin gency idea of fit as het ero-per for mance leads to a new operationalization of fit. Through out this chap ter, the em pha sis is on turning the theoretical dis cussion in this book into future re search practice by emphasizing the hypotheses that would allow the theoretical ideas to be tested empirically. The hypotheses given are illustrative of the pos si bil i ties rather than ex haus tive.

273

274

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

Fit and Bureaucracy Theory As we dis cussed in Chap ter 4, the pre vail ing causal mod els of the rela tion ships among size and the bu reau cratic struc tural vari ables leave open the ques tion of whether their un der ly ing cause is con tin gencyfit or some other causal pro cess, for ex am ple, uni ver sal ist ic cau sa tion. In order to dem on strate that the un der ly ing causal pro cesses that generate the relationships among size and the bu reau cratic struc tural vari ables are contingency fits, each relationship needs to be examined em pir ically. The relationships in volved are those between size and each of horizontal and vertical structural differentiation, between horizontal structural differentiation and formalization, and between vertical struc tural dif fer en ti a tion and de cen tral iza tion. For each of these four relationships the examination needs to proceed in two stages. First, show that there is fit of each of the structural variables (horizontal structural differentiation, vertical structural dif ferentiation, formal iza tion, and de cen tral iza tion) to size that pos i tively af fects per for mance. Sec ond, show that each of the struc tural vari ables co mes into fit with size, af ter a size change, through the in di rect pro cess of mis fit caus ing low per for mance and re sult ing struc tural ad ap ta tion. For for mal iza tion, the mis fit be tween it and size al lows spe cial iza tion to in crease the level of formalization, through administrative specialists, who are newly cre ated by size in crease(s), pro mot ing their new manage ment sys tems. For cen tral iza tion, the mis fit be tween it and size is cre ated by the newly cre ated hi er ar chi cal level(s), due to re cent size in creases, so that hi er archical levels affect decentralization through misfit. Thus the effect of spe cial iza tion on for mal iza tion and of hi er ar chy on de cen tral iza tion is conditional upon misfit leading to low performance. In this way, the apparent di rect relationships among these bureaucratic structural vari ables are re ally de pend ent upon in di rect ef fects of mis fit of struc ture to size. Some spec i men hy poth e ses are: There is a fit of formalization to size that pos i tively af fects per for mance. Size in crease causes mis fit be tween the old level of for mal iza tion and the new size level. The mis fit of for mal iza tion to size low ers per for mance and re sults in an in crease in for mal iza tion to fit the new size level. When an or ga ni za tion is in mis fit of its for mal iza tion to its size, spe cialization acts to in crease for mal iza tion, through newly cre ated ad min is tra tive specialists pro mot ing their new man age ment sys tems.

Fu ture Re search Op por tu nities

275

Researching Neo-Contingency Theory In the pre vi ous chap ter, a num ber of new the o ret i cal de vel op ments for con tin gency the ory were ad vanced. These ideas need to be tested in em pir i cal study and there fore of fer av e nues for fu ture re search. To provide a framework for empirical in vestigation, these theoretical ideas need to be turned into hy poth e ses that can be tested. The fol low ing discussion seeks to facilitate that process by suggesting hypotheses and addressing other re search is sues such as the op er a tional def i ni tion of vari ables im plied by the newer ap proaches. Dis equi lib rium The ory Above (in Chap ter 9), it was sug gested that high per for mance leads or ga ni za tions that are in fit to in crease the level of their con tin gency so that they move into mis fit. Donaldson (1987, p. 14) showed that or ga ni za tions in fit move into mis fit by in creas ing their con tin gency. In that study the contingency was diversification, so the fit was a functional struc ture to an un di ver si fied strat egy, and the mis fit was a func tional struc ture to a di ver si fied strat egy. There is scope for test ing whether in creas ing the contingency causes misfit for other contingencies and their re lated as pects of struc ture. These would in clude, but not be restricted to, size and bu reau cratic struc ture, task un cer tainty and or ganic struc ture, and so on. What ever the con tin gency and its re lated as pect of or ga ni za tional struc ture, the hy poth e ses are: For an organization in fit, the higher its per for mance, the higher the probability that it will in crease the level of its con tin gency vari able. For an organization in fit, the higher its per for mance, the higher the probability that it will move from fit into misfit. For a firm, or ga ni za tional per for mance could be mea sured by profit. High profit gives surplus (or slack) resources that could be used to increase con tin gen cies, such as size (i.e., em ploy ees), tech nol ogy (i.e., new cap i tal in vest ments), geo graphic spread, ver ti cal in te gra tion, or diver si fi ca tion. Wher ever pos si ble, or ga ni za tional per for mance should be mea sured by the ob jec tive per for mance; but where this is not pos si ble sub jec tive es ti mates of per for mance could be used, and these have been shown to correlate quite highly with the objective measures in some or ga ni za tions (Dess and Rob in son 1984).

276

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

The The ory of Per for mance-Driven Or ga ni za tional Change A cen tral idea of the the ory of per for mance-driven change is that or gani za tions in mis fit fail to make the adap tive changes that would move them into fit, if they have per for mance above the satisficing level. As we have seen, there is case study evidence support ing this idea, in that corpo ra tions failed to adapt their struc ture to their new strat egy un til they underwent a crisis of low performance (Chandler 1962). Donaldson (1987) shows that diversified large corporations in misfit (i.e., hav ing a func tional struc ture) were more likely to adopt the di vi sional struc ture, thereby mov ing into fit with their di ver si fied strat egy, the lower their antecedent performance. There is a need to replicate these quantitative findings on samples of other corporations and es tab lish the gen er al ity of the find ings with other types of or ga ni za tions, such as small firms, gov ern men tal or ga ni za tions, and so on. The test ing also needs to be ex tended to other com bi na tions of con tin gen cies and structure, such as size and bureaucracy and uncertainty and organic structure. While the the ory of per for mance-driven orga ni za tional change as serts that low per formance produces an adap tive re sponse, the theory of threat-ri gidity asserts the contrary, that low performance induces ri gid ity and fail ure to adapt. The sug ges tion has been that both theories may be true but ap ply at dif fer ent lev els of per for mance. The level of performance that induces threat-rigidity is lower than that which induces adaptive change. Thus the testable hypoth e sis is that there is a curvilinear re la tion ship be tween or ga ni za tional per for mance and or ga ni za tional ad ap ta tion: Or ga ni za tional ad ap ta tion fails to oc cur at higher and very low lev els of performance, and occurs at an intermediate point (i.e., below the satisficing level). The satisficing level depends upon the performance expectations held by se nior man ag ers, board directors, and owners, but for op er ational, re search pur poses it is prob a bly per for mance be low the in dustry average (Donaldson 1994; Hilmer 1998). At least that suggestion may enable empirical research to proceed, lead ing to a more refined location of the level of per for mance deemed to be un sat is fac tory and thus in duc ing adap tive change.

Fu ture Re search Op por tu nities

277

Or ga ni za tional Port fo lio The ory The theory that or ganizational change is driven by organizational per for mance has been for mal ized into or ga ni za tional port fo lio the ory (as we discussed in the previous chapter). Organizational portfolio the ory is a new the ory and there fore needs to be em pir i cally re searched. There are numerous theo retical propositions that lend themselves to testing. Perhaps the most fundamental one is that fluctuations in performance are re quired so that ad ap ta tion and growth are trig gered re cur rently, pro duc ing over the long run an or ga ni za tion that is structur ally elab o rated and suc cess ful. The hy poth e sis is that: Some level of organizational risk (i.e., variation over time in perfor mance) nontrivially above zero is re quired for long-term suc cess. This is not, of course, to say that the higher the risk, the greater the success, but rather that zero or a very low level of risk will tend not to pro duce ep i sodes of ei ther ad ap ta tion or growth, so that the or ga ni zation will be stable but stagnant. Organizational perfor mance is affected neg atively by mis fit and posi tively by fit. But it is also af fected by other causes of or ga ni za tional performance. Their interaction with misfit con trols ad aptation and with fit controls growth. Organizational portf olio theory identifies eight such causes as or ga ni za tional port fo lio fac tors: the busi ness cy cle, competition, debt, di ver si fi ca tion, di vest ment, divisionalization, di visional risk, and di rec tors. Each of these fac tors gives rise to the o ret i cal propositions about adap tive change and growth that can be the basis for research hypotheses. Without being exhaustive, the following provides some il lus tra tions of the hy poth e ses that might be de duced from or ga ni za tional port fo lio the ory. Business Cycle. The up swing of the busi ness cy cle cre ates growth into misfit and also buoys up the performance of a firm, so that misfit persists. The downswing creates economic stringency that leads to adaptive changes into fit. Some hypotheses are: Or ga ni za tions are more likely to go from fit to mis fit in the up swing of the busi ness cy cle than in its down swing. Or ga ni za tions are more likely to go from mis fit to fit in the down swing of the busi ness cy cle than in its up swing.

278

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

Competition. Competition re duces slack and so induces adaptive change. It can take many forms, in clud ing hav ing com pet i tors that are them selves in fit. The gen eral prop o si tion is: An or ga ni za tion in mis fit is more likely to move into fit, the greater its competition. A spe cific der i va tion is the hy poth e sis that: A di ver si fied firm in mis fit (i.e., func tion ally struc tured) is more likely to divisionalize, the more of its com pet i tors that are in fit (e.g., un di ver si fied and func tional or di ver si fied and di vi sional). Fit applies to any combination of contingencies and organizational struc tures that con sti tute fit in con tin gency the ory, and each de serves to be stud ied a hy poth e sis in em pir i cal re search. Debt. Debt reduces free cash flow because of the large, periodic pay ments required to ser vice it. Thus per for mance (e.g., prof it abil ity) is re duced that it may more eas ily drop be low the satisficing level, thereby in duc ing adap tive change. A hy poth e sis is: An or ga ni za tion in mis fit is more likely to move into fit, the more it is indebted. Diversification. Diversification spreads the firm across the business cycles of industries and thereby stabilizes corporate performance, reducing the probability of low performance and hence adaptive change when the or ga ni za tion is in mis fit. A hy poth e sis is: The more a firm is di ver si fied, the lower the prob a bil ity that it will make needed adap tive change. Divisionalization. Divisionalization augments di versification, re inforc ing its ef fects. A hy poth e sis is: Divisionalization and di ver si fi ca tion in ter act to lower the prob a bil ity that a firm will make needed adap tive changes. Divisional Risk. Divisional risk (i.e., variation in performance over time), like or ga ni za tional risk, makes di vi sional ad ap ta tion and growth more likely. Some hypotheses are: The higher the risk of a division in misfit, the greater the probability that it will move into fit. The higher the risk of a di vi sion in fit, the greater its rate of growth. Directors . Nonexecutive directors induce con servatism and reduce organizational risk, which has neg a tive con se quences for or ga ni za tional adaptation. Some hypotheses are:

Fu ture Re search Op por tu nities

279

The higher the pro por tion of a board of di rec tors who are non executive di rec tors, the lower the prob a bil ity that the firm will make needed adap tive changes. A nonexecutive chair per son of a board of di rec tors low ers the prob a bility that the firm will make needed adap tive changes. Divestment . Divestment creates new, slack resources that can bolster otherwise flagging performance. The more of its assets that the firm sells and the lon ger the pe riod over which it makes any se rial di vest ment, the greater this ef fect. A hy poth e sis is: Di vest ment low ers the prob a bil ity that a firm will make needed adaptive changes. The typ i cal large cor po ra tion in many coun tries, such as Australia, Canada, France, Germany, Italy, Japan, New Zealand, the United Kingdom, and the United States, is di versified and divisionalized (Ca pon, et al.t 1987; Khandwalla 1977; Dyas and Thanheiser 1976; Pavan 1976; Suzuki 1980; Ham il ton and Shergill 1993; Channon 1973; Rumelt 1974, re spec tively). An im pli ca tion of the fore go ing or ga ni zational portfolio theory is that a large firm that is diversified and divisionalized will tend to fail to make needed adap tive changes, so that the typical large firms in these countries are pre dicted to be be com ing chron i cally maladapted and suboptimal performers. This is es pe cially true if the firm has a nonexecutive dominated board and re sorts to di vestment. In the United States, n otwithstanding the high lev els of di versification among many large firms (Fligstein 1985; Rumelt 1974), there has been some thing of a trend in some large firms to downscope (Da vis, Diekmann, and Tinsley 1994; Hoskisson and Hitt 1994), that is, re duce di ver si fi ca tion. This means that sub stan tial di vest ment is quite common in the United States, thereby reinforcing tendencies toward chronic maladaptation and suboptimal performance in those firms. This is a longer-run ef fect that co mes into play af ter the firm re ceives the ben e fits of adopt ing a divisional struc ture that fits its strat egy. Thus there are many test able hy poth e ses that flow from or ga ni za tional portfo lio the ory (for a fuller list of the o ret i cal prop o si tions, see Donaldson 1999). Quasi-Fit While con tin gency the ory states that or ga ni za tions adapt and move into fit, they may not move into full fit. They may move into only

280

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

quasi-fit, that level of fit that is suf fi cient to raise per for mance enough that contin gen cies in crease again (e.g., growth). This in turn means that the or ga ni za tion be gins to move to ward a new mis fit. In this way the or gani za tion can cy cle to ward and away from fit with out ever en ter ing full fit. Thus the organization may develop over time in both its contingency and struc tural vari ables with out ever mov ing along the line of full fits. There are al ter nate in cre ments of in crease in the or ga ni za tional structural variable and the contingency (e.g., size). Therefore, a growing organization could in crease its lev els on both the struc tural and con tin gency vari ables in a zig zag path that is par al lel to, but lies below, the full-fit line. Con versely, a de clin ing or ga ni za tion could repeatedly decrease its lev els of or ga ni za tional struc ture and the con tin gency, so that it at tains only quasi-fit and fol lows an other zig zag path that is par al lel to, but above, the full-fit line. Some hypotheses are: Organizations that move from mis fit to ward fit tend not to at tain full fit, but rather only quasi-fit. Or ga ni za tions that are re peat edly in creas ing the level of their contin gency move par al lel to and be low the full-fit line. Or ga ni za tions that are re peat edly de creas ing the level of their con tin gency move par al lel to and above the full-fit line. An Operationalization of Fit as Het ero-per for mance In the pre vi ous chap ter, the fit line was reconceptualized from a line of iso-per for mance to one of het ero-per for mance. The het ero-per for mance con cept was ar gued to ex plain or ga ni za tional change more ad e quately and, in particular, to be compatible with the structural adaptation to re gain fit (SARFIT) the ory of or ga ni za tional dy nam ics. As we have seen, the re la tion ship be tween fit and per for mance is a key one in con tin gency the ory. There fore re search ers need to study fit-per for mance for many reasons. An important reason is to validate postulated fits between con tin gen cies and or ga ni za tional struc tures. An other rea son is to explain organizational change as adaptation into fit. Yet another motive for identifying fits is to offer managerial prescriptions. Hence research needs to have an operational definition of the hetero-performance concept of fit to use in empirical studies. Such an operationalization will now be offered.

Fu ture Re search Op por tu nities

281

A het ero-per for mance fit line would have a higher level of per for mance be ing pro duced by each suc ces sive point along that line. There fore an organization has an incentive to increase the level of its contingency over time and so move along the line, thereby in creas ing its struc ture. We will first pos tu late the per for mance pro duced by be ing in fit, that is, in ly ing on the fit line, and then pos tu late the per for mance pro duced by be ing in mis fit, that is, ly ing off the fit line. Fit and Per for mance For the line of hetero-performance, each point along the fit line produces a higher level of per for mance than the pre vi ous point, go ing from the lowest to the highest levels of the contingency. How much increase is pro duced is not given by the ory and needs to be found em pir i cally, by ex am in ing the ac tual per for mance of each point along the fit line. However, in the interest of stimulating and guiding future research, we will pos tu late the fol low ing. Assume that the fit line runs through the origin (i.e., contingency variable is zero, structural variable is zero; see Figure 10.1). This is reasonable because otherwise the organization could have a positive level of the organizational structural variable when the contingency vari able is zero, or zero level of struc ture when the level of the con tin gency is positive. Examples would be some formalization with zero employees, or zero for mal iza tion with some em ploy ees, re spec tively. Either of these scenarios seems implausible, so that it seems more sensible to posit that when the con tin gency is zero, the struc ture is also zero. As sume fur ther that the or ga ni za tional struc ture and the con tingency vari ables are each mea sured on scales that vary from zero to ten. As sume also that the slope of the fit line is 45 de grees, so its slope is one (i.e., the ra tio of a unit in cre ment in the or ga ni za tional struc ture to a unit in cre ment in the con tin gency is 1). Thus the first point on the fit line lies at level one of the or ga ni za tional struc ture and level one of the contingency variable. The next point on the fit line lies at organiza tional structure two and contingency variable two. Similarly, the successive points on the fit line lie at struc ture three and con tin gency vari able three, and so on, up to struc ture ten and con tin gency vari able ten. Thus for any level of the contingency variable, it is fit ted by the same level of the or ga ni za tional struc ture vari able.

282

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

The key pos tu late is that, for a line of het ero-per for mance, each point on the fit line is one unit of performance greater than the next lower point. At the or i gin the per for mance is 0, so then the first fit point is 1 unit of per for mance (see Fig ure 10.1). This first fit is at level one of the con tin gency vari able and level one of the or ga ni za tional struc ture variable. The second fit point produces 2 units of performance and is at level two of the con tin gency and struc tural vari ables. The third fit point pro duces 3 units of per for mance and is at level three of the con tin gency and struc ture vari ables. A sim i lar pat tern holds for the fourth through the ninth points on the fit line. The final, tenth fit point produces 10 units of per for mance and is at level ten of the con tin gency and structure vari ables. Thus the fit points along the fit line range in their per for mances from 1 to 10. For a firm, performance could be measured by profit, so that the pos tu late would be that a high fit would be ten times greater in amount of profit pro duced than a low fit, cet eris pa ri bus. This provides the organization with an incentive to progress along the fit line and thereby to grow through successive increments from low to high val ues on both the con tin gency and the or ga ni za tional struc ture variables. Mis fit and Per for mance What are the im pli ca tions of this fit model for the re la tion ship between misfit and performance? Clearly, an organization in misfit has lower per for mance than an or ga ni za tion that is in fit at the same level of the con tin gency vari able. How ever, as we have just seen, the same de gree of mis fit does not pro duce the same per for mance. Be cause performance is greater at higher points along the fit line (i.e., those that cor re spond to higher lev els of the con tin gency vari able), a mis fit is creating greater per for mance at higher lev els of the con tin gency than at lower. Thus the per for mance of a de gree of mis fit (i.e., de vi a tion off the fit line of one level of ei ther the con tin gency or struc tural vari ables) var ies ac cord ing to the level of the con tin gency vari able. A de gree of mis fit cre ates higher per for mance at higher lev els of the con tin gency than at lower lev els. Of course, the greater the distance (i.e., deviation) from the fit line, the lower the per for mance, so that each suc ces sive degree of greater mis fit di min ishes per for mance. Nevertheless, an organization that is in sub stantial mis fit, be ing some way off the fit line (i.e., is sev eral de grees in

Fu ture Re search Op por tu nities

283

Fig ure 10.1. Het ero-Per for mance: Fit, Mis fit, and Per for mance

mis fit), at high val ues of the con tin gency, while not hav ing per for mance as high as it would be if were in fit, still has quite high per for mance. Thus an or ga ni za tion in some mis fit at high con tin gency has per for mance equal to anor ga ni za tion in fit at a me dium level of the con tin gency. The next step is to spec ify the ex act per for mance of each of the mis fits. Tra di tional ap proaches such as the Eu clid ean dis tance for mula (Drazin and Van de Ven 1985) seek to spec ify math e mat i cally the re la tion ship be tween the de gree of mis fit and per for mance, that is, by a for mula that applies to every misfit. This is not possible with hetero-performance mis fit, be cause, as we have just seen, the per for mance each pro duces var ies ac cord ing to the lo ca tion of the mis fit, that is, to the point along the fit line, to which that mis fit is near est. Thus con sid er ing the two-di men sional space formed by the contin gency and the structure variables, the

284

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

perfor mance of each point in that space needs to be de fined. This can be done by fol low ing an al go rithm. Starting with the points immediately be low the fit line, that is, just one space be low the fit line, for each of them their degree of misfit is mild (see Fig ure 10.1). An or ga ni za tion can en ter that mild mis fit by increasing its level of the con tin gency by one unit, so that it leaves the old fit. The or ga ni za tion may then re gain fit by in creas ing the level of its struc ture by one unit so that it en ters the new fit. These two moves bring the or ga ni za tion to the next high est point along the fit line from the previous point, that is, an increase of one performance unit. As argued in the pre vi ous chap ter, to have an in cen tive to make these two moves, the performance while in misfit must be less than that of the two fits, the old and the new. Given that the per for mance of the old fit is one unit less than the new fit, the mild mis fit must be one per for mance unit less than the old fit and two per for mance units less than the new fit. Using this algorithm or decision rule we can specify the per formance of the mild fits in Fig ure 10.1. For ex am ple, a mild mis fit ex ists at con tin gency level two and structural level one (see Fig ure 10.1). An or ga ni za tion would move into this mild misfit by increasing its contingency level by one unit, thereby mov ing out of the fit at con tin gency and struc ture of one that yielded a performance of 1. Thus the misfit produces a performance of 0. An or ganization there has an incentive to in crease its struc tures to level two and so move up into the new fit that cor re sponds to its con tin gency level of two, be cause that new fit has per for mance of 2. A sim i lar logic ap plies to each of the mild mis fit points that lie im me di ately below the fit line. Each of these mis fits has a per for mance that is two units less than the fit that corresponds to its level of the contingency variable. Thus these eight mild misfits (corresponding to contingency levels two to ten) range in their performances from 0 to 8 units. A sim i lar logic ap plies to each mis fit point that is far ther away from the fit line (be low that line). Thus for any level of the con tin gency, each mis fit that is one more de gree far ther from the fit line is two units less in its per for mance. This ful fills the con di tion that it be one per for mance unit less than the point to its left, which is nearer to the line, be cause it is more of a misfit and so produces less performance. The condition is also ful filled that the fo cal mis fit point be less in its per for mance than the point above it, which is less of a mis fit. Thus this value of per for mance for the fo cal mis fit main tains the con di tion that an or ga ni za tion

Fu ture Re search Op por tu nities

285

moving into a higher degree of misfit by increasing its contingency level will have consequently less performance. It also maintains the condition that an or ga ni za tion that in creases its struc ture will re duce its mis fit. Fur ther more, it main tains the con di tion that it will re duce its misfit by moving closer to a new fit, because there is an incentive to move to the new fit rather than to the old fit. The greatest de gree of misfit occurs at level ten of contingency and one of structure, where per for mance is –8 units (see Fig ure 10.1). Thus mis fit var ies from +8 to –8, re spec tively, at lev els nine and one of the struc ture, both for level ten of the contingency variable. The causal pattern is that misfit produces lower per for mance, the greater the dis tance from the fit line, which is pos si ble only at the lower lev els of the struc tural vari able. For mis fits above the fit line, the logic is sim i lar. A mild mis fit above the fit line pro duces lower per for mance than its ad ja cent old and new fits, be cause it is a mis fit. The or ga ni za tion moves out from the old fit by de creas ing its con tin gency level and so en ters the mis fit. The or ga niza tion may then de crease its struc ture so that it at tains the new fit that cor re sponds to its level of the con tin gency vari able. Be cause the or ga ni za tion has thereby moved down along the fit line, the new fit has less performance than the old. The new fit must produce more per formance than the mild mis fit, in or der that there is an in cen tive for the organization to move into the new fit rather remaining in the misfit. There fore the mild mis fit is one per for mance unit less than the new fit. For ex am ple, an or ga ni za tion is in fit at level two of the struc ture and contingency variables, so that its performance is 2 units (see Figure 10.1). It then de creases its con tin gency by one unit, so en ter ing mis fit, which low ers its per for mance to 0. The or ga ni za tion then de creases its struc ture by one unit to at tain the new fit (at level one of con tin gency and struc ture), which pro duces a per for mance of 1 unit. Thus the or ganization has fol lowed the SARFIT model, by changing first its con tin gency and then ad just ing its struc ture so that con tin gency causes struc ture. In so do ing the or ga ni za tion has moved down the fit line. Lower points along the fit line produce less performance than the higher point from which the organization started. Hence there is no incentive for an or ga ni za tion to move down the fit line. There fore an or ga ni za tion will tend only re luc tantly to move down the fit line, usu ally as a re sult of be ing forced to. This could be caused by some factor other than fit, for ex am ple, such as through cuts to its bud get by a par ent or ga ni za tion. Al ter na tively, the or ga ni za tion might be driven to re duce its

286

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

con tin gency by ide ol ogy, such as to re duce its size be cause of the pos i tive value placed on down siz ing at cer tain his tor i cal pe ri ods (Budros 1997; Cascio 1993; Mc Kin ley, Sanchez, and Schick 1995). An or ga ni zation that has de clined to a new, lower level of the con tin gency vari able (e.g., size) will have to re gain fit by adopting the new structural level that fits that lower level. Given a choice, man age ment would prefer to move back up to the old, higher level of the con tin gency and re gain fit that way, because it would produce higher performance. How ever, in contingency decline that option will often be infeasible, because the rea son that caused the de cline may still be op er a tive. For in stance, if the de cline was caused by re duc tion in re sources, man age ment may be unable to re store those re sources. Or again, the down siz ing ide ol ogy that caused shrink age in em ployee num bers may be still dom i nant. Be cause of the in cen tives, growth may be more fre quent than de cline, de pend ing upon the eco nomic con di tions. There fore more mis fits would be ex pected to be be low than above the fit line. How ever, SARFIT ap plies in de cline as in growth and there fore con tin gency-driven change can oc cur there. Thus we have de picted the re la tion ship between mis fit and per for mance above as well as be low the fit line. The greatest degree of misfit above the line occurs at level one of contingency and ten of struc ture, where per for mance is –8 units. Thus mis fit (above the line) var ies from +8 to –8 per for mance units, re spectively, at lev els nine and one of the con tin gency, both for level ten of the structural variable. The causal pattern is that misfit produces lower per for mance, the greater the dis tance from the fit line, which is pos si ble above the fit line only at the higher lev els of the struc tural vari able. The pattern of performance for each misfit is sym metrical for the mis fits above and be low the line (Fig ure 10.1). This is to be ex pected given that the same theory was used to create both spec i fi ca tions, the SARFIT model of organizational dynamics. Combining both above and be low the fit line, the great est de gree of mis fit oc curs at level ten of each of the con tin gency and struc ture vari ables, where per for mance is –8 units. Thus, overall, misfit ranges from +8 to –8. Performance is higher, the far ther along the fit line, which is only pos si ble at high lev els of the con tin gency and struc tural vari ables si mul ta neously. Per for mance is lower the greater the mis fit, that is, the greater the dis tance from the fit line, which is pos si ble only at the lower lev els of the con tin gency or struc tural vari ables.

Fu ture Re search Op por tu nities

287

We have stated the neces sary relationships among fit, misfit, and performance to mo ti vate struc tural ad ap ta tion to re gain fit in re sponse to contingency changes. However, there is no presumption that all organizations will ac tu ally move up and down the fit line, which is the line of full fit. For the rea sons ad vanced in Chap ter 9, many or ga ni zations will move along the paths of the quasi-fit lines that lie alongside of, but off, the full-fit line. The performances ascribed to the misfit points here pro vide them with in cen tives to adapt struc ture to changes in the con tin gency vari able. Of course, these relationships of performance to fit and misfit are theoretical postulates at pres ent. Whether they are valid is a task for empirical re search to in ves ti gate. The fit line is held to be one in which performance is higher at higher levels of contingency than at lower levels, that is, higher performance at the top and the bottom. Thus organizations ly ing on the fit line should be ex am ined to see whether their per for mances dif fer in this way. The hy poth e sis is that: Organizations lying at each successive point on the fit line produce higher per for mance at each higher level of the con tin gency vari able than at the level immediately be low. The con trast ing hy poth e sis from the iso-per for mance con cept is that: Organizations lying on the fit line produce the same performance at each level of the con tin gency vari able. The test ing of the over all model, con tain ing both fits and mis fits, would be as fol lows. For each or ga ni za tion in the study, first find its po si tion in the ma trix (Fig ure 9.1) and thus its per for mance as pre dicted by the ma trix. The pre dicted per for mance of that or ga ni za tion should corre spond to its actual per for mance rel a tive to the other or ga ni za tions in the study. The cor re la tion would then be cal cu lated be tween the pre dicted and actual performances. If this correlation is positive, then some support ex ists for the model. The higher the pos i tive cor re la tion, the greater the em pir i cal sup port for the model. Thus the hy poth e sis is that the correlation be tween the pre dicted and ac tual per for mances would be pos i tive and sub stan tial. Given that the model is in tended to be an advance on other, more traditional measures of fit, the hypothesis would be that the correlation be tween the pre dicted and ac tual per formances would be greater than that be tween other mea sures of mis fit and ac tual per for mance. Other mea sures of mis fit would be ex em pli fied by the Euclidean deviation scores. These misfit-performance

288

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

rela tion ships are ex pected to be neg a tive, whereas the pre dicted per formance re la tion ship from the het ero-per for mance con cept is ex pected to be pos i tive. Thus a hy poth e sis would be that: The pos i tive cor re la tion be tween ac tual per for mance and that predicted by the hetero-performance fit con cept is higher than the neg ative mis fit-per for mance cor re la tion of the Eu clid ean de vi a tion method. Clearly many factors other than fit or mis fit af fect organizational per for mance. The po ten tial con found ing af fect of these can be con trolled for sta tis ti cally in the anal y ses dis cussed above. For ex am ple, the ac tual performance that is being compared with that predicted from the hetero-performance con cept could be per for mance (e.g., prof its) af ter controlling for other con found ing vari ables (such as in dus try, etc).

Conclusions The in ten tion of this chap ter has been to help fu ture em pir i cal re search address some of the structural contingency the ory issues that have been dis cussed in this book. Guid ance has been of fered in the form of hypotheses to be tested and the operationalization of the key re la tion ship be tween fit and or ga ni za tional per for mance. Re gard ing bu reau cracy, there is a need to test hy poth e ses em pir i cally about the effect of fit on performance. There is also a need to show t h a t o r ga ni za tional change conforms to the dy namic pat tern of con tingency change lead ing to struc tural change, by in ves ti gat ing size change and bureaucratic structural change. More over, it needs to be shown empirically that these changes occur through the intervening pro cesses of mis fit low er ing per for mance. The new the o ret i cal de vel op ments of neo-con tin gency the ory provide many op por tu ni ties for fresh em pir i cal work in struc tural contingency theory. There is a need to test the hypotheses from disequilibrium the ory that the higher the per for mance of a firm in fit, the greater the probability that it will increase the level of its contingency variables and so move into mis fit. While such high per for mance of a firm in fit stim u lates in creases in the con tin gen cies, low per for mance of a firm in misfit stim ulates struc tural ad ap ta tion. This idea that organizational change is driven by or ga ni za tional per for mance has been re fined into or ga ni za tional port fo lio the ory. This yields many the o ret i cal prop o sitions and test able hy poth e ses, il lus tra tive ex am ples of which have been

Fu ture Re search Op por tu nities

289

pro vided in this chap ter. Putting many in di vid ual ef fects to gether, an organization is more likely to make adaptive struc tural change if it is un di ver si fied in a com pet i tive in dus try with a deep busi ness cy cle and it has a functional structure, mainly executive directors, much debt, and does not di vest. Each of these ef fects needs to be stud ied sep a rately in or der to es tab lish the va lid ity of each com po nent part of or ga ni za tional port fo lio the ory. Again, the new con cept of quasi-fit de serves to be examined, such as by ascertaining whether organizations em pir ically tend to at tain only quasi-fit, rather than full fit, as they struc tur ally adapt and re com mence grow ing. Throughout this book we have stressed that the idea of fit is at the heart of con tin gency the ory. There fore it is ap pro pri ate that we should end by sug gest ing a new operationalization of the re la tion ship be tween fit and per for mance that is con sis tent with the change dy nam ics that have been ar gued, that is, struc tural ad ap ta tion to re gain fit (SARFIT). Thus con tin gency the ory con tin ues to de velop, such as in the di rections con sid ered here un der the head ing of neo-con tin gency the ory. In this way work con tin ues in the tra di tion that has seen so much valu able re search work over the years. Contingency theory has a rich past. It also has a prom is ing fu ture.

THEerCON Ref encesTIN GENCY THEORY OF OR GA NI ZA TIONS

References

Abdel-khalik, A. Rashad. 1988. “Hi er ar chies and Size: A Prob lem of Iden ti fi ca tion.” OrganizationStudies 9:237-251. Aldrich, Howard E. 1972a. “Re ply to Hilton: Se duced and Aban doned.” Administrative Sci ence Quar terly 17:55-57. Aldrich, Howard E. 1972b. “Tech nol ogy and Or ga ni za tional Struc ture: A Re ex am i na tion of the Find ings of the Aston Group.” Ad min is tra tive Sci ence Quar terly 17:26-43. Al ex an der, Ju dith W., and W. Alan Randolph. 1985. “The Fit Be tween Tech nol ogy and Struc ture as a Pre dic tor of Per for mance in Nursing Sub units.” Acad emy of Man age ment Jour nal 28:844-859. Al-Jibouri, Sadia Jabouri Joudi. 1983. “Size, Tech nol ogy, and Or ga ni za tional Struc ture in the Man u fac turing In dus try of a De veloping Coun try: Iraq.” Ph.D. diss. Mis sis sippi State Uni ver sity. Al len, Ste phen A. 1978. “Or ga ni za tional Choices and Gen eral Man age ment In flu ence Net works in Divisionalized Com panies.”Acad emy of Man age ment Jour nal 21:341-365. Amburgey, Terry L., and Tina Dacin. 1994. “As the Left Foot Fol lows the Right? The Dynamics of Stra te gic and Struc tural Change.” Academy of Man age ment Jour nal 37:1427-1452. An der son, Stu art. 1996. “Un ravelling the ‘Publicness Puz zle’ in Or ga ni za tion The ory: A New Use for the Aston Ap proach.” Pa per pre sented at the an nual con fer ence of the Brit ish Acad emy of Man age ment. Aston Uni ver sity, Bir ming ham, UK, Sep tem ber. Argote, Linda. 1982. “In put Un cer tainty and Or ga ni za tional Co or di na tion in Hos pi tal Emer gency Units.” Ad min is tra tive Sci ence Quar terly 27:420-434. 291

292

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

Argyris, Chris. 1964. In te grating the In di vid ual and the Or ga ni za tion. New York: John Wiley. Argyris, Chris. 1972. The Ap pli ca bil ity of Or ga ni za tional So ci ol ogy. Lon don: Cam bridge University. Ar mour, Henry Ogden, and Da vid J. Teece. 1978. “Or ga ni za tional Struc ture and Eco nomic Per formance: A Test of the Multidivisional Hy poth e sis.” Bell Jour nal of Economics 9:106-122. Astley, W. Gra ham 1985. “Or ga ni za tional Size and Bu reau cratic Struc ture.” Organiza tion Studies 6:201-228. Astley, W. Gra ham, and An drew H. Van de Ven. 1983. “Cen tral Per spec tives and De bates in Or ga ni za tion The ory.” Ad min is tra tive Sci ence Quar terly 28:245-273. Astley, W. Gra ham, and Ed ward J. Zajac. 1990. “Be yond Dyadic Ex change: Func tional In ter de pen dence and Sub-unit Power.” OrganizationStudies 11:481-501. At-Twaijri, Mohamed Ibrahim Ahmad, and John R. Montanari. 1987. “The Im pact of Con text and Choice on the Bound ary-span ning Pro cess: An Em pir i cal Ex ten sion.” Hu man Re la tions 40:783-798. Ayer, A. J. 1936. Lan guage, Truth and Logic . Lon don: Gollancz. Ayoubi, Z. M. 1981. “Tech nology, Size and Or ganization Struc ture in a De veloping Coun try: Jor dan.” Pp. 95-114 in Or ga ni za tion and Na tion: The Aston Programme IV, ed ited by D. J. Hickson and C. J. McMillan. Farnborough, Hants, UK: Gower. Azumi, K., and C. J. McMillan. 1981. “Man age ment Strat egy and Or ga ni za tion Struc ture: A Jap a nese Com par a tive Study.” Pp. 155-172 in Or ga ni za tion and Na tion: The Aston Programme IV, ed ited by D. J. Hickson and C. J. McMillan. Farnborough, Hants, UK: Gower. Bacharach, Sam uel B. 1989. “Or ga ni za tional The ories: Some Cri te ria for Eval u a tion.” Acad emy of Man age ment Re view 14:496-515. Badran, M., and C. R. Hinings. 1981. “Strat egies of Ad min is tra tive Con trol and Con tex tual Con straints in a Less De veloped Coun try: The case of Egyp tian Pub lic En ter prise.” Pp. 115-131 in Or ga ni za tion and Na tion: The Aston Programme IV, ed ited by D. J. Hickson and C. J. McMillan. Farnborough, Hants, UK: Gower. Bar ney, Jay B. 1990. “The De bate Be tween Tra di tional Man age ment The ory and Or ga ni za tional Eco nom ics: Sub stan tive Dif fer ences or In ter group Con flict?” Acad emy of ManagementReview 15:382-393. Bartlett, C. A., and S. Ghoshal. 1989. Managing Across Bor ders: The Transnational Solution . Boston: Har vard Busi ness School Press. Baysinger, Barry D., and Henry N. But ler. 1985. “Cor po rate Gov er nance and the Board of Di rec tors: Per for mance Ef fects of Changes in Board Com po si tion.” Jour nal of Law, Eco nom ics and Or ga ni za tion 1:101-124. Baysinger, Barry D., R. T. Kosnik, and T. A. Turk. 1991. “Ef fects of Board and Own er ship Structure on Cor porate R&D Strat egy.” Academy of Man agement Journal 34:205-214. Birkinshaw, Julian. 1999. Contingency The ory and the Char acteristics of Knowl edge: Stra te gic, En vi ron men tal and Knowl edge Pre dic tors of In ter na tional R&D Or ga ni za tion. Lon don: Lon don Busi ness School. Blau, Pe ter M. 1970. “A For mal The ory of Dif fer en ti a tion in Or ga ni za tions.” American SociologicalReview 35:201-218.

References

293

Blau, Pe ter M. 1971. “Com ments on Two Math e mat i cal For mu la tions of the The ory of DifferentiationinOrganizations.” AmericanSociologicalReview 36:304-307. Blau, Pe ter M. 1972. “In ter de pen dence and Hi er ar chy in Or ga ni za tions.” SocialScience Research 1:1-24. Blau, Pe ter M., Ce cilia McHugh Falbe, Wil liam Mc Kin ley, and Phelps K. Tracy. 1976. “Tech nol ogy and Or ga ni za tion in Man u fac turing.” Ad min is tra tive Sci ence Quar terly 21:21-40. Blau, Pe ter M., and Mar shall W. Meyer. 1987. Bu reau cracy in Mod ern So ci ety., 3rd ed. New York: Ran dom House. Blau, Pe ter M., and P. A. Schoenherr. 1971. The Struc ture of Or ga ni za tions. New York: Ba sic Books. Bour geois, L. J., III. 1984. “Stra te gic Man age ment and De ter min ism.” Acad emy of Man age ment Re view 9:586-596. Boyd, B. 1995. “CEO Du al ity and Firm Per for mance: A Con tin gency Model.” Strategic ManagementJournal 16:301-312. Brealey, Rich ard A., and Stew art C. Myers. 1996. PrinciplesofCorporateFinance , 5th ed. New York: McGraw-Hill. Brech, E. F. L. 1957. Or gani sa tion: The Frame work of Man age ment. Lon don: Longmans, Green. Bryman, A., A. D. Beardsworth, E. T. Keil, and J. Ford. 1983. “Re search Note: Or ga ni za tional Size and Spe cial iza tion.” OrganizationStudies 4:271-277. Budros, Art. 1997. “The New Cap i tal ism and Or ga ni za tional Ra tio nal ity: The Adop tion of Down sizing Pro grams, 1979-94.” So cial Forces 76:229-249. Buhner, R., and P. Möller. 1985. T “ he In for ma tion Con text of Cor po rate Dis clo sures of Divisionalization De ci sions.” Jour nal of Man age ment Studies 22:309-326. Burns, Lawton R. 1989. “Ma trix Man age ment in Hos pi tals: Testing The ories of Ma trix Struc ture and De vel op ment.” Ad min is tra tive Sci ence Quar terly 34:349-368. Burns, Tom. 1963. “In dus try in a New Age.” New So ci ety 18:17-20. Burns, Tom, and G. M. Stalker. 1961. The Man age ment of In no va tion . Lon don: Tavistock. Bur rell, Gib son, and Mor gan, Gareth. 1979. SociologicalParadigmsandOrganisational Anal y sis: El e ments of the So ci ol ogy of Cor po rate Life . Lon don: Heinemann. Ca ble, John, and Manfred J. Dirrheimer. 1983. “Mar kets and Hi er ar chies: An Em pir i cal Test of the Multidivisional Hy poth e sis in West Ger many.” In ter na tional Jour nal of IndustrialOrganization 1:43-62. Ca ble, John, and Hirohiko Yasuki. 1985. “In ter nal Or ga ni za tion, Busi ness Group and CorporatePerformance: An Em pirical Test of the Multidivisional Hy poth e sis in Japan.” In ter na tional Jour nal of In dus trial Or ga ni za tion 3:401-420. Cameron, Kim S., M. U. Kim, and Da vid A. Whetten. 1987. “Or ga ni za tional Ef fects of De cline and Tur bu lence.” Ad min is tra tive Sci ence Quar terly 32:222-240. Cameron, Kim S., Rob ert I. Sutton, and Da vid A. Whetten, eds. 1988. Readings in Organizational De cline: Frame works, Re search, and Pre scrip tions . Cam bridge, MA: Ballinger. Ca pon, N., C. Christodolou, John U. Farley, and James M. Hulbert. 1987. “A Com par a tive Anal y sis of the Strat egy and Struc ture of United States and Aus tra lian Cor po ra tions.” Jour nal of In ter na tional Busi ness Studies 18:51-74.

294

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

Cascio, Wayne F. 1993. “Down sizing: What Do We Know? What Have We Learned?” Acad emy of Man age ment Ex ec u tive 7:95-104. Caufield, Clyde Curtis. 1989. “An In tegrative Re search Re view of the Re lationship Between Tech nology and Struc ture: A Meta-An alytic Syn thesis.” Ph.D. diss. Univer sity of Iowa. Chandler, Al fred D., Jr. 1962. Strategy and Struc ture: Chap ters in the His tory of the IndustrialEnterprise. Cam bridge: MIT Press. Chan dler, Al fred D., Jr. 1977. The Vis i ble Hand: The Man a ge rial Rev o lu tion in Amer i can Business . Cam bridge, MA: Belknap Press. Channon, Derek F. 1973. The Strat egy and Struc ture of Brit ish Enterprise. Lon don: Macmillan. Channon, Derek F. 1978. The Ser vice In dus tries: Strat egy Struc ture and Fi nan cial Per for mance. Lon don: Macmillan. Chenhall, Rob ert H. 1979. “Some El e ments of Or ga ni za tional Con trol in Aus tra lian Divisionalized Firms.” Aus tra lian Jour nal of Man age ment Sup ple ment to 4:1-36. Child, John. 1972a. “Or ga ni za tion Struc ture and Strat egies of Con trol: A Rep li ca tion of the Aston Study.” Ad min is tra tive Sci ence Quar terly 17:163-177. Child, John. 1972b. “Or ganizationalStructure,Environment and Per for mance: The Role of Stra te gic Choice.” Sociology 6:1-22. Child, John. 1973a. “Par kin son’s Prog ress: Ac count ing for the Num ber of Spe cial ists in Organizations.” Ad min is tra tive Sci ence Quar terly 18:328-348. Child, John. 1973b. “Pre dicting and Un der stand ing Or ga ni za tion Struc ture.” Ad min istra tive Sci ence Quar terly 18:168-185. Child, John. 1974. “Man a ge rial and Or ga ni za tional Fac tors As so ci ated With Com pany Per for mance.” Jour nal of Man age ment Studies 11:174-189. Child, John. 1975. “Man a ge rial and Or ga ni za tional Fac tors As so ci ated With Com pany Performance, Part 2: A Con tingency Anal y sis.” Journal of Man agement Studies 12:12-27. Child, John. 1977. “Or ga ni za tional De sign and Per for mance: Con tin gency The ory and Be yond.” Or ga ni za tion and Ad min is tra tive Sci ence 8:169-183. Child, John. 1982. “Dis cus sion Note: Divisionalization and Size: A Com ment on the Donaldson/Grinyer De bate.” OrganizationStudies 3:351-353. Child, John. 1984. Organization: A Guide to Prob lems and Prac tice, 2nd ed. London: Harper and Row. Child, John. 1997. “From the Aston Programme to Stra te gic Choice: A Jour ney From Con cepts to The ory.” Pp. 45-71 in Ad vance ment in Or ga ni za tional Be hav iour: Essays in Hon our of Derek S. Pugh, ed ited by Tim othy Clark. Aldershot, Hants, UK: Ashgate. Child, John, and Al fred Kieser. 1979. “Or ga ni za tional and Man a ge rial Roles in Brit ish and West Ger man Com panies: An Ex am i na tion of the Cul ture-free The sis.” Pp. 51-73 in Or ga ni za tion and Na tion: The Aston Programme IV, ed ited by D. J. Hickson and C. J. McMillan. Farnborough, Hants, UK: Gower. Child, John, and Roger Mansfield. 1972. “Tech nol ogy, Size and Or ga ni za tion Struc ture.” Sociology 6:369-393.

References

295

Childers, Grant W., Bruce H. Mayhew, Jr., and Louis N. Gray. 1971. “Sys tem Size and Struc tural Dif fer en ti a tion in Mil i tary Or ga ni za tions: Testing a Base line Model of the Di vi sion of La bor.” Amer i can Jour nal of So ci ol ogy 76:813-830. Cibin, R., and R. M. Grant. 1996. “Re structuring Among the World’s Leading Oil Companies, 1980-92.” Brit ish Jour nal of Man age ment 7:283-307. Clark, P. 1990. Aston Programme: De scribing and Ex plaining the Struc ture of Ca na dian Tex tile Firms . Bir ming ham, UK: Aston Uni ver sity, Aston Programme Press. Co hen, J., and P. Co hen. 1983. Ap plied Mul ti ple Re gres sion/Cor re la tion Anal y sis for the Be hav ioral Sci ences, 2nd ed. Hillsdale, NJ: Law rence Erlbaum . Co hen, Wes ley M., and Rich ard C. Levin. 1989. “Em pir i cal Studies of In no va tion and Mar ket Struc ture.” Pp. 1059-1107 in Hand book of In dus trial Or ga ni za tion, Vol. 2, ed ited by R. Schmalensee and R. D. Willig. Am ster dam: Elsevier Sci ence Pub lishers B. V. Col lins, Paul D., and Frank Hull. 1986. “Tech nol ogy and Span of Con trol: Wood ward Re visited.” Jour nal of Man age ment 23:143-164. Conaty, J., H. Mahmoudi, and G. A. Miller. 1983. “So cial Struc ture and Bu reau cracy: ACom par i son of Or ga ni za tions in the United States and Prerevolutionary Iran.” OrganizationStudies 4:105-128. Cool, K., and I. Dierickx. 1993. “Ri valry, Stra tegic Groups and Firm Prof itability.” Strategic Man age ment Jour nal 14:47-59. Coo per, Da vid J., Bob Hinings, Royston Green wood, and John L. Brown. 1996. “Sed i men ta tion and Trans for ma tion in Or ga ni za tional Change: The Case of Ca na dian Law Firms.” OrganizationStudies 17:623-647. Corey, E. Raymond, and Ste ven H. Star. 1971. Organization Strategy: A Marketing Approach . Boston: Har vard Uni ver sity, Grad u ate School of Busi ness Ad min is tra tion, Di vi sion of Re search. Crawford, Mi chael Arthur. 1983. “The Char acter, De terminants and Per formance Effects of Inter-Unit In ter ac tions Within Or gani sa tions: A Disaggregated Sys tems Ap proach.” Ph.D. diss. Uni ver sity of New South Wales, Syd ney, New South Wales, Australia. Cronbach, Lee J., and Lita Furby. 1970. “How Should We Mea sure ‘Change’—Or Should We?” PsychologicalBulletin 74:68-80. Cro zier, Michel. 1964. The Bu reau cratic Phe nom e non. Lon don: Tavistock. Cul len, J. B., K. S. An der son, and D. D. Baker. 1986. “Blau’s The ory of Struc tural Dif fer en ti a tion Re visited: A The ory of Struc tural Change or Scale.” Acad emy of Man age ment Jour nal 29:203-229. Daft, Rich ard L., and Pa tri cia Bradshaw. 1980. “The Pro cess of Hor i zon tal Dif fer en ti a tion: Two Models.” Ad min is tra tive Sci ence Quar terly 25:441-456. Daily, Catherine M., and Dan R. Dal ton. 1994a. “Bank ruptcy and Cor porate Governance: An Em pir i cal As sess ment.” Stra te gic Man age ment Jour nal 15:643-654. Daily, Catherine M., and Dan R. Dal ton. 1994b. “Cor porate Gov ernance and the Bank rupt Firm: The Im pact of Board Com position and Struc ture.” Academy of ManagementJournal 37:1603-1617. Dal ton, Dan R., Catherine M. Daily, Jon a than L. John son, and Alan E. Ellstrand. 1999. “Num ber of Di rec tors and Fi nan cial Per for mance: A Meta-Anal y sis.” Acad emy of ManagementReview 42:674-686.

296

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

Dalton, Dan R., Wil liam D. Todor, Michael J. Spendolini, Gordon J. Fielding, and Lyman W. Por ter. 1980. “Or ga ni za tion Struc ture and Per for mance: A Crit i cal Re view.” Acad emy of Man age ment Re view 5:49-64. Damanpour, Fariborz. 1992. “Or ganizational Size and In no va tion.” Organization Studies 13:375-402. Daniels, John D., Rob ert A. Pitts, and Marietta J. Tretter. 1984. “Strat egy and Struc ture in U.S. Mul tinationals: An Ex ploratory Study.” Academy of Man agement Jour nal 7:292-307. Daniels, John D., Rob ert A. Pitts, and Marietta J. Tretter. 1985. “Or ga nizing for Dual Strat egies of Prod uct Di ver sity and In ter na tional Ex pan sion.” StrategicManagement Journal 6:223-237. Da vis, Ger ald F., Kristina A. Diekmann, and Catherine H. Tinsley. 1994. “The De cline and Fall of the Con glom er ate Firm in the 1980s: The Deinstitutionalization of an Or ga ni za tional Form.” AmericanSociologicalReview 59:547-570. Da vis, James H., F. Da vid Schoorman, and Lex Donaldson. 1997. “To ward a Stew ard ship The ory of Man age ment.” Acad emy of Man age ment Re view 22:20-47. Da vis, Stan ley M. 1972. “Ba sic Struc tures of Mul ti na tional Cor po ra tions.” Pp. 193-211 in Man aging and Or ga nizing Mul ti na tional Cor po ra tions, ed ited by Stan ley M. Da vis. Elmsford, NY: Pergamon. Da vis, Stan ley M., and Paul R. Law rence. 1977. Matrix. Read ing, MA: Ad di son-Wes ley. Dawson, San dra, and Dor o thy Wedder burn. 1980. “In tro duc tion.” Pp. xiii-xl in In dustrial Or ganization: The ory and Prac tice, 2nd ed., by Joan Wood ward. Ox ford: Oxford Uni ver sity Press. Delery, John, and D. Har old Doty. 1996. “Modes of The o rizing in Stra te gic Hu man Re source Man age ment: Tests of Uni ver sal ist ic, Con tin gency, and Con fig u ra tional Per for mance Pre dic tions.” Acad emy of Man age ment Jour nal 39:802-835. Dess, Greg ory G., and R. B. Rob in son. 1984. “Mea suring Or ga ni za tional Per for mance in the Ab sence of Ob jec tive Mea sures: The Case of the Pri vately-held Firm and the Con glom er ate Busi ness Unit.” Stra te gic Man age ment Jour nal 5:265-273. Dess, Greg ory G., and Don ald W. Beard. 1984. “Di men sions of Or ga ni za tional Task Environments.” Ad min is tra tive Sci ence Quar terly 29:52-73. Dess, Greg ory G., R. Duane Ire land, and Mi chael A. Hitt. 1990. “In dus try Ef fects and Stra te gic Man age ment Re search.” Jour nal of Man age ment 16:7-27. Dewar, Rob ert, and Jerald Hage. 1978. “Size, Tech nol ogy, Com plex ity, and Struc tural Differentiation: To ward a The o ret i cal Syn the sis.” Ad min is tra tive Sci ence Quar terly 23:111-136. Dewar, Rob ert, and James Werbel. 1979. “Uni ver sal ist ic and Con tin gency Pre dic tions of Em ployee Sat is fac tion and Con flict.” AdministrativeScienceQuarterly 24:426-448. DiMaggio, Paul J., and Wal ter W. Powell. 1983. “The Iron Cage Re visited: In sti tu tional Isomorphism and Col lec tive Ra tio nal ity in Or ga ni za tion Fields.” Amer i can So cio logicalReview 48:147-160. Donaldson, Gordon. 1994. Corporate Restructuring: Man aging the Change Pro cess From Within . Boston: Har vard Busi ness School Press. Donaldson, Lex. 1976. “Wood ward, Tech nol ogy, Or gani sa tional Struc ture and Per for mance—A Cri tique of the Uni versal Gen eralisation.” Journal of Management Studies 13:255-273.

References

297

Donaldson, Lex. 1979. “Re gaining Con trol at Nipont.” Jour nal of Gen eral Man age ment 4:14-30. Donaldson, Lex. 1982a. “Divisionalization and Di ver si fi ca tion: A Lon gi tu di nal Study.” Acad emy of Man age ment Jour nal 25:909-914. Donaldson, Lex. 1982b. “Divisionalization and Size: A The oretical and Em pirical Critique.” OrganizationStudies 3:321-337. Donaldson, Lex. 1985. In De fence of Or ganization The ory: A Re ply to the Critics . Cambridge UK: Cam bridge Uni ver sity Press. Donaldson, Lex. 1986. “Size and Bu reau cracy in East and West: A Pre lim i nary Meta Anal y sis.” Pp. 67-91 in The En ter prise and Man age ment in East Asia, ed ited by S. R. Clegg, D. Dunphy, and S. G. Redding. Hong Kong: Uni ver sity of Hong Kong Press. Donaldson, Lex. 1987. “Strat egy and Struc tural Ad just ment to Re gain Fit and Per for mance: In De fense of Con tin gency The ory.” Jour nal of Man age ment Studies 24:1-24. Donaldson, Lex. 1990a. “The Ethe real Hand: Or ga ni za tional Eco nom ics and Man age ment The ory.” Acad emy of Man age ment Re view 15:369-381. Donaldson, Lex. 1990b. “A Ra tio nal Ba sis for Crit i cisms of Or ga ni za tional Eco nom ics: A Re ply to Bar ney.” Acad emy of Man age ment Re view 15:394-401. Donaldson, Lex. 1995a. Amer i can Anti-Man age ment The ories of Or ga ni za tion: A Cri tique of Par a digm Pro lif er a tion . Cam bridge UK: Cam bridge Uni ver sity Press. Donaldson, Lex. 1995b. ContingencyTheory . Vol. 9 in His tory of Man age ment Thought Series. Aldershot, Hants, UK: Dartmouth Pub lishing. Donaldson, Lex. 1996a. For Posi tivist Or ganization The ory: Proving the Hard Core . London: Sage. Donaldson, Lex. 1996b. The Nor mal Sci ence of Struc tural Con tingency The ory.” Pp. 57-76 in The Hand book of Or ga ni za tion Studies, ed ited by S. R. Clegg, C. Hardy, and W. Nord. Lon don: Sage. Donaldson, Lex. 1997. “Derek Pugh: Sci entific Revolutionary in Or ganization Studies.” Pp. 23-43 in Ad vance ment in Or ga ni za tional Be hav iour: Es says in Hon our of Derek S. Pugh, ed ited by Tim o thy Clark. Aldershot, Hants, UK: Ashgate. Donaldson, Lex. 1999. Per for mance-Driven Or ga ni za tional Change: The Or ga ni za tional Portfolio . Thou sand Oaks, CA: Sage. Donaldson, Lex, and Clyde C. Caufield. 1989. “Eco nom ics of Scale in Pub lic and Pri vate Ad min is tra tion: Is Mar ket Dis ci pline Mythic?” Syd ney, NSW: Aus tra lian Grad u ate School of Man age ment in the Uni ver sity of New South Wales (Working Pa per, 89-003). Donaldson, Lex, and James H. Da vis. 1991. “Stew ard ship The ory or Agency The ory: CEO Gov ernance and Share holder Re turns.” Australian Journal of Man agement 16:49-64. Donaldson, Lex, and Malcolm Warner. 1974. “Struc ture of Or ga ni za tions in Oc cu pa tional In ter est As so ci a tions.” Hu man Re la tions 27:721-738. Dotsey, Mi chael, and Rob ert G. King. 1987. “Busi ness Cycles.” Pp. 302-310 in The New Palgrave: A Dic tio nary of Eco nom ics, ed ited by John Eatwell, Murray Milgate, and Pe ter Newman. Lon don: Macmillan. Doty, D. Har old, and Wil liam H. Glick. 1994. “Typologies as a Unique Form of The ory Build ing: To wards Im proved Un der stand ing and Modeling.” Acad emy of Man age ment Re view 19:230-251.

298

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

Doty, D. Har old, Wil liam H. Glick, and George P. Huber. 1993. “Fit, Equifinality, and Or ga ni za tional Ef fec tive ness: A Test of Two Con fig u ra tional The ories.” Acad emy of ManagementJournal 36:1196-1250. Drazin, Rob ert, and An drew H. Van de Ven. 1985. “Al ter na tive Forms of Fit in Con tin gency The ory.” Ad min is tra tive Sci ence Quar terly 30:514-539. Duncan, Rob ert B. 1972. “Char ac ter is tics of Or ga ni za tional En vi ron ments and Per ceived En vi ron men tal Un cer tainty.” Ad min is tra tive Sci ence Quar terly 17:313-327. Dunning, John H., Bruce Kogut, and Magnus Blomström. 1990. GlobalizationofFirms and the Com petitiveness of Na tions . Lund, Swe den: Lund Uni versity, Institute of Eco nomic Re search. Dunphy, D. C., and D. Stace. 1988. “Transformational and Co ercive Strat egies for Planned Or ga ni za tional Change: Be yond the O. D. Model.” OrganizationalStudies 9:317-334. Durkheim, Emile. 1964. The Di vi sion of La bor in So ci ety . New York: Free Press. Dyas, Gareth P., and Heinz T. Thanheiser. 1976. The Emerging Eu ropean Enterprise: Strat egy and Struc ture in French and Ger man In dus try . Lon don: Macmillan. Ed wards, Jeffrey R., and Mark E. Parry. 1993. “On the Use of Poly no mial Re gres sion Equa tions as an Al ter na tive to Dif fer ence Scores in Or ga ni za tional Re search.” Academy of Man age ment Jour nal 36:1577-613. Egelhoff, Wil liam G. 1982. “Strat egy and Struc ture in Mul ti na tional Cor po ra tions: An In for ma tion-Pro cessing Ap proach.” Ad min is tra tive Sci ence Quar terly 27:435-458. Egelhoff, Wil liam G. 1988a. Or ga nizing the Mul ti na tional En ter prise: An In for ma tion ProcessingPerspective . Cam bridge, MA: Ballinger. Egelhoff, Wil liam G. 1988b. “Strat egy and Struc ture in Mul ti na tional Cor po ra tions: ARe vi sion of the Stopford and Wells Model.” StrategicManagementJournal 9:1-14. Eilon, S. 1977. “Struc tural De ter min ism.” Omega, The In ter na tional Jour nal of Man age ment Studies 5:499-504. Etzioni, Amitai. 1961. A Com par a tive Anal y sis of Com plex Or ga ni za tions: On Power, Involvement and Their Cor re lates. New York: Free Press. Etzioni, Amitai. 1968. The Ac tive So ci ety: A The ory of So ci etal and Po lit i cal Pro cesses . Lon don: Col lier-Macmillan; New York: Free Press. Ezzamel, M. A., and K. Hilton 1980. “Divisionalisation in Brit ish In dus try: A Pre lim i nary Study.” Ac count ing and Busi ness Re search 10:197-214. Ezzamel, M. A., and R. Wat son. 1993. “Or ga ni za tional Form, Own er ship Struc ture and Corporate Performance: A Con textual Empirical Anal ysis of UK Com panies.” British Jour nal of Man age ment 4:161-176. Faas, F. A. M. J. 1985. “How to Solve the Com munication Prob lems on the R and D Interface.” Jour nal of Man age ment Studies 22:83-102. Feyerabend, P. 1975. Against Method: Out line of an An ar chis tic The ory of Knowl edge . Lon don: New Left Books. Fiedler, F. E. 1967. A The ory of Lead er ship Ef fec tive ness . New York: McGraw-Hill. Fiegenbaum, A., and H. Thomas. 1990. “Stra te gic Groups and Per for mance.” Strategic ManagementJournal 11:197-215. Finkelstein, Syd ney, and Don ald C. Hambrick. 1996. Stra te gic Lead er ship: Top Ex ec u tives and Their Ef fects on Or ga ni za tions . Min ne ap o lis/St Paul, MN: West.

References

299

Fisher, Cynthia D., and Rich ard Gitelson. 1983. “Meta-Anal y sis of the Cor re lates of Role Con flict and Role Am bi gu ity.” Jour nal of Ap plied Psy chol ogy 68:320-333. Flan ders, Allan, Ruth Pomeranz, and Joan Wood ward. 1968. ExperimentinIndustrial De moc racy: A Study of the John Lewis Part ner ship . Lon don: Faber. Fligstein, Neil. 1985. “The Spread of the Multidivisional Form Among Large Firms, 1919-1979.” AmericanSociologicalReview 50:377-391. Fligstein, Neil. 1990a. “Or ga ni za tional, De mo graphic and Eco nomic De ter mi nants of the Growth Pat terns of Large Firms, 1919-1979.” Pp. 45-76 in BusinessInstitutions, ed ited by Craig Calhoun, Vol. 12, ComparativeSocialResearch, Green wich, CT: JAI. Fligstein, Neil. 1990b. The Trans formation of Cor porate Con trol. Cam bridge, MA: Har vard Uni ver sity Press. Fligstein, Neil. 1991. “The Struc tural Trans for ma tion of Amer i can In dus try: An In sti tu tional Ac count of the Causes of Di ver si fi ca tion in the Largest Firms, 1919-1979.” Pp. 311-336 in The New Institutionalism in Or ganizational Anal ysis, ed ited by Wal ter W. Powell and Paul J. DiMaggio. Chi cago: Uni ver sity of Chi cago Press. Fligstein, Neil, and Pe ter Brantley. 1992. “Bank Con trol, Owner Con trol, or Or ga ni za tional Dy nam ics: Who Con trols the Large Mod ern Cor po ra tion?” Amer i can Jour nal of So ci ology 98:280-307. Franko, Law rence G. 1974. “The Move To ward a Multidivisional Struc ture in Eu ro pean Organizations.” Ad min is tra tive Sci ence Quar terly 19:493-506. Frederickson, J. W. 1984. “The Com prehensiveness of Stra tegic Decision Processes: Extensions, Ob servations, Future Di rec tions.” Academy of Man agement Jour nal 27:445-466. Free man, J., and M. T. Hannan. 1975. “Growth and De cline Pro cesses in Or ga ni za tions.” AmericanSociologicalReview 40:215-228. Freeman, John H., and Jerrold E. Kronenfeld. 1973. “Prob lems of Defi nitional De pend ency: The Case of Ad min is tra tive In ten sity.” So cial Forces 52:108-121. Fry, Louis R., and Deborah A. Smith. 1987. “Con gruence, Con tin gency, and The ory Build ing.” Acad emy of Man age ment Re view 12:117-132. Galbraith, Jay R. 1973. DesigningComplexOrganizations . Read ing, MA: Ad di son-Wes ley. Galbraith, Jay R., and Rob ert K. Kazanjian. 1988. “Strat egy, Tech nol ogy, and Emerging Or ga ni za tional Forms.” Pp. 29-41 in Fu tures of Or ga ni za tion: In no vating to Adapt Strat egy and Hu man Re sources to Rapid Tech no log i cal Change, ed ited by Jerald Hage. Lexington, MA: Lexington Books. Galtung, Johan. 1967. Theory and Methods of So cial Research . Oslo, Nor way: Universitetsforlaget. Galunic, D. C., and K. M. Eisenhardt. 1994. “Re newing the Strat egy-Struc ture-Per for mance Par a digm.” Pp. 215-255 in ResearchinOrganizationalBehavior, Vol. 16, ed ited by L. L. Cummings and B. M. Staw. Green wich, CT: JAI. Gerwin, Don ald. 1979a. “The Com par a tive Anal y sis of Struc ture and Tech nol ogy: A Crit i cal Ap praisal.” Acad emy of Man age ment Re view 4:41-51. Gerwin, Don ald. 1979b. “Re lationships Be tween Struc ture and Technology at the Organizational and Job Levels.” Jour nal of Man age ment Studies 16:70-79. Gerwin, Don ald. 1981. “Re la tion ships Be tween Struc ture and Tech nol ogy.” Pp. 3-38 in Handbook of Or ga ni za tion De sign: Vol. 2. Re modeling Or ga ni za tions and Their

300

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

Environments, ed ited by P. Nystrom and W. Starbuck. New York: Ox ford Uni ver sity Press. Gerwin, Don ald, and Wade Christoffel. 1974. “Or ga ni za tional Struc ture and Tech nol ogy: A Com puter Model Ap proach.” ManagementScience 20:1531-1542. Ghoshal, Sumantra, and P. Moran. 1996. “Bad for Prac tice: A Cri tique of the Trans ac tion Cost The ory.” Acad emy of Man age ment Re view 21:13-47. Ghoshal, Sumantra, and Nitin Nohria. 1993. “Horses for Courses: Or ganizational Forms for Mul ti na tional Cor po ra tions.” SloanManagementReview, Win ter:23-35. Giddens, An thony. 1984. The Con sti tu tion of So ci ety . Cam bridge: Pol ity. Goldman, P. 1973. “Size and Dif ferentiation in Or ganizations:A Test of a The ory.” Pacific So cio log i cal Re view 16:89-105. Gouldner, Alvin W. 1954. Pat terns of In dus trial Bu reau cracy. Glencoe, IL: Free Press. Grant, Rob ert M. 1993. Re struc turing and Stra te gic Change in the Oil In dus try . Milan, It aly: FrancoAgneli. Greenwood, Royston, and C. R. Hinings. 1976a. “Cen tralization Revisited: Fur ther Discussion.” Ad min is tra tive Sci ence Quar terly 21:151-155. Greenwood, Royston, and C. R. Hinings. 1976b. “Con tingency The ory and Public Bureaucracies.” Pp. 87-101 in Or ga ni za tional Struc ture: Ex ten sions and Rep li ca tions: The Aston Programme II, edited by D. S. Pugh and C. R. Hinings. Farnborough, Hants, UK: Saxon House. Gresov, Chris to pher. 1989. “Ex ploring Fit and Mis fit With Mul ti ple Con tin gencies.” Ad min is tra tive Sci ence Quar terly 34:431-453. Gresov, Chris topher. 1990. “Ef fects of De pendence and Tasks on Unit De sign and Efficiency.” OrganizationStudies 11:503-529. Gresov, Chris to pher, and Rob ert Drazin. 1997. “Equifinality: Func tional Equiv a lence in Or ga ni za tion De sign.” Acad emy of Man age ment Jour nal 22:403-428. Grinyer, Pe ter H. 1982. “Dis cus sion Note: Divisionalization and Size—A Re join der.” OrganizationStudies 3:339-350. Grinyer, Pe ter H., and Masoud Yasai-Ardekani. 1980. “Di men sions of Or ga ni za tional Struc ture: A Crit i cal Rep li ca tion.” Acad emy of Man age ment Jour nal 23:405-421. Grinyer, Pe ter H., and Masoud Yasai-Ardekani. 1981. “Strat egy, Struc ture, Size and Bureaucracy.” Acad emy of Man age ment Jour nal 24:471-486. Grinyer, Peter H., Masoud Yasai-Ardekani, and Shawki Al-Bazzaz. 1980. “Strat egy, Structure, the En vironment, and Fi nancial Per formance in 48 United King dom Com panies.” Acad emy of Man age ment Jour nal 23:193-220. Hage, Jerald. 1965. “An Ax i om atic The ory of Or ga ni za tions.” AdministrativeScience Quarterly 10:289-320. Hage, Jerald. 1974. CommunicationsandOrganizationalControl:CyberneticsinHealth and Wel fare Set tings . New York: Wiley InterScience. Hage, Jerald. 1980. The ories of Or ga ni za tion: Form, Pro cess and Trans for ma tion . New York: John Wiley. Hage, Jerald. 1988. Fu tures of Or ga ni za tions: In no vating to Adopt Strat egy and Hu man Re sources to Rapid Tech no log i cal Change . Lexington, MA: Lexington Books. Hage, Jerald, and Mi chael Aiken. 1967a. “Pro gram Change and Or ga ni za tional Prop erties: A Com par a tive Anal y sis.” Amer i can Jour nal of So ci ol ogy 72:503-519.

References

301

Hage, Jerald, and Mi chael Aiken. 1967b. “Re lationship of Cen tralization to Other Struc tural Prop erties.” Ad min is tra tive Sci ence Quar terly 12:72-92. Hage, Jerald, and Mi chael Aiken. 1969. “Rou tine Technology, So cial Struc ture and Organizational Goals.” Ad min is tra tive Sci ence Quar terly 14:366-376. Hage, Jerald, and Mi chael Aiken. 1970. So cial Change in Com plex Or ga ni za tions. New York: Ran dom House. Hage, Jerald, and Rob ert Dewar. 1973 “Elite Values Ver sus Or ga ni za tional Struc ture in Pre dicting In no va tion.” Ad min is tra tive Sci ence Quar terly 18:279-290. Hall, D. J., and M. D. Saias. 1980. “Strat egy Fol lows Struc ture!” StrategicManagement Journal 1:149-163. Hall, Rich ard H. 1963. “The Con cept of Bu reaucracy: An Em pirical As sess ment.” Administrative Sci ence Quar terly 69:32-40. Hambrick, Don ald C. 1981. “En vi ron ment, Strat egy, and Power Within Top Man age ment Teams.” Ad min is tra tive Sci ence Quar terly 26:253-275. Ham il ton, R. T., and G. S. Shergill. 1992. “The Re la tion ship Be tween Strat egy-Struc ture Fit and Fi nan cial Per for mance in New Zea land: Ev i dence of Gen er al ity and Va lid ity With En hanced Con trols.” Jour nal of Man age ment Studies 29:95-113. Ham il ton, R. T., and G. S. Shergill. 1993. The Logic of New Zea land Busi ness: Strat egy, Struc ture, and Per for mance . Auckland, New Zea land: Ox ford Uni ver sity Press. Hannan, Mi chael T., and John Free man. 1977. “The Pop u la tion Ecol ogy of Or ga ni za tions.” Amer i can Jour nal of So ci ol ogy 82:929-964. Hannan, Mi chael T., and John Free man. 1984. “Struc tural In er tia and Or ga ni za tional Change.” AmericanSociologicalReview 49:149-164. Hannan, Mi chael T., and John Free man. 1989. Organizational Ecol ogy. Cam bridge, MA: Har vard Uni ver sity Press. Hannan, Mi chael T., J. Ranger-Moore, and J. Banaszak-Holl. 1990. “Com pe ti tion and the Evo lu tion of Or ga ni za tional Size Dis tri bu tions.” Pp. 246-68 in Organizational Evo lu tion: New Di rec tions, ed ited by J. V. Singh. Newbury Park, CA: Sage. Har ris, Barry C. 1983. Or ga ni za tion: The Ef fect on Large Cor po ra tions . Ann Ar bor, MI: UMI Re search. Heise, D. R. 1972. “How Do I Know My Data? Let Me Count the Ways.” Administrative Sci ence Quar terly 17:58-61. Hickson, Da vid J., C. R. Hinings, C. A. Lee, R. E. Schneck, and J. M. Pennings. 1971. “A Stra te gic Con tin gencies The ory of Intraorganizational Power.” Administrative Sci ence Quar terly 16:216-229. Hickson, Da vid J., C. R. Hinings, C. J. McMillan, and J. P. Schwitter. 1974. “The Cul tureFree Con text of Or ganization Struc ture: A Trinational Com par i son.” Sociology 8:59-80. Hickson, David J. and C. J. McMillan, eds. 1981. Or ga ni za tion and Na tion: The Aston Programme IV . Farnborough, Hants, UK: Gower. Hickson, Da vid J., Derek S. Pugh, and Di ana G. Pheysey. 1969. “Op er a tions Tech nol ogy and Or ga ni za tion Struc ture: An Em pir i cal Re ap praisal.” AdministrativeScience Quarterly 14:378-397. Hill, Charles W. L. 1985a. “In ter nal Or ga ni za tion and En ter prise Per for mance: Some UK Ev i dence.” Man a ge rial and De ci sion Eco nom ics 6:210-216.

302

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

Hill, Charles W. L. 1985b. “Ol i ver Wil liam son and the M-form Firm: A Crit i cal Re view.” Jour nal of Eco nomic Is sues 19:731-751. Hill, Charles W. L. 1988a. “Cor porate Con trol Type, Strat egy, Size and Fi nancial Performance.” Jour nal of Man age ment Studies 25:403-417. Hill, Charles W. L., Mi chael A. Hitt, and Rob ert E. Hoskisson. 1992. “Co operative Versus Com pet i tive Struc tures in Re lated and Un re lated Di ver sified Firms.” Or ganizationScience 3:501-521. Hill, Charles W. L., and J. F. Pickering 1986. “Divisionalization, De cen tral iza tion and Performance of Large United King dom Com panies.” Journal of Man agement Studies 23:26-50. Hill, Charles W. L., and S. A. Snell. 1988. “Ex ter nal Con trol, Cor po rate Strat egy, and Firm Per for mance in Re search-In ten sive In dus tries.” Stra te gic Man age ment Jour nal 9:577-590. Hilmer, Fred er ick G. 1998. Strictly Board room: Im proving Gov er nance to En hance Com pany Per for mance , 2nd ed. Mel bourne, Vic to ria, Aus tra lia: In for ma tion Aus tra lia. Hilton, G. 1972. “Causal In ference Anal ysis: A Se ductive Pro cess.” Administrative Science Quar terly 17:44-55. Hinings, C. R., Da vid J. Hickson, J. M. Pennings, and R. E. Schneck. 1974. “Struc tural Con di tions of Intraorganizational Power.” Ad min is tra tive Sci ence Quar terly 19:22-44. Hinings, C. R., and Glo ria Lee. 1971. “Di men sions of Or ga ni za tion Struc ture and Their Con text: A Rep li ca tion.” Sociology 5:83-93. Hinings, C. R., Derek S. Pugh, Da vid J. Hickson, and C. Turner. 1967. “An Ap proach to the Study of Bu reau cracy.” Sociology 1:61-72. Hinings, C. R., S. Ranson, and A. Bryman. 1976. “Churches as Or ga ni za tions: Struc ture and Con text.” Pp. 102-114 in OrganizationalStructure:ExtensionsandReplications: The Aston Programme II, edited by D. S. Pugh and C. R. Hinings. Farnborough, Hants, UK: Saxon House. Hirsch, Paul. 1975. “Or ga ni za tional Ef fec tive ness and the In sti tu tional En vi ron ment.” Ad min is tra tive Sci ence Quar terly 20:327-344. Hoffman, James J., Nancy M. Carter, and John B. Cul len. 1994. “The Ef fect of LagStructure Iden ti fi ca tion When Testing for Fit.” OrganizationStudies 15:829-848. Holdaway, Ed ward A., John F. Newberry, Da vid J. Hickson, and R. Pe ter Heron. 1975. “Dimensions of Or ganizations in Com plex Societies: The Ed ucational Sec tor.” Administrative Sci ence Quar terly 20:37-58. Hopkins, H. Don ald. 1988. “Firm Size: The Interchangeability of Mea sures.” Human Relations 41:91-102. Hoskisson, R. E. 1987. “Multidivisional Struc ture and Per for mance: The Con tin gency of Di ver si fi ca tion Strat egy.” Acad emy of Man age ment Jour nal 30:625-644. Hoskisson, R. E., and C. S. Galbraith. 1985. “The Ef fect of Quan tum Ver sus In cre men tal M-form Re or ga ni za tion on Per for mance: A Time-se ries Ex plo ra tion of In ter ven tion Dy nam ics.” Jour nal of Man age ment, 11:55-70. Hoskisson, Rob ert E., and Mi chael A. Hitt. 1994. Downscoping: How to Tame the Di ver sified Firm . New York: Ox ford Uni ver sity Press. Hull, Frank. 1988. “In ven tions From R&D Or ga ni za tional De signs for Ef fi cient Re search Per for mance.” Sociology 22:393-415.

References

303

Hull, Frank, and Paul D. Col lins. 1987. “High-Tech nol ogy Batch Pro duc tion Sys tems: Wood ward’s Missing Type.” Acad emy of Man age ment Jour nal 30:786-797. Hummon, Nor man P. 1971. “A Math e mat i cal The ory of Dif fer en ti a tion in Or ga ni za tions.” AmericanSociologicalReview 36:297-303. Hunter, John E., Frank L. Schmidt, and Gregg B. Jack son. 1982. Meta-Analysis: Cumulating Re search Find ings Across Studies. Beverly Hills, CA: Sage. Inkson, J. H. K., Derek S. Pugh, and Da vid J. Hickson. 1970. “Or ga ni za tion Con text and Struc ture: An Ab bre vi ated Rep li ca tion.” Ad min is tra tive Sci ence Quar terly 15:318-329. Isajew, Wsevolod. 1968. CausationandFunctionalisminSociology . Lon don: Routlege & Kegan Paul. Jaques, Elliott. 1976. A Gen eral The ory of Bu reau cracy . Lon don: Heinemann. Jarley, Paul, Jack Fiorito, and John Thomas Delaney. 1997. “A Struc tural Con tin gency Ap proach to Bu reau cracy and De moc racy in U.S. Na tional Un ions.” Academy of ManagementJournal 40:831-861. Jennings, Dan iel F., and Sam uel L. Sea man. 1994. “High and Low Levels of Or ga ni za tional Ad ap ta tion: An Em pir i cal Anal y sis of Strat egy, Struc ture and Per for mance.” Stra te gic Man age ment Jour nal 15:459-475. Jensen, Mi chael C., and Wil liam H. Meckling. 1976. “The ory of the Firm: Man a ge rial Be hav ior, Agency Costs and Own er ship Struc ture.” Jour nal of Fi nan cial Eco nom ics 3:305-360. Johns, Gary. 1981. “Dif fer ence Score Mea sures of Or ga ni za tional Be hav ior Vari ables: A Cri tique.” Or ga ni za tional Be hav ior and Hu man Per for mance 27:443-463. Johnson, G. 1987. Strategic Change and the Man agement Pro cess. Oxford UK: Ba sil Blackwell. Joyce, William F. 1986. “Matrix Organization: A So cial Experiment.” Academy of Management Jour nal 29:536-561. Keller, Rob ert T. 1994. “Tech nol ogy-In for ma tion Pro cessing Fit and the Per for mance of R&D Pro ject Groups: A Test of Con tin gency The ory.” Acad emy of Man age ment Journal 37:167-179. Kenny, Gra ham K., and Da vid C. Wil son. 1984. “The In ter de part men tal In flu ence of Managers: In dividual Sub-Unit Per spec tives.” Journal of Man agement Studies 21:409-427. Kesner, Idalene F. 1987. “Di rec tors, Stock Own er ship and Or ga ni za tional Per for mance: An In ves ti ga tion of For tune 500 Com panies.” JournalofManagement 13:499-508. Ketchen, Da vid J., James G. Combs, Craig J. Rus sell, Chris Shook, Michelle A. Dean, Janet Runge, Franz T. Lohrke, Stefanie E. Naumann, Dawn Ebe Haptonstahl, Rob ert Baker, Brenden A. Beckstein, Charles Han dler, Heather Honig, and Ste phen Lamoureux. 1997. “Or ga ni za tional Con fig u ra tions and Per for mance: A Meta-Anal y sis.” Acad emy of Man age ment Jour nal 40:223-240. Ketchen, Da vid J., James B. Thomas, and Charles C. Snow. 1993. “Or ganizational Configurations and Per formance: A Com parison of The oretical Ap proaches.” Acad emy of Man age ment Jour nal 36:1278-1313. Khandwalla, Pradip N. 1973. “Vi a ble and Ef fec tive Or ga ni za tional De signs of Firms.” Acad emy of Man age ment Jour nal 16:481-495. Khandwalla, Pradip N. 1974. “Mass Out put Ori en ta tion of Op er a tions Tech nol ogy and Or ga ni za tional Struc ture.” Ad min is tra tive Sci ence Quar terly 19:74-97.

304

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

Khandwalla, Pradip N. 1977. The De sign of Or ga ni za tions . New York: Har court Brace Jovanovich. Kimberly, John R. 1976. “Or ga ni za tional Size and the Structuralist Per spec tive: A Re view, Cri tique and Pro posal.” Ad min is tra tive Sci ence Quar terly 21:571-597. Knight, Ken neth. 1977. Ma trix Man age ment. New York: PBI-Petrocelli Books. Kolodny, Harvey F. 1979. “Evo lu tion to a Ma trix Or ga ni za tion.” Acad emy of Man age ment Re view 4:543-553. Kraft, Ken neth L., George M. Puia, and Jerald Hage. 1995. “Struc tural Con tin gency Theory Re visited: Main Ef fects Versus In teractions in Child’s Na tional Study of Manufacturing and Ser vice Sec tors.” Revue Canadienne des Sci ences de l’Administration—Ca na dian Jour nal of Ad min is tra tive Sci ences 12:182-194. Kuc, B., D. J. Hickson, and C. J. McMillan. 1981. “Cen trally Planned De vel op ment: A Comparison of Pol ish Fac tories With Equiv a lents in Brit ain, Ja pan and Swe den.” Pp. 75-91 in Organization and Na tion: The Aston Programme IV, ed ited by D. J. Hickson, and C. J. McMillan. Farnborough, Hants, UK: Gower. Kuczynski, M. G. 1986. “Re cent De vel op ments in Busi ness Cy cle The ory.” Jour nal of Eco nomic Dy nam ics and Con trol 10:255-260. Kuhn, Thomas S. 1970. TheStructureofScientificRevolutions , 2nd ed. Chicago: Uni ver sity of Chi cago Press. Lakatos, I. 1974. “Fal si fi ca tion and the Meth od ol ogy of Sci en tific Re search Programmes.” Pp. 91-196 in Crit i cism and the Growth of Knowl edge, ed ited by I. Lakatos and A. Musgrave. Cam bridge: Cam bridge Uni ver sity Press. Lammers, Cornelis J., and Da vid J. Hickson, eds. 1979. Or ga ni za tions Alike and Un like: International and Inter-In stitutional Studies in the So ciology of Or ganization. London: Routledge & Kegan Paul. Law rence, Paul R. 1993. “The Con tin gency Ap proach to Or ga ni za tional De sign.” Pp. 9-18 in Hand book of Or ga ni za tional Be hav ior, ed ited by Rob ert T. Golembiewski. New York: Mar cel Dekker. Law rence, Paul R., and Da vis Dyer. 1983. Re newing Amer i can In dus try: Or ga nizing for Ef fi ciency and In no va tion . New York: Free Press. Law rence, Paul R., and Jay W. Lorsch. 1967. Or ga ni za tion and En vi ron ment: Man aging Differentiation and In tegration . Boston: Har vard University, Grad uate School of Busi ness Ad min is tra tion, Di vi sion of Re search. Leavitt, Har old J. 1951. “Some Ef fects of Cer tain Com mu ni ca tion Pat terns on Group Per for mance.” Jour nal of Ab nor mal and So cial Psy chol ogy 46:38-50. Lewin, Arie Y., and Carroll U. Stephens. 1994. “CEO At ti tudes as De ter mi nants of Or ga ni za tion De sign: An In te grated Model.” OrganizationStudies 15:183-212. Likert, Rensis. 1961. New Pat terns of Man age ment. New York: McGraw-Hill. Lin coln, James R., Mitsuyo Hanada, and Kerry McBride. 1986. “Or ga ni za tional Struc tures in Jap a nese and U.S. Man u fac turing.” AdministrativeScienceQuarterly 31:338-364. Lioukas, S. K., and D. A. Xerokostas. 1982. “Size and Ad min is tra tive In ten sity in Or ga ni za tional Di vi sions.” ManagementScience 28:854-868. Littler, Craig R., and Thomas Bram ble.1995.“ConceptualisingOrganizational Re struc turing in the 1990s.” Jour nal of the Aus tra lian and New Zea land Acad emy of Management 1:45-56.

References

305

Locke, E. A., and G. P. Latham. 1990. A The ory of Goal-Set ting and Task Per for mance. Englewood Cliffs, NJ: Prentice Hall. Lockwood, D. 1964. “So cial Integration and Sys tem Integration.” Pp. 244-257 in Exploration in So cial Change, edited by George K. Zollschan and Wal ter Hirsch. Lon don: Routledge & Kegan Paul. Lorsch, Jay W., and Ste phen A. Al len.1973. Man aging Di ver sity and Inter-De pend ence: An Or ganizational Study of Multidivisional Firms. Boston: Har vard University, Grad u ate School of Ad min is tra tion, Di vi sion of Re search. Lorsch, Jay W., and Paul R. Law rence. 1972. “En vi ron men tal Fac tors and Or ga ni za tional In te gra tion.” Pp. 38-48 in Or ga ni za tional Planning: Cases and Con cepts, ed ited by Jay W. Lorsch and Paul R. Law rence. Home wood, IL: Irwin/Dorsey. Lorsch, Jay W., and John J. Morse. 1974. Or ga ni za tions and Their Mem bers: A Con tin gency The ory Ap proach . New York: Harper and Row. MacMillan, Al ex an der, and Rich ard L. Daft. 1979. “Ad min is tra tive In ten sity and Ra tio Vari ables: The Case Against Defi ni tional De pend ency.” So cial Forces 58:228-248. MacMillan, Al exander, and Rich ard L. Daft. 1984. “In ferences About Eco nomics of Scale DO De pend on the Form of Sta tistical Anal ysis: A Rec onciliation.” Social Forces 62:1059-1067. Mahajan, Vijay, S. Sharma, and R. A. Bettis. 1988. “The Adop tion of the M-form Organizational Struc ture: A Test of Im itation Hy poth e sis.” Management Science 34:1188-1201. Mahoney, Jo seph T. 1992. “The Adop tion of the Multidivisional Form of Or ga ni za tion: A Con tin gency Model.” Jour nal of Man age ment Studies 29:49-72. Mansfield, Roger. 1973. “Bu reau cracy and Cen tral iza tion: An Ex am i na tion of Or ga ni za tional Struc ture.” Ad min is tra tive Sci ence Quar terly 18:477-488. March, James G., with Rob ert I. Sutton. 1999. “Or ga ni za tional Per for mance as a De pend ent Vari able.” Pp. 338-353 in James G. March, The Pur suit of Or ga ni za tional In telligence . Ox ford UK: Ba sil Blackwell. Marsden, Peter V., Cynthia R. Cook, and Arne L. Kalleberg. 1994. “Or ganizational Struc tures: Co or di na tion and Con trol.” AmericanBehavioralScientist 37:911-929. Marsh, Rob ert M. 1992. “Re search Note: Cen tral iza tion of De ci sion-Making in Jap a nese Fac tories.” OrganizationStudies 13:261-274. Marsh, Rob ert M., and Hiroshi Mannari. 1976. Mod ern iza tion and the Jap a nese Fac tory. Prince ton, NJ: Prince ton Uni ver sity Press. Marsh, Rob ert M., and Hiroshi Mannari. 1980. “Tech no log i cal Im pli ca tions The ory: A Jap a nese Test.” OrganizationStudies 1:161-183. Marsh, Rob ert M., and Hiroshi Mannari. 1981. “Tech nol ogy and Size as De ter mi nants of the Or ganizational Struc ture of Jap anese Fac tories.” Administrative Sci ence Quarterly 26:33-57. Marsh, Rob ert M., and Hiroshi Mannari. 1989. “The Size Im perative?Longitudinal Tests.” OrganizationStudies 10:83-95. Mar tin, Jo anne. 1992. Cul tures in Or ga ni za tions: Three Per spec tives . New York: Ox ford Uni ver sity Press. Mayhew, Bruce H., Jr., Thomas F. James, and Grant W. Childers. 1972. “Sys tem Size and Structural Dif ferentiation in Mil itary Organizations: Testing a Harmonic Se ries Model of the Di vi sion of La bor.” Amer i can Jour nal of So ci ol ogy 77:750-765.

306

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

McKelvey, Bill, and Howard E. Aldrich. 1983. “Pop ulations, Nat ural Se lection and Applied Or ga ni za tional Sci ence.” Ad min is tra tive Sci ence Quar terly 28:101-128. McKinley, William, Carol M. Sanchez, and Al len G. Schick. 1995. “Or ganizational Downsizing: Con straining, Cloning and Learning.” Academy of Man agement Executive 9:32-44. McMillan, Charles J., Da vid J. Hickson, Chris topher R. Hinings, and Rodney E. Schneck. 1973. “The Struc ture of Work Or ga ni za tions Across So ci eties.” Acad emy of ManagementJournal 16:555-569. Mer ton, R. K. 1949. So cial The ory and So cial Struc ture . Chi cago: Free Press. Meyer, Alan D., Anne S. Tsui, and C. R. Hinings. 1993. “Con fig u ra tional Ap proaches to OrganizationalAnalysis.” Acad emy of Man age ment Jour nal 36:1175-1195. Meyer, John W., and Brian Ro wan. 1977. “In stitutionalized Organizations: For mal Struc ture as Myth and Cer e mony.” Amer i can Jour nal of So ci ol ogy 83:340-363. Meyer, John W., and W. Rich ard Scott, with the as sis tance of B. Ro wan and T. E. Deal. 1983. OrganizationalEnvironments:RitualandRationality. Beverly Hills, CA: Sage. Meyer, Mar shall W. 1971. “Some Con straints in An a lyzing Data on Or ga ni za tional Struc tures: A Com ment on Blau’s Pa per.” AmericanSociologicalReview 36:294-297. Meyer, Mar shall W. 1972. “Size and the Struc ture of Or ga ni za tions: A Causal Anal y sis.” AmericanSociologicalReview 37:434-441. Meyer, Mar shall W. 1979. Change in Pub lic Bu reau cracies . Cam bridge, UK: Cam bridge University Press. Meyer, Mar shall W., and Vipin Gupta. 1994. “The Per for mance Par a dox.” Re search in OrganizationalBehavior 16:309-369. Meyer, Mar shall W., and Lynne G. Zucker. 1989. Permanently Fail ing Organizations. Newbury Park, CA: Sage. Miles, Ray mond E., and Charles C. Snow. 1978. Or ga ni za tional Strat egy, Struc ture, and Process . To kyo: McGraw-Hill Kogakusha. Miller, C. Chet, Wil liam H. Glick, Yau-De Wang, and George P. Huber. 1991. “Un der standing Technology-Structure Re lationships: The ory De velopment and MetaAn a lytic Testing.” Acad emy of Man age ment Jour nal 34:370-399. Miller, Danny. 1986. “Con fig u ra tions of Strat egy and Struc ture: To wards a Syn the sis.” Stra te gic Man age ment Jour nal 7:233-249. Miller, Danny. 1990. The Icarus Par a dox: How Ex cep tional Com panies Bring About their Own Down fall: New Les sons in the Dy namics of Cor porate Success, De cline, and Renewal . New York: HarperCol lins. Miller, Danny, and Cornelia Droge. 1986. “Psy chological and Tra ditional De ter minants of Struc ture.” Ad min is tra tive Sci ence Quar terly 31:539-560. Miller, Danny, Cornelia Droge, and Jean-Ma rie Toulouse. 1988. “Stra te gic Pro cess and Con tent as Me di a tors Be tween Or ga ni za tional Con text and Struc ture.” Acad emy of ManagementJournal 31:544-569. Miller, Danny, and Jean-Ma rie Toulouse. 1986. “Chief Ex ec u tive Per son al ity and Cor po rate Strat egy and Struc ture in Small Firms.” ManagementScience 32:1389-1409. Miller, George A. 1987. “Meta-Anal y sis and the Cul ture-Free Hy poth e sis.” Organization Studies 8:309-326.

References

307

Mindlin, Sergio E., and Howard Aldrich. 1975. “Interorganizational De pend ence: A Review of the Con cept and a Re ex am i na tion of the Find ings of the Aston Group.” Ad min is tra tive Sci ence Quar terly 20:382-392. Miner, J. B. 1982. The ories of Or ga ni za tional Struc ture and Pro cess . Chi cago: Dryden. Mintzberg, H. 1979. The Struc turing of Or ganizations A Syn thesis of the Re search. Englewood Cliffs, NJ: Prentice Hall. Mintzberg, H. 1983. Power In and Around Or ganizations. Englewood Cliffs, NJ: Prentice Hall. Mintzberg, H., and James A. Wa ters. 1982. “Tracking Strat egy in an En tre pre neur ial Firm.” Acad emy of Man age ment Jour nal 25:465-499. Mohr, Law rence B. 1971. “Or ga ni za tional Tech nol ogy and Or ga ni za tional Struc ture.” Ad min is tra tive Sci ence Quar terly 27:420-434. Montanari, John R., 1979. “Stra tegic Choice: A The oretical Anal y sis.” Journal of Management Studies 16:202-221. Montanari, John R., and Philip J. Adelman. 1987. “The Ad min is tra tive Com po nent of Organizations and the Ratchet Ef fect: A Critique of Cross-Sec tional Studies.” Journal of Man age ment Studies 24:113-123. Moran, Pe ter, and Sumantra Ghoshal. 1996. “The ories of Eco nomic Or ga ni za tion: The Case for Re al ism and Bal ance.” Acad emy of Man age ment Re view 21:58-72. Muth, Melinda, and Lex Donaldson. 1998. “Stew ard ship The ory and Board Struc ture: A Con tin gency Ap proach.” CorporateGovernance:AnInternationalReview 6:2-28. Nohria, Nitin, and Sumantra Ghoshal. 1997. The Dif fer en ti ated Net work: Or ga nizing Mul ti na tional Cor po ra tions for Value Cre ation . San Fran cisco: Jossey-Bass. Nunnally, J. C. 1978. Psychometric The ory . New York: McGraw-Hill. Ocasio, Wil liam. 1995. “The En act ment of Eco nomic Ad ver sity: A Reconcept ual ization of The ories of Fail ure-In duced Change and Threat-Ri gid ity.” Re search in Or ga ni za tional Be hav ior 17:287-331. Palmer, D., R. Friedland, P. Devereaux Jennings, and M. E. Powers. 1987. “The Eco nom ics and Pol i tics of Struc ture: The Multidivisional Form and Large U.S. Cor po ra tion.” Administrative Sci ence Quar terly 32:25-48. Palmer, D., P. Devereaux Jennings, and Xuegang Zhou. 1993. “Late Adop tion of the Multidivisional Form by Large U.S. Cor porations: In stitutional, Political, and Economic Ac counts.” Ad min is tra tive Sci ence Quar terly 38:100-131. Par kin son, C. North cote. 1957. Par kin son’s Law and Other Studies in Ad min is tra tion . Boston: Hought on Mifflin. Par sons, Talcott. 1951. The So cial Sys tem . Lon don: Routledge & Kegan Paul. Parsons, Talcott. 1961. “Sug gestions for a So ciological Ap proach to the The ory of Organizations.” Pp. 32-47 in Com plex Or ga ni za tions: A So cio log i cal Reader, ed ited by Amitai Etzioni. New York: Holt, Rinehart and Winston. Pavan, Rob ert J. 1976. “Strat egy and Struc ture: The Ital ian Ex perience.” Journal of Economics and Busi ness 28:254-260. Payne, Roy L., and Roger Mansfield. 1973. “Re la tion ships of Per cep tions of Or ga ni za tional Cli mate to Or ga ni za tional Struc ture, Con text, and Hi er ar chi cal Po si tion.” Ad min is tra tive Sci ence Quar terly 18:515-526. Pennings, J. M. 1975. “The Rel e vance of the Struc tural Con tin gency Model for Or ga ni za tional Ef fec tive ness.” Ad min is tra tive Sci ence Quar terly 20:393-410.

308

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

Pennings, J. M. 1987. “Struc tural Con tin gency The ory: A Multivariate Test.” Organiza tion Studies 8:223-240. Pennings, J. M. 1992. “Struc tural Con tingency The ory: A Reap praisal.” Research in OrganizationalBehavior 14:267-309. Penrose, Edith T. 1959. The The ory of the Growth of the Firm . Ox ford, UK: Ba sil Blackwell. Perrow, Charles. 1967. “A Frame work for the Com par a tive Anal y sis of Or ga ni za tions.” AmericanSociologicalReview 32:194-208. Perrow, Charles. 1970. “De part men tal Power and Per spec tives in In dus trial Firms.” Pp. 59-89 in Power in Or ga ni za tions, ed ited by M. N. Zald. Nash ville, TN: Vanderbilt Uni ver sity Press. Perrow, Charles. 1986. Complex Organizations: A Crit ical Es say, 3rd ed. New York: Random House. Pfeffer, Jeffrey. 1972. “Size and Com position of Cor porate Boards of Di rec tors: the Organization and its En vi ron ment.” Ad min is tra tive Sci ence Quar terly 17:218-228. Pfeffer, Jeffrey. 1982. OrganizationsandOrganizationTheory. Marshfield, MA: Pitman. Pfeffer, Jeffrey. 1997. New Di rec tions for Or ga ni za tion The ory: Prob lems and Pros pects . New York: Ox ford Uni ver sity Press. Pfeffer, Jeffrey, and Ger ald R. Salancik. 1978. The Ex ter nal Con trol of Or ga ni za tions: A Re source De pend ence Per spec tive . New York: Harper and Row. Pickle, Hal, and F. Friedlander. 1967. “Seven so ci etal cri te ria of or ga ni za tional suc cess. ” PersonnelPsychology 20:165-178. Pitts, Rob ert A. 1974. “In cen tive Com pen sa tion and Or ga ni za tion De sign.” Personnel Journal 53:338-344. Pitts, Rob ert A. 1976. “Di ver si fi ca tion Strat egies and Or ga ni za tional Pol icies of Large Di ver sified Firms.” Jour nal of Eco nom ics and Busi ness 28:181-188. Pitts, Rob ert A. 1977. “Strat egies and Struc tures for Di ver si fi ca tion.” Acad emy of Man age ment Jour nal 20:197-208. Pop per, K. R. 1945. The Open So ci ety and Its En emies: Vol. 2. The High Tide of Proph ecy: Hegel, Marx and the Af ter math. Lon don: Routledge & Kegan Paul. Pop per, K. R. 1959. The Logic of Sci en tific Dis cov ery . New York: Harper and Row. Por ter, Mi chael. 1980. CompetitiveStrategy . New York: Free Press. Powell, Thomas C. 1992. “Or ganizational Align ment as Com petitive Ad vantage.” Strategic Man age ment Jour nal 13:119-134. Powell, Wal ter W., and Paul DiMaggio, eds. 1991. The New Institutionalism in Or ga ni zationalAnalysis . Chi cago: Uni ver sity of Chi cago Press. Price, James L. 1972. Hand book of Or ga ni za tional Mea sure ment . Lexington, MA: D. C. Heath. Also, up dated as, 1997. “Hand book of Or ga ni za tional Mea sure ment.” In ter na tional Jour nal of Man power 18:303-558. Priem, R. L. 1992. “An Ap pli ca tion of Met ric Con joint Anal y sis for the Eval u a tion of Top Man agers’ In di vid ual Stra te gic De ci sion Making Pro cesses: A Re search Note.” Stra te gic Man age ment Jour nal 13:143-151. Priem, R. L. 1994. “Ex ecutive Judge ment, Organizational Congruence, and Firm Performance.” OrganizationScience 5:421-437. Pugh, Derek S. 1981a. “The Aston Pro gram Per spec tive: The Aston Pro gram of Re search: Ret ro spect and Pros pect.” Pp. 135-166 in Per spec tives on Or ga ni za tion De sign and Be hav ior, ed ited by An drew Van de Ven and Wil liam Joyce. New York: John Wiley.

References

309

Pugh, Derek S. 1981b. “Re join der to Starbuck.” Pp. 199-203 in Per spec tives on Or ga ni za tion De sign and Be hav ior, ed ited by Andrew Van de Ven and William Joyce. New York: John Wiley. Pugh, Derek S., and Da vid J. Hickson. 1972. “To the Ed i tor: Causal In fer ence and the Aston Studies.” Ad min is tra tive Sci ence Quar terly 17:273-276. Pugh, Derek S., and Da vid J. Hickson. 1976. Or ga ni za tional Struc ture in Its Con text: The Aston Programme I. Farnborough, Hants, UK: Saxon House. Pugh, Derek S., David J. Hickson, and C. R. Hinings. 1969. “An Em pir i cal Tax on omy of Structures of Work Or ga ni za tions.” Ad min is tra tive Sci ence Quar terly 14:115-126. Pugh, Derek S., Da vid J. Hickson, C. R. Hinings, K. M. Mac don ald, C. Turner, and T. Lupton. 1963. “A Con cep tual Scheme for Or ga ni za tional Anal y sis.” Administrative Sci ence Quar terly 8:289-315. Pugh, Derek S., Da vid J. Hickson, C. R. Hinings, and C. Turner. 1968. “Di men sions of Or ga ni za tion Struc ture.” Ad min is tra tive Sci ence Quar terly 13:65-105. Pugh, Derek S., Da vid J. Hickson, C. R. Hinings, and C. Turner. 1969. “The Con text of Or ga ni za tion Struc tures.” Ad min is tra tive Sci ence Quar terly 14:91-114. Pugh, Derek S., and C. R. Hinings. 1976. OrganizationalStructure: Ex tensions and Rep li ca tions: The Aston Programme II. Farnborough, Hants, UK: Saxon House. Pugh, Derek S., and R. L. Payne, eds. 1977. Or ga ni za tional Be hav iour in Its Con text: The Aston Programme III . West mead, Farnborough, Hants, UK: Saxon House, Teakfield. Randolph, W. Alan, and Greg ory G. Dess. 1984. “The Con gruence Per spective and Organization De sign: A Con cep tual Model and Multivariate Re search Ap proach.” Acad emy of Man age ment Re view 9:114-127. Reeves, T. K., and B. A. Turner. 1972. “A The ory of Or ga ni za tion and Be hav ior in Batch Pro duc tion Fac tories.” Ad min is tra tive Sci ence Quar terly 17:81-98. Reimann, Ber nard C. 1973. “On the Di men sions of Bu reau cratic Struc ture: An Em pir i cal Re ap praisal.” Ad min is tra tive Sci ence Quar terly 18:462-476. Reimann, Ber nard C. 1974. “Di mensions of Struc ture in Effective Or ganizations: Some Em pir i cal Ev i dence.” Acad emy of Man age ment Jour nal 17:693-708. Reimann, Ber nard C. 1977. “Di men sions of Or ga ni za tional Tech nol ogy and Struc ture: An Ex plor atory Study.” Hu man Re la tions 30:545-566. Reimann, Ber nard C. 1979. “Par kin son Re visited: A Com po nent Anal y sis of the Use of Staff Spe cial ists in Man u fac turing Or ga ni za tions.” Hu man Re la tions 32:625-641. Reimann, Ber nard C. 1980. “Or ga ni za tion Struc ture and Tech nol ogy in Man u fac turing: System Versus Work Flow Level Per spec tives.” Academy of Man agement Journal 23:61-77. Reimann, Ber nard C. and Giorgio Inzerilli. 1979. “A Com par a tive Anal y sis of Em pir i cal Re search on Tech nol ogy and Struc ture.” Jour nal of Man age ment 5:167-192. Richards, V. G. 1980. “Research Note: The Aston Data bank.” Organization Studies 1:271-278. Routamaa, Vesa. 1985. “Or ga ni za tional Struc turing: An Em pir i cal Anal y sis of the Re la tion ships and Di men sions of Struc tures in Cer tain Finn ish Com panies.” Journal of Man age ment Studies 22:498-522. Roy, Wil liam G. 1990. “Func tional and His tor i cal Logics in Ex plaining the Rise of the American Industrial Cor poration.” Pp. 19-44 in Comparative Social Research. Green wich, CT: JAI.

310

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

Rumelt, Rich ard P. 1974.Strat egy, Struc ture and Eco nomic Per for mance . Boston: Harvard Uni ver sity, Grad u ate School of Busi ness Ad min is tra tion, Di vi sion of Re search. Salter, Malcolm S., and Wolf A. Weinhold. 1979. Diversification Through Ac qui si tion: Strat egies for Cre ating Eco nomic Value . New York: Free Press. Samuelson, Paul A. 1980. Economics, 11th ed. New York: McGraw-Hill. Schlevogt, Kai-Al ex an der, and Lex Donaldson. 1999. “Eclec tic Fit: Com bining The o ret i cal and Em pir i cal Ap proaches to Ex plain Or ga ni za tional Per for mance.” Pa per pre sented at the an nual meet ing of the Acad emy of Man age ment, Or ga ni za tion and Man age ment The ory Di vi sion, Chi cago, Au gust. Schoonhoven, C. B. 1981. “Prob lems With Con tin gency The ory: Testing As sump tions Hidden Within the Lan guage of Con tingency The ory.” Administrative Sci ence Quarterly 26:349-377. Schreyögg, Georg. 1980. C “ on tin gency and Choice in Or ga ni za tion The ory.” Organization Studies 1:305-326. Scott, Bruce R. 1971. Stages of Cor porate De velopment. Boston: Har vard Busi ness School. Scott, W. Rich ard. 1992. Organizations: Rational, Nat u ral and Open Sys tems , 3rd ed. Englewood Cliffs, NJ: Prentice Hall. Scott, W. Rich ard. 1995. InstitutionsandOrganizations . Thou sand Oaks, CA: Sage. Scott, W. Rich ard, and John J. Meyer. 1994. In sti tu tional En vi ron ments and Or ga ni za tions. Thou sand Oaks, CA: Sage. Selznick, Philip. 1949. TVA and the Grass Roots . Berke ley: Uni ver sity of Cal i for nia Press. Shenoy, S. 1981. “Or ganization Struc ture and Con text: A Rep lication of the Aston Study in In dia.” Pp. 133-154 in Or ga ni za tion and Na tion: The Aston Programme IV, ed ited by D. J. Hickson and C. J. McMillan. Farnborough, Hants, UK: Gower. Silverman, Da vid. 1968. “For mal Organizations or In dustrial Sociology: Towards a Social Ac tion Anal y sis of Or ga ni za tions.” Sociology 2:221-238. Silverman, Da vid. 1970. The The ory of Or ga ni za tions . Lon don: Heinemann. Si mon, Her bert A. 1976. Ad min is tra tive Be hav ior: A Study of De ci sion-Making Pro cesses in Ad min is tra tive Or ga ni za tion, 3rd ed. New York: Free Press. Singh, Jitendra V. 1986. “Tech nol ogy, Size and Or ga ni za tional Struc ture: A Re ex am i na tion of the Okayama Study Data.” Acad emy of Man age ment Jour nal 29:800-812. Slappendel, Carol. 1996. “Per spec tives on In no va tion in Or ga ni za tions.” Organization Studies 17:107-129. Smith, Chris, John Child, and Mi chael Rowlinson. 1990. Re shaping Work: The Cad bury Experience . Cam bridge, UK: Cam bridge Uni ver sity Press. Smith, M. R. 1978. “Profits and Ad ministrativeIntensity: A Longitudinal Anal ysis.” Sociology 12:509-521. Sprecht, Da vid A. 1973. “Sys tem Size and Struc tural Dif fer en ti a tion in For mal Or ga ni za tio ns:An Al ter na tive Base line Gen er a tor.”AmericanSociologicalReview 38:479-480. Starbuck, Wil liam H. 1981. “A Trip to View the El e phants and Rat tle snakes in the Gar den of Aston.” Pp. 167-98 in PerspectivesonOrganizationDesignandBehavior, ed ited by Andrew Van de Ven and William Joyce. New York: John Wiley. Staw, Barry M., Lance E. Sandelands, and Jane E. Dutton. 1981. “Threat-Ri gidity Effects in Or ga ni za tional Be hav ior: A Mul ti level Anal y sis.” AdministrativeScience Quarterly 26:501-524.

References

311

Steer, Pe ter, and John Ca ble. 1978. “In ter nal Or ga ni za tion and Profit: An Em pir i cal Anal y sis of Large U.K. Com panies.” The Jour nal of In dus trial Eco nom ics 27:13-30. Stopford, John M., and L. T. Wells, Jr. 1972. ManagingtheMultinationalEnterprise . New York: Ba sic Books. Suzuki, Y. 1980. “The Strat egy and Struc ture of Top 100 Jap a nese In dus trial En ter prises 1950-1970.” Stra te gic Man age ment Jour nal 1:265-291. Tauber, I. 1968. “A Yard stick of Hos pi tal Or ga ni za tion.” Di ploma The sis, Uni ver sity of Aston, Bir ming ham, Eng land. Tayeb, Monir. 1987. “Con tin gency The ory and Cul ture: A Study of Matched Eng lish and the In dian Man u fac turing Firms.” OrganizationStudies 8:241-261. Tay lor, F. W. 1947.ScientificManagement . Lon don and New York: Harper. Teece, Da vid J. 1981. “In ter nal Or ga ni za tion and Eco nomic Per for mance: An Em pir i ca l Anal y sis of the Prof it abil ity of Prin ci pal Firms.” The Jour nal of In dus trial Eco nom ics 30:173-199. Thomp son, James D. 1967. Or ga ni za tions in Ac tion . New York: McGraw-Hill. Thompson, R. S. 1981. “In ternal Organization and Profit: A Note.” The Jour nal of IndustrialEconomics 30:201-211. Thomp son, R. S. 1983. “Dif fu sion of the M-form Struc ture in the U.K.: Rate of Im i ta tion, Inter-Firm and Inter-In dustry Dif fer ences.” International Journal of Industrial Organization 1:297-315. Tolbert, Pamela S., and Lynne G. Zucker. 1983. “In sti tu tional Sources of Change in the For mal Struc ture of Or ga ni za tions: The Dif fu sion of Civil Ser vice Re form, 18801935.” Ad min is tra tive Sci ence Quar terly 28:22-39. Tosi, Henry, Ramon Aldag, and Ron ald Storey. 1973. “On the Mea sure ment of the En vi ron ment: An As sess ment of the Law rence and Lorsch En vi ron men tal Un cer tainty Scale.” Ad min is tra tive Sci ence Quar terly 18:27-36. Trist, E. L., and K. W. Bamforth. 1951. “Some So cial and Psy cho log i cal Con se quences of the Long wall Method of Coal-Getting.” Hu man Re la tions 4:3-38. Tung, Rosalie. 1979. “Di men sions of Or ga ni za tional En vi ron ments: An Ex plor atory Study of Their Im pact on Or ganization Struc ture.” Academy of Management Journal 22:672-693. Tushman, Mi chael L. 1977. “Spe cial Bound ary Roles in the In novation Pro cess.” Administrative Sci ence Quar terly 22:587-605. Tushman, Mi chael L. 1979. “Work Char ac ter is tics and Sub unit Com mu ni ca tion Struc ture: A Con tin gency Anal y sis.” Ad min is tra tive Sci ence Quar terly 24:82-98. Van de Ven, An drew H., A. L. Delbecq, and R. Koenig, Jr. 1976. “De ter mi nants of Co or di na tion Modes Within Or ga ni za tions.” AmericanSociologicalReview 41:322-338. Van de Ven, An drew H. and Rob ert Drazin. 1985. “The Con cept of Fit in Con tin gency The ory.” Pp. 333-365 in ResearchinOrganizationalBehaviour, Vol. 7, ed ited by B. M. Staw and L. L. Cummings. Green wich, CT: JAI. Van de Ven, An drew H., and Di ane L. Ferry. 1980. Mea suring and As sessing Or ga ni za tions. New York: John Wiley. Venkatraman, N. 1989. “The Con cept of Fit in Strat egy Re search: To ward Ver bal and Sta tis ti cal Cor re spon dence.” Acad emy of Man age ment Re view 14:423-444. Wag ner, John A., III, Aaron Buchko, and Rich ard Z. Gooding. 1988. “Aston Re search on Or ga ni za tional Struc ture: A Meta-An a lytic Ex am i na tion of Generalizability.” Pa per

312

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

pre sented at the an nual meet ing of the Acad emy of Man age ment, Or ga ni za tion and Man age ment The ory Di vi sion, An a heim, Cal i for nia, Au gust. Wall, Toby D., J. Mar tin Corbett, Robin Mar tin, Chris W. Clegg, and Paul R. Jack son. 1990.“AdvancedManufacturingTechnology,WorkDesign,andPerformance: AChange Study.” Jour nal of Ap plied Psy chol ogy 75:691-697. Wall, Toby D., and Roy Payne. 1973. “Are De ficiency Scores De fi cient?” Journal of Applied Psy chol ogy 58:322-326. Weber, Max. 1964. The The ory of So cial and Eco nomic Or ga ni za tion, trans lated by A. M. Henderson and Talcott Par sons, edited by Talcott Par sons. New York: Free Press. Weber, Max. 1968. Econ omy and So ci ety: An Out line of In ter pre tive So ci ol ogy, ed ited by Guenther Roth and Claus Wittich. New York: Bedminster Press. Werts, Charles E., and Rob ert L. Linn. 1970. “A Gen eral Lin ear Model for Studying Growth.” PsychologicalBulletin 73:17-22. Whetten, Da vid A. 1980. “Sources, Re sponses, and Ef fects of Or ga ni za tional De cline.” Pp. 342-374 in The Or ga ni za tional Life Cy cle: Is sues in the Cre ation, Trans for ma tion and De cline of Or ga ni za tions, ed ited by John R. Kimberly and Rob ert H. Miles. San Fran cisco: Jossey-Bass. Whisler, Thomas L., Harald Meyer, Ber nard H. Baum, and Pe ter F. Sorensen, Jr. 1967. “Cen tral iza tion of Or ga ni za tional Con trol: An Em pir i cal Study of Its Mean ing and Mea sure ment.” Jour nal of Busi ness 40:10-26. Whittington, Richard 1989. Corporate Strat egies in Re cession and Re covery: Social Struc ture and Stra te gic Choice . Lon don: Al len and Unwin. Whittington, Rich ard, Mi chael Mayer, and Francesco Curto. 1999. “Chandlerism in Post-War Eu rope: Stra te gic and Struc tural Change in France, Ger many and the UK, 1950-1993.” In dus trial and Cor po rate Change 8:519-550. Wil liam son, Ol i ver E. 1970. Cor po rate Con trol and Busi ness Be hav ior: An In quiry Into the Effects of Or ganization Form on En terprise Be havior. Englewood Cliffs, NJ: Prentice Hall. Wil liam son, Ol i ver E. 1975. Mar kets and Hi er ar chies: Anal y sis and An ti trust Im pli ca tions . New York: Free Press. Wil liam son, Ol i ver E. 1985. The Eco nomic In sti tu tions of Cap i tal ism: Firms, Mar kets, RelationalContracting. New York: Free Press. Wil liam son, Ol i ver E. 1996. “Eco nomic Or ga ni za tion; The Case for Can dor.” Academy of Man age ment Re view 21:48-57. Wil liam son, Ol i ver E., and Narottam Bhargava. 1972. “As sessing and Clas sifying the Inter nal Struc ture and Con trol Ap pa ra tus of the Mod ern Cor po ra tion.” Pp. 125-148 in Mar ket Struc ture and Cor po rate Be hav iour: The ory and Em pir i cal Anal y sis of the Firm, ed ited by Keith Cowl ing. Lon don: Gray-Mills. Wong, Gilbert Y. Y., and Philip H. Birnbaum-More. 1994. “Cul ture, Con text and Struc ture: A Test on Hong Kong Banks.” OrganizationStudies 15:99-123. Wood, R. E., and E. A. Locke. 1990. “Goal Set ting and Strat egy Ef fects on Com plex Tasks.” Pp. 73-109 in Re search in Or ga ni za tional Be hav ior, Vol. 12, ed ited by B. M. Staw and L. L. Cummings. Green wich, CT: JAI. Wood ward, Joan. 1958. Man age ment and Tech nol ogy. Lon don: H.M.S.O. Woodward, Joan. 1965. Industrial Or ganization: The ory and Prac tice. Oxford, UK: Ox ford Uni ver sity Press.

References

313

Wood ward, Joan. 1970. In dus trial Or ga ni za tion: Be hav ior and Con trol . Ox ford, UK: Ox ford Uni ver sity Press. Young, Ruth C. 1988. “Is Pop u la tion Ecol ogy a Use ful Par a digm for the Study of Or ga ni za tions?” Amer i can Jour nal of So ci ol ogy, 94:1-24. Yuchtman, E., and S. E. Sea shore. 1967. “A Sys tems Re source Ap proach to Or ga ni za tional Ef fec tive ness.” AmericanSociologicalReview 32:891-903. Zeffane, Rachid M. 1989a. “Cen tral iza tion or For mal iza tion? In dif fer ence Curves for Strat egies of Con trol.” OrganizationStudies 10:327-352. Zeffane, Rachid M. 1989b. “Or ga ni za tion Struc tures and Con tin gencies in Dif fer ent Na tions: Al ge ria, Brit ain, and France.” So cial Sci ence Re search 18:331-369. Zucker, Lynne G. 1977. “The Role of Institutionalization in Cul tural Persistence.” AmericanSociologicalReview 42:726-743. Zucker, Lynne G. 1987. “In sti tu tional The ories of Or ga ni za tion.” Amer i can Re view of Sociology 13:443-464. Zwerman, Wil liam L. 1970. New Perspectives on Or ganization The ory: An Empirical Reconsideration of the Marxian and Clas si cal Anal y ses. Westport, CT: Green wood.

THE In dex CON TIN GENCY THEORY OF OR GA NI ZA TIONS

Index

Achievement need, 90 Adaptive change, 12-16, 161 and quasi-fit, 258-262 and range restriction, 231 as caused by business cycle, 252-253 as caused by competition, 253 as caused by debt, 253-254 as caused by divisional risk, 254 as caused by low performance, 249-257 forestalled by diversification, 254-255 forestalled by divestment, 255-256 forestalled by directors, 256-257 forestalled by divisionalization, 255 theory of, 9-16, 268 Ad hoc variation structure, 83-85 Administrative intensity: and definitional dependency, 172 and divisionalization, 175 and size, 72-75, 102-122, 170-173, defined, 72 Administration, experts in, 103, 111, 120 Administrators, 71-75 Analyzability, of task, 48-49

Association between contingency and structure, 8-9, 11 Autonomy of divisions, 80, 183, 255

Boards of directors, 155-156, 256-257 Bureaucratic structure, 2-3, 23-28, 62-64 and diversification, 81 and divisionalization, 27 and size, 65-67, 69, 71, 92, 95 and structural differentiation, 27 contingency theory of, 115-122 curvilinear relationships with size, 106-110 dimensions of, 23, 64 dysfunctions of, 76 fit of to size, 24, 68, 225 fit of to task interdependence, 28 functions of, 76 future study of, 274-275 measures of, 64, 92 meta-analyses of, 66-67 replication studies of, 66 315

316

THE CON TINGENCY

size as cause of, 102-122, 137-138 Weberian ideal-type of, 62-63 Bureaucracy theory, 23-30, 61-100, 102- 122 and computerization, 129-130 and divisionalization, 27, 77-81 and fit, 24 and matrix structures, 87-88 and size, 24, 102-122 and structural differentiation, 27, 69-76 causal connections among structural variables, 102-122 contingency theory of, 115-122 future study of, 274-275 in multinational corporations, 85 merger of with organic theory, 95 reconciling with organic theory, 92-95 synthesis with organic theory, 28-30 Business cycle, 252-253, 269, 278

Cartesianism, 141-152 Causality: of size and bureaucracy, 101-122 reverse, 138-141 Centralization, 22, 23-25, 39-40, 46, 49, 63-64 and centrally planned economies, 91 and environmental hostility, 89 and multinational corporations, 83-85 and personality of CEO, 90 and public accountability, 67, 88-89 and size, 65-67 and task interdependence, 80 fit to size, 225 Challenges to contingency theory, 161-177 Choice, 54 and equifinality, 194-195 versus determinism, 131-137 Classical management theory, 3, 38, 46, 51, 52, 53 Cognitive capacity of managers, 81, 175 Cognitive processing, 48, 50 Cognitive configurations, 151-152

THEORY OF OR

GANIZATIONS

Competition, 163, 239, 253, 259, 269, 278 Complexity: environmental, 19-20 of contingency theory, 30, 95 organizational, 39, 71, 103 Configurationalism, 141-152 cognitive versus existential configurationalism, 151-152 critique of, 144-152 divisionalized conglomerate configuration, 150 Conflict resolution, 43-45 Confounds, 233-239, 288 Congruence, 186-189 Constituents of fit, 236-239 Contingencies, 2-3, 17-21 association with organizational structure, 8-9 causal connection of size and task contingencies, 96-98 cause of changes in, 247-248 causes versus contingencies, 88-91 centrality, 155 changes in, 9, 11 changing contingencies to fit structure, 133-134 contingency change caused by performance, 248 contingency change causes misfit, 248 contradictory contingencies, 197 critical challenge, 153-155 definition of, 6, 153-154 environmental hostility, 89 innovation, 18-19, 39-41, 46, 47, 94-95, 223 intensity of interaction, 57-58 interdependence, 19-21, 27-30, 42-46, 50-52, 55-58, 77-87, 95, 126 leverage, 155-156 members’ predispositions, 47 multiple contingencies, 197-198 of organizational power, 152-156 personality of CEO, 90 public accountability, 88-89

Index raw materials, 48 regulation, 155-156 size, 2-3, 21, 24, 28-30, 65-67, 69-75, 80-81, 91-98, 102-122, 126-131, 137-138, 141, 170-173, 175 strategy, 3, 11-15, 77-79, 138-141, 154, 221-225 task contingencies, 18-21, 22-23, 27-30, 36-49, 53, 55-58, 81-86, 91-98, 126, 217-221, 225 task interdependence, 19-21, 27-30, 42-46, 50-52, 55-58, 77-87, 95, 126 task uncertainty, 18, 20, 22-23, 29-30, 36-49, 53, 55-58, 96, 126, 225 technology, 18, 48-55, 126-131, 219 technology versus size, 126-131 uncertainty, 18, 20, 22-23, 29-30, 36-49, 53, 55-58, 96, 126, 225 Contingency functionalist explanations, 114-118 Contingency theory, 2-3 and organizational change, 9-10 as distinguished from universalistic theories, 4 as dynamic, 9-10 as static, 9 challenges to from other theories, 161-177 conceptual integration of, 16-30 controversies in, 125-157 core paradigm of, 7-16 definition of, 1-2, 5-7 integration of, 16-30 making more dynamic, 248-249 modest complexity of, 30, 95 neo-contingency theory, 245-271 of board of directors, 155-156 of bureaucracy, 115-122 of co-optation, 155-156 of group decision making, 36 of leadership, 36 of organizational structure, 4 problem of reverse causality for, 139 problems in, 246, 258, 263-265

317 structural contingency theory, 4 theoretical integration of, 16-30 Control, 22-23, 37, 39 direct and indirect, 65, 69, 103, 104, 130 in multinational corporations, 83-85 over means or ends, 78 substitute or alternative means of control, 65, 84 Controversies in contingency research: configurations versus Cartesianism, 141-152 contingencies of organizational power, 152-156 determinism versus choice, 131-137 reverse causality, 138-141 size as cause of bureaucracy, 137-138 technology versus size, 126-131 Co-optation, 155-156 Core paradigm of contingency theory, 7-16 Cost reduction, 18, 223 Craft technology, 48-50 Crisis, 15, 176, 249-257, 276 Cultures: of nations, 90-91 of organizations, 84 matrix, 88 Curvilinear relationship, 8, 65, 69-70, 72-73, 106-110, 130, 190

Debt, 253-254, 269, 278 Decentralization, 22-27, 39-40, 46, 49, 64, 93 and diversification, 81 and divisionalization, 77 and size, 29-30, 65-67, 69, 81, 92, 95, 102-122 and task interdependence, 29-30 and task uncertainty, 29-30 caused by low performance, 176 Declining organizations, 173, 261-262, 280 Defender strategy, 18, 224

318

THE CON TINGENCY

Definitional dependency, 172, 236 Determinism, 131-137 Deviation analysis, 210-212, 217, 219, 225 Differentiation, 41-48, 83-85 Differentiated fit structure, 83-85 Directors, 256-257, 279 Disequilibrium, 247-249, 275-276 Diversification, 3, 11, 13-16, 19, 269, 278 and bureaucratic structure, 81 and divestment, 256 and divisionalization, 77-79, 82, 166 and task interdependence, 29, 77-81 caused by divisionalization, 139-141 fit of structure to, 3, 13, 77, 79, 183 fit of divisionalization to, 13, 221-225 preserves misfit, 254-255 reduces risk, 254-255 studies of fit and performance of, 13, 221-225 Divestment, 255-256, 269, 279 Divisional risk, 254, 279 Divisionalization, 3, 11, 13-16, 77-81, 278, 279 and cognitive capacity of managers, 81 and decentralization, 77 and diversification, 77-79 and documentation, 77 and institutional theory, 163-168 and opportunism by managers, 80, 174-176 and plant size, 80 and population-ecology, 168-170 and size, 80-81, 175 and specialization, 77 and structural differentiation, 27 and task interdependence, 29-30 and threat-rigidity, 176 and vertical span, 77 as caused by competition, 253 by geography, 82 caused by diversification, 139-141 definition of, 77 divisionalized conglomerate configuration, 150 evidence of quasi-fit in, 259

THEORY OF OR

GANIZATIONS

fit to diversification, 3, 13, 77, 79, 166, 183, 221-225 fit to diversification affects performance, 13, 79, 221-225 fit to task interdependence, 28 international division, 82 M-form, 170-171, 174-176, 223 part of bureaucratic structure, 27, 77 preserves misfit, 255 rational, 176 reduces risk, 255 Documentation, 63-64 and size, 67 and divisionalization, 77 Dynamics, 9-16, 248-249, 268 Dysco-optation, 156

Early adopters, 165, 223 Ecology. See Population-ecology Economics, 71, 244-245, 268-269 Economies of scale in administration, 71-75 Effectiveness of organization, 6, 53, 71, 74, 217 Efficiency, 38, 220 Emergent view of organizations, 76 Engineering technology, 48-50 Environment, 2, 17-18, 22, 38, 39, 46, 47, 217-218 and members’ predispositions, 47 complexity, 19-20 dynamism, 19-20 engineering of, by organization, 134- 135 environmental hostility, 89 munificence, 19-20 of multinational corporations, 83, 86 Equifinality, 143, 148 and choice, 194-195 problems with, 193-195 Exceptions, number of, 48 Existential configurationalism, 151-152

Index Falsification, 183-185 Finance, 244-245, 268-269 Fit, 181-213 affects performance in multinational corporations, 83 alternation of with misfit, 248-249 and hetero-performance, 265-268, 281-283, 287-288 and performance, 10-16, 68, 134, 185-200, 215-243 as caused by business cycle, 252-253 as caused by competition, 253 as congruence, 186-189 as interaction term, 189-192 as iso-performance, 192-193 as tautology, 182-183 constituents of, 236-239 effect of on quality, 218-220 in future research, 274-275, 277-283, 287-288 in multinational corporations, 83-84 institutional theory and, 163-168 interaction fit, 189-192, 195 line of, 200-202 mediates between size and structure, 115-118 multiple fits, 198-200 negative effect of performance on, 233-236 of bureaucracy to size, 68, 225 of divisionalization to diversification, 183, 79 of strategy and structure, 11, 13-15, 79 of structure to multinational strategy, 82-85 quasi-fit, 257-262 selection fit, 195-196 systems fit, 195, 199-200 to defender strategy, 224 to members’ predispositions, 47 to multiple contingencies, 197-198 to prospector strategy, 224 to size and uncertainty, 225 to technology, 53 validation of, 202-212

319 Fit line: for hetero-performance, 265-268, 281-282, 287 identification of, 200-202 See also quasi-fit line, 260-262 Formalization, 22-25, 63-64 and internal labor markets, 75 and multinational corporations, 83-85 and organic structure, 39-40 and quasi-fit, 260-262 and size, 29-30, 65-67, 73, 75, 92, 96, 102-122, 225 and task interdependence, 29-30 and task uncertainty, 29-30 caused by specialization, 96 reduces uncertainty, 96 Founding, effect of, 74 Functional background of manager, 135 Functional structure, 11, 13-14, 77-80, 82 Functionalist explanations, 10, 68, 89, 110-118, 161-164, 167 Future research, 239-242, 273-289

Geographically extensive operations, 82 Global integration, 82-86 Global strategy, 83-85 Group decision making, 36, 55 Grouping of organizational sub-units, 57, 95 Growing organizations, 173, 261, 280

Hetero-performance, 263-268, 281-288 Hierarchy, 41-44, 69, 94 and delegation, 104 and matrix structures, 87-88 and size, 69-71, 73, 75, 102-122 location of task interdependence in, 57

Icarus Paradox, 144-145 Ideology, 54, 111, 132 Incremental change, 145-146, 249

320

THE CON TINGENCY

Inertial tendencies of organizations, 168-170 Information-processing, 82, 183, 220 Information technology, 129-130 Innovation, 18-19, 39-41, 46, 47, 94-95, 223 magnitude of innovation, 40-41 rate of innovation, 39-41 Institutional theory: and divisionalization, 163-168 and range restriction, 232 critique of, 162-168 Integrated variety structure, 83-85 Integration, structural, 41-48 fit of to size, 225 in multinational corporations, 83-85 of contingency theory, 16-30 Integrators, 43-45, 57 Intensive technology, 50-51 Interaction: intensity of, 57-58, 80 interaction fit, 189-192, 195 Interdependence, See Task interdependence Interests, 111, 114, 170-176 Internal labor markets, 75 International strategy, 83-84 Iso-performance, 192-193, 263, 287

Knowledge, 37, 39, 74, 104, 112, 114-115, 166, 257-258, 260

Lags, 13-14, 136 Late adopters, 165, 223 Local responsiveness, 82-86 Long-linked technology, 50-51

M-form, 170-171, 174-176 and administrative intensity, 175 fit and performance of, 223 mediocrity of, 255 See also Divisionalization

THEORY OF OR

GANIZATIONS

Macro-organization structure, 94, 126 Managerial decision, 195-196 Managers: and size, 71-75 cognitive capacity of, 81, 175 opportunism by, 80, 174-176 shared values among, 84, 86 Materials, perceived nature of, 48 Material factors, 54-55 Matrix structures, 86-88 and task interdependence, 87 Mechanistic structure, 2, 37-40, 47, 49, 51, 94 and technology, 53 definition of, 21-22, 24-25 dimension of, 22 fit to defender strategy, 224 fit to task uncertainty, 22-23, 28 fit to task interdependence, 27-28, 52 Mediating technology, 50 Micro-organization structure, 94, 126 Mimetic isomorphism, 163-164 Misfit: alternation of with fit, 248-249 and hetero-performance, 265-268, 282-288 and performance, 10-16 and unreliability, 227-228 as caused by business cycle, 252-253 as caused by contingency change, 247-248 as caused by high performance, 248 fostered by directors, 256-257 fostered by diversification, 254-255 fostered by divestment, 255-256 fostered by divisionalization, 255 future research on, 274-275, 278, 280, 282-288 in multinational corporations, 83-84 of strategy and structure, 11, 13-15 Multinational corporations, 81-86 ad hoc variation structure in, 83-85 and global integration, 82-86 and global strategy, 83-85 and international strategy, 83-84

Index and local responsiveness, 82-86 and multinational strategy, 83-85 and transnational strategy, 83-84 differentiated fit structure in, 83-85 integrated variety structure in, 83-85 structural uniformity structure in, 83-85 Multinational strategy, 83-85 Multiple contingencies, 197-198, 225 Multifits, 196-200, 225 Munificence, 19-20 Mutual adjustment, 51, 56, 57

Neo-contingency theory, 245-271, 275-288 Neo-human relations theory, 4, 38, 46, 51, 217 Nonroutine technology, 48-50 Normative integration, 83-86 Normative isomorphism, 163, 166

Opportunism by managers, 80, 174-176 Organic structure, 2, 37-40, 46-47, 49, 52, 93-94 and technology, 53 definition of, 22, 24-25 dimension of, 22-23 fit of to prospector strategy, 224 fit of to task uncertainty, 22-23, 220 fit of to task interdependence, 27-28, 52 Organic theory, 21-30, 35-59 and matrix structures, 87-88 and task interdependence, 27-28 and task uncertainty, 22-23, 28 in multinational corporations, 86 measures of, 93 merger with bureaucracy theory, 95 reconciling with bureaucracy theory, 92-95 synthesis with bureaucracy theory, 28-30 underlying connections in, 55-58 Organizational change, 1, 9-16 and disequilibrium, 247-249

321 and hetero-performance, 263-268 and quasi-fit, 259-262 as performance-driven, 249-257 factors that forestall, 254-257 factors that promote, 252-254 Organizational portfolio theory, 250-257, 277-280 Overhead cost, 71

Paradigm of contingency theory, 7-16 Participation, 23, 25, 36, 38, 46, 53, 62, 93, 98, 132, 217, 219-220 Performance: affected by fit, 10-13, 53, 68, 79, 84, 134, 215-243 affected by fit of bureaucracy to size, 68 affected by fit of divisionalization to diversification, 13, 79 affected by fit in multinational corporations, 83 affected by fit with technology, 53 affects fit, 11, 14-16, 233-236 and configurations, 147-149 and effectiveness, 6 and functionalist explanations, 110-118 and quasi-fit, 259 as cause of contingency change, 248 as cause of misfit, 248 causes of, 13, 68, 216-226, 250-257 causes change, 177 causes decentralization, 176 causes rational adaptation, 176 causes strategic adaptation, 176 causes structural adaptation, 14-16, 176 causes time lags, 136 crisis of, 15, 176, 249-257, 276 fluctuations in, 250-257 future research of, 275-276 hetero-performance, 263-268, 281-288 per manager, 74 wider measures of, 225 Performance-driven organizational change, 12, 14-16, 136, 233-234, 249-257, 276-280

322

THE CON TINGENCY

Permanently failing organizations, 177 Personality of manager, 90, 135 Philosophical issues in contingency theory research, 182-185 Plant size, 80, 175 Political explanation, 111, 114, 170-177 Polycentralized structure, 49 Pooled interdependence, 50, 52 Population change, 168-170 Population-ecology, 168-170, 232 Portfolio theory, 251, 269 Positivist, 137 Power, 152-156 Predispositions of organizational members, 47 Problem-solving, 16, 249, 260 Profit accountability, 80, 255 Prospector strategy, 18, 224 Public accountability, 67, 88-89

Quality, effect of fit on, 218-220 Quasi-fit, 257-262, 280-288

Range restriction, 229-233, 236-237 Rationality, 10, 161-167, 170-177, 249, 260 Reciprocal interdependence, 50-52 Regression analysis, 209-210 Related products or services, 80, 223 Relativity perspective, 253 Resource scarcity, 20 Resource dependence theory, 152-156 Reverse causality, 138-141 Risk, organizational, 250-251, 254-257, 269, 277 divisional, 254 Routine technology, 48-50

SARFIT, 11-16, 116-118, 281, 286-287 and range restriction, 232 and hetero-performance, 266-268 Satisficing, 12, 16, 249, 260, 276-277

THEORY OF OR

GANIZATIONS

Selection fit, 195-196 Sequential interdependence, 50-52 Shared values among managers, 84, 86 Simple structure, 2-3, 23-24 Size, 2-3: and administrators, 71-75, 170-173 and bureaucracy, 24, 29, 65-67, 69, 71, 102-122, 137-138 and centralization, 65-67 and decentralization, 65-67, 75 and divisionalization, 80-81, 175 and documentation, 67 and formalization, 65-67, 75 and hierarchy, 69-71, 73, 75 and internal labor markets, 75 and managers, 71-75, 170-173 and number of levels, 65, 70, 75 and participation, 98 and specialization, 65-67, 71 and standardization, 65-67 and structural differentiation, 69-71, 73-75 and span of control, 71, 73 causal connection with task contingencies, 96-98 caused by bureaucracy, 137, 141 causes diversification, 97 causes innovation, 97 causes task interdependence, 97 definition of, 21 synthesis with task contingency, 2830, 91-98 versus technology, 126-131 Span of control, 70, 71, 73 Specialization, 22-25, 63-64 and divisionalization, 77 and size, 29-30, 65-67, 71, 92, 95, 102-122 and task interdependence, 29-30 as structural differentiation, 75 by occupation versus task, 39, 93 causes formalization, 96 reduces task uncertainty, 96 Spurious relationships, 236-239 Stakeholders, 225

Index Standardization, 63-67 Strategic choice, 132-137 Strategic contingencies theory of organizational power, 153-155 Strategy, 3, 154 and structure, 11-15, 77-79 changes in, 11-15 reverse causality of strategy by structure, 138-141 studies of fit to, and performance, 13, 221-225 Structural adaptation to regain fit, 1116, 116-118 Structural contingency theory. See Contingency theory Structural differentiation, 69 and bureaucratic structure, 27 and divisionalization, 27 and size, 29-30, 69-71, 73-75, 95, 102-122 and specialization, 75 and task interdependence, 29-30 critique of theory of, 106-107, 112-114 universalistic nature of theory, 112-114 Structural uniformity structure, 83-85 Subgroup analysis, 204-208 Synthesis of size with task contingency, 28-30, 91-98 Systems fit, 195, 199-200

Task analyzability, 48-49 Task contingencies, 18-21, 126 and multinational corporations, 81-86 causal connection with size contingency, 96-98 fit to, and performance, 217-221 synthesis with size contingency, 28-30, 91-98 task analyzability, 48-49 task interdependence, 19-21, 27-28, 81-86,126 task uncertainty, 18, 20, 22-23, 29-30, 36-49, 53, 55-58, 96, 126, 225

323 Task interdependence, 19-21, 42-46, 55- 58, 126 and diversification, 29, 79-81 and divisionalization to, 28, 29-30, 77-81, 95 and intensity of interaction, 57-58 and matrix structures, 87 and multinational corporations, 81-86 and uncertainty, 56-57 fit of bureaucratic structure to, 28 fit of mechanistic and organic structures to, 27-28, 52 location of interdependence in hierarchy, 57 pooled interdependence, 50, 52, 56-57, 80 reciprocal interdependence, 50-52, 56-57 sequential interdependence, 50-52, 56-57, 79 Task uncertainty, 18, 20, 23, 29, 36-49, 53, 55-58, 225 and decentralization, 29-30 and formalization, 29-30, 96 and group decision making, 36 and interdependence, 56-57 fit of mechanistic and organic structures to, 22-23 reduced by specialization, 96 Tautology, 182-183 Technology, 18, 48-55, 219 craft technology, 48-50 engineering technology, 48-50 fit with affects performance, 53, 126- 131 generalizability of effects of, 130-131 information technology, 129-130 intensive technology, 50-51 large batch and mass production, 52-54 long-linked technology, 50-51 mediating technology, 50 nonroutine technology, 48-50 production continuity, 54 routine technology, 48-50 versus size, 126-131

324

THE CON TINGENCY

Threat-rigidity, 89, 176-177, 276 Tolerance for ambiguity, 47 Transnational strategy, 83-84

Unanalyzability, of task, 48-49 Uncertainty, See Task uncertainty Universalistic relationships, 4, 38, 39, 112-114, 139, 217, 237-238, 260, 274

THEORY OF OR

GANIZATIONS

Unrelated products or services, 80, 223 Unreliability, 227-229

Validation of fit model, 202-212 Vertical integration, 19, 80, 223 Ver ti cal span, 63, 65, 75, 77

THE CON About the TIN Au thor GENCY THEORY OF OR GA NI ZA TIONS

About the Author

Lex Donaldson is Pro fes sor of Or ga ni za tional Design at the Austra lian Graduate School of Management, which is a joint venture of the Universities of New South Wales and Syd ney. He holds a B.Sc. in Be havioral Sci ences from the Uni ver sity of Aston (1968) and a Ph.D. from the University of Lon don (1974). His re search in ter ests are or ga ni za tional the ory, or ga ni za tional struc ture, and corporate governance. Donaldson is the au thor of six books, in clud ing Performance-Driven Or ga ni za tional Change: The Or ga ni za tional Port fo lio (1999), For Posi tiv ist Organization Theory: Proving the Hard Core (1996), American AntiMan age ment The ories of Or ga ni za tion: A Cri tique of Par a digm Pro lif er a tion (1995), and In De fence of Or ganization The ory: A Re ply to the Critics (1985). He is coauthor (with Freder ick G. Hilmer) of Management Redeemed: Debunking the Fads That Undermine Our Cor porations (1996). He also ed ited a col lec tion of key ar ti cles un der the ti tleCon tingency The ory (1995). Donaldson is on the editorial boards of Academy of Management Review, Or ganization Studies, and the Strategic Management Journal. He has served as a guest editor of a spe cial issue of the Academy of Management Review on “Market Discipline and the Discipline of 325

326

THE CON TINGENCY

THEORY OF OR

GANIZATIONS

Management” (1990). He has held visiting appointments at the Uni versities of Aston, Iowa, London, Mary land, Northwestern, Oxford, and Stan ford.