booksclub.org making and breaking governments cabinets and legislatures in parliamentary democracies...

7/29/2019 Booksclub.org Making and Breaking Governments Cabinets and Legislatures in Parliamentary Democracies Politica

1/315

Making and Breaking Governments offers a theoretical argument abouthow parliamentary democracy works. The heart of a parliamentary re-gime is the process by which the elected representatives choose a govern-ment, consisting of a cabinet and ministers, to serve as the executive armof the regime. Strategic interaction among parliamentary parties createsnew governments and either maintains them in office or, after a resigna-tion or vote of no confidence, replaces them with some alternative govern-ment. The authors formulate a theoretical model of this strategic interac-tion, derive consequences, formulate empirical hypotheses on the basis ofthese, and test the hypotheses with data drawn from the postwar Euro-pean experience with parliamentary democracy.


2/315


3/315

MAKING AND BREAKING GOVERNMENTS


4/315

POLITICAL ECONOMY OF INSTITUTIONS AND DECISIONSEditorsJames E. Alt, Harvard UniversityDouglass C. North, Washington University of S t. Louis

Other books in the seriesAlberto Alesina and Howard Rosenthal, Partisan Politics, Divided Government,and the EconomyJames C. Alt and Kenneth Shepsle, eds., Perspectives on Positive PoliticalEconomyJeffrey S. Banks and Eric A. Hanushek, eds., Modern Political Economy: OldTopics, New DirectionsYoram Barzel, Economic Analysis of Property RightsRobert Bates, Beyond the Miracle of the Market: The Political Economy ofAgrarian Development in KenyaPeter Cowhey and Mathew D. McCubbins, Structure and Policy in Japan andthe United StatesGary W. Cox, The Efficient Secret: The Cabinet and the Development ofPolitical Parties in Victorian EnglandJean Ensminger, Making a Market: The Institutional Transformation of anAfrican SocietyMurray Horn, The Political Economy of Public Administration: InstitutionalChoice in the Public SectorJack Knight, Institutions and Social ConflictMichael Laver and Kenneth Shepsle, Cabinet Ministers and ParliamentaryGovernmentLeif Lewin, Ideology and Strategy: A Century of Swedish Politics (Englishedition)Gary Libecap, Contracting for Property RightsMathew D. McCubbins and Terry Sullivan, eds., Congress: Struct ire and PolicyGary J. Miller, Managerial Dilemmas: The Political Economy of HierarchyDouglass C. North, Institutions, Institutional Change, and EconomicPerformanceElinor Ostrom, Governing the Com mons: The Evolution of Institutions forCollective ActionMark Ramseyer and Frances Rosenbluth, The Politics of Oligarchy:Institutional Choice in Imperial JapanJean-Laurent Rosenthal, The Fruits of Revolution: Property Rights, Litigation,and French AgricultureCharles Stewart III, Budget Reform Politics: The Design of the AppropriationsProcess in the House of R epresentatives, 18651921John W aterbury, Exposed to Innumerable Delusions: Public Enterprise andState Power in Egypt, India, Mexico, and Turkey


5/315

MAKING AND BREAKINGGOVERNMENTSCABINETS AND LEGISLATURES INP AR LIAMENTAR Y DEMOC R AC IES

MICHAEL LAVERUniversity of Dublin

KENNETH A. SHEPSLEHarvard University

CAMBRIDGEUNIVERSITY PRESS


6/315

Published by the Press Syndicate of the University of C amb ridgeThe Pitt Building, Trumpington Street, Cambridge CB2 1RP

40 West 20th Street, New Y ork, NY 100 11-4 211 , USA10 Stamford Road, Oakleigh, Melbourne 3166, Austral ia

Cambridge University Press 1996First published 1996

Library of Congress Cataloging-in-Publication DataLaver , Michael , 1949-

M ak ing and break ing governm ents : cabinets and legislatures inparliamentary democracies / Michael Laver, Kenneth A. Shepsle.

p. cm. - (Political eco nom y of institu tion s and decisions)Includes bibliographical references and index.

ISBN 0-521-43245-6. - ISBN 0-521-43836-5 (pbk.)1. Cab inet system. 2. Cabinet officers. 3. Represen tative

governm ent and representation. 4. Com parative government.I. Shepsle, Ken neth A. II. Title. III. Series.

JF331.L39 19968 2 1 . 8 - d c 2 0 9 5 - 8 7 5 7

CIPA catalog record for this book is available from the British Library.

ISBN 0-521-432 45-6 H ardb ackISBN 0-521-43 836-5 Paperback

Transferred to digital printing 2004


7/315

Contents

Series editors' preface page viiAcknowledgments ixPART IT H E C O N T E X T

1 Theory, institutions, and government formation 32 The social context of government formation 183 The government formation process 43PART IIT H E M O D E L

4 Government equilibrium 615 Strong parties 90PART IIIEMPIRICAL INVESTIGATIONS

6 Two cases: Germany, 1987 ; Ireland, 19 92 -199 3 1257 Theoretical implications, data , and operationalization 1478 Exploring the model: A comparative perspective 1589 A multivariate investigation of portfolio allocation 174PART IVA P P L I C A T I O N S , E X T E N S I O N S , A N D C O N C L U S I O N S

10 Party systems and cabinet stability 19511 Making the model more realistic 22312 Government formation, intraparty politics, andadministrative reform 24613 Governments and parliaments 277References 289Index 295


8/315


9/315

Series editors' preface

The Cambridge series on the Political Economy of Institutions and Deci-sions is built around attempts to answer two central questions: How doinstitutions evolve in response to individual incentives, strategies, andchoices, and how do institutions affect the performance of political andeconomic systems? The scope of the series is comparative and historicalrather than international or specifically American, and the focus is posi-tive rather than normative.Laver and Shepsle's theoretically innovative book pushes the study ofgovernment formation a big step forward. Rather than concentrating onthe equilibrium collective policy that emerges from the balance and distri-bution of party strengths in the parliament, they treat governments ascollections of ministers with individual jurisdictions and policy as a bun-dle of individual party-preferred policies, depending on which party re-ceives which portfolio. Jurisdiction is the key concept, in their view. Itmakes a party 's promises credible to its electors, for it can carry them out(only) if it receives the relevant ministry. At the same time portfolioallocation m onitors party behavior, for the party receiving a portfolio hasno excuse for not carrying out its preferred policy.Their portfolio-based model of government formation is based onconstitutional features carefully documented in a massive 14-countrystudy, published in the companion volume Cabinet Ministers and Parlia-mentary Government. The model bases its predictions on a sequence ofproposals from an historically determined status quo. Bargaining takesplace among rationally foresighted policy-motivated parties, each with anexplicit veto over every cabinet in which it could participate, in a latticeof feasible governments. Laver and Shepsle's predicted importance for"strong" parties opens up significant possibilities for comparative empiri-cal evaluations of theirs and other contending models.Moreover, in their model, institutions limit the number of enforceablepolicy outcomes, providing the possibility of equilibrium results for the

vii


10/315

Series editors' prefacegovernment formation game where cycling would otherwise be inherent.Indeed, many other features of their model, including indivisible policies,explicit sequences, and the absence of exogenous enforcement, combinedwith rationally foresighted actors, also mark the work as a significantadvance in the creation of theories about institutions as political solutionsin situations involving politically costly transactions.

v n i


11/315

Acknowledgments

This project began in 1988 at Harvard University, where one of us(Shepsle) was a member of the faculty of the Department of Governmentand the other (Laver) was a visitor for the 19881989 academic year (onleave from University College, Galway). We taught together, playedsquash together, and even, from time to time, discussed research ideastogether. We were especially fascinated by the fact that coalition theory one of Laver's specialties had lost its head of steam in American model-ing circles, whereas the formal study of institutions one of Shepsle'sfields of concentration - had never really acquired a head of steam on theother side of the Atlantic. We believed tha t this confluence of possibilities,namely, the application of American-style institutional models to Euro-pean-style coalition and governance arrangements, constituted a majoropportunity, one that we investigated for the rest of that academic year.During this very fertile period, we wrote two articles (Laver and Shepsle,1990a, 1990b) and managed to persuade the National Science Foundationto support further work (SES-8914294). At the very outset, then, we in-curred a debt to University College, Galway, Harvard University, and theU.S. National Science Foundation for research leaves and financial supportfor which we are forever grateful.Very early in the life of the project that culminates in this volume, wealso had the good fortune to interact with three formal theorists whoshared many of our interests. David Austen-Smith and Jeffrey Banks ofthe University of Rochester and Norman Schofield of Washington Univer-sity, St. Louis, have been close friends, constructive critics, and generallyhave taken the time to steer us right when we've gone wrong (or at leastto tell us in no uncertain terms that we had gone wrong!). Numerousothers have, along the way, been sources of insight and encouragement,and we cannot name them all. But we would be remiss if we didn'tacknowledge and thank Kaare Strom of the University of California, SanDiego, who proved a wise sounding board, given his general sympathy

IX


12/315

Acknowledgmentsfor formal theory approaches and his considerable knowledge of thesubstance of parliamentary politics; Gary King of Harvard University forvery useful advice on methodological issues; Paul Doyle of UniversityCollege, Galway, for outstanding programming assistance; and AlisonAlter of Harvard University for data collection and statistical assistance.We also wish to thank those who read and commented on earlier drafts ofthis volume. In addition to those already mentioned, these include JohnAldrich, Patrick Dunleavy, Avner Greif, Paul Mitchell, Walter Petersen,Itai Sened, Steve Solnick, Paul Warwick, and Barry Weingast.Many universities and organizations provided forums for us to developand test out our ideas. These include the American Political Science As-sociation (1990 meetings in San Francisco), the European Consortium forPolitical Research (1991 meetings in Bochum, Germany, and 1992 meet-ings in Limerick, Ireland), the European Public Choice Society (1993 meet-ings in Portrush, Northern Ireland), the European Science Foundation(1992 conference in Sesimbra, Portugal), the Office of Research and Devel-opment of the Central Intelligence Agency (which also provided supportfor our computer model), and political science or economics or politicaleconomics departments and programs at Duke University, the Universityof North Carolina, the University of Rochester, Columbia University,Princeton University, Australia National University, George Mason Univer-sity, Carnegie-Mellon University, Boston University, the University of Cali-fornia, Berkeley, the University of California, San Diego, Stanford Univer-sity, London School of Economics, University College, Dublin, 1994Polybios Program at the Netherlands Institute of Advanced Study, MaxPlanck Institute, Cologne, the University of Iowa, Washington University,St. Louis, Texas A & M University, and the University of Illinois. We thankthose who invited us and took our ideas seriously at all these fine institu-tions. We wish to single out the Rockefeller Foundation's Villa Serbelloniin Bellagio, Italy, for providing us with elegant surroundings and stimulat-ing colleagues for afive-weekperiod in the summer of 1992, during whichtime nearly all of the work for this volume was completed in its initialform; and Duke University, which provided a congenial and welcomingbase for Laver in the fall of 1994 as we put the finishing touches to themanuscript. And, of course, our home universities - Harvard University,University College, Galway, and Trinity College, Dublin provided uswith research support and sabbatical leaves, and even when they expectedus to work for them, they still tolerated our occasional absences to hopacross the Atlantic.

Cambridge University Press has made our lives easy, not only by pub-lishing this volume but also by publishing an earlier companion volume(Laver and Shepsle, 1994). Alex Holzman, political science editor atCambridge, made sure that all the trains ran on time throughout the


13/315

Acknowledgmentsproduction process. And series editors, James Alt and Douglass North,provided advice and counsel. Thanks to them all.Finally, we thank one another and our families. Aside from the oneyear that we were together in Cambridge, Massachusetts, this project hasbeen sustained mostly by e-mail, fax machine, and transatlantic traveland telephone. Blocks of time have been spent away from kith and kinand we appreciate their support, encouragement, and generosity in toler-ating these absences. Those same blocks of time were spent in the house-hold of a coauthor, so each would like to thank the other's family for finehospitality. Finally, there is the tip of the hat from one coauthor to theother. The collaboration has also been a friendship.

Michael LaverDalkey, IrelandKenneth A. ShepsleWellesley, M assachusetts

XI


14/315


15/315

P ART I

The context


16/315


17/315

Theory, institutions, and government formation

GOVERNMENT FORMATION AND PARLIAMENTARYDEMOCRACYThis book is about the making and breaking of governments in parlia-mentary democracies. The essence of parliamentary democracy is theaccountability of the government (or cabinet or executive or administra-tion) to the legislature. In most working parliamentary democracies, thisrelationship is enshrined in a constitutional provision that the govern-ment must retain the support of a majority of legislators, tested in alegislative motion of confidence (or no confidence) in the executive. Agovernment that loses such a vote is defeated and obliged to resign.1Although formal mechanisms for installing new governments differ con-siderably between countries, a government, however installed, is immedi-ately exposed in the legislature to a potential vote of no confidence. It isthe ability of the executive to win this vote and thus maintain theconfidence of the legislature that is the universal acid test of governmentviability in a parliamentary democracy.Even though the executive is responsible in this sense to the legislature,the cabinet and its ministers nevertheless retain wide-ranging power togovern the country as long as they remain in office. This power comprisesboth effective political control over the administrative departments ofstate and a firm grip on the day-to-day activities of the legislature. Apartfrom the opposition's ability to seek dismissal of an administration byproposing a motion of no confidence, it is difficult in most parliamentarydemocracies for anyone outside the executive to have a significant impacton the process of legislation. Executive control of the entire legislative

Hn practice, an executive may well resign in anticipation of losing a motion of noconfidence, so that we may only rarely actually see the constitutional procedurepushed to its ultimate conclusion. This does not, of course, take away from thefundamental importance of the confidence vote as one of the key institutional founda-tions of parliamentary democracy.


18/315

The contextprocess, from the drafting of bills to the determination of the order ofbusiness, combined with the almost universal provision that the executivemay recommend the dissolution of the legislature and the holding of newelections,2 gives any incumbent administration considerable power overthe legislature.3The impact of government formation extends far beyond legislative-executive relations, however. Government formation is at the very heartof representative politics, as well as of a large part of economic and sociallife, in Western parliamentary democracies. If representative politics iswhat provides the vital connections between the preferences of ordinaryindividuals and the ambitions of politicians, between elections and partycompetition, and between legislative politics, policy making, and the gov-ernance of the country, then the formation of a government is the focalpoint at which all of these things come together in a parliamentary democ-racy. Elections have meaning for voters because they provide a choicebetween different packages of possible political outcomes, offered bydifferent political parties. Parties themselves are important because theyare key actors that take part in the government formation process. Thismeans that party competition in parliamentary democracies is typicallycouched in terms of what different parties would and would not do ifthey were in government and thereby had a chance to put their ideas intopractice.After the election has been held in most parliamentary democracies, itis often the case that no party will have won an overall majority. Subse-quent legislative politics is much more about building and maintaining agovernment than it is about legislating. Even when legislation is impor-tant, the legislative agenda in most parliamentary democracies is verymuch in the hands of the government.Government formation also impinges fundamentally upon economiclife. Indeed the claim that a change of government implies at least apotential change of public policy is well illustrated by the market turbu-lence often caused by uncertainties about the stability of a government orthe outcome of an election. Good examples of this can be found in thegeneral instability of European currency markets in the run-up to theFrench legislative election of 1993, given uncertainties about the impactof a change of government on official policy toward the French franc, and

2An important counterexample is Norway, in which the government cannot inter-fere with a fixed three-yearly timetable of legislative elections.3Obviously, the extent to which the executive controls the business of the legislaturevaries from country to cou ntry w ith, for exam ple, variations in the role and power oflegislative committees. Nonetheless, in nearly all parliamentary democracies, thepower of the executive over the legislature is very considerable. This point is made bynearly every country specialist describing the interactions between legislature andexecutive in the parliamentary democracies covered in Laver and Shepsle (1994).


19/315

Theory, institutions, and government formationthe intense speculation against the peseta prior to the Spanish election ofthe same year. This turbulence arises precisely because major shifts in thepartisan composition of the government can imply major changes in therules of the economic game. And uncertainties about these rules of thegame can feed back in a very volatile way onto expectations about keyeconomic parameters, whether these relate to the money supply, the bud-get deficit, exchange and interest rates, or whatever.

OUR BASIC PREMISESIn what follows, therefore, we set out to explore as systematically aspossible the complicated business of building and maintaining agovern-ment in a parliamentary democracy. We base our argument upon a num-ber of premises and it is as well to be quite clear about the essential thrus tof these from the very start, though we will elaborate on them in thefollowing chapter.

The role of modelsOur most fundamental premise is that it is possible to make generalstatements about the politics of building and maintaining a government,and that such general statements can give us valuable insights into thepolitical processes involved. Obviously, even a casual look at parliamen-tary democracies around the world reveals huge diversity. Every countryhas a different constitution and a different historical, geographical, andcultural setting. Once we look at the formation of any individual govern-ment, the sheer welter of activity by large numbers of politicians, offi-cials, lobbyists, interest groups, voters, and commentators in the making,running, and breaking of governments is so rich and complex as to makeevery particular government appear to be an utterly distinctive politicalentity, quite unlike any other. The purpose of our enterprise, however, isto cut through all of this variation inan attempt to see whether, at thecore of it all, there are interesting and important features of governmentformation that are common to all parliamentary democracies.We do this by developing a generic model of government formation inparliamentary democracies. Our model consists of a set of general state-ments about the process of government formation, statements that wetake to be at least somewhat plausible, and a set of logical linkagesbetween these. Our hope and intention is that systematic use of thismodel allows us to develop some interesting and nonobvious insights intothe business of building and maintaining a government. Obviously, themodel we develop can be no substitute for the careful description of anyparticular case. But at the same time there are on the face of things strong


20/315

The contextsimilarities in the government formation process in parliamentary democ-racies at different places and times, and we do believe that our modelsheds light on these. It can do this in two different ways.

First, by allowing us at least logically to hold a wide range ofinstitutional and historical variations constant, while systematically ma-nipulating key variables, the model may be used as a discovery tool.This allows us to enhance our understanding by attempting to answer"what if?" questions about the politics of a particular situation. Whatif no party wins an overall majority after the next British election?What if, after dominating every Italian government in the postwar era,the Christian Democrats suffer electoral catastrophe, as they did inMarch 1994? What if, after over 40 years as a major player in Ger-man governments, the Free Democrats were to fall below the 5 percentelectoral threshold?Country specialists are typically called on to answer such questionsand they rarely shrink from doing so, drawing upon their detailed knowl-edge of a particular situation. To the extent that they can justify state-ments about what is likely to happen in circumstances that have not yetarisen, whether to others or merely to themselves, they must have in theirminds some private model of parliamentary politics, however vague andimplicit. Effectively they are saying "if we could hold these things con-stant and change tha t thing in such a way, then the outcome is likely to bethis." Such statements, of course, have exactly the same form as thosegenerated by more explicit models. We strongly believe that being moreprecise and explicit about models of politics can permit a scholar both tocheck his or her intuitions more systematically and to tease out furtheranalysis of the problem at hand - possibly counterintuitive but nonethe-less useful and even, sometimes, accurate. Explicit models of politics,therefore, are not enemies of intuition. On the contrary, they are enginesof intuition.

The second important intellectual role of models of politics is as anessential prerequisite for systematic and informative empirical research. Itis of course possible to collect facts about a particular situation in thebelief that the situation itself is self-evidently important, that the factsspeak for themselves, and that more facts are always better than fewer.Many of those who study politics unfortunately adopt this approach,either explicitly or implicitly. We believe, however, that this approach isinefficient at best and downright misleading more often than not. Theessential role of empirical research, as with pure theory, is to enhance ourunderstanding of the world. Empirical research in the social sciences doesthis by revealing the presence or absence of systematic patterns in actualhuman behavior. But there is of course a more or less infinite number ofbehaviors that we could study, and patterns in these that we could look


21/315

Theory, institutions, and government formationfor. Only a theoretical model of the situation under analysis can tell uswhich activities to study and what patterns to look out for.We can put some flesh on the bones of this point by taking a politicalexample closely related to our own work, but not a part of it. All parlia-mentary democracies have a constitutionally m andated maximum periodbetween elections. In many countries, however, the government may callan election before this period has expired. In these countries, elections arein fact almost always called before the obligatory date. Since we neverobserve it to be a binding constraint in practice, does this mean that wecould abolish the constitutionally mandated maximum period betweenelections with no impact on the actual frequency of elections? Mostpeople would intuitively feel that to change the constitution in this waywould make a big difference to electoral politics. Yet, in the absence ofany democratic country that has in practice abolished the constitutionalrequirement that elections be held every so many years, "pure" empiricalresearch on the calling of elections, with no model to structure it, wouldreveal that this constitutional provision was virtually never used. It takesa model to give shape to our intuitions. In this case it might be a modeldealing with the way in which the maximum interelectoral period acts asa constraint on politicians, a constraint that they anticipate by callingearly elections on dates that they expect to be more favorable to themthan the ultimate constitutional deadline.4 Without such a model, empiri-cal research on the calling of elections is likely to be misleading.Models are thus the essential intellectual equipment with which weboth give structure to our intuition and organize our empirical under-standing of the world. Of course, even the dogs in the street know that amodel of a cat's brain is no substitute for the cat's brain itself. Our simplemodels of parliamentary democracy are mere shadows of any particularinstance of the real thing. Indeed we quite consciously seek to abstractaway what Milton Friedman (1953) referred to as the "attendant circum-stances" of any case. To be sure one person's mere details are the heart ofthe matter for somebody else. This is another way of saying that what weinclude in a model, and what we exclude from it, are matters of choicethat we must be prepared to defend.But the reason we abstract from certain details by design is to allow usto formulate generalizations about a world in which we cannot conductcarefully controlled experiments. We want to find out whether things aretrue not only about one particular cat's brain but about the brains ofother cats. In the same way we want to find out whether things tend to betrue about parliamentary democracies as a whole, even if every parliamen-

4Such a model, used to make empirical estimates of the "censoring" effect ongovernment durations of constitutionally mandated maximum interelectoral periods,has been developed by King, Alt, Burns, and Laver (1990).


22/315

The contexttary democracy differs from every other in myriad ways. Models are notonly discovery tools for use in a single country, but are also tools that canhelp us to highlight, and to learn from, patterns underlying politics in arange of different countries.

Rationality and the role of policy in politicsEvery model of politics is based on some underlying assumption aboutwhat motivates the various actors and characterizes their decision mak-ing. The model we elaborate in this book is based on the assumption thatpoliticians are at least implicitly rational decision makers. By this wemean that the decisions they make are in some sense (to be elaborated inthe following chapter) expected to further their essential aims and aspira-tions. We do not assume that politicians are human calculating engines.We do assume, however, that most politicians do not behave at randombut rather act as if they have some agenda, and that the political decisionsthey make serve to further this agenda.For the purposes of this book we go one stage further and conceive ofthe agenda of individual politicians in terms of the public policies thatthey attempt to pursue. Thus a basic premise of this book is that policy isimportant for politics in general and for government formation in particu-lar. This may seem a somewhat controversial idea, given at least onepopular image of politicians as power-hungry egomaniacs concernedwith nothing beyond their own personal well-being, and prepared to sayand do almost anything in order to advance this. Fortunately, we do notactually need to challenge this rather depressing view of politicians whenwe claim that policy is important. Even if politicians are power-hungryegomaniacs who are not at all intrinsically interested in policy, policyseeking may nonetheless serve as their driving force. So long as policy isimportant for voters, activists, and other electorally relevant groups, poli-ticians will be impelled by the incentives of party competition, operatingoutside the government formation process, to heed the policy priorities ofvoters.5

Of course, policy is importan t if politicians actually do care about theeffects of implementing different policies but, even if they care not a whitfor the substance of policy, we assume that they do care about the conse-quences of failing to implement their policy promises. For this reason,politicians engaged in government formation can be characterized interms of the policy promises with which they are associated, and can be5This may be, for example, because politicians compete at elections by makingpolicy promises to voters and anticipate damage to their reelection chances if they areseen not to make an effort to implement these promises if given the chance to do soafter the election.


23/315

Theory, institutions, and government formationassumed to be concerned, as agents of electoral interests, about what anyprospective government will do once it has taken office. This premise, ofpolicy-driven politics and commonly known partisan policy positions,which we elaborate in the following chapter, is a cornerstone of ouranalysis.

GOVERNMENT FORMATION: THE STORY SO FARA review of what we might think of as traditional theories of governmentformation can be found in Laver and Schofield (1990). Rather surprisingly,and despite the fact that this is of the essence of parliamentary democracy,political scientists have tended not to take account of the intimate interac-tion between legislature and executive when modeling the life and death ofgovernments. Government formation and maintenance are typicallytreated as problems of legislative coalition building and many coalitiontheories implicitly assume that parliamentary democracies are governeddirectly by their legislatures. Many recent accounts of government forma-tion, therefore, are ineffect spatial models of legislative voting.Spatial conceptualizations of politics have a long history, extendingback at least as far as the seating arrangements of the French ConstituentAssembly after the Revolution of 1789, in which radicals sat themselvesto the left of the chair andconservatives to the right. In contemporaryscholarship, the patron saints of spatial analysis are Harold Hotelling(1929), Duncan Black (1958), and Anthony Downs (1957). In their analy-ses, politics is conducted in terms of a set of issues that can adequately bedescribed interms of a single dimension of policy this is the same left-right dimension as structured politics inpostrevolutionary France. Themost important result from this body of work is Black's Median VoterTheorem. Making the assumption that voters can be modeled solely interms of their most preferred policy (ideal point) on a single dimension ofpolicy, Black showed tha t the ideal point of the median voter has a vitallyimportant p roperty that has informed a huge body of subsequent work. Itis the only point that is preferred by some majority of voters to any otherpoint on the policy dimension.6

This result underlies the very important centripetal political tendenciesthat characterize the Hotelling-Downs-Black approach. If their approach6Formally, Black's Theorem isstated for the circumstance inwhich preferences are"single-peaked" and the number of voters isodd. "Single peakedness" describes theshape of each voter's preference curve on the left-right dimension: The preference

curve is at a maximum just over the voter's ideal point and declines from this peak forpoints farther away from the voter's ideal ineither direction. The theorem may easilybe extended toeven numbers of voters. In this case, there are two median voters andthe theorem says that any point in the closed interval bracketed by the ideal points ofthese voters has the prop erty that no other point is preferred to it by a majority.


24/315

The contextis applied to elections between two competing parties, each intent uponwinning majority voter support, then the implication is that each partywill gravitate toward the ideal policy position of the median voter. If onthe other hand their approach is applied to legislative policy making, inwhich choices are typically made between some substantive motion andamendments to this, then the implication is that the winning version ofthe motion will represent the policy position most preferred by the me-dian legislator. In either event, the final outcome of the political process isat the median of the distribution of ideal points. And this central outcomeis an equilibrium in the sense that, once arrived at, there will be nomovement away from it.There are good theoretical and empirical reasons to believe that poli-tics is not always one-dimensional, however. The natural generalizationof the unidimensional approach, with its well-behaved equilibrium at thecenter of the distribution of policy preferences, is a multidimensionalspatial model. In this, policy is characterized by as many dimensions asare necessary to capture the different world views of key political actors.Each voter's ideal political outcome may be thought of as resulting fromthe implementation of a bundle of ideal policies, one for each salientdimension. The famous "chaos" theorems of McKelvey (1976) andSchofield (1978) imply that, when two or more dimensions of policy areimportant, it is nearly certain that majority preferences between policyproposals will cycle indefinitely, with no proposal defeating all others.Any proposed bundle of policy positions will be majority-defeated byanother, which in turn will be majority-defeated by another, with thisprocess continuing until it cycles back to the original proposal and beginsall over again. This suggests either that policy is in perpetual flux, withevery policy proposal displaced by something else, or that the music willstop at some point and the actual proposal implemented will be selectedarbitrarily from those in the cycle. Except under the most highly unusualcircumstances, there is no equilibrium.It is worth noting tha t, in those special circumstances in which there isan equilibrium in multidimensional majority voting games, this equilib-rium is still at the center of things, in the sense that it is at the multidimen-sional median policy position.7 Kadane (1972) has shown that, whenpolicies are selected by majority rule in a multidimensional policy space,if an equilibrium exists (and it may not) it must be the m ultidimensionalmedian*

7The multidimensional median position is the position such that the median voteron each dimension gets his or her ideal policy on that dimension.8Kadane's theorem applies when voters have preferences that are single-peaked bydimension and separable by dimension. By "single-peaked by dimension" we meanthat on any specific dimension (say the j th ), hold ing preferences on the other n 110


25/315

Theory, institutions, and government formationThis result, of which we shall make use in Chapter 4, implies qualifiedcentripetality. If there is a multidimensional equilibrium, then it will becentral in precisely the same way that Black's Theorem yields a central

result for the unidimensional case. The stark contrast between the twocases, however, rests on the issue of existence: An equilibrium alwaysexists in the unidimensional case and very rarely does in its multidimen-sional counterpart.A central feature of these results, however, is that they apply to elec-toral or legislative decision making on particular policy proposals andthat the policies agreed upon are assumed to be implemented automati-cally as soon as they pass a majority vote. The results have nothing at allto say about decision making in the po litical executive, or indeed aboutany other aspect of the actual implementation of real policy decisions.This means that they do not apply directly to analyses of the making andbreaking of political executives in parliamentary democracies.Even the best recent work at the cutting edge of coalition theorycontinues to illustrate this lack of concern both for policy implementationand for the role of the executive. Very good examples are provided bySchofield (1993), working within the traditions of cooperative gametheory, and Baron (1991), who proposes a noncooperative theory ofgovernment formation.9dimensions fixed, the ith vo ter's m ost preferred position on this dimension is Vj'and herpreferences decline as points are considered further away from this location along thej t h dimension. By "separable by dimension" we mean that i's ideal point and prefer-ences on th e j t h dimension do n ot vary as compo nents on the other n 1 dimensions arechanged. In this sense dimensions may be treated as "inde pend ent" of one another. Wemay give Kadane's result some precision in a form that will prove useful to us insubse quen t analysis. Suppose there is an n-dimensional space of policies (n ^ 2), and aset V = {1, 2, . . . , v} of v oters. Let the ith voter's ideal be written y1 = (yj1, y-J, . . . , yj),where, generally, y 1 is the ith voter's ideal policy on the j t h dimension. Define themultidimensional median as the point m = (m b m2, . . . , m n), where irij is the medianon the j t h policy dimension (rrij = median ie V y-1). For convenience, we assume that thenum ber of voters is odd so tha t there is a unique median on each dimension. Kadane'sTheorem states that either m is an equilibrium or there is no equilibrium. Kadane'sproof uses an "imp rovem ent alg orithm " showing that any point not at the multidimen-sional median position will be beaten in a majority vote by a point that is identical inevery other respect, except that it is at the median position on one additional dimen-sion. Only for m can there be no such point (though m itself can be beaten by a pointdiffering from it on two or more dimensions).

Cooperative game theory is different from game-theoretic accounts of cooperation(see, e.g. Axelrod, 198 4; Taylor, 19 76, 1 987). Cooperative game theory is a theory ofstrategic interaction in which it is assumed that any deal that is consummated isenforceable. Although no details of the enforcement mechanism are part of this theory,implicit is some exogenous mechanism or third party (like a court) that enforces dealsby punishing violators with a certainty and severity sufficient to discourage reneging.Theories of cooperation, in contrast, are more interested in how individuals maysecure joint advantages, even in contexts in which there is no exogenous enforcementmechanism. In these circumstances, deals struck stay stuck because they are self-

11


26/315

The contextAccording to Baron (1991: 138), for example, governments areformed on the basis of a legislative vote of confidence and "[i]f themotion of confidence is successful, the policy program is implemented

and remains in effect until the next mandated election." Policy, oncesettled by the legislature, is implemented automatically and effortlessly.This is despite the fact that, in his noncooperative model, "[pjarties areunable to commit to how they will act in the future" (1991: 139). EvenBaron's definition of who is in the government makes no reference to thecabinet, but only to the legislature: "[t]he term government will be usedto refer to the parties that do not vote against a motion of confidence on apolicy proposal" (1991: 138, italics in original). Supporting parties whoreceive no cabinet seats at all are thus counted by Baron as governmentmem bers, as he in effect models a regime of "governance by legislature."For Schofield (1993:17), the forecast policy outcome of a coalition is aunique point chosen by coalition members as a compromise betweentheir ideal points. Implementation of this policy position is not modeled,and Schofield is silent on precisely what it means, in strategic terms, to be"in" government. As a consequence, in both these models the policyoutputs of governments are determined solely by the strategic situation inthe legislature, unmediated by the cabinet, the bureaucracy, or any otherexecutive institution.

When the executive is considered at all by conventional coalition theo-ries, it is seen not as a decision-making body in its own right, but rather asa prize to be shared out by a winning legislative coalition a set of perksof office, the most important of which are seats at the cabinet table(Browne and Feste, 1975; Browne and Franklin, 1973; Browne andFrendreis, 1980; Budge and Keman, 1990; Gamson, 1961). Thus theselatter theorists also see parliamentary government as a legislative process,not as a continual strategic interaction between legislature and executive.Cabinet portfolios, once allocated, are simply consumed as benefits bythe legislative parties that hold them.The conventional coalition theorists' view of governance by legislatureflies in the face of the undoubtedly firm control of legislatures by thegovernment of the day, a generalization recently reviewed and supportedin a series of country studies (Laver and Shepsle, 1994). As we shall see,conceiving of the structure of policy making and implementation in amore realistic manner provides an altogether richer and more suggestiveenforcing. The study of self-enforcing strategic interaction is associated w ith noncoop-erative game theory. This latter class of models is appropriate for circumstances inwhich deal-making violations may not be subsequently adjudicated (except by nailingthe SOB next time around). In general a priori terms, it might well be thought thatdeals among the parties of a national government are not susceptible to "external"enforcement and therefore that noncooperative game theory is the most appropriatetool for modeling them.

12


27/315

Theory, institutions, and governm ent formationaccount of government formation. A legislature makes and breaks govern-ments, to be sure, but it does not, in our view, rule the country.

THE INSTITUTIONS OF GOVERNMENT DECISIONMAKING (CABINETS, DEPARTMENTS, AND MINISTERS)In response to w hat we see as a major opportunity for theoretical innova-tion in the analysis of government formation, we present an alternativeapproach in this book. This approach takes more explicit account of theinstitutional, and more specifically the departmental, structure withinwhich policy decisions are made and implemented in the real world ofparliamentary democracy.Perhaps the most distinctive feature of our approach, therefore, is theassumption that most important policy decisions are taken by the execu-tive. Effective government decisions on a particular issue, furthermore,are assumed not to be selections of abstract positions in a continuousissue space, but rather to be choices between a very limited number ofwell-developed and implementable proposals for particular courses ofaction. Developing such proposals in the real world of public administra-tion is a major task requiring specialist skills. Taken together with thesheer volume of business that real governments must conduct, this im-

plies that only the government department with jurisdiction over a par-ticular policy area is effectively equipped to develop feasible and imple-mentable policy proposals in that area and present these to the cabinet fordecision. And this in turn means that the entire process of policy forma-tion on any given issue is very heavily influenced by whoever has politicalcontrol over the relevant government department the cabinet ministerin charge of the department concerned.10 In effect, the development offully elaborated policy proposals that real governments can actuallychoose between is an activity strongly structured along departmentallines and heavily conditioned by the views of the political departmenthead. Government departments and their political masters, in short, pos-sess considerable agenda power (Burch, 1993).

This specialization and division of labor in the development of real-world policy proposals implies tha t the most credible signal of governmentpolicy on a particular issue is the policy associated with the minister incharge of the department with jurisdiction over this issue. Thus a hard-linedefense policy is signaled, for example, by a politician with a reputation as10

We shall, most of the time, only be concerned with the party label of an incumb entminister, on the hypothesis that, personal idiosyncracies aside, politicians can be ex-pected to toe their pa rty's line. In C hapter 12 we examine intraparty politics, at whichtime we entertain the possibility that different ministerial candidates from the sameparty may have different goals.13


28/315

The contexta ha w k serving as minister for defense. A mo re conciliatory defense policyis signaled by replacing this defense minister with a politician, perhapsfrom a different party, who has a reputation as a dove.

As evidence of the plausibility of this general assumption, consider theanalyses typically produced in quality newspapers when a new cabinettakes office. Much of this analysis concerns the implications for govern-ment policy of having particular individuals or parties, with particularrepu tations , in charge of particular governmen t depa rtme nts. Analysis bythe Paris correspondent of the Economist (April 3, 1993, p. 29) on theappointment of a new French cabinet in early 1993 serves to illustratethis point:Mr Balladur's new . . . go ve rn m en t... is a model offine-tuningand fairness. Allof the main factions within the conservative coalition are represented. A leadinganti-Maastricht campaigner, Charles Pasqua, returns to his old post at the Minis-try of the Interior. But the tone of the government is strongly pro-European andpro-Maastricht, as is Mr. Balladur himself.Alain Juppe, secretary-general of the RPR, who helped to persuade a hesitantJacques Chirac, leader of the Gaullists, to campaign for a Yes vote in theMaastricht referendum, takes over as foreign minister. Alain Lamassoure, anardent European, takes the European affairs portfolio. Edmond Alphandery, an-other Maastricht supporter and passionate advocate of the "strong franc" policy(as is Mr. Balladur) becomes financeminister . . . . More worrying is the appoint-ment of Senator Jean Puech, a farm lobbyist, to agriculture. This is a sensitivepost, given the right's campaign promise to try to veto the GATT farm deal.Appointing a farm lobbyist bodes ill for world trade.The new cabinet (from which the three main right-wing leaders, Mr. Chirac,Valery Giscard d'Estaing and Raymond Barre, are all notably absent) won Mr.Mitterand's immediate approval. That was despite the appointment of FrancoisLeotard, former leader of the Parti Republicain . . . as defence minister. Mr.Mitterand is reported to have rejected Mr. Leotard's appointment to the samepost in 1986. The president has promised "scrupulously to respect" the will of thepeople for a change of government policy.

Equivalent passages, in a wide range of sources, can be found after theappointment of almost any new cabinet in a parliamentary democracy.This reflects a strong tendency for those who need to be able to forecastgovernment policy and the businessmen and -women who read theEconomist are good examples of these to base their forecasts on thesignals that are sent as a result of the allocation of cabinet portfoliosbetween politicians with well-known positions of key issues.Thus the logic of departmentalism in government decision makingimplies that the overall policy position associated with any given govern-ment is determined by the allocation of cabinet portfolios among govern-ment participants. For us, then, a government consists of an allocation of

14


29/315

Theory, institutions, and government form ationauthority in particular policy jurisdictions to particular political parties(see note 10) with well-known policy reputations in these areas. Becausethere is only a handful of key policy jurisdictions and only a limitednumber of parties with ministerial-caliber politicians who can credibly benominated to these, the number of different potential cabinets is alsolimited. Thus, government policy outputs are selected from a finite set ofpolicy forecasts, each forecast being associated with a particular portfolioallocation. The finite nature of the set of credible potential governmentsmeans that the business of building and maintaining a government isexplicable, as we shall see, in a more straightforward manner than thegeneral spatial model suggests.

P L A N O F C A M P A I G NThe argument in this book is developed in four parts. In the remainder ofthis first part, we lay the foundations for a model of government forma-tion in parliamentary democracies. We do this initially in Chapter 2,where we set out and defend our basic ideas and assumptions. In Chapter3 we combine our ideas and assumptions into a generic model of govern-ment formation.Part II is devoted to refining this model. In Chapter 4, we elaborate themodel by using it to explore the notion of an equilibrium cabinet. One ofthe reasons why government formation need not be chaotic is tha t there isoften the possibility of an equilibrium cabinet at the generalized medianpolicy position. Another source of stability in government formation isthat there may be a party in a distinctively powerful bargaining positionable to dominate the progress of coalition bargaining, a theme we de-velop in Chapter 5.In the third part of the analysis we use our model to structure anempirical exploration of government formation in the real world. InChapter 6 we use the model to motivate two brief case studies of realgovernment formation - Germany in 1987 and Ireland in 1 99 2-19 93 . InChapters 79 we conduct a more systematic empirical analysis, an analy-sis that we feel shows the value of our approach in a very convincingmanner.The fourth part of the book develops our model further, exploring arange of intriguing methodological and substantive problems associatedwith analyses of party competition and government formation. We begthe reader's indulgence until this point, for it is here that we entertainextensions of our model's domain and relaxations of its assumptions. InChapter 10, we use the model to help us understand factors affecting thestability and instability of governments. We use simulations to examinethe likelihood of particular cabinets being thrown into disequilibrium a

15


30/315

The contextfirst step on the road to developing a systematic theoretical account ofcabinet stability. In Chapter 11 we treat three conceptual issues that bearon our model in particular and party competition in general. These haveto do with the structure of the issue space in which politics is conducted.First, we look at correlations between positions on different issue dimen-sions, and the ways in which higher correlations greatly enhance theprospects for and stability of equilibrium cabinets. Second, we look atwhat happens when preferences over public policy on one issue dimen-sion are affected by public policy on a quite different dimension - whenviews abou t a country 's foreign policy, for example, are affected by whicheconomic policy it is pursuing.11 Third, we look at what happens when,as is typically the case in the real world, government departments havejurisdiction over a bundle of different policy dimensions, and jurisdic-tions are in our sense complex.In Chapter 12 are three extensions of our model to interesting substan-tive questions related to the making and breaking of governments. First,we look at intraparty politics, and the prospect that what we have up tothis point considered as a unitary party is in fact a coalition of factions,each supporting a set of cabinet-rank politicians. Second, we look atminority cabinets, in which the cabinet comprises fewer parties thannecessary to control a legislative majority, and surplus majority cabinets,from which at least one party could be expelled, while still leaving thecabinet controlling a majority. Although such cabinets appear as anoma-lies in many accounts of government formation, both types are not onlyquite common empirically, but can easily be accounted for within ourgeneral approach as well. Third, we look at administrative reform, andthe prospect that alterations in the allocation of policy areas to govern-ment departments may be an integral part of what is on the table when agovernment is formed, rather than a rigid administrative structure thatmust be taken as given.

In the final chapter of the book, we review the main themes that havebeen highlighted by our analyses most notably legislativeexecutiverelations, departmentalism, and centripetal policy tendencies - and dis-cuss the extent to which these can be reconciled with some of the receivedwisdom of political science. Throughout the volume we are rather catho-lic in our approach, combining the deductive tools of formal politicaltheory with multivariate statistical analysis, case studies, and simulationtechniques.The single most important lesson to be learned from the discussion tofollow concerns the institutional arrangements of parliamentary democ-1 formally, what we do at this stage is to relax the assumption of separablepreferences.

16


31/315

Theory, institutions, and government formationracy. In particular, the departmental organization of governmental deci-sion making structures the environment in which governments are born,live, and die. The model that we develop takes this structure as its basicpremise. On our account, when politicians consider the making andbreaking of governments, they look ahead both to the most likely conse-quences of putting a particular cabinet in charge of running the country'saffairs and to the most likely consequences of replacing that cabinet withsome viable alternative. This leads to a theory of cabinet equilibrium.Majority-rule models that ignore the structure of government decisionmaking often fail to identify equilibrium outcomes. Their conclusionstands in stark contrast to the real political world, where governmentformation appears to be more stable and is not characterized by theendless churning of alternative administrations that disequilibrium im-plies. A striking feature of our approach is that cabinet equilibriums arequite common and tend to be close to the center of the configuration ofparty policy positions.The search for an account of equilibrium cabinets was our originalmotivation; its theoretical development and empirical evaluation areprobably the main finished products of this volume. This search, how-ever, proved unexpectedly fruitful. What became clear to us as we pro-ceeded is that the making and breaking of governments is at the crux ofmuch of what is important about parliamentary democracy. We believethe model throws light on a range of substantively important features ofthe political process, including government stability, minority and surplusmajority government, intraparty politics, and administrative reform. Wehope tha t even a reader initially skeptical about our general approach willagree that the model offers the opportunity to visit some interestingsubstantive destinations.

17


32/315

The social context o f governm ent formation

The political process that dominates our discussion in the rest of thisbook involves a set of politicians in a parliamentary democracy, eachmotivated to achieve some objective or another, competing and cooperat-ing among themselves to form a government. The hopes and fears, aimsand aspirations of each politician are fulfilled to a greater or lesser extentdepending on the outcome of this process.Before we can develop any systematic analysis of this crucial aspect ofdemocratic politics, we must be explicit about our assumptions. In thischapter, therefore, we elaborate assumptions on a range of matters rele-vant to the making and breaking of governments in parliamentary democ-racies. These include the aims, aspirations, and rationality of key actors;the institutional process by which a government is formed; the manner inwhich actors forecast the likely consequences of having different govern-ments in power; and collective decision making both between and withinparties . We begin with perhaps the most fundamental assumptions of all,which have to do with the hopes and fears of politicians and the rationalcalculus that they use to make decisions.

THE MOTIVATIONS OF POLITICIANSThose writing on the politics of government formation tend to assumeone of two things about the fundamental motivations of politicians in-volved in political bargaining.1 Some assume that politicians are con-cerned above all else to get into office - and that they will say and dowhatever is necessary to achieve this. In their search for power, politiciansmay make policy promises either to each other or to the electorate but, onthis interpretation, such policies are promoted for purely instrumental

xFor a review of writings on the motivations of politicians engaged in governmentformation, see Laver and Schofield (1990: chap. 3).18


33/315

Social context of government formationreasons. If the chances of getting into power are enhanced by changingpolicies, then an "office-seeking" politician will see no cost in changing.Nonetheless, to assume that a politician is an office seeker is notnecessarily to imply that he or she will change policies just to suit the timeof day. The need to maintain long-term credibility, both with fellowpoliticians and with voters, may well provide politicians with incentivesto stick with a policy that in the short term has come to look like aliability. Thus even office-seeking politicians may promote distinctive poli-cies, and the need to maintain subsequent credibility may ensure thatthese policies are quite stable. Indeed, once office-seeking politicians haveprom oted a particular policy, a strong need for credibility may even resultin them behaving just as if they were sincerely trying to achieve the policyin question.Perhaps the most influential exponent of the office-seeking assumptionabout the motivations of politicians has been Anthony Downs, whoseposition is summed up in his famous statement that "parties formulatepolicies in order to win elections, rather than win elections in order toformulate policies" (Downs, 1957: 28). Much of what Downs has subse-quently become famous for is concerned with policy competition betweenparties . As his basic motivational assumption shows, however, the funda-mental driving force behind such competition is the desire to get intooffice; policy advocacy for Downs is purely instrumental. AlthoughDowns does have a brief discussion of multiparty politics, he was writingfor the most part with a two-party system in mind. For him, "winning"an election automatically implies going into office and coalition buildingplays no part in government formation. He also assumed that partiesbehave as if they are unitary actors.An alternative influential assumption about the fundamental motiva-tion of key political actors is that they are intrinsically concerned aboveall else with public policy. De Swaan (1973: 88), for example, assumedthe "considerations of policy are foremost in the minds of the actors . . .the parliamentary game is, in fact, about the determination of majorgovernment policy." This is not, of course, to deny that it may be neces-sary to get into power in order to enact policy. Getting into power in-volves both making promises to the electorate and, in coalition systems,making deals with fellow politicians. Both in elections and in governmentformation, it may be necessary for a party to make compromises changing ideal policies in order to make them more acceptable to somegroup or another. Just as office-seeking politicians may stick to certainpolicies that they don't actually care about, policy-seeking politiciansmay change policies that they do care about. A policy-seeking politiciantrying to get into power in order to implement certain ideals may evenbehave as if he or she is a pure office seeker.

19


34/315


35/315

Social context of government formationStage: 1 2 3 4 5

Figure 2.1. A decision tree

how ever, our kn ow ledge of geometry allows an analyst explicitly, compre -hensively, and realistically to model the decisions involved.Similarly, we need not assume that rational politicians are continu-ously performing explicit costbenefit calculations when they make stra-tegic decisions designed to fulfill their basic objectives. What we doassume is that, just as people cross a crowded room as if they areperforming a complex series of interlocking geometric calculations, poli-ticians make strategic decisions as if they have the ability to perform acontinuous costbenefit calculus. And it is this calculus that we model.An essential part of the process of making effective strategic decisionsis looking ahead. Each course of action chosen leads to another set ofdecisions and, at any particular time, each decision maker stands at anode on a complex decision tree such as that shown in Figure 2.1. 2 For aperson to get from A, where they are now, to B, where they want to be,typically involves decisions arising in a number of distinct stages. If no-body else is involved, then these decisions may all be taken by the sameper son . Ge tting from A to B in Figure 2.1 involves choosing option r overopt ion q at Stage 1, optio n s over optio n t at Stage 2, option v over o ptionu at Stage 3, option w over opti on x at Stage 4, and o ption z over option y

2Another way of thinking of Figure 2.1 is as the representation of an extensive formgame generated by the decision process being modeled.21


36/315

The contextat Stage 5. In this way an actor navigates her way toward her objectiveacross a decision-making landscape that may be physical, social, or psy-chological. Knowing an actor's preference for B over any other outcomeand assuming her to be rational, one can predict her choice at any stage inthe decision process. In this sense the outcome of the entire process is aforegone conclusion.Most political decisions, of course, involve interdependent choices outcomes depend on a sequence of choices made by a number of actorswhose fates are intertwined with one another. Thus, for example, thedecision tree in Figure 2.1 may involve two actors who make their choicesin sequence. They could, for example, be two presidential candidatesacting and reacting to each other during an election campaign. Actor 1decides between q and r at Stage 1. Actor 2 decides between s and t atStage 2, if Actor 1 has already chosen option r, and between other optionsif Actor 1 chose option q. If Actor 2 chose option s at Stage 2, then Actor1 may choose between options u and v at Stage 3, otherwise she mustchoose between other options, and so on.If the preferences of the various actors over the possible final outcomesare well known, then it is once more possible to predict the course of theirinteraction quite straightforwardly. This is done by working backwardfrom each potential final outcome. To continue with the example inFigure 2.1, it will be known that, at Stage 5, if the game is at the nodeleading to outcome B, then Actor 1 will choose option z over option y,yielding her most-preferred outcome. Since this is the final move in thegame, she will surely make such a choice if she is rational. Similarly, it ispossible to predict with certainty the choice of Actor 1 at every one of theother decision nodes that might be reached in the final stage of thedecision process. In effect, we can replace each decision nc ie at Stage 5with the outcome we surely know will be chosen by the actor with thepower to do so. We can think of these outcomes as the "strategic equiva-lents" of the decision nodes we know will lead to them.Having replaced every final decision node with its strategic equivalentand using the same logic, we can now predict which option Actor 2 willchoose at each of the decision nodes in Stage 4 of the process. This isbecause, knowing what will happen at Stage 5, we know the particularfinal outcome generated by any choice Actor 2 makes at Stage 4. Know-ing Actor 2's preferences over these outcomes, we can predict whichoption she would choose if she were to find herself at any of the Stage 4decision nodes. We can therefore replace each of these decision nodeswith its strategic equivalent.In this Avay, we work back up the decision tree, at each stage replacingdecision nodes w ith their strategic equivalents. Once we have determinedthe strategic equivalent of the very first decision node, we have in effect

22


37/315

Social context of government formationforecast the entire course of the game and predicted the final outcome. Ifthe strategic equivalent of being at the first decision node of the game inFigure 2.1 is outcome B, for example, then the strategic equivalent of theentire decision process is B.Another way of putting this is that, given sufficiently rich knowledge,everyone will have rational foresight concerning the outcome of thisdecision process, if it is embarked upon. One important implication ofthis is that, since everyone can easily forecast this outcome, it is notnecessary for the actors actually to work their way through the decisionprocess in order to find out what the outcome will be. In a world ofperfect and complete information, once the actors are at A, they might aswell be at B.

For the same reason it is not necessary for two chess players actually toplay a particular game out to a checkmate, once both recognize that thisis the inevitable outcome. Once it is clear in the final move in the gamethat Cassie will checkmate Seth, then Seth will resign before this. Once itis clear that Seth can do nothing to avoid allowing Cassie into this desir-able penultimate position, he will resign. Once it is clear that Cassie canmake an antepenultimate move that puts Seth in this vulnerable position,then he will resign. Once it is clear that Seth cannot make a move thatprevents Cassie's antepenultimate strategic stroke, Seth resigns. In short,the players will work back up the game tree from the inevitable endingand anticipate it. As soon as the ending can be seen as inevitable - andthe better the players, the further in advance this can be seen the endingis assumed and the losing player resigns.It should be clear that the logic of these examples assumes that eachactor has perfect information about all aspects of the decision process,and complete information about the preferences of other actors. In thearguments that follow, we do for the most part assume that the rules ofthe government formation process and the preferences, expressed as pol-icy positions, of all relevant actors are common knowledge in this sense.3The knowledge allows each actor to forecast the choice of every relevantdecision maker at each stage in the government formation process.Obviously, the assumption that all actors have perfect and completeinformation about both the rules of the government formation processand the preferences of all other relevant actors can never be realized inpractice. The big wide world is much messier than this, and bargainingover the formation of real governments will in practice be much morefluid than we assume. The same problem does of course apply to any

3However, it is important to bear in mind, as we saw in the previous section, thatthese policies do not necessarily represent sincere preferences, but may represent pub-lished positions from which parties, for reasons originating outside the governmentformation process, are unwilling to diverge.23


38/315

The contextcarefully specified assumption that we or anybody else might care tomake about how politicians interact, and it is not an argument for notmaking assumptions or constructing models of the world. N otwithstand-ing this, the assumption that all actors' policy positions are commonknowledge is probably more realistic in a model of government forma-tion than it is in models of many other political processes. Governmentformation, after all, is a game played by a very limited number of eliteactors who have typically interacted over a long period and thereforeknow each other rather well. The assumption isuseful analytically be-cause it allows us to develop a relatively parsimonious model of govern-ment formation.

POLITICS WITHIN PARTIESIn all parliamentary democracies, politicians belong topolitical partiesthat are more or less disciplined, in the sense that most members, for themost part, follow party directives on both policy and strategy. It has thusbeen quite common for people writing about the politics of governmentformation to treat political parties as if they were unitary actors. Indeed itis common in this literature for references to political parties to be heavilyanthropomorphic and inparticular to treat each party as if it thinkswith a single mind.Yet parties can and do both split and fuse. It is also quite clear thatvigorous and sometimes bitter debate about both policy and strategy doesgo on inside political parties. Internal party turmoil, such as that whichresulted in the replacement of Margaret Thatcher as British Conservativeleader by John Major, can lead to dramatic changes at the governmentallevel. So are we justified in treating parties as unitary actors when wemodel the making and breaking of governments?There are sound theoretical reasons to suppose that political parties inparliamentary democracies will behave in a disciplined manner. The ratio-nale is straightforward: If individual party politicians were permitted topursue their own private desires at every opportunity, then their party'sreputation in the wider political process would constantly be put at risk.People would not know what the party stood for, so that voters, interestgroups, and others who might see the party as acting on their behalfwould have nothing to rely on. As a consequence, the party would beseverely hampered in its ability to attract support. Thus, the ultimateenforcers of party discipline are voters who want to know what they arevoting for at election time, and who will not vote for parties if theycannot be confident that party politicians will stick to the party line afterthe election is over.Empirically, the unitary actor assumption has been extensively re-

24


39/315


40/315

The contextment if given the opportunity to do so. Note that an ideal policy positionin these terms is a realistic policy. It is not the policy tha t the party wouldlike to see implemented in a perfect world, but rather the policy that theparty would implement in the real world in which it finds itself.It has become conventional in recent years to describe the policy posi-tions of political actors in terms of a set of key dimensions of policy. Formany people, the most familiar of these is the leftright dimension ofsocioeconomic policy. Those who use this dimension in effect assume acorrelation between party positions on economic issues such as publicspending and on social issues such as capital punishment or abortion. Theleftright dimension can then be used to describe a typical Europeanparty system by placing a communist or left socialist party at the left-hand end, followed by a socialist or social democratic party, perhaps byan agrarian or center party, a Christian democratic party, a secular conser-vative party, and possibly a neofascist party on the extreme right. The useof a single dimension of policy to describe a party system is simple andintuitive, has a tradition stretching back at least to the French Revolution,and of course loses a lot of detailed information about the party system.Using a larger number of independent dimensions of policy separatingeconomic and social policy, for example - enables richer descriptions ofparty politics,4 but can greatly increase analytical complexity.

This leaves us with some extremely difficult questions to answer. Howmany independent dimensions of policy should we use to describe ade-quately the ideal policy position of a given political party? Taking allparties in a given system together, how many dimensions of policy do weneed for an adequate description of party competition? These are deepissues tha t we cannot explore fully here (for some thoughts on the matter,see Laver and Hunt, 1992). Nonetheless, since it is of the essence of ourapproach tha t party competition and government behavior are describedin terms of positions on key policy dimensions, we cannot ignore themcompletely. The basic problem is one of deciding upon how to choose themost appropriate dimensionality of the policy space we use to describeparty competition, in the knowledge that the choice of different dimen-sionalities can yield quite different descriptions. In concrete terms, howdo we decide whether the most appropriate description of party competi-tion in Germany, say, is in terms of one, two, three,five, en, or even fiftykey dimensions of policy?Obviously, depending upon the level of detail we are prepared to use incollecting information on policy positions, any number of independentdimensions of policy could be detected. For example, one research project

4For example, a liberal party may now be placed toward the right of an economicpolicy dimension and toward the left of a social policy dimension.26


41/315

Social context of government formationhas estimated party positions by content analyzing party election manifes-tos for a range of European party systems on a total of 54 policy dimen-sions. Various techniques of dimensional analysis, such as factor analysisor multidimensional scaling, were used to collapse this large number ofpolicy dimensions into a smaller number of underlying dimensions. Ulti-mately, all of these data were also used to estimate positions on a singleleft-right dimension. Thus the same dataset could in this case be used togenerate anything from a 1-dimensional to a 54-dimensional representa-tion of policies in a given party system (Budge, Robertson, and Head,1987). Furthermore, there is no single "correct" representation. Just asthere is no map of a given piece of territory that is perfect for everypossible purpose we need one map for locating mountain ranges oroceans and another for rinding a needle in a haystack there is no unam-biguously correct dimensionality for the policy space we use to describeparty competition. Different applications call for different levels of detailin our description.At the same time, party positions on sets of policy dimensions can behighly correlated with one another. Thus it may well be the case thatparty positions on the public sector borrowing dimension, for example,can be very accurately predicted from their positions on the personaltaxation and/or inflation dimensions. Positions on abortion might well behighly correlated with those on divorce, and so on. This implies that areasonably parsimonious and intuitive description of party policy posi-tions can be given in terms of a limited number of independent underlyingdimensions of policy, a matter that we consider in more detail in Chapter11. If this is true, almost any new salient policy area that we might thinkof will generate a set of party positions that is highly correlated with oneor more of the underlying policy dimensions. The problem now becomesone of estimating the number of underlying policy dimensions necessaryto generate party positions on any salient issue in party competition.

Estimating the dimensionality of a policy space in these terms is stillnot a straightforward task, once it is accepted that published policy posi-tions may in some sense be as much a product of party competition as anexogenous input to it. Normal techniques of dimensional analysis allassume tha t, if positions on two dimensions are perfectly correlated, thenthere is a single dimension underlying them both. But in the presentcontext it may equally be the case that the correlated party positions are astrategic response to a particular circumstance of party competition. Par-ties could in principle choose positions on the two dimensions quiteindependently; in a particular set of circumstances they do not. Thus, isthere one underlying dimension, or two?We offer no magic formula for determining the most appropriate di-mensionality of any given representation of party competition. We have

27


42/315

The contextraised the issue in the preceding discussion, however, to highlight the factthat the choice of any given spatial representation of a particular partysystem is no more than one among many that could have been chosen.What we offer in the following chapters, therefore, is a set of tools thatcan be applied to a particular spatial representation of party competitionthat the analyst feels happy with for reasons that are quite outside ourterms of reference. We have nothing to say about which particular spatialrepresentation might be appropriate. But we must very firmlydraw atten-tion to the fact that the choice of one spatial representation over anotherhas fundamental consequences for the ensuing analysis. This implies that,for obvious reasons, the analyst should be very sure that he or she feelscomfortable that the spatial representation to be used is valid before theanalysis is commenced and should resist all temptations to fiddle withthe space once the analysis is under way.As we shall see when we come in subsequent chapters actually to applyour model to particular real-world cases, however, settling upon a particu-lar set of policy dimensions on which to base an analysis of governmentformation is not the open-ended problem that it might appear to be whenconsidering the problem in purely abstract terms. One of the interestingfeatures of our approach is that we come to most real-world cases withmuch of the dimensional analysis already done for us by the politicalsystem itself. This is because particular policy dimensions have been for-mally identified as being salient by virtue of having been designated aspart of the official jurisdiction of a cabinet ministry. Indeed the formaldetermination of the policy responsibilities of particular cabinet portfo-lios, and hence the identification of the policy dimensions that are "inplay," is something that we can typically take as given, even if somewhatfuzzy around the edges, at the beginning of the government formationprocess.5

THE GOVERNMENTThe government as the cabinet

People living in parliamentary democracies typically think of "the govern-ment" of their country in terms of the cabinet. Of the many arms ofgovernment the civil service, the judiciary, publicly owned corpora-tions, and the various tiers of local administration, for example it is the5Obviously, the allocation of policy dimensions to the jurisdiction of different cabi-

net portfolios may change as a result of the government formation process, a possibil-ity to which we return in Chapter 12. Nonetheless, the set of portfolios that character-izes the outgoing gov ernment the status quo in the government formation process can be taken as identifying, almost in the manner of a giant living factor analysis, aparticular set of policy dimensions as being salient.28


43/315

Social context of governm ent formationcabinet that symbolizes the apex of political responsibility. It is the cabi-net that is expected to guide affairs of state by making and overseeing theimplementation of policy on important issues. It is the cabinet that isexpected to react to and deal with major crises and emergencies. And,perhaps most significantly of all, it is the cabinet whose survival is on theline at election times. If voters in parliamentary democracies feel that theyhave any control over their political destinies, it is because they can passjudgment on the cabinet, albeit indirectly, when they vote in nationalelections. For many people, the terms cabinet and government are moreor less synonymous. Although many parliamentary democracies do notexplicitly use the term cabinet, all have a functional equivalent, whether itis called a council of ministers, an executive council, or whatever.

The typical process of forming a government in a parliamentary de-mocracy, to which we will return in more detail, involves the selection ofa prime minister, who must win a majority vote (actual or implicit) in thelegislature before he or she can take office. The prime minister thennominates a cabinet of ministers, which must also be capable of winninga majority vote (actual or implicit) in the legislature. The cabinet is free toset policy on any matter, subject only to the law of the land and to theconstraint that it can be dismissed on the basis of a majority vote in thelegislature. In practice, as we might expect, a rational anticipation ofdefeat may lead a cabinet to resign in advance of losing such a vote. Inthis way the cabinet is held collectively responsible for government policydecisions.In parliamentary democracies, therefore, changes in government de-pend on changes in the cabinet generated by changes in the balance offorces in the legislature. It is by having an effect on the balance of forcesin the legislature when voting at election time that voters in parliamen-tary democracies can have some control over affairs of state. The fact thatmany voters are in effect voting for potential governments rather than forindividual candidates has profound effects upon party competition, manyof which are far beyond our remit here. One, at least, is very importantfor the government formation process, however. If voters vote for partiesand not candidates, then the party label is a very valuable commodity and being denied the party label is a very serious sanction. This sanctionis a powerful weapon in the hands of the party organization, and thethreat of it reduces the incentives of individual legislators to defy theirparty line. This in turn contributes greatly to party discipline, and theconsequent ability of parties to function as if they were unitary actors.Closing the circle, it is because parties

booksclub.org making and breaking governments cabinets and legislatures in parliamentary democracies...

Documents