A Unified Theory of Party Competition: A Cross-National Analysis Integrating Spatial and Behavioral Factors

by James F. Adams, Samuel Merrill III, Bernard Grofman

Overview

The authors explain how parties and candidates position themselves on the Left-Right ideological dimension and other issue dimensions. Their unified theoretical approach to voter behavior and party strategies takes into account voter preferences, voter's partisan attachments, expected turnout, and the location of the political status quo. The approach, tested through extensive cross-national analysis, includes studies of the plurality-based two-party contests in the U.S. and multiple-party competition in France, Britain, and Norway.

Product Details

ISBN-13:	9780521836449
Publisher:	Cambridge University Press
Publication date:	03/21/2005
Pages:	332
Product dimensions:	0.87(w) x 9.02(h) x 5.98(d)

About the Author

James Adams is Associate Professor of Political Science at the University of California, Santa Barbara. His primary research interest is the application of spatial modeling to real world elections, and the insights this approach can provide into theories of political representation. He is the author of Party Competition and Responsible Party Government: A Theory of Spatial Competition Based upon Insights from Behavioral Voting Research (2001), as well as articles in the American Journal of Political Science, the Journal of Politics, the British Journal of Political Science, and Public Choice.

Samuel Merrill, III, is a Professor of Mathematics and Computer Science at Wilkes University, Pennsylvania. He received a PhD from Yale University, Connecticut. His current research involves mathematical and statistical modeling, particularly in political science. He is the author (with Bernard Grofman) of A Unified Theory of Voting (Cambridge, 1999) and Making Multicandidate Elections More Democratic (1988) and has published in a number of journals including the American Political Science Review, the American Journal of Political Science, and the Journal of the American Statistical Association. He has been a visiting professor at Yale University and a visiting scholar at the University of Washington.

Bernard Grofman is Professor of Political Science (and adjunct Professor of Economics) at the University of California, Irvine. He received his PhD in Political Science from the University of Chicago in 1972. He is an expert on comparative election systems and models of voting, and social choice theory. He has published over 200 articles in journals such as the American Political Science Review, the American Journal of Political Science, Electoral Studies, and Public Choice, and he has authored or co-edited seventeen books. He has been a Fellow of the Center for Advanced Study in the Behavioral Sciences at Stanford, a visiting professor at the University of Michigan, the University of Washington, and the University of Mannheim, and a scholar-in-residence at the University of Bologna, Kansai University, Osaka, the German Science Center, Berlin, Pompeu Fabra University, Barcelona and the Brookings Institution, Washington DC. He is past president of the Public Choice Society and a Fellow of the American Academy of Arts and Sciences.

Read an Excerpt

Cambridge University Press
0521836441 - A Unified Theory of Party Competition - A Cross-National Analysis Integrating Spatial and Behavioral Factors - by James F. Adams, Samuel Merrill III and Bernard Grofman
Excerpt

CHAPTER 1

Modeling Party Competition

1.1. Introduction to the Unified Theory of Party Competition

Despite the insights provided by the standard models of party competition, these models have failed to account for some fundamental empirical regularities found throughout the democratic world. Regardless of electoral system, political parties tend to be located across a wide ideological spectrum, and parties tend to exhibit considerable consistency in their ideological locations. Many models of party competition - particularly, spatial models that emphasize electoral incentives - generate predictions that are at variance with empirical observation. Either these theories predict convergence of party positions to similar, centrist positions (Enelow and Hinich 1984; de Palma, Hong, and Thisse 1990; Hinich and Munger 1994; Lin, Enelow, and Dorussen 1999), or they predict instability in party positions and the absence of an equilibrium (Schofield 1978; McKelvey 1986). With important exceptions to be discussed later,¹ models that follow the tradition of Anthony Downs (1957) highlight centripetal forces in party competition - forces that draw party positions toward the center of the policy spectrum. On the other hand, the main alternative approaches to spatial models - sociological models that are driven by voter demographics, and social psychological theories that rest entirely on notions of partisan loyalties - customarily fail to take into account the importance of policy issues and campaign strategies.

At the heart of this book is a desire to reconcile theory and evidence: to explain the theoretical puzzle of party divergence and to identify the factors that most strongly affect the extent of that divergence.² In order to address this puzzle, we offer a unified theory that implies that centrifugal as well as centripetal forces affect party competition, and we show that the theory accounts for much of the divergence of party policy strategies observed in a wide range of countries. Empirically, we investigate the strategic aspects of party locational choices vis-à-vis policy issues in four countries: the United States, Britain, France, and Norway. These countries differ both in their institutional arrangements - for example, parliamentary versus presidential democracies, plurality versus proportional representation voting systems - and in the number of competing parties, ranging from two-party competition in the case of the United States to seven-party competition in the case of Norway, with Britain and France representing intermediate cases.

Throughout most of the book (Chapters 2-10), we assume that parties are engaged in vote-maximizing behavior. Using only this assumption, we show that three factors, each linked to voter choice, can generate strong centrifugal pressures on the positioning of parties: (1) the existence of nonpolicy considerations in voter decision making - most especially, party loyalty; (2) the capacity of voters to discount the claims of parties or candidates concerning the policy changes they could achieve; and (3) an unwillingness of citizens to participate in the electoral process when they find that none of the competing parties or candidates are sufficiently attractive (abstention due to alienation). We refer to models incorporating these three factors as our unified theory of party competition.³ In later chapters, we will explicate exactly why each of these factors can be expected to promote party divergence.

In illuminating the theoretical importance of centrifugal factors, we show how to bridge the gap between two major schools of thought. On the one hand are spatial models in the tradition of Anthony Downs, which emphasize policy-oriented voters who support parties based on ideological or issue proximity to party platforms. On the other hand are voter-choice models that emphasize the existence and persistence of strong partisan loyalties or that view voting as rooted in class, ethnicity, or other sociodemographic factors. To put it simply (in terms familiar to many political scientists), in this book we show how to reconcile the spatial-modeling traditions of the University of Rochester with the behavioral traditions of the University of Michigan and Columbia.

We believe that a unified theory of party competition, which integrates the behavioralist's perspective on voting into the spatial-modeling framework, provides insights into party/candidate policy positioning and policy representation that neither approach provides by itself. For example, although spatial modelers increasingly recognize the critical role that voters' non-policy-related motivations can play in shaping party strategies, we believe that the extent of this impact on spatial strategies and mass-elite policy linkages is still imperfectly understood. We argue that when we incorporate voters' measured non-policy-related motivations into the spatial model, this changes the strategic logic of spatial competition in an important way: namely, it shows that parties and candidates have electoral incentives to appeal on policy grounds to voters who are biased toward them for nonpolicy reasons.

Because voters' non-policy-related characteristics often correlate with their policy beliefs (so that, for instance, working-class voters tend to support more liberal policies than do middle-class voters), this strategic incentive tends to draw vote-seeking parties and candidates to divergent regions of the policy space. Thus voters' non-policy-related considerations - even if parties cannot manipulate them in the course of an election campaign - are nonetheless crucial for understanding party strategies. A second, related argument is that because parties have electoral incentives to appeal to voters who support them for nonpolicy reasons, we should expect a strong connection between the policy positions that parties present and the beliefs of their supporters, even if voter support for a given party is only in part due to policies.

For each of the four countries we examine, we find that the policy positions projected by the unified theory - based only on the assumption of vote maximization by parties - correlate well with the observed data; and in two of these countries (Norway and France), our projections also anticipate the actual configurations of the parties' policy positions. However, we find that the policies of major parties and candidates in real-world elections in the other two countries (Britain and the United States) are similar to, but more extreme than, their vote-maximizing positions as computed using our unified models of party competition.

In order to complete our theory, in Chapters 11 and 12 we move beyond the effects of voter-centered influences on party strategy to incorporate a crucial missing link: political elites' policy-seeking motivations. We show that by incorporating reasonable assumptions about elites' policy preferences and the uncertainty associated with national election outcomes, we can explain the distinctly noncentrist policies that British and American parties and candidates have proposed in recent elections.

Because it sheds light on the puzzle of mass-elite linkage, our work also has normative implications. The responsible party model specifies that parties' policies should reflect the views of the voting constituencies that support them (see Sartori 1968; Dalton 1985). In the theoretical chapters, we present results showing that parties have electoral motivations to reflect their supporters' beliefs - expectations that are confirmed by our empirical analyses.

Elections plausibly provide a vehicle through which citizens influence the policy behavior of governments only if (1) the parties present a diverse array of policy positions that reflect the spectrum of public opinion and (2) citizens base their voting decisions in significant part on a comparison of these policy platforms. Neither the spatial modeling nor the behavioral approach alone, however, supports the expectation of strong mass-elite policy linkages. If voters are purely policy oriented, as pure spatial modeling assumes, then we should expect that the policies of every party will be similar, so that voters are not afforded the option of choosing between truly distinct policy visions. If voters deemphasize policies, as behavioral researchers suggest, then it would appear that there is little reason to expect election outcomes to reflect public opinion, even that of the supporters of a winning party.

The unified theory of party competition also portrays the impact of turnout decisions on party competition in a fashion quite different from what is commonly taken to be the main Downsian insight about turnout - namely, that few will be interested enough in political outcomes to be willing to bear the cost of voting. If this view were right, or if, for whatever reason, a substantial number of citizens failed to vote, it is hard to see why elections would necessarily reflect citizens' policy preferences. It would appear, on the face of it, that a citizen who does not cast a ballot would have no influence on the choice of parties and hence on the policies that they might advocate while seeking election or implement if elected. To the contrary, by emphasizing the role of alienated voters, we will show how the implicit threat of a refusal to vote can lead parties to be more faithful to their own supporters.

1.2. Data and Methodology

The primary data sources that combine information on voters' policy preferences and data relevant to their non-policy-related motivations are national election studies. Such studies are available from only a relatively small number of countries. As noted earlier, we will employ our unified theory of party competition to analyze party/candidate strategies and mass-elite policy linkages in historical elections in Britain, France, Norway, and the United States. We employ survey data drawn from national election studies in these countries to estimate the distributions of voters' policy preferences and the parameters of our unified models of party competition in each election. These empirical estimates are used to compute the electoral effects of possible changes in the parties'/candidates' policy positions. Such computations form the basis for analyzing the strategic policy decisions of parties and candidates in these elections.

The four countries we have chosen to study permit us to apply and test the unified theory of party competition in diverse settings. These polities differ along a variety of important dimensions, including their institutional features, their electoral laws, the number of major parties, and the party composition of their governments. Britain and Norway are parliamentary democracies, while the United States features a presidential system and France a mixed presidential-parliamentary system. In comparing the two parliamentary democracies, we find that the British "two-and-a-half " party system, in conjunction with the plurality-voting rule, typically awards a single party a parliamentary majority.⁴ By contrast, the seven significant parties of the Norwegian system, in which seats are awarded via proportional representation, typically produce a splintered seat allocation that creates incentives for coalition governments. In comparing presidential elections in France and the United States, we find that U.S. presidential contests typically feature only two major candidates (the Democratic and Republican Party nominees) but also feature a large proportion of nonvoting citizens. The outcome depends on a single stage, with the winner determined by electoral votes, which in turn depend on plurality voting by citizens in each state. By contrast, French presidential elections typically feature several competitive candidates, with outcomes determined over two ballots, the winner being selected by direct popular vote.⁵ By applying our methodology in these diverse settings, we explore whether the unified theory of party competition can provide a general explanation for party behavior and mass-elite policy linkages that applies both to two-party and to multiparty systems, to parliamentary and to presidential democracies, and to single-party and to coalition governments.

Our analyses have both empirical and theoretical components. We present results on the optimal policy strategies that parties and candidates should pursue under each of several versions of the unified model for the actual voting model parameters (and the distributions of respondents' policy preferences) estimated from election survey data. Specifically, in each case we seek a Nash equilibrium - a set of positions that vote-seeking or policy-seeking candidates/parties would present in order to win elections and that would be stable in the sense that no party would have an incentive to deviate unilaterally from its position. These analyses are developed in Chapters 2-10 using only the assumption of vote-maximizing parties, but later, in Chapters 11 and 12, we incorporate alternative assumptions about policy-seeking motivations. The optimal strategies obtained are compared to observed strategies. Such comparisons permit us to ask: In these real-world elections, were the candidates/parties simply vote maximizers, or did their behavior reflect both vote-maximizing and policy-seeking motivations?

1.3. Justifying Our Theoretical Focus: Why Assume a Unified Model of Party Competition with Vote-Maximizing Parties?

Although our theoretical focus entails integrating the behavioral and spatial-modeling perspectives, we retain the Downsian assumption that politicians seek to maximize votes in the general election. Of course, we recognize that there are many alternative explanations for the phenomenon of party divergence, which invoke factors falling outside the scope of our model (we will discuss these factors later). This raises the questions: Why do we emphasize the importance of integrating behavioral models of the vote into spatial-modeling theory? Why do we omit numerous alternative factors that plausibly illuminate party behavior? A clearly relevant concern is the need to develop a parsimonious theory, and by focusing our attention on the implications of behavioral voting research for party strategies, and by focusing in particular on our three central variables - party identification, voter discounting of the parties' positions, and voter abstention - this is exactly what we have done.

Our modeling is intended to be applicable over a wide range of cases without the need to build in special institutional features of particular political systems and, perhaps most importantly, without the need, at least initially, to assume any motivation on the part of candidates or parties other than winning elections. In addition, we believe that in exploring the implications of our three key variables, we highlight factors that are quite important for understanding party strategies and that are underemphasized in the existing spatial-modeling literature. Simply put, we believe that our unified approach highlights strategic considerations that are extremely important to politicians, considerations that in and of themselves go a long way toward illuminating parties' and candidates' policy strategies in real-world elections.

The preceding observations do not imply that we discount spatial-modeling work that explores the many alternative considerations that plausibly influence politicians' policy strategies. Indeed, we believe that these models - many of which have been developed in the past five years - identify numerous factors that illuminate party competition and, in particular, the strategic factors that may motivate party divergence. These factors include:

•	The possibility of strategic entry by new parties. Theoretical work by Palfrey (1984) and Callander (2000) shows that existing parties in two-party systems may be motivated to present divergent policies in order to deter entry by new competitors who might siphon off votes from the existing parties. The threat of new parties - particularly small ones - is related to the threat of abstention, which we study extensively in Chapters 7 through 9. In either case, the potential loss of support by alienated citizens affects party positioning.
•	The role of nomination processes - for example, the potentially polarizing effects of primary elections in the United States (see, e.g., Coleman 1971; Aranson and Ordeshook 1972; Fiorina 1973, 1974; Polsby 1980; Owen and Grofman 1995; Gerber and Morton 1998; Burden 2000a; Grofman and Brunell 2001).
•	The influence of party activists and campaign donors. Theoretical and empirical work by Aldrich (1983a,b; 1995) and Baron (1994) suggests that, even in the absence of party primaries, policy-motivated party activists (and donors) often pressure party elites to present divergent, noncentrist policy positions. McGann (2002) develops theoretical results suggesting that similar processes operate in multiparty systems. Important recent papers by Moon (in press) and Schofield (2003; see also Miller and Schofield 2003) on U.S. and British elections present theoretical and empirical arguments that candidates gain scarce campaign resources (notably, campaign funds and increased efforts from party activists) by presenting noncentrist policies, and that these resources can be used to enhance the candidates' images along "valence-related" dimensions of evaluation (competence, integrity, etc.). These papers suggest that when a mixture of valence-related and policy-related considerations moves voters, especially those most concerned with policy outputs, candidates maximize votes by diverging from the center of the policy space.
•	Information-centered models of elections. Recent papers by Snyder and Chang (2002) and Callander and Wilkie (2002) argue that in certain circumstances, candidates and parties can decrease voter uncertainty about their policy positions - thus rendering their platforms more attractive to risk-averse voters - by presenting noncentrist policies. There is a related theoretical literature that explores the implications of restrictions on candidate positioning in terms of movement from previously announced positions (Kramer 1977; Wuffle, Feld, and Owen 1989; Kollman, Miller, and Page 1992; see also Downs 1957: 107-13).
•	Studies on the geographic distribution of party loyalties and voter ideologies across constituencies (see, e.g., Austen-Smith 1986; Grofman et al. 2000; Grofman in press). For plurality-based competition, when there are multiple constituencies and these constituencies differ in the location of their median voter, a risk-averse party may opt for positions that make it nearly certain to win some given set of constituencies (say, those where the median voter in the constituency is considerably to the left of the overall median voter), rather than taking moderate positions that may render the party more widely competitive but that do not give it a decisive advantage in any given set of constituencies. The latter option bears the risk that, given unfavorable national trends favoring the other party (caused by exogenous factors, such as differences in the popularity of the parties' candidates for national office, or times of national crisis that might favor the incumbent), a party might be largely shut out of the legislature.6
•	Studies on the role of electoral laws. There is a voluminous literature on contrasts between list proportional representation (PR) and plurality or runoff methods in matters such as the incentives for party proliferation and proportionality of seats-votes relationships (see, e.g, Taagepera and Shugart 1989; Lijphart 1994). Here the most important finding is that electoral rules that favor multiple parties, such as list methods of proportional representation, also foster a wider ideological range among the parties. Only a limited number of studies have looked directly at the relationship between electoral rules and ideological differences among parties (see, e.g., Dow 2001). In this context, Gary Cox has made many of the most important contributions (see, especially, Cox 1990, 1997). Some of the most interesting recent work has been done by Anthony McGann. For example, McGann has shown that when ideological distributions are nonsymmetric, sequential elimination runoffs that require a majority vote for a candidate to win tend to elect candidates who are more extreme than the median voter - candidates located between the median and the mode of the ideological distribution (McGann, Koetzle, and Grofman 2002; McGann, Grofman, and Koetzle 2003; see also McGann 2002).
•	Politicians' policy motivations. There is extensive theoretical work (Wittman 1977, 1983; Cox 1984; Calvert 1985; Chappel and Keech 1986; Londregan and Romer 1993; Roemer 2001; Groseclose 2001; Schofield 1996) suggesting that policy-motivated politicians - that is, politicians who are seeking office as a means of enacting their proposed policies rather than proposing policies as a means of attaining office - may propose divergent, noncentrist policies. A key theoretical result derived from this approach is that policy divergence emerges only when the competing politicians are uncertain about the election outcome, a condition that is quite plausible in real-world elections. We investigate policy-seeking motivations in depth in Chapters 11 and 12 .

We believe that the studies just cited identify important determinants of candidate/party strategies. However, in this book our primary goal is to determine how much progress we can make toward understanding politicians' policy strategies without recourse to any of these explanations except the last one. We hope to convince the reader that the progress we can thereby achieve, relying entirely on the unified theory with vote-maximizing parties, is considerable. But as noted earlier, in order to complete our theory we will incorporate political elites' policy-seeking motivations - an augmentation of the model that is particularly pertinent in our explanation of the distinctly noncentrist policies offered by British and American parties and candidates.

1.4. Plan of the Book

Chapters 2 though 4 introduce the general theoretical ideas that are the foundation of this volume.

Chapter 2 provides the statistical framework for the unified, conditional logit model for voter choice that is used throughout the book - a (multivariate) random-utility model that incorporates policy dimensions, measured nonpolicy factors such as partisanship and voter and candidate characteristics, and a random component representing unmeasured factors that render the vote choice probabilistic. The development begins with the traditional proximity model of voter choice, after which discounting of party positioning is introduced. Appendix 2.1 discusses alternatives to the conditional logit model - in particular, the generalized extreme value (GEV) model and the multinomial probit (MNP) model. Appendix 2.2 discusses the empirical status of party identification as an explanatory variable in empirical voting research, while Appendix 2.3 discusses the status of policy-discounting theory as an alternative to the standard proximity model and the link between the discounting model and directional voting.

In Chapter 3, we present four simple, nontechnical examples designed to convey the logic of party competition under successively more complex versions of the unified theory of party competition. When only policy is included in a deterministic model, parties are drawn toward the center. When a partisan component is added, however, policy divergence may occur as long as there are three or more parties. Specifically, we present reasons why parties have electoral incentives to present policies that reflect the beliefs of voters who are biased toward them in part for nonpolicy reasons. Incorporating the response of candidates to voter discounting of their claimed platforms substantially increases divergence of party strategies. Finally, adding a probabilistic component reflecting unmeasured components of voter utility to a model that also incorporates measured nonpolicy factors leads to a more completely specified model and, as in the case of a comparable deterministic model, to divergence of party strategies. Not only is voter choice more realistic under a probabilistic model, but also - unlike the results of a deterministic approach, in which there are dramatic differences in the one-dimensional and multidimensional cases - the results under a probabilistic model are similar for one and more than one dimension.

---

© Cambridge University Press

Table of Contents

1. Modeling party competition; 2. How voters decide: the components of the unified theory of voting; 3. Linking voter choice to party strategies: illustrating the role of non-policy factors; 4. Factors influencing the link between party strategy and the variables; 5. Policy competition under the unified theory: empirical applications to the 1988 French Presidential Election; 6. Policy competition under the unified voting model: empirical applications to the 1989 Norwegian parliamentary election; 7. The threat of abstention: candidate strategies and policy representation in US presidential elections; 8. Candidate strategies with voter abstention in US presidential elections: 1980, 1984, 1988, 1996, and 2000; 9. Policy competition in Britain: the 1997 general election; 10. The consequences of voter projection: assimilation and contrast effects; 11. Policy-seeking motivations of parties in two-party elections: theory; 12. Policy-seeking motivations of parties in two-party elections: empirical analysis; 13. Concluding remarks.