Party Policy Strategies and Valence Issues: An Empirical Study of Ten Post-Communist European Party Systems

Transcription

1 Party Policy Strategies and Valence Issues: An Empirical Study of Ten Post-Communist European Party Systems Zeynep Somer-Topcu Department of Political Science University of California at Davis Davis, CA Prepared for the Conference on the Dynamics of Party Position Taking, Binghamton, NY, March 23-24, All errors are the sole responsibility of the author.

2 Abstract: An analysis of parties ideological shifts in ten Eastern and Central European post- Communist countries reveals empirical patterns that support expectations generated by the Adams-Merrill model of policy-seeking parties in parliamentary democracies with proportional representation. Based upon analyses showing that governing parties systematically lose votes for reasons that appear unrelated to their policy positions, we surmise that being in government depresses parties valence images, and we evaluate whether governing parties tend to moderate their Left-Right positions, as predicted by the Adams-Merrill centripetal effects of valence result. We find strong support for this centripetal governing party hypothesis. We also find that opposition parties whose valence images plausibly improve relative to those of governing parties systematically shift towards more radical ideologies, a centrifugal opposition party effect that again supports the Adams-Merrill model. Finally, we find that these effects are mediated by the level of unemployment: specifically, as unemployment increases so do the centripetal/centrifugal effects of being in government or opposition. Given that governing parties electoral support and therefore, plausibly, their valence images responds to the level of unemployment, this mediating economic effects is again consistent with expectations generated by the Adams-Merrill model. 1

3 Introduction: The spatial modeling literature first incorporated valence in Stokes s (1963) famous article, where he defined valence issues as those on which parties or leaders are differentiated not by what they advocate but by the degree to which they are linked in the public s mind with conditions or goals or symbols of which almost everyone approves or disapproves (Stokes, 1992, 143). Since Stokes, the literature on valence has addressed two, related, questions: How do valence issues affect parties and candidates electoral support?, and, How does valence affect parties /candidates strategic policy positioning? The economic voting literature constitutes an important part of the first strand of valence research (see Lewis-Beck and Stagmaier, 2000 for a review of the economic voting literature). Personal images of party leaders (King et al. 2002) and media coverage of political scandals and intra-party divisions (Clark, 2006) have also been used to explain the effect of valence on party/candidate support. Studies on how valence affects the strategic policy positioning of parties/candidates are mostly theoretical in approach (MacDonald and Rabinowitz, 1998; Ansolabehere and Snyder, 2000; Groseclose, 2001; Schofield, 2003; Adams, Merrill and Grofman, 2005; but see Schofield and Sened 2005, 2006; Grose 2005). These formal models basically state that parties/candidates have strategic incentives to position themselves differently depending on their valance advantages/disadvantages vis-à-vis their electoral opponents. The expectations of spatial modelers have differed, however, about how valence-related considerations will affect parties policy strategies. This paper is an attempt to empirically evaluate hypotheses suggested by Adams and Merrill, who present theoretical results that valence-disadvantaged, policy-seeking, parties in parliamentary democracies with Proportional Representation will moderate their policies the Adams-Merrill centripetal valence effects 2

4 (CVE) result. The CVE results also states that when a policy-seeking party s valence improves it will shift away from the center, thereby trading off some of its enhanced election prospects (i.e. its probability of becoming the median parliamentary party) in pursuit of policy. On the other hand, when a policy-seeking party loses valence, it has incentives to make additional policy compromises and therefore to shift toward the center of the policy space, in order to compensate for its diminished electoral prospects. In order to empirically evaluate the Adams-Merrill CVE hypothesis, I make use of two widely-recognized empirical regularities that plausibly reflect on parties valence images: first, that compared to opposition parties, governing parties vote shares tend to decrease over time (Paldam, 1991; Whitten and Palmer, 1999; McDonald and Budge, 2006; Clark, 2006); second, that this government vote loss effect is mediated by economic conditions, so that governing parties losses tend to increase as the state of the economy declines (Kramer, 1971, Lewis-Beck and Bellucci, 1982; Lewis-Beck, 1986; Paldam, 1991; Remmer, 1991; Powell and Whitten, 1993; Przeworski, 1996; Bell, 1997; Whitten and Palmer, 1999). These patterns suggest that governing parties valence images tend to deteriorate relative to the images of opposition parties, and furthermore, that the magnitudes of these valence effects depend on economic conditions. This implies that the Adams-Merrill CVE result has the following, empirical implications: that governing parties will moderate their ideologies, a prediction I label the centripetal governing party hypothesis; that opposition parties will shift to more radical ideologies, which I label the centrifugal opposition party hypothesis; that these effects are mediated by the state of the economy, an effect I label the mediating economic effects hypothesis. I evaluate the above hypotheses via empirical analyses of party policy shifts in ten post- Communist party systems. These Eastern and Central European post-communist countries offer 3

5 a promising context in which to test these hypotheses, for several reasons. First, these democratic party systems are quite new, so that there exist few, if any, empirical analyses of parties policy positioning strategies in these countries. Second, empirical analyses that I report below suggest that in elections in these post-communist systems my two central independent variables governing party status and the state of the economy strongly influence parties vote shares, an effect which suggests that these variables are closely tied to parties valence images. This implies that these post-communist systems constitute an appropriate setting in which to evaluate my hypotheses. Empirically, I find support for all of the hypotheses. I find a statistically significant tendency for parties in government to subsequently shift to more moderate positions, thereby supporting the centripetal governing party hypothesis, and also that opposition parties systematically shift away from the center, a finding that supports the centripetal opposition party hypothesis. Finally, I conclude that these centripetal/centrifugal effects are mediated by the level of unemployment, a finding that supports the mediating economic effects hypothesis. These results have interesting implications for party strategies and for the connections between spatial modeling and empirical research. The confirming results for the centrifugal and centripetal hypotheses, and also the mediating effects of unemployment, imply that parties change their policy positions when their valence images shift. These are supportive findings for the spatial model proposed by Adams and Merrill. As Aldrich and Alt (2003) state, it is important to develop empirical implications of theoretical models. In the political science field, we have had either strong theories with few empirical tests, or empirical tests with weaker theories. This research provides a nice empirical analysis of a spatial model, which fills a gap in the comparative party positioning literature. 4

6 In section 2, I introduce the empirical literature on valence effects, and the hypotheses associated with the Adams-Merrill Centrifugal Valence Effects (CVE) result. Section 3 answers the question of why I chose post-communist Europe to empirically examine the hypotheses. Next, in Section 4, I present the research design and discuss the variables and the statistical models that I use to evaluate my valence-related hypotheses. Section 5 reports empirical analyses of parties policy strategies in the Eastern and Central European post-communist parliamentary democracies, which bear on the hypotheses. Section 6 concludes. Valence Effects: Literature and Hypotheses Several recent spatial modeling studies have incorporated valence into empirical analyses of American Congressional and Senate elections (Londregan and Romer, 1993; Grose, 2005, 2006) as well as multiparty elections held outside the United States (Schofield, 2005, 2006, 2007; Schofield and Sened, 2005, 2006; Schofield and Cataife, 2006; Adams, Merrill, and Grofman, 2005). Nevertheless, we are unaware of any cross-national analysis of the effects of valence on parties strategic policy positioning that evaluates hypotheses on comparative statics - - that is, how parties adjust their positions in response to fluctuations in their own (and their opponents ) valence images. That is what I attempt here. Empirically evaluating the effects of valence on party positioning is difficult, because we have no reliable measures of parties valence images over long time periods and/or across countries. Schofield and his coauthors have examined a small number of countries using a very sophisticated measure of valence, which is based on complex analyses of election survey data. Because these analyses are demanding, however, they cover only a handful of elections. Clark (2006), meanwhile, constructs measures of parties valence images based on media reports 5

7 compiled in Keesing s record of World Events. However, the reliability of this method of measuring valence is also questionable. With no universally accepted comparative valence measure available, it is difficult to empirically test the Adams-Merrill CVE hypothesis. Nevertheless, we may be able to construct proxy measures for valence, based on the distinction between governing parties and opposition parties. Several previous studies conclude that governing parties tend to systematically lose votes, compared to opposition parties (Paldam, 1971; Rose and Mackie, 1983; McDonald and Budge, 2006; Clark, 2006; Powell and Whitten, 1993; Whitten and Palmer, 1999). 1 This loss plausibly arises because being in government tends to hurt parties valence images, since governing parties have disproportionate tendencies to become embroiled in scandals, and they also tend to be blamed by voters for failing to solve problems such as crime, unemployment, corruption, etc. As Clark (2006) states, governing parties tend to face closer scrutiny for their actions, given their role in managing the affairs of the country. Clark s empirical analyses of media reports on political scandals and intra-party divisions in eight Western European democracies confirm that these reports focus disproportionately on governing parties. 2 Clark also finds that negative media reports are associated with vote losses in the subsequent election a connection that supports the hypothesis that negative media coverage depresses the targeted party s valence image. 3 1 Paldam (1971) finds that government parties on average lose about 1.6 percent of their vote shares in the subsequent election, while Whitten and Palmer (1999) estimate that this percentage is above 3 percent. 2 We note that Clark finds no similar tendency for positive media reports to focus on governing parties. 3 An additional plausible explanation for governing parties vote losses in the American politics context is the surge and decline thesis (Campbell, 1991). According to this argument, parties are likely to be in the government because they were unusually successful in the previous election, due to the effects of short-term valence forces that are unlikely to operate in the current election. This effect, which may also operate outside the U.S., again implies that governing parties valence scores will tend to decline over time, as governing parties images regress towards the party s long-term valence equilibrium level. 6

8 The above summary suggests that there are good reasons to expect that governing parties tend to see their valence images deteriorate, compared to opposition parties. Therefore a party s governing status may serve as a useful proxy for its valence image, and hence analyzing the policy trajectories of governing parties versus the trajectories of opposition parties seems like a promising way to evaluate the Adams-Merrill predictions. Although being in government or opposition is surely not a perfect measure of valence, it provides an avenue for empirically evaluating the centripetal and centrifugal effects of valence, cross-nationally and over time. These considerations motivate our first two hypotheses: H1a: The Centripetal Governing Party Hypothesis: Governing parties tend to moderate their Left-Right positions over time.. H1b: The Centrifugal Opposition Party Hypothesis: Opposition parties tend to shift to more radical Left-Right positions over time. The hypotheses listed above provide a promising jumping-off point for evaluating the Adams-Merrill CVE hypothesis. However, a problem that arises is that even if hypotheses H1a and H1b are supported empirically, there are plausible alternative explanations for these patterns that have nothing to do with parties valence-related considerations. For instance, governing parties may differ from opposition parties in terms of their underlying motivations (i.e. officeseeking versus policy-seeking), and furthermore governing parties recognize that their policies have real world consequences, unlike opposition parties who can engage in cheap talk. Therefore we must be cautious about concluding that empirical support for hypotheses H1a and H1b provides convincing evidence in favor of the Adams-Merrill model. 7

9 The preceding considerations raise the question: Are there additional variables that plausibly mediate valence effects relating to governing and opposition parties that we can use to construct more demanding tests of Adams and Merrill s model? One such potential mediating variable is the state of the economy. There is extensive previous research suggesting that governing parties support is tied to the strength of the economy. 4 The main thesis of this economic voting literature is that voters punish governing parties when the economy is weak, and reward governing parties when the economy is strong. 5 Furthermore, survey-based research suggests that the causal process underlying economic voting is that citizens use the state of the economy to make inferences about the competence of the government i.e. that the state of the economy is a valence issue that reflects on the governing parties. (Below I show that these economic effects operate in the Eastern and Central European post-communist elections that I analyze in this paper.) This literature thereby suggests that governing and opposition parties valence images are mediated by the strength of the economy, an effect that suggests two additional hypotheses on the relationship between unemployment and parties policy strategies 6 : 4 See Lewis-Beck and Stagmaier, 2000 for a nice review of the economic voting literature. 5 Kramer (1971), for instance, concludes that election outcomes are responsive to objective changes in national economic conditions occurring under the incumbent party in the US, while Lewis-Beck and Bellucci (1982) find support for the economic voting thesis in France and Italy. Then, in his extended analysis that covers Britain, France, Germany, and Italy, Lewis-Beck (1986) finds a clear and consistent pattern of economic voting in these countries. On the other hand, Paldam s (1991) 17-nation study does not identify consistent, cross-national effects of economic conditions on governing party support. More recently, the literature has focused on how electoral effects related to the national economy differ between systems in which responsibility can be clearly attributed to the government and those in which clarity of responsibility is low. Powell and Whitten s (1993) seminal work on clarity of responsibility shows that economic voting works especially for clear systems. Economic voting in the developing world has not been examined as extensively as it has in the developed world. However, Remmer (1991) examines the political impact of the economic crisis in Latin America in the 1980s and finds that economic voting exists when exchange rates, GDP, and inflation rates are examined. In post-communist Europe, Bell (1997) examines the 1993 election in Poland and identifies strong economic voting effects, especially for unemployment. Tavits (2005) examines electoral volatility in post-communist Europe and also finds economic voting effects. 6 The focus will be on unemployment rather than inflation or GDP growth due to circumstances that are peculiar to the post-communist context. As Pacek (1994) and Przeworski (1996) state, post-communist countries have faced an enormous increase in the unemployment rates with the collapse of communism. While unemployment was nonexistent under communism, it immediately jumped to above 10 percent in almost all post-communist countries. 8

10 H2a: The unemployment effects hypothesis governing parties: The higher the level of unemployment, the stronger governing parties tendencies to moderate their Left-Right positions. H2b: The unemployment effects hypothesis opposition parties: The higher the level of unemployment, the stronger opposition parties tendencies to shift towards more radical policies. Why Eastern and Central Europe? I plan to evaluate the hypotheses presented in Section 2 in the context of Eastern and Central European party systems. There are two reasons for this substantive focus. First, the cases of Eastern and Central Europe are intrinsically interesting, since there are as yet few empirical analyses of the factors that influence parties policy strategies in these regions. Second, my empirical analyses, as I will show below, suggest that my key independent variables namely being in government and unemployment are closely tied to parties valence images in Eastern Europe. Until very recently, data limitations in post-communist countries limited scholars abilities to study many interesting questions that they have examined in the context of advanced industrial societies. However as the time since the first competitive independent election in post- Communist systems has lengthened we have begun to acquire sufficient data to systematically analyze election outcomes and voting behavior in these countries. At the same time these party systems have stabilized, and the major parties have (for the most part) developed clearly identifiable ideological positions. With the publication of the new Comparative Manifesto High unemployment rates together with minimal security provisions make employment-related economic concerns paramount in these countries. 9

11 Project (CMP) data in the beginning of 2007, we now have comparable policy positions of the post-communist parties based on their party manifestos. This allows us to examine many interesting questions that we have not been able to ask in this context before. Moreover, as I document below, the two independent variables that are central to my hypotheses governing status and the unemployment rate are closely tied to Central and Eastern European parties vote shares, which suggests that these variables plausibly serve as useful proxies for parties valence images in these countries. With respect to this point, we note that there are also good theoretical reasons to expect governing status and the state of the economy to strongly influence parties valence images in post-communist systems. First of all, these are not advanced industrial societies and governments have much to do to satisfy all citizens in such an economically-torn environment. Corruption tends to be high, the welfare state cannot function efficiently, and therefore we might expect voters to hold governing parties accountable for these problems and to punish them accordingly. Second, we might expect economic voting to be particularly salient in post-communist societies, in part because, as Tucker (2005) has argued, the incidence of party identification is low in these newly-competitive systems and therefore the electoral salience of retrospective economic voting should be correspondingly higher. In support of this hypothesis, Blondel and Muller-Rommel (2001) document that in Lithuania Slezevicius single-party government, although a majority government, was forced to resign as a result of economic troubles. In Czechoslovakia, the fiscal crisis in 1997 led to protests and the resignation of the government. In 1998 the Romanian prime minister Ciorbea resigned as a result of economic problems. He could not achieve what he promised: economic development. In Poland, Hungary, and Bulgaria, we see ex-communists returning to power with the second competitive elections. Although anti-communists were 10

12 elected in the first competitive elections in each of these countries, they could not maintain power in subsequent elections. The main reason is stated as economic failure. Research Design: To evaluate my hypotheses on the centripetal/centrifugal effects of valence (governing vs. opposition status), along with the mediating effects of the economy, I analyzed data on parties policy shifts in ten post-communist parliamentary democracies for the period since their first competitive elections until the end of These are Bulgaria, Czech Republic, Estonia, Hungary, Lithuania, Latvia, Poland, Romania, Slovakia, and Slovenia. All of these countries use proportional systems to elect their representatives, although Lithuania and Hungary have mixed systems with some seats elected by plurality rule. 7 Our data on party positions is based on the Comparative Manifesto Project (CMP) leftright party placements, which have been published for all major parties at each election up to 2003 ( Klingemann et al, 2006). These data provide longitudinal, cross-national measurements of party policy positions based on their published manifestos. More specifically, these data include information for each party on the proportion of manifestos dedicated to several issues. The coders match up quasi-sentences in the policy programme with a category of policy (e.g. the environment, international relations, welfare, etc.) and take the percentages of each category as a measure of the party s priorities. CMP-based party placements generally correlate with other widely used measures of party positioning such as expert surveys, party placements by survey respondents, and other word-scoring techniques. 7 In Hungary, in addition to 152 members elected by PR system and 176 members elected in single-seat constituencies, 58 parliamentarians are elected to so-called compensation seats, which are distributed based on the number of votes lost in either the single-seat or the multi-seat constituencies. In Lithuania, 70 of the members of the legislative body are elected ina a nationwide vote by PR system and the other 71 are elected in single-member districts. 11

13 The authors of the CMP dataset have also developed an index that measures the overall left-right ideology for each party s manifesto in each election year, which ranges from -100 to +100 and for which high numbers represent a more right-wing emphasis. This Left-Right index has been employed by scholars to examine the changes of party ideologies and the reasons behind these changes (Adams et al, 2004, 2006; Adams and Somer-Topcu, 2006; McDonald and Budge, 2005; McDonald et al, 2004; Meguid, 2005, Nagel, 2001). To test the mediating effect of the economy I rely on yearly unemployment data for the post-communist European countries, from the World Development Indicators dataset. Evaluating the governing party and economic variables as valence proxies: Model specification We specify multivariate regression models in order to evaluate the extent to which the two crucial independent variables, namely being in government and unemployment, influence party vote shares effects that are critical if these variables are to be considered reliable proxies for parties valence images. In these analyses the dependent variable is the change in the party s vote share in the current election compared to the previous election. With respect to governing party status as a proxy for valence, the crucial independent variable is a dummy variable denoting whether or not the focal party was in government at any time since the previous election. 8 With respect to the effect of unemployment as a mediating variable, the crucial independent variable is the unemployment rate in the country at the time of the current election, interacted with the dummy variable for governing party status: 8 A different coding rule, which coded the parties that were in government within the last year of the current election period, supported substantive conclusions that are identical to those I report below. 12

14 Party s vote change (t) = b 1 + b 2 [governing party] + b 3 [party s vote change (t 1)] (1) Party s vote change (t) = b 1 + b 2 [unemployment] + b 3 [governing party] + b 4 [governing party*unemployment] + b 5 [party s vote change (t 1)] (2) where Party s vote change (t) = the change in the party s vote share in the current election (election t), compared with its vote share in the previous election (election t-1). Similarly, Party s vote change (t-1) represents the changes in the party s vote share in the earlier election. Governing party = 1 if the party was ever in government between election t-1 and t, 0 otherwise. Unemployment = the percentage of unemployment at the time of the current election. 9 In equation 1 the independent variable [governing party] is constructed so that, if the estimated coefficient b 2 on this variable is negative, this indicates that governing parties tend to lose votes in the current election, compared to opposition parties. This will support the hypothesis that the governing party variable is a useful proxy for valence. In equation 2 the independent variable [governing party*unemployment] is constructed so that, if the estimated coefficient on this variable (b 4 in equation 2) is negative, then governing parties vote losses increase with the unemployment rate. This will support the hypothesis that governing parties valence images are mediated by the level of unemployment. The variable party s vote change (t-1) is included because, as the surge and decline thesis also states, governments usually come to power as a result of short-term shift in voter support. When the factors that generated this swing moderate in the subsequent elections, voters regress to more normal patterns of behavior and parties that gained votes in the previous election tend 9 If the election was held in the second half of the calendar year, that year s unemployment figure was used; if the election was held in the first half of the year I used the previous year s unemployment figure. 13

15 to suffer electoral reverses in the current election, and vice versa (Paldam, 1986, Strom and Lipset, 1984, Powell and Whitten, 1993). To control for this effect, a lagged measure of previous vote gain or loss is included. 10 Model specification for the hypotheses: To evaluate the hypotheses on the centripetal and centrifugal tendencies for governing and opposition parties (hypotheses H1a-H1b), along with the hypotheses on how these effects are mediated by the unemployment rate (Hypotheses H2a-H2b), I specify two multivariate regression models. The dependent variable now becomes the change in the party s Left-Right position in the current election, compared to the previous election. Being in government is still the crucial independent variable. With respect to hypotheses H1a-H1b, the expectation is that governing parties will moderate their Left-Right positions over time (the Centripetal Governing Party Hypothesis), i.e. that leftist governing parties will shift to the right and rightist governing parties will shift to the left; and, that opposition parties will shift towards more radical positions (the Centrifugal Opposition Party Hypothesis), i.e. that leftist opposition parties will shift farther left and rightist opposition parties will shift farther to the right. With respect to hypotheses H2a- H2b, the expectation is that governing parties centripetal tendencies will increase with the rate of unemployment (the Unemployment Effects Hypothesis Governing Parties) and that opposition parties centrifugal tendencies will also increase with unemployment (the Unemployment Effects Hypothesis Opposition Parties). The following specifications are intended to evaluate these hypotheses: Party s policy shift (t) = b 1 + b 2 [governing party* party ideology] + b 3 [governing party] 10 Analyses by excluding the vote change (t-1) variable returned the same findings, with even stronger effects of the key independent variables on the vote change (t) dependent variable. 14

16 + b 4 [ideology] + b 5 [party s policy shift (t-1)] (3) Party s policy shift (t) = b 1 + b 2 [governing party* party ideology* unemployment] + b 3 [party ideology*unemployment] + b 4 [governing party] + b 5 [unemployment] + b 6 [party ideology] + b 7 [party s policy shift (t-1)] (4) where Party s policy shift (t) = the change in the party s Left-Right policy position in the current election (election t), compared with its position in the previous election (election t-1), as computed using the CMP codings of parties policy programmes. Similarly, Party s policy shift (t-1) represents the changes in the party s Left-Right position in the earlier election. Party Ideology = +1 if the party is left-wing, -1 if the party is right-wing, 0 otherwise The coding decisions on which parties were considered to be left-wing and right-wing were based upon the party classifications reported in Klingemann et al (2006). In the CMP data, each party is coded as belonging to a particular party family. Among these families, the Communist, Green, and Social Democratic party families are coded as left-wing, while parties that are classified as being members of the Conservative, Nationalist, and Christian Democratic party families are coded as right-wing. 11 If the estimated coefficient b 2 on the interaction variable [governing party*party ideology] in equation 3 is positive, this indicates that governing parties that are left-of-center tend to shift to the right, and right-wing governing parties tend to shift to the left, compared to 11 Agrarian, special interest, and Liberal parties are coded 0 and therefore the hypotheses do not make any specific predictions about these party families. 15

17 opposition parties. To the extent that this coefficient estimate is positive and statistically significant, this will support hypothesis H1a (the Centripetal Governing Party Hypothesis). The movement of opposition parties is captured by the party ideology variable, and a negative estimated coefficient for this variable (coefficient b 4 in equation 3) would support the Centrifugal Opposition Parties Hypothesis (H1b). Equation 4 is intended to evaluate hypotheses H2a-H2b, on the mediating effects of unemployment on parties centripetal/centrifugal tendencies. The independent triple interaction variable [governing party* party ideology* unemployment] is constructed so that, if the estimated coefficient b 2 on this variable is positive, then this indicates that unemployment indeed mediates governing parties centripetal tendencies in the manner posited by the Unemployment Effects Hypothesis Governing Parties (H2a). In order to evaluate how unemployment mediates opposition parties centrifugal tendencies, I incorporate the variable [party ideology*unemployment]. If the estimated coefficient b 4 on this variable is negative, then this indicates that unemployment also mediates opposition parties centrifugal tendencies as posited by the Unemployment Effects Hypothesis Opposition Parties (H2b). Empirical Findings: The analyses encompass 60 vote and policy shifts for parties in ten post-communist countries from the date of their first competitive elections until Our data should be regarded as cross-sectional time-series and therefore using simple OLS regression is problematic since there may be unobserved differences between parties (Green et al., 2001). Autocorrelation and heteroscedasticity (differences in the variance of the error term across parties) may also be a concern for these analyses if there are reasons to believe that the set of parties competing in a 16

18 given election are influenced by unobserved, election-specific variables, and if being in government or opposition influences some parties more strongly than others. By using panel corrected standard errors clustered by the election date we address these concerns (Beck and Katz, 1995). Evaluating the electoral effects of being in government and the mediating effects of unemployment. Table 1 reports the parameter estimates for the governing party (columns 1 and 2) and economic voting assumptions (columns 3-6). While the first, third, and fifth columns test the assumptions for all ten post-communist countries, I dropped Lithuania and Hungary from the analyses reported in columns 2, 4, and 6. This is because these two countries feature mixed PRplurality voting systems, while the Adams-Merrill model is intended to apply to fully proportional systems. The first two columns in Table 1 show that being in government is associated with vote losses, although this effect is statistically significant only for the eight-country analysis (column 2). Substantively, the coefficient estimates on the [governing party] variable imply that being in government depresses a party s vote share by about 2.5 percentage points. 12 The economic voting assumption is tested first only on governing parties (columns 3 and 4). It appears from the results that unemployment has a negative effect on vote share changes for the governing parties. When all parties are examined in the last two columns, the negative coefficient for the interaction variable [governing party*unemployment] shows that the economic voting assumption is again supported. 12 If the [vote change (t-1)] independent variable is omitted from the specification (and note that this variable does not achieve statistical significance) and the model is then re-estimated, the coefficient on the [governing party] variable is statistically significant in both the ten-country and the eight-country analyses, and the estimated effect of being in government is to depress vote share by 3.5 to 4.5 percentage points. 17

19 In toto, the analyses summarized above support the hypotheses that governing parties vote shares tend to decline, and that this effect is mediated by the level of unemployment. Substantively, these results suggest that the governing party variable and the unemployment variable are useful proxies for changes in parties valence images. [Table 1 about here] Evaluating the relationships between governing status, unemployment, and parties policy shifts. The results for the centripetal and centrifugal hypotheses as well as the hypotheses concerning the mediating effect of unemployment on party shifts are reported on Table 2. The first two columns test Hypotheses H1a-H1b, first for all ten post-communist countries (column 1) and then for the eight proportional systems (column 2). Consistent with expectations, the coefficient estimate on the interaction variable [governing party *ideology] is positive and statistically significant, indicating that, ceterus paribus, a leftist (rightist) governing party would tend to shift to the right (left) in the upcoming election, compared to the opposition parties. Moreover, the coefficient estimate on the [party ideology] variable is negative and statistically significant, indicating that opposition parties tend to shift away from the center. These estimates support the Centripetal Governing Party Hypothesis (H1a) and the Centrifugal Opposition Parties Hypothesis (H1b). [Table 2 about here] The third and fourth columns test the mediating unemployment effects hypotheses (H2a and H2b). The estimated coefficient on the triple interaction variable [governing party* ideology* unemployment] is positive, indicating that governing parties tendencies to moderate their policy positions increase with the rate of unemployment. It is statistically significant and in 18

20 the expected direction predicted by the Unemployment Effects Hypothesis Governing Parties (H2a). Furthermore, the coefficient estimate on the [ideology* unemployment] variable is negative and statistically significant, indicating that the higher the level of unemployment, the stronger is opposition parties tendency to shift to more radical Left-Right positions. This coefficient estimate supports the Unemployment Effects Hypothesis Opposition Parties (H2b). Sensitivity Analyses: Several additional analyses were conducted to evaluate the cross-national comparability of our data and the robustness of the results. To evaluate the reliability of the CMP s left-right scores across countries, we tested the results by dropping one country at a time from the data. If the criticisms of the CMP data are valid, we should see changes in the substantive effects of the key independent variables when we conduct these analyses. The analyses for both of the hypotheses when dropping one country at a time (10 analyses in all for each hypothesis) reveal the same substantive results for our hypotheses, and convince us that our results are not driven by any factors specific to a single country. Moreover, we re-estimated our models while using the actual party positions rather than changes in parties positions as the dependent variable, and our results continued to support our hypotheses. Finally, columns 5 and 6 on table 2 report parameters for an alternative specification that is estimated to assure ourselves that our conclusions were robust in the presence of other control variables. More specifically, we tested the effect of past election results, following the previous work of Budge (1994). He concludes that if parties gained (lost) votes in the previous elections they would move to the same (opposite) direction in the upcoming elections (see also Nagel 2001). To control for this effect, we included two variables into our two main models: the difference between the party s vote share at the previous election (t-1) and its vote share at the 19

21 election before that (t-2), which is labeled as vote change (t-1) as before; and the interaction variable [vote change (t-1)* party policy shift (t-1)]. When these variables are included into the analyses, we could not find support for past election results in the post-communist countries. However, our results regarding the main hypotheses of the paper still hold even when these variables are included into the analyses. Conclusion: The spatial model of policy-seeking parties contesting parliamentary elections under Proportional Representation developed by Adams and Merrill is tested here, via empirical analyses of party policy shifts in ten post-communist countries. Specifically, I present arguments that the Adams-Merrill model implies the Centripetal Governing Party hypothesis: namely, that parties whose valence images deteriorate have policy-seeking motivations to moderate their policies, compared with valence-advantaged parties. Those valence-advantaged parties, on the other hand, have incentives to move farther away from the center, which supports the Centrifugal Opposition Party hypothesis. In addition to these hypotheses, the mediating effect of unemployment is also tested, namely that the centripetal and centrifugal effects of valence are magnified by worsening economic conditions. We do not claim that being in government or opposition perfectly captures parties valence images. But absent a reliable, cross-national measure of valence, we use governing party status as a proxy in order to assess the effects of valence, using the assumption that governing parties lose votes in the upcoming elections. Moreover, to be able to argue that unemployment works as a mediating force between valence and policy shifts, we also assumed that economic voting exists. 20

22 Post-Communist countries provide a fertile ground to test our hypotheses. Eastern and Central Europe are areas for which comparative analyses of party policy strategies was unfeasible until very recently. With the publication of the CMP data for this region, it has become possible to test many interesting questions. Moreover, the two assumptions that we made to test our main Centrifugal and Centripetal hypotheses work very well in this region: As I showed empirically, governing parties in Eastern and Central Europe are punished in upcoming elections, and this effect is significantly mediated by the level of unemployment. The empirical analyses revealed that the predictions of Adams and Merrill hold in post- Communist Europe, if we measure valence disadvantage with governing party status. Governing parties move to center in upcoming elections, as they lose valence, while opposition parties shift away from the center as they gain valence relative to governing parties. Furthermore, I find that this effect is mediated by the level of unemployment, a result that again supports the Adams- Merrill model, given that governing parties valence images plausibly fluctuate in response to this economic indicator. Sensitivity analyses reveal that these estimated effects are robust over different specifications. These analyses can be extended to examine other regions, especially the Western European countries. As Paldam (1991) and Palmer and Whitten (1999) show, governing parties are consistently punished at the ballot box in Western Europe. And, although the evidence is mixed, the literature starting with Powell and Whitten (1993) shows that if the clarity of governing party responsibility is high, economic voting is also a significant factor in Western European elections. With these conditions satisfied, an analysis of the Western European countries is also possible. 21

23 Appendix I: Descriptive Statistics Table: Descriptive Statistics of the Variables Used: Min Max Mean St Dev Vote change (t) Governing Party Unemployment Left-Right ch (t) Ideology

24 References: Adams, James Policy Divergence in Multicandidate Probabilistic Spatial Voting, Public Choice 100, Adams James, Michael Clark, Lawrence Ezrow, and Garrett Glasgow Understanding Change and Stability in Party Ideologies: Do Parties Respond to Public Opinion or to Past Election Results? British Journal of Political Science 34(4): Adams, James, Samuel Merrill III, and Bernard Grofman A Unified Theory of Party Competition: A Cross-National Analysis Integrating Spatial and Behavioral Factors. Cambridge: Cambridge University Press. Adams James, Michael Clark, Lawrence Ezrow, and Garrett Glasgow Are Niche Parties Fundamentally Different from Mainstream Parties? The Causes and the Electoral Consequences of Western European Parties policy Shifts, American Journal of Political Science 50(3): Adams, James and Zeynep Somer-Topcu Do Parties Adjust Their Policies in Response to Rival Parties Policy Shifts? Spatial Theory and the Dynamics of Party Competition in Twenty- Five Postwar Democracies. Presented at the Annual Meeting of the American Political Science Association, Philadelphia, PA, August 31- September 3. Aldrich, John and James Alt Introduction to the Special Issue. Political Analysis 11: Beck, Nathaniel, and Jonathan N. Katz What to Do (and Not to Do) with Time-Series Cross-Section Data. American Political Science Review 89(3): Blondel, Jean and Ferdinand Muller-Rommel Cabinets in Eastern Europe. Hampshire and New York: Palgrave. Budge, Ian A New Theory of Party Competition: Uncertainty, Ideology, and Policy Equilibria Viewed Comparatively and Temporally. British Journal of Political Science 24(4): Clark, Michael Valence Issues and Electoral Outcomes in Western Europe, Working paper. Green, Donald P., Soo Yeon Kim, and David H. Yoon Dirty Pool. International Organization 55(2): Grose, Christian R Do Legislators Use Pork Projects to Deviate from Constituents Interests? Valence Advantages and Position-taking in Congress. Working paper. 23

25 Grose, Christian R Valence Advantages and Ideological Shirking in the US Senate: Why Do Senators Take Positions That Are Different From Their Constituents Preferences? Working Paper. Groseclose, Timothy A Model of Candidate Location When One Candidate Has a Valence Advantage, American Journal of Political Science, 45, King, Anthony Do Leaders Personalities Really Matter?, in Anthony King ed., Leaders Personalities and the Outcomes of Democratic Elections (Oxford, U.K.: Oxford University Press). Klingemann, Hans-Dieter; Andrea Volkens, Judith Bara, Ian Budge, and Michael McDonald Mapping Policy Preferences: Estimates for Parties, Electors, and Governments in Eastern Europe, European Union and OECD Oxford: Oxford University Press. Lewis-Beck, Michael S. and Mary Stegmaier Economic Determinants of Electoral Outcomes. Annual Review of Political Science 3: Londregan, John, and Romer Thomas Polarization, Incumbency, and the Personal Vote, in William A. Barnett, Melvin J. Hinich, and Norman J. Schofield ed., Political Economy: Institutions, Competition, and Representation, (New York: Cambridge University Press). MacDonald, Stuart E., and George Rabinowitz Searching the Paradox of Nonconvergence: Valence, Position, and Direction in Democratic Politics, Electoral Studies, 17, No.3, McDonald, Michael, and Ian Budge Elections, Parties, Democracy: Conferring the Median Mandate. Oxford: Oxford University Press. McDonald, Michael, Sylvia Mendes, and Ian Budge What Are Elections For? Conferring the Median Mandate. British Journal of Political Science. Meguid, Bonnie Competition between Unequals: The Role of Mainstream Party Strategy and Niche Party Success. American Political Science Review 99(3): Nagel, Jack Center-Party Strength and Major-Party Polarization in Britain. Presented at the Annual Meeting of the American Political Science Association, San Francisco, CA, August 30-September 2. Pacek, Alexander C Macroeconomic Conditions and Electoral Politics in East Central Europe. American Journal of Political Science 38(3): Paldam, Martin The Distribution of Electoral Results and the Two Explanations of the Cost of Ruling. European Journal of Political Economy 2:

26 Paldam, Martin How Robust is the Vote Function?: A Study of Seventeen Nations Over Four Decades, in Norputh, H., Lewis-Beck, M.S., Lafay, J. (Eds.), Economics and Politics: The Calculus of Support (Ann Arbor: University of Michigan Press), Powell, G. Bingham Elections as Instruments of Democracy. Majoritarian and Proportional Visions. New Haven: Yale University Press. Powell, G. Bingham and Guy D. Whitten A Cross-National Analysis of Economic Voting: Taking Account of the Political Context. American Journal of Political Science 37.2: Przeworski, Adam Public Support for Economic Reforms in Poland. Comparative Political Studies 29(5): Rose, Richard, and Thomas T. Mackie Incumbency in Government: Asset or Liability?, in Hans Daalder and Peter Mair ed., Western European Party Systems: Continuity and Change, (Beverly Hills, CA: Sage Publications), Schofield, Norman Valence Competition in the Spatial Stochastic Model, Journal of Theoretical Politics, 15, No.4, Schofield, Norman A Valence Model of Political Competition in Britain: Electoral Studies 24: Schofield, Norman Equilibria in the Spatial Stochastic Model of Voting with Party Activists. Review of Economic Design 10(3): Schofield, Norman Political Equilibria with Electoral Uncertainty. Social Choice and Welfare 28(3): Schofield, Norman and Guido Cataife A Model of Political Competition with Activists Applied to the Elections of 1989 and 1995 in Argentina. Working Paper. Schofield, Norman and Itai Sened Modeling the Interaction of Parties, Activists and Voters: Why is the Political Center so Empty? European Journal of Political Research 44: Schofield, Norman and Itai Sened Multiparty Democracy: Elections and Legislative Politics. Cambridge: Cambridge University Press. Stokes, Donald Spatial Models and Party Competition, American Political Science Review, 57, Stokes, Donald Valence Politics, in Dennis Kavanagh, ed., Electoral Politics (Oxford: Clarendon Press),

27 Strom, Kaare and Seymour Martin Lipset Macroeconomics and Macropolitics: The Electoral Performance of Democratic Governments. Presented at the annual meeting of the American Political Science Association, Washington, DC. Tucker, Joshua Regional Economic Voting: Russia, Poland, Hungary, Slovakia and Russia from Cambridge, MA: Cambridge University Press. Whitten, Guy D. and Harvey D. Palmer Cross-National Analyses of Economic Voting. Electoral Studies 18: