Does Partisanship Hurt Electoral Accountability? Individual-Level and Aggregate-Level Comparisons of Western and Postcommunist Democracies

Abstract

While the negative effect of partisanship on electoral accountability has been assumed in established Western democracies, its empirical test has been rare, especially for postcommunist democracies whose democratization processes are so distinct that their partisanship might deliver different political impacts from their Western counterparts. Through individual- and aggregate-level regressions, I find that partisanship, in general, does hurt electoral accountability at both levels. From a comparative perspective, the individual-level tests reveal that such a negative effect of partisanship has been more salient among the postcommunist voters with fewer electoral resources to rely on than among the Western voters whose abilities to hold the government accountable are stronger. At the aggregate level, however, this tendency becomes reversed as any increase in partisanship beyond its current relatively high level may induce the over-institutionalization of the Western party systems restricting electoral accountability, while it could rather help to stabilize the postcommunist party systems making it easier for voters to discern which political group is accountable for policy outcomes.

Keywords

partisanship electoral accountability postcommunist voting behavior party system institutionalization

Introduction

Government accountability is an essential feature of representative democracy.¹ The degree to which politicians are constrained from pursuing private interests and compelled to abide by the rules of the democratic game largely rests on the ability of citizens to hold them responsible for their political actions.² Where incumbents are not fairly judged on the basis of their performance, citizens cannot be well represented by their governments. In this regard, particularly, the reward–punishment mechanism of elections constitutes the backbone of government accountability.³ However, as a large number of empirical analyses demonstrates, even voters in mature democracies do not always use elections as an instrument to hold the government accountable. This is because, at least in part, incumbent performance is not the only voting determinant; voter decision making is driven by many other factors, such as partisanship.⁴

Partisanship, as one of the strongest predictors of voting decision widely discussed in the literature,⁵ functions to restrict electoral accountability largely in two fashions. First, partisanship serves to provide “political cues to voters who feel themselves in need of guidance because they must make political decisions under confusing circumstances.”⁶ It is impractical to expect ordinary citizens to become fully informed when participating in elections. This cognitive limitation leads them to rely on partisanship as a heuristic to help them process political information more efficiently. In this process, they tend to filter out information along their partisan identities and thus have biased views regarding incumbent performance. Second, partisans, even though their judgment is based on correct information, are predisposed to vote for their preferred party in the hope that the party’s electoral victory will bring them concrete benefits.⁷ Parties build on social constituencies they claim to represent and choose policies to court the favor of those groups. It is therefore logical for voters to support the party representing their interests even though they are well aware that the party’s performance in retrospect has not been up to the mark.

As such, it can be posited that partisanship restrains electoral accountability. However, this inverse relationship between partisanship and electoral accountability has been mostly implicitly theorized and rarely empirically tested in the previous literature. A handful of comparative studies test the inverse relationship,⁸ but most focus on established Western democracies, rather than comparing old and young democracies. This omission is striking given the popularity of comparative research between established and emerging democracies in matters of partisanship and voting behavior.⁹

In particular, young postcommunist democracies, given their distinct paths to democracy, should diverge from Western democracies in certain ways. First of all, their lack of democratic history meant their democratic institutions had to “be constructed virtually from scratch.”¹⁰ Massive socioeconomic reform coincided with a broad spectrum of political reforms, making their democratization processes more complicated. This unique path of democratization has been associated with certain pathologies such as low levels of civil society participation¹¹ and unstable party systems,¹² making them distinct from even other groups of new democracies. This postcommunist distinctiveness offers reasons to expect the relationship between partisanship and electoral accountability to differ, at least in degree, from established Western democracies.¹³

I further assume that such differences in the moderating effects of partisanship on accountability between the two regions depend on the level of analysis, whether they are gauged at the individual or aggregate level. This is predicated on the idea that partisanship plays different roles depending on the level of analysis.¹⁴ At the individual level, partisanship frames one’s views to make sense of the complex political world shaping political attitudes and behaviors.¹⁵ For this reason, partisanship is often used as a heuristic that lowers the cognitive bar to settle on political opinions and issue positions and, most importantly in this research, to participate in elections and how to vote. At the aggregate level, meanwhile, partisanship functions as “the central organizing principle of mass politics”¹⁶ that contributes to the formation of ties between political parties and society. Such connections can, when sufficiently established, serve to promote electoral stability and party system institutionalization.¹⁷ At this aggregate level, partisanship is considered one of the prerequisites for democratic development, especially for newly democratizing societies. Where social cleavages have yet to crystallize into political divisions along party lines, growing partisanship can strengthen electoral accountability by stabilizing party systems and making responsibility for policy outcomes clear. If the connections are too strong, however, it might lead to rigid party systems in which parties are not judged on policy performance. Given the different roles partisanship plays depending on the level of analysis as such, it is suggested that any political impact of partisanship needs to be tested at both levels.

In sum, the questions explored through this article are as follows: Does partisanship have a negative impact on electoral accountability? If so, how does the negative effect vary between Western and postcommunist democracies? And, is the pattern of regional variation the same across the level of analysis? I begin this article by summarizing the debate on this topic and show how my research advances this research agenda. I will then construct individual-level and aggregate-level hypotheses based on existing theories and test them by turn. The last section will summarize and discuss the findings of the study.

Partisanship and Electoral Accountability

A long line of literature mostly focused on Western advanced democracies has regarded partisanship and government performance as major voting determinants. This literature can be divided into three streams. The first stream addresses which of the two is the more fundamental predictor of vote choice by locating them along with an array of voting determinants. A classical sociopsychological model of vote choice, the “funnel of causality”¹⁸ shows that partisanship is a deep and long-term factor shaping voter decision in contrast to short-term factors such as candidate images, issue positions, and policy performance. Given that distant causes condition the extent to which more proximate causes account for outcomes,¹⁹ partisanship, in comparison to short-term performance issues, can be said to determine a larger portion of variation in voters’ electoral decisions.

Second, there has been a discussion over the relative importance of partisanship and performance as determinants. The heart of the discussion lies in whether partisanship has weakened during the past half century and how such a trend has strengthened the other determinants like government performance. Some argue that partisanship has lost its power to determine vote choice due to social changes that took place in the 1960s and 1970s.²⁰ They contend that the decay of this partisan-electoral linkage has led to the rise of independent voters who judge candidates in accordance with their policy performance. This partisanship decline thesis, however, has been challenged by other scholars claiming that the relative importance of party identification and economic performance in voting behavior has remained more or less the same.²¹

Finally, a third group has challenged the notion of economic voting by contending that voters’ economic perceptions themselves are shaped by their partisan orientations, so any causal effects economic perceptions exert on vote choice are in fact spurious or at best overestimated.²² This argument concerning endogeneity in models of economic voting provoked a debate on the validity of such findings. The advocates of economic voting, such as Michael Lewis-Beck and his colleagues,²³ have responded with new research that continues to show that the economy still plays a decisive role in voting choice.

Figure 1 illustrates these complex relations among variables in a path diagram. Previous scholarship in the Western context has either examined the separate or relative effects of partisanship and party performance on vote choice (arrows A1 and A2²⁴) or analyzed how performance perception is shaped by partisanship (arrow B), but little empirical work has tested the effect of partisanship on the performance–voting linkage (arrow C in bold), not to mention simultaneous testing of all the causal arrows. When it comes to the abovementioned third stream of the literature regarding the simultaneous relations between perceptions, partisanship, and vote choice, in particular, Wlezien et al.²⁵ pioneered in revealing a reciprocal relationship between economic evaluations and vote choice, but their analysis almost equated vote choice with partisanship—which party they support—and did not put great emphasis on the mechanisms by which vote choice influences economic perceptions, one of which will be partisanship. Anderson et al.²⁶ well acknowledged the theoretical possibility of partisanship acting as a determinant of both economic perceptions and vote choice, but their empirical analysis only looked at how economic perceptions are shaped by vote choice and partisanship. Meanwhile, a series of studies conducted by Evans and his colleagues²⁷ focused on a feedback loop between economic perceptions and party identification, but took vote choice as one of the measures of partisanship failing to disentangle complex relations between perceptions, partisanship, and vote choice. Van der Brug et al.²⁸ also provided a thorough analysis that breaks into two stages in which the effect of the economy on party support and that of party support on vote choice are tested in turn, but did not go further to examine the opposite causality going from party support to performance evaluations.

Figure 1

Relationships between partisanship and electoral accountability

To my knowledge, only a couple of studies in the economic voting literature have conducted empirical analyses of how partisanship moderates the linkage between government evaluation and vote choice (arrow C). First, Tillman,²⁹ through his individual-level analysis, found that partisans make less use of economic performance in judging the government relative to non-partisans. Kayser and Wlezien³⁰ conducted both individual- and aggregate-level tests revealing that partisanship reduces the strength of economic voting at both levels. In their tests of the interactive effect between partisanship and economic performance in vote choice, they acknowledge an endogenous relationship, that partisanship may determine economic evaluations. To avoid biased estimates, specifically, Kayser and Wlezien chose to use only objective economic data such as GDP growth rather than subjective perceptions of the economy.³¹ Their use of an objective measure has a point in that, as they claimed, subjective evaluations should involve observable and unobservable attitudes and behaviors whose effects are hard to separate out as well as in that their goal of study was rather to analyze the effect of the real economy, not its perceptions, on the strength of partisanship as a voting determinant. Admitting that every model is bound to have its own limitations, though, I seek to distinguish my analysis from theirs by examining how one’s subjective evaluations of the government affect vote choice, and such an individual-level analysis requires to take into account the endogeneity that subjective judgment may be shaped by political preferences, which is why I employed the simultaneous modeling technique.³²

In the postcommunist context, it is hard to find studies that make direct reference to the relationship between partisanship and accountability. Instead, partisanship and electoral accountability have been studied in separate streams of research. With respect to postcommunist partisanship, there has been a debate as to whether and how much viable partisanship has developed in the region. Early studies contended that postcommunist partisanship has not been fully developed, mainly because partisanship formation is a time-consuming process that cannot be done overnight.³³ However, later work made a counterargument that a sufficient level of partisanship comparable to Western democracies has already emerged in this region.³⁴ More recent studies focusing on the substantial sense of partisanship claim that postcommunist partisanship with time has come to play a role in shaping citizens’ political positions and voting decisions.³⁵

In terms of electoral accountability, the first decade of postcommunist studies were largely devoted to answering the question of whether postcommunist voters are economic voters.³⁶ Following these studies, a series of attempts were made to identify factors that account for the variation in the strength of economic voting across postcommunist countries such as a voter’s level of political information,³⁷ the position of the incumbent’s party in relation to reforms (i.e., a pro-reform party or not),³⁸ the degree of party system institutionalization,³⁹ and types of economic performance.⁴⁰

As such, while there has been a plenty of work on postcommunist countries that investigates partisanship and electoral accountability in isolation, little has examined how they function together. I attempt to fill this gap by conducting more comprehensive empirical analyses that cover both Western and postcommunist democracies. The contribution of this study to the existing literature is thus threefold. First, going beyond Western established democracies, this study expands the scope of analysis to postcommunist democracies in analyzing the relationship between partisanship and electoral accountability for the first time. Second, this study directly compares the two regions, Western advanced and postcommunist emerging democracies, in terms of the degree to which partisanship constrains electoral accountability. Finally, this article makes a methodological contribution by employing various statistical techniques that can untangle complex causal relations at both individual and aggregate levels. At the individual level, especially, this study will provide a rare opportunity to disentangle complex causal relations between partisanship, government evaluations, and vote choice by testing multiple equations simultaneously.

Theories and Hypotheses

On the basis of the studies discussed above, it is possible to construct a number of hypotheses on the relationship between partisanship and electoral accountability. The Michigan model of partisanship predicts that an increase in partisanship will diminish the effect of short-term performance-based factors in explaining voting results, thereby weakening electoral accountability.

However, subsequent theories challenge the negative effect of partisanship. As noted in the previous section, a group of scholars claim that the linkages between social cleavages and political preferences have attenuated.⁴¹ Since Lipset and Rokkan⁴² set the groundwork for the translation of social divisions into political divisions, the 1960s and 1970s saw a series of socioeconomic changes that weakened both the size and in-group cohesion of traditional social groups, notably the working class. Instead, more diverse issue-oriented factors have come to shape people’s voting decisions. If this is the case, more recent analyses should observe a weaker effect of partisanship, which could keep us from finding such inverse relations. Also, with respect to endogeneity between incumbent evaluation and partisan ties, if partisanship shapes one’s evaluation of incumbents, the effects of the evaluation on voting may not be influenced by partisanship since partisan effects are already absorbed into the effects of the evaluation. In this case, therefore, the hypothesis that partisanship impedes the evaluation–voting linkage may not hold up.

Additionally, I expect the negative effect of partisanship on performance voting to manifest itself both at the individual and aggregate levels of analysis. At the individual level, partisans of the governing government parties will be more likely to vote for them even under conditions of poor performance. Similarly, the connection between aggregate-level performance variables and electoral results for the incumbents should weaken as the number of partisans increases in a society. Thus, I will test the following hypotheses:

Hypothesis 1A: At the individual level, partisanship moderates the effect of government evaluation on vote choice.

Hypothesis 1B: At the aggregate level, an increase in partisanship moderates the effect of government performance on electoral results.

If these hypotheses are supported, the next question is whether Western and postcommunist democracies diverge in terms of this inverse relationship. The expectation is that the results will vary according to the level of analysis. On the individual level, the negative linkage between partisanship and accountability should be more salient in the Western than the postcommunist sample. As Western partisanship has been built for generations in long-established party systems, it should be more deeply engrained in citizens’ minds than its postcommunist counterpart. Western parties were built on social cleavages that had developed for centuries as a result of a set of historical junctures, and such close connections between parties and social groups allowed the citizens to establish psychological attachments towards their preferred parties making partisanship a form of social identity that runs through families.⁴³ This social identity tends to work as an affective orientation that sometimes goes beyond rational political decision making.⁴⁴ If this is the case, Western partisans should be less likely to base their voting decisions on the rational evaluation of government performance compared to non-partisans. Similarly, the lack of a long history of party system development in the postcommunist region will mean that partisanship has not acted as strongly as a social and affective identity. Moreover, the massive economic reforms in the region will mean that the socioeconomic identities on which partisan orientation are built have been fluid, complicating and disrupting the formation of party–society linkages. If postcommunist partisanship is less a cleavage-based identity than a shallow expression of party preference, it should play a weaker role in restricting electoral accountability than Western partisanship.

However, it is also plausible to argue otherwise. Since Western voters have been exposed to democratic elections for a longer period, they could be more sophisticated voters whose ability to hold the government accountable cannot be so easily eroded. A long history of democracy should have made Western voters more educated on how democracy works and more politically informed than those in young democracies. Also, well-developed civil societies and social networks in Western societies provide abundant resources voters can rely on when voting other than partisanship. If this is the case, the effect of partisanship that moderates the evaluation–vote connection could be restrained. By contrast, postcommunist voters who lack experience with democratic elections might not get used to linking government performance with their vote choices as a means of giving reward or penalty to the incumbent. In addition, given that extant levels of education and resources are lower, they may resort to partisanship to a greater extent compared to their Western counterparts.⁴⁵ As I expect the latter mechanisms to overpower the former, I construct the hypothesis:

Hypothesis 2A: At the individual level, the moderating effect of partisanship on electoral accountability is weaker among Western voters than among postcommunist voters.

At the aggregate level, the same hypothesis as the individual-level one may be forged in that in Western democracies where citizens are well experienced in utilizing elections as a mechanism of accountability, the correlation between national-level policy performance and electoral results for the incumbents could be so strong that it will not so easily be moderated by an increase in partisanship. In the postcommunist context, contrarily, voters may not place the whole blame upon their governments for economic problems they suffer during the early stage of economic reforms, if they regard it as “the consequence of the mediocre state of the economy after the collapse of communism rather than bad economic policies of the post-communist governments.”⁴⁶ If this is the case, the moderating effect of partisanship would appear more strongly in postcommunist democracies than in Western democracies at the aggregate level as well. However, there is a reason to presume that it will not be the case.

As previously noted, aggregate partisanship establishes ties between citizens and parties that may foster electoral accountability by contributing to party system stability, especially for new democracies. However, if partisanship exceeds a certain level it may harm accountability by locking voters into partisan orientations and keep them from judging governments on the basis of their performance. Especially in old democracies where a sufficient level of partisanship is already in place, an additional increase in the aggregate level of partisanship may lead to an overinstitutionalization of the party system. In overinstitutionalized party systems, political parties secure a stable support base from partisans who are not easily affected by how the parties perform, and this clearly hurts overall electoral accountability.⁴⁷

Meanwhile, in new democracies where social ties between voters and parties are not established, an increase in partisanship at the society level could rather help to stabilize party systems by making it easier for voters to discern which party is responsible for policy outcomes. This mechanism is pertinent to postcommunist democracies that have suffered higher levels of electoral volatility than other democracies over the past two decades. Thus, any increase in the aggregate level of partisanship in the region could promote electoral stability, which in turn contributes to electoral accountability by making electoral competition more predictable.⁴⁸ If this is the case, even if the relationship between partisanship and accountability is not positive as such, the role of partisanship moderating the evaluation–voting linkage should be weaker in postcommunist democracies than in Western democracies. In line with this reasoning, I construct the hypothesis below:

Hypothesis 2B: At the aggregate level, the moderating effect of partisanship on electoral accountability is stronger in Western democracies than in postcommunist democracies.

Individual-Level Analysis

Data and Method

Given that data availability restrains the conduct of any individual-level comparative analysis, the Comparative Study of Electoral Systems (CSES) survey data has a merit because of its wide temporal and spatial coverage with a consistent set of items being asked. This analysis uses the second and third waves of the data⁴⁹ collected from thirty-six electoral studies in seventeen Western democracies and seventeen studies in eleven postcommunist democracies⁵⁰ from 2001 to 2011.

Using the CSES data requires caution because of its multilevel nature. As individual respondents are nested in both countries and survey years (which are in most cases equivalent to election years), running a simple ordinary least squares regression without taking into account within-group similarities may violate its core assumptions such as homoscedasticity and independence of error variances.⁵¹ In this regard, I employ a multilevel modeling technique taking country-year as upper level and individual respondents as lower level.⁵² Another issue that needs to be addressed is the complex relations among variables tested. As mentioned above, voting determinants are so intertwined with one another that they can be placed on either side of the voting equation. Government evaluation is one such variable as it may function as a predictor of vote choice along with partisanship, but at the same time, as noted above, can be an outcome influenced by partisanship. Without addressing such endogenous relations, the effect of government evaluation on voting may be overestimated. To address both the multilevel nature of the data analyzed and the recursive relations of the variables tested, therefore, I test multilevel (random intercept) structural equation (recursive) models.⁵³

Variables

Incumbent vote, one of the two main endogenous variables, is coded as 1 if the respondent voted for the incumbent party or parties in the last election, and as 0 otherwise. For parliamentary systems into which most countries in my analysis are categorized, a party or parties that constitute the cabinet before the elections count as incumbent. The CSES asks the respondent’s vote choice on presidential elections in presidential or semi-presidential systems for five cases: France (2002), Romania (2004, 2009), and the United States (2004, 2008). For these, I regard a party that holds the presidency before the elections as incumbent.⁵⁴ Another endogenous variable, incumbent evaluation, is constructed using a five-point scale asking the respondents to assess the general performance of the government at that time between 1 (very bad job) and 5 (very good job). A common set of demographic variables used for an individual-level analysis such as age, gender, and education are included in both equations as control variables.

The respondent’s partisanship is introduced into both equations as the main exogenous variable that influences government evaluations as well as vote choices. As the two endogenous variables above gauge how the “incumbent” parties are evaluated and whether they are voted into office, I create two dummies indicating whether the respondents in question identify with incumbent or opposition parties—incumbent partisanship (1 for incumbent partisans and 0 for opposition partisans or non-partisans) and opposition partisanship (1 for opposition partisans and 0 for incumbent partisans or non-partisans)—leaving non-partisans as the comparison group. The CSES items used for these variables are: “Do you usually think of yourself as close to any particular political party?” (Yes/No); if answered yes, “Which party is that?”⁵⁵ I expect the effect of incumbent partisanship on the incumbent evaluation and voting for the incumbent to be positive, and the effect of opposition partisanship to be negative.

To see how partisanship moderates the evaluation–vote linkage, in order to test hypothesis 1A, I multiply each of the two partisanship dummies by incumbent evaluation.⁵⁶ I expect both interaction effects to be negative since partisanship, either incumbent or non-incumbent, should impede the connection between one’s performance evaluation and vote choice (i.e., to the extent that incumbent partisans vote for the incumbents despite a negative judgment on performance, or opposition partisans would not vote for them even though their evaluation is positive). In order to test for variation across the regions for the negative partisanship–accountability linkage, hypothesis 2A, I create a three-way interaction term by multiplying the abovementioned two-way interaction with a postcommunist regional dummy. The resulting equations for the main model testing for regional variation are presented below (with irrelevant constitutive terms omitted), and the relationships between the main endogenous and exogenous variables can be depicted as a path diagram in Figure 2.⁵⁷

\begin{array}{l} I n c u m b e n t E v a l u a t i o n_{i j} = β_{10 j} + γ_{11} A g e_{i j} + γ_{12} G e n d e r_{i j} + γ_{13} E d u c a t i o n_{i j} + \\ γ_{14} I n c u m b e n t P a r t i s a n s h i p_{i j} + γ_{15} O p p o s i t i o n P a r t i s a n s h i p_{i j} + ζ_{1 j} + ε_{1 i j} \end{array}

\begin{array}{l} \begin{array}{l} I n c u m b e n t V o t e_{i j} = β_{20 j} + γ_{21} A g e_{i j} + γ_{22} G e n d e r_{i j} + γ_{23} E d u c a t i o n_{i j} + \\ β_{21} I n c u m b e n t E v a l u a t i o n_{i j} + γ_{24} I n c u m b e n t P a r t i s a n s h i p_{i j} + \\ β_{22} Incumbent Evaluation \times Incumbent Partisanshi p_{i j} + γ_{25} O p p o s i t i o n P a r t i s a n s h i p_{i j} + \end{array} \\ β_{23} Incumbent Evaluation \times Opposition Partisanshi p_{ij} + γ_{26} P o s t c o m m u n i s t D u m m y_{j} + \\ β_{24} Incumbent Evaluation \times Incumbent Partisanship \times Postcommunist Du m m y_{ij} + \\ β_{25} Incumbent Evaluation \times Opposition Partisanship \times Postcommunist Dumm y_{i j} + \\ γ_{27} A n n u a l G D P g r o w t h {(%, o n e y e a r l a g g e d)}_{j} + ζ_{2 j} + ε_{2 i j} \end{array}

(where i = each respondent, j = each country-year, ζ = upper-level error term, ε = lower-level error term. No correlations exist among the error terms across equations at both levels; variables of main interest are in bold. Those sets of the composite variables of the interaction term irrelevant to the analysis are not reported.)

Figure 2

A path diagram on the recursive relationships

Analysis and Results

Before turning to the regressions, a simple descriptive analysis should give readers a sense of how voter choices and incumbent evaluations are distributed according to partisanship types across the two democracies. Restricting the sample to the second and third CSES studies, Table 1 presents how partisanship biases one’s evaluation and vote choice. In the Western democracies, those attached to governing parties evaluate them more positively (3.70) than non-partisans (3.11), whereas those with opposition partisanship give more negative ratings (2.69). There is a similar pattern among the postcommunist voters. Partisanship also exerts a huge impact on one’s vote choice, as most incumbent partisans cast ballots for the incumbent parties while opposition partisans rarely did so in both. To see the indirect effect of partisanship on vote choice through party evaluation, I calculate mean percentages of the votes for the incumbents across positive or negative government ratings, and compare them according to partisan types. In Table 2, the numbers highlighted in bold particularly show that the incumbent partisans are predisposed to vote for the governing parties even though they evaluate them negatively, while the opposite is the case for the opposition partisans. Comparing the two groups of democracies, it is not clear in which group partisanship is more likely to restrain the evaluation–vote linkage; while more biased vote choices for the incumbents are made among the postcommunist incumbent partisans with negative ratings (84.3 percent compared to 81.6 percent among their Western counterparts), the opposition partisans in the West are more biased than the postcommunist ones in terms of their vote choices for the incumbents (6.6 percent compared to 7.5 percent).

Table 1

Mean of Incumbent Evaluations and Incumbent Votes across Partisan Types and Regions

		Incumbent Evaluation		Incumbent Votes
		Individuals	Mean	Individuals	Mean
Western	Incumbent partisans	13589	3.70	12923	90.0%
	Opposition partisans	13914	2.69	13388	4.0%
	Non-partisans	29719	3.11	24361	38.5%
Post-communist	Incumbent partisans	3473	3.69	2794	90.9%
	Opposition partisans	6527	2.27	5989	3.8%
	Non-partisans	12312	2.72	8148	37.9%

Table 2

Mean of Incumbent Votes by Incumbent Evaluation across Partisanship and Regions

		Incumbent Evaluation	Incumbent Votes
	Partisanship	(Positive: 4 or 5, Negative: 1 or 2)	Individuals	Mean
Western	Incumbent partisans	Positive	10131	92.1%
	Incumbent partisans	Negative	2274	81.6%
	Opposition partisans	Positive	4985	6.6%
	Opposition partisans	Negative	7328	2.4%
	Non-partisans	Positive	13424	49.7%
	Non-partisans	Negative	8830	22.7%
Post-communist	Incumbent partisans	Positive	2460	92.4%
	Incumbent partisans	Negative	568	84.3%
	Opposition partisans	Positive	1446	7.5%
	Opposition partisans	Negative	3988	2.5%
	Non-partisans	Positive	2806	58.2%
	Non-partisans	Negative	3105	20.1%

To estimate their causal relations, I ran a set of regression models. Table 3 presents the results of the full models that simultaneously test the two equations listed above. In model 1.1, the effects of the variables of primary interests turn out to be all significant and correctly signed in accordance with my expectations. In the first equation with incumbent evaluation as an endogenous variable, it is found that incumbent partisanship leads to more positive incumbent ratings while opposition partisanship has the opposite effect. In the second equation taking incumbent vote as an endogenous variable, incumbent partisanship is positively correlated to voting for the incumbents, while opposition partisanship has a negative impact on it. Also, incumbent evaluation, even after taking into account its endogeneity with partisanship and controlling for macroeconomic performance indicated by GDP growth rates, is shown to have a significant positive effect on vote for the incumbent. These results indicate that one’s subjective perception of government performance, though it is influenced by partisan orientation, has its own independent effect on vote choice.

Table 3

Multilevel Structural Equation Models of the Individual-Level Relationship

Model	1.1		1.2
Model	Coefficient	SE	Coefficient	SE
Fixed Effects
DV: Incumbent Evaluation
Age	−0.001*	0.001	−0.001***	0.000
Male	−0.019	0.017	−0.018**	0.008
Education	0.008	0.005	0.003	0.002
Incumbent partisanship	0.707***	0.021	0.723***	0.010
Opposition partisanship	−0.381***	0.021	−0.382***	0.009
Constant	2.906***	0.040	3.104***	0.019
DV: Incumbent Vote
Age	0.002***	0.001	0.002***	0.000
Male	−0.035	0.027	−0.030**	0.014
Education	−0.005	0.008	−0.021***	0.004
Incumbent evaluation	0.433***	0.014	0.351***	0.009
Incumbent partisanship	1.837***	0.093	1.874***	0.063
Opposition partisanship	−0.919***	0.105	−0.965***	0.068
Incumbent Evaluation × Incumbent Partisanship	−0.127***	0.027	−0.128***	0.018
Incumbent Evaluation × Opposition Partisanship	−0.148***	0.033	−0.140***	0.021
Postcommunist dummy			−0.010	0.060
Incumbent Evaluation × Incumbent Partisanship × Postcommunist Dummy			−0.108***	0.040
Incumbent Evaluation × Opposition Partisanship × Postcommunist Dummy			−0.051	0.039
Annual % change in GDP	0.016***	0.005	−0.003	0.002
Constant	−1.732***	0.080	−2.094***	0.051
Random Effects: Standard Deviation (Constant)
Country-level
DV: Incumbent evaluation	0.168***		0.385***
DV: Incumbent vote	0.173***		0.146***
Individual-level
DV: Incumbent evaluation	1.037***		1.034***
DV: Incumbent vote	1 (constrained)		1 (constrained)
Individuals	71,699		71,699
Country-years	53		53
Log likelihood	–127484.370		–130656.550

Note: The variances of the constant terms of both equations are found significant, indicating that the data examined contain unobserved group-level effects that cannot all be explained by the independent variables in the models. DV = dependent variable. Two-tailed tests; ***p < .01, **p < .05, *p < .1.

More importantly, both interaction terms in the second equation highlighted in bold produce significant negative coefficients, suggesting that having partisanship at the individual level, whether incumbent or opposition, reduces the effect of incumbent evaluation on incumbent vote. Figure 3 illustrates these findings as a path diagram. It shows that partisanship not only has direct and indirect effects on voting (via incumbent evaluation) but also exerts a conditioning impact on the evaluation–vote linkage as depicted in dashed lines. Such declines in the effect of the incumbent evaluation among partisans can be seen in Figure 4, which depicts the extent to which the average predicted probabilities of the incumbent vote⁵⁸ change according to ratings of government performance between partisans and non-partisans. The effects of incumbent evaluation for both types of partisans are flatter than that of non-partisans. To be more specific, non-partisans who rated their government more positively are more likely to vote for the government parties while such a tendency is much less salient among partisans. Those who are attached to government parties are predisposed to vote for the parties even when they evaluate the governments negatively. Conversely, those who oppose government parties are not inclined to vote for them even though their evaluations of the incumbents are positive. These results thus support hypothesis 1A that partisanship weakens electoral accountability at the individual level.⁵⁹

Figure 3

Estimated coefficients on the recursive relationships (based on model 1.1)

Figure 4

Changes in the average predicted probabilities of incumbent vote on incumbent evaluation across partisanship types

Turning to the regional comparison, in model 1.2, I include two three-way interactions that now include a postcommunist dummy. In line with hypothesis 2A, that negative effect is greater among postcommunist voters, the signs on both three-way interaction terms should be negative. The results largely support the hypothesis; the two interaction terms highlighted in bold all exert negative effects on the incumbent vote. Although the interaction with opposition partisanship just misses the conventional level of statistical significance with its actual p value of 0.187, its incumbent counterpart exerts a strongly significant negative effect, showing that the postcommunist voters are more susceptible to their partisan identities in linking government evaluations to voting decisions. In short, all of these individual-level results can be said to confirm the hypothesis that partisanship leads to weaker electoral accountability to a greater degree among postcommunist voters than among Western voters.

Aggregate-Level Analysis

Data and Method

Turning next to the aggregate-level analysis, I ran a separate set of time-series cross-sectional (TSCS) models⁶⁰ using country-level electoral and macroeconomic data. To compare the results of the individual-level analyses with their aggregate-level counterparts, it would be best to use the same models and variables whose values are aggregated by election years. However, the small number of election-level cases used (n=53) complicated structural equation modeling because the degree of freedom of the models was constrained to the point where they could not converge. For this reason, I employed the TSCS models and constructed a new set of country-year-level variables that have been frequently used in previous aggregate-level analyses. Given the possibility of endogeneity bias in single-equation modeling, following the lead of earlier studies,⁶¹ I used objective macroeconomic data as a proxy for national-level government performance. In an attempt to relate the aggregate-level analyses to the individual-level ones, I restricted the sample to the elections used in the individual-level tests (a total of thirty-seven elections in seventeen Western countries and eighteen elections in eleven postcommunist countries).⁶²

In TSCS models, one needs to address group-specific effects. I chose to use random effects estimation over fixed effects for two reasons. First, the results of the Hausman test (used to check if group effects are correlated with the regressors) suggest that testing via random effects is more appropriate for the data used in this analysis. Second, even after taking into account recent concerns about using the Hausman test as a proper measure of model selection,⁶³ using random effects can be justified on the ground that the main goal of this study is to uncover cross-regional differences. Fixed effects is mainly used to explore within-group causes by holing group-level effects at fixed levels, making the test of the effects of time-invariant variables infeasible. Since my aggregate-level models contain a regional dummy that does not vary over time, using random effects is preferred.

Variables

The dependent variable is the electoral performance of incumbent party or parties in national-level parliamentary elections.⁶⁴ I construct two types of incumbent electoral performance variables based on previous studies: (1) share of votes received by the incumbent party or parties in t-th elections⁶⁵; (2) incumbent gains or losses compared to the vote share of the previous elections.⁶⁶

The main independent variable is aggregate partisanship. As partisanship is an individual and subjective notion, it is hard to find a measure for partisanship at the aggregate level. Some studies employ electoral volatility as a proxy for the aggregate-level partisanship,⁶⁷ mostly measured by the Pederson index.⁶⁸ This is built on the idea that a society in which more people hold partisan orientations and cast ballots in line with them would yield less electoral volatility. As this analysis takes the incumbent’s electoral results as a dependent variable, however, using electoral volatility based on the same electoral data might expose the analysis to endogeneity. For this reason, I rely on the same CSES item used in the individual-level analysis as an aggregate measure by calculating the percentage of the respondents who answered that there is a party they feel close to.

Another important independent variable is macro-level policy outcomes. While there are a number of issues for which the government takes responsibility, the economy has been one of the most important and widely talked issues,⁶⁹ including in the postcommunist context.⁷⁰ Thus, I use annual growth rate in GDP (one-year lagged) to measure the government’s general performance.⁷¹ To see whether its impact on electoral results is moderated by partisanship, I interact it with aggregate partisanship. For partisanship to hurt electoral accountability at the aggregate level as hypothesized in hypothesis 1B, the interactive effect should be negative. For regional comparison, I include the postcommunist regional dummy and construct its three-way interaction term with the above-mentioned interaction between GDP growth and partisanship. In line with hypothesis 2B that the moderating effect of partisanship should be weaker in the postcommunist democracies, I expect the three-way interaction to exert a positive effect. Table 4 provides the summary statistics of the variables tested.

Table 4

Summary Statistics of the Main Variables Used in the Aggregate-Level Analysis

	Western(37 election years)		Postcommunist(18 election years)
	Mean	SD	Mean	SD
Incumbent vote share (t, %)	40.0	10.5	37.5	17.8
Incumbent vote margin (between t and t – 1, %)	−5.2	8.0	−9.6	15.3
Partisanship (% of CSES partisans)	48.7	13.4	49.4	18.1
GDP growth (one-year lagged, %)	2.3	1.7	2.6	6.2

Analysis and Results

Table 5 presents the results of the TSCS models. Overall, the independent effect of economic performance on incumbent election results is shown to be robust across the models tested, confirming the economic voting theory at the aggregate level as well. When it comes to the interaction between partisanship and the economy, models 2.1 (incumbent vote share as outcome) and 2.3 (incumbent vote margin as the outcome) testing the moderating effect of partisanship in general are consistent with hypothesis 1B. In both models, a one-unit increase in partisanship significantly reduces the effects of the previous year’s GDP growth on the incumbent electoral performance. This negative effect of partisanship on electoral accountability is registered whether incumbent electoral performance is measured as incumbent vote share or as incumbent vote margins between consecutive elections. Figure 5 presents the results more clearly. As the level of aggregate partisanship increases, the average marginal effect of GDP growth drops in both models, suggesting that where higher levels of partisanship are in place, economic performance plays a weaker role in predicting electoral results.

Table 5

TSCS Models of Incumbent Electoral Performance on Aggregate-Level Partisanship

Model	2.1		2.2		2.3		2.4
	DV: Incumbent Vote Share (t)				DV: Incumbent Vote Margins (t and t – 1)
	Coefficient	SE	Coefficient	SE	Coefficient	SE	Coefficient	SE
Incumbent vote share (t – 1)	0.66***	0.02	0.66***	0.02
Partisanship	0.41***	0.07	0.63***	0.10	0.39***	0.09	0.58***	0.12
GDP growth (one-year lagged)	3.47***	1.33	7.36***	1.79	3.40**	1.35	6.90***	2.07
GDP growth × Partisanship	−0.07***	0.02	−0.17***	0.04	−0.07***	0.02	−0.15***	0.04
Postcommunist			28.54***	10.52			20.23*	11.42
GDP growth × Partisanship × Postcommunist			0.15***	0.05			0.12**	0.05
Constant	−9.94***	3.16	−18.24***	4.35	−24.93***	5.27	−32.36***	6.21
Number of elections	55		55		55		55
Number of countries	28		28		28		28

Note: DV = dependent variable. Two-tailed tests; ***p < .01, **p < .05, *p < .1.

Figure 5

Changes in the average marginal effect of GDP growth on aggregate-level incumbent electoral results

As for the regional comparison, models 2.2 and 2.4 produce results in line with hypothesis 2B; partisanship moderates the economy–vote linkage to a greater extent in Western democracies than in postcommunist democracies. The three-way interaction terms in both models highlighted in bold exert a significant positive effect, meaning that the moderating effect of partisanship for the Western countries is greater than that of the postcommunist countries. The visual presentation in Figure 6 displays how the two regions differ in the extent to which increases in aggregate partisanship lead to declines in the average marginal effect of GDP growth. Notably sharper slopes for the Western countries compared to the postcommunist ones in both graphs indicate that partisanship restrains electoral accountability more strongly in the former than in the latter. While economic performance has less impact on the incumbent’s electoral performance as partisanship strengthens in the Western societies, this relationship is almost absent in the postcommunist societies. In contrast to individual-level, the aggregate-level tests show that an increase in partisanship in the Western democracies is more likely to hurt electoral accountability than in the postcommunist democracies.

Figure 6

Changes in the average marginal effect of GDP growth across the regions

Conclusions

This article explores if partisanship decreases the extent to which the incumbent’s electoral prospects depend on her performance, and examines if its restrictive effect manifests itself differently in Western and postcommunist democracies. Even though one might easily consider partisanship as an individual-level attribute, I examined its effect at the aggregate level as well. In this way, I was able to avoid the “individualistic fallacy” of “incorrectly imputing to the higher order unit the aggregation of values for individuals.”⁷² After all, how individual vote choices differ between partisans and non-partisans is one thing, and how electoral outcomes vary across different degrees of partisanship in a society is another. The individual-level political effects are not always scaled up to the aggregate-level ones.

With respect to the relationship between partisanship and electoral accountability, the results of this analysis show that partisanship does have a negative impact on electoral accountability regardless of the level of analysis, largely confirming the previous findings by Tillman⁷³ and Kayser and Wlezien.⁷⁴ At the individual level, party identification is found to crowd out voting based on performance evaluations. This holds even after addressing the endogeneity raised by the recent economic voting literature that partisanship biases individual performance evaluations. This relationship holds at the aggregate level as well, revealing that the election’s reward–punishment mechanism is weaker where the level of partisanship is higher. These results suggest that despite the recent arguments regarding the weakening ties between partisanship and social cleavages, partisanship still plays a role in limiting the range of voters’ electoral decisions and dampens electoral accountability. To my knowledge, this is the first empirical finding on the interactive effect of partisanship and government performance on the vote in both the Western and postcommunist contexts.

Once this relationship is established, the next question becomes how the relationship varies when different political contexts are taken into account. My individual-level tests find that the negative relationship has been more marked among postcommunist voters than among Western voters. These findings show that although Western partisanship has been built over generations in contrast to more recent postcommunist affiliations, its effect can be counteracted by the Western voter’s ability to hold their governments accountable. As theorized earlier, longer experience with democratic systems and elections as well as a denser level of social networks may enable Western voters to take advantage of elections as an instrument to reward or punish their representatives more effectively compared to postcommunist voters.

However, this pattern is reversed when it comes to the aggregate-level analysis, showing that a one-unit increase in partisanship in the Western societies has exerted a more powerful effect in weakening electoral accountability than in the postcommunist ones. Such a regional pattern may have resulted from the differences in the existing level of aggregate partisanship as well as the degree of party system institutionalization between the two regions. In postcommunist, new democracies where social ties between voters and political parties are weak and party systems are not fully institutionalized, increasing partisanship at the society level can help to stabilize party systems, which in turn makes it easier for the voters to discriminate which political group is accountable for policy outcomes. Meanwhile, in established democracies in which there already exists close connections between citizens and parties under institutionalized party systems, any increase in the level of partisanship could result in the over-institutionalization of party systems and prevent elections from functioning as a reward–punishment mechanism.

The findings of this study suggest that as democracy becomes established, the political impact of partisanship changes, but the change may occur in a different direction depending on the level of analysis. At the individual level, as democracy matures, partisans become more sophisticated and gain the ability to judge the government more on the basis of policy outcomes. Of course, this does not mean that partisans use government performance as a voting criterion as often as non-partisans do; partisans still put less emphasis on performance issues and rely more on their partisan attachments than non-partisans. As they become more experienced with how elections are supposed to work, however, the gap between partisans and non-partisans in the strength of electoral accountability should narrow.

Conversely, at the aggregate level, the degree to which partisanship hurts accountability increases as democracy becomes consolidated. In new democracies where the translation of social cleavages into political cleavages has not been established and party supply is unstable, increasing partisanship could rather enhance electoral accountability by contributing to party system institutionalization and electoral stability. For new democracies, thus, it is important to build strong ties between citizens and parties and make party elites aware of the risk of losing their support bases in the aftermath of such strategic decisions as dissolving existing parties and creating new ones in pursuit of their political interests. As democracy matures and partisanship at the society level reaches a certain point, however, any increase in partisanship could rather weaken the performance–vote linkage by allowing political parties to rely overly on their consolidated supporters who cast ballots irrespective of party performance. In this case, therefore, preventing party systems from being overly institutionalized is also an important political challenge.

Footnotes

Appendix A

1.

Anderson, C. J. “The End of Economic Voting? Contingency Dilemmas and the Limits of Democratic Accountability,” Annual Review of Political Science 10 (2007): 271–96; V. O. Key, The Responsible Electorate (New York: Vintage, 1966); A. Przeworski, S. Stokes, and B. Manin, Democracy, Accountability, and Representation (Cambridge: Cambridge University Press, 1999).

2.

Key, The Responsible Electorate.

3.

Anderson, “The End of Economic Voting?”

4.

In a nutshell, this article follows the classical Michigan model of partisanship, which is defined as a psychological sense of attachment that one has to a particular party (A. Campbell, P. Converse, W. Miller, and D. Stokes, The American Voter, University of Michigan, Survey Research Center [New York: Wiley, 1960]). Electoral accountability is defined as the extent to which government performance accounts for the government’s electoral fortunes.

5.

Campbell et al., The American Voter; R. J. Dalton, “Political Cleavages, Issues, and Electoral Change,” in Comparing Democracies: Elections and Voting in Global Perspective, ed. L. LeDuc, R. G. Niemi, and P. Norris (Thousand Oaks, CA: Sage, 1996); R. J. Dalton, Citizen Politics: Public Opinion and Political Parties in Advanced Industrial Democracies, 5th ed. (Washington, DC: CQ Press, 2008); R. J. Dalton, “Ideology, Partisanship, and Democratic Development,” in Comparing Democracies: Elections and Voting in the 21st Century, ed. L. LeDuc, R. G. Niemi, and P. Norris (London: Sage, 2010); M. N. Franklin, T. T. Mackie, and H. Valen, Electoral Change: Responses to Evolving Social and Attitudinal Structures in Western Countries (Cambridge: Cambridge University Press, 1992); D. P. Green, B. Palmquist, and E. Schickler, Partisan Hearts and Minds: Political Parties and the Social Identities of Voters (New Haven, CT: Yale University Press, 2002); D. Sanders, “Party Identification, Economic Perceptions, and Voting in British General Elections, 1974-97,” Electoral Studies 22, no. 2 (2003): 239–63.

6.

P. W. Shively, “The Development of Party Identification among Adults: Exploration of a Functional Model,” American Political Science Review 73, no. 4 (1979): 1040.

7.

M. S. Lewis-Beck, R. Nadeau, and A. Elias, “Economics, Party, and the Vote: Causality Issues and Panel Data,” American Journal of Political Science 52, no. 1 (2008): 84–95; S. J. Rosenstone and J. M. Hansen, Mobilization, Participation, and Democracy in America (New York: Maxwell Macmillan, 1993); E. R. Tillman, “Economic Judgments, Party Choice, and Voter Abstention in Cross-national Perspective,” Comparative Political Studies 41, no. 9 (2008): 1290–309.

8.

M. A. Kayser and C. Wlezien, “Performance Pressure: Patterns of Partisanship and the Economic Vote,” European Journal of Political Research 50 (2011): 365–94; Tillman, “Economic Judgments, Party Choice.”

9.

To name a few: Dalton, “Political Cleavages, Issues”; Dalton, “Ideology, Partisanship”; A. Gallego, G. Rico, and E. Anduiza, “Disproportionality and Voter Turnout in New and Old Democracies,” Electoral Studies 31 (2012): 159–69; Tillman, “Economic Judgments, Party Choice”; W. Van der Brug, M. Franklin, and G. Tóka, “One Electorate or Many? Differences in Party Preference Formation between New and Established European Democracies,” Electoral Studies 27 (2008): 589–600; A. Walczak, W. Van der Brug, and C. E. de Vries, “Long- and Short-Term Determinants of Party Preferences: Inter-generational Differences in Western and East Central Europe,” Electoral Studies 31 (2012): 273–84.

10.

V. Bunce, “Comparing East and South,” Journal of Democracy 6, no. 3 (1995): 91–92.

11.

M. Bernhard, “Civil Society after the First Transition: Dilemmas of Post-communist Democratization in Poland and Beyond,” Communist and Post-communist Studies 29, no. 3 (1996): 309–30; M. Bernhard and E. Karakoç, “Civil Society and the Legacies of Dictatorship,” World Politics 59, no. 4 (2007): 539–67; M. M. Howard, The Weakness of Civil Society in Post-communist Europe (Cambridge: Cambridge University Press, 2003); G. Pop-Eleches, “Communist Development and the Postcommunist Democratic Deficit,” in Historical Legacies of Communism, ed. M. K. Beissinger and S. Kotkin (New York: Cambridge University Press, 2014); Rosenstone, Mobilization, Participation, and Democracy in America; J. A. Tucker, Regional Economic Voting: Russia, Poland, Hungary, Slovakia and the Czech Republic, 1990-1999 (New York: Cambridge University Press, 2006).

12.

J. Bielasiak, “The Institutionalization of Electoral and Party Systems in Postcommunist States,” Comparative Politics 34 (2002): 189–210; B. Epperly, “Institutions and Legacies: Electoral Volatility in the Postcommunist World,” Comparative Political Studies 44, no. 7 (2011): 829–53; E. N. Powell and J. A. Tucker, “Revisiting Electoral Volatility in Post-communist Countries: New Data, New Results and New Approaches,” British Journal of Political Science 44, no. 1 (2014): 123–47; M. Tavits, “The Development of Stable Party Support: Electoral Dynamics in Post-communist Europe,” American Journal of Political Science 49, no. 2 (2005): 283–98.

13.

There is a concern using the concept of partisanship across different political contexts, especially between two-party presidential systems and multi-party parliamentary systems. First, Garry (“Making ‘Party Identification’ More Versatile: Operationalising the Concept for the Multiparty Setting,” Electoral Studies 26, no. 2 [2007]: 346–58) and Weisberg (“Political Partisanship,” in Measures of Political Attitudes, ed. J. Robinson, P. Shaver, and L. Wrightsman [San Diego, CA: Academic Press, 1999], 681–729) contend that voters for the multi-party setting may be involved with multiple identities leading to multiple partisan attachments. Reflecting this consideration, I count up to three parties, if available, listed by the CSES respondents for the parties they feel close to as their party identities. Second, Rosema (“Partisanship, Candidate Evaluations, and Prospective Voting,” Electoral Studies 25 [2006]: 533–46) argues that partisanship conceptualized by the Michigan scholars under the US system may not be applicable to the European context because it is “doubted whether party identification could be distinguished meaningfully from vote choice” in parliamentary systems where “the competition is not primarily between candidates, but between parties” (469–71). However, my own research based on the first three modules of the CSES shows that even taking into account all three choices for the attached party in the survey, a smaller percentage of the respondents in the Western parliamentary systems answered their preferred parties as their vote choice for the party-list elections (15.9 percent) than those in the United States for the presidential elections (24.4 percent). This reveals that party attachments and vote choices may be different constructs even in parliamentary systems.

14.

J. Aldrich, Why Parties? The Origin and Transformation of Political Parties in America (Chicago: University of Chicago Press, 1995); Campbell et al., The American Voter; L. M. Bartels, “Beyond the Running Tally: Partisan Bias in Political Perceptions,” Political Behavior 24, no. 2 (2002): 117–50; T. Brader and J. A. Tucker, “Pathways to Partisanship: Evidence from Russia,” Post-Soviet Affairs 24, no. 3 (2008): 263–300; T. Brader, J. A. Tucker, and D. Duell, “Which Parties Can Lead Opinion? Experimental Evidence on Partisan Cue Taking in Multiparty Democracies,” Comparative Political Studies 46, no. 11 (2013): 1–33.

15.

Bartels, “Beyond the Running Tally”; A. S. Gerber and G. A. Huber, “Partisanship, Political Control, and Economic Assessments,” American Journal of Political Science 54, no. 1 (2010): 153–73; M. S. Lewis-Beck, “Does Economics Still Matter? Econometrics and the Vote,” Journal of Politics 68, no. 1 (2006): 208–12; M. S. Lewis-Beck, R. Nadeau, and A. Elias, “Economics, Party, and the Vote: Causality Issues and Panel Data,” American Journal of Political Science 52, no. 1 (2008): 84–95.

16.

Brader and Tucker, “Pathways to Partisanship: Evidence from Russia,” 264.

17.

T. Brader and J. A. Tucker, “The Emergence of Mass Partisanship in Russia, 1993-1996,” American Journal of Political Science 45, no. 1 (2001): 69–83; Dalton, “Ideology, Partisanship, and Democratic Development”; Franklin et al., Electoral Change; S. Mainwaring and T. Scully, Building Democratic Institutions: Party Systems in Latin America (Stanford, CA: Stanford University Press, 1995).

18.

Campbell et al., The American Voter, chap. 2.

19.

Dalton, Citizen Politics.

20.

Dalton, “Political Cleavages, Issues, and Electoral Change”; Dalton, “Ideology, Partisanship, and Democratic Development”; Franklin et al., Electoral Change; M. P. Wattenberg, The Decline of American Political Parties: 1952-1964 (Cambridge: Harvard University Press, 1996).

21.

L. M. Bartels, “Partisanship and Voting Behavior, 1952-1996,” American Journal of Political Science 44 (2000): 35–50; Sanders, “Party Identification, Economic Perceptions, and Voting in British General Elections, 1974-97.”

22.

C. Wlezien, M. Franklin, and D. Twiggs, “Economic Perceptions and Vote Choice: Disentangling the Endogeneity,” Political Behavior 19, no. 1 (1997): 7–17; C. Anderson, S. Mendes, and Y. Tverdova, “Endogenous Economic Voting: Evidence from the 1997 British Election,” Electoral Studies 23 (2004): 683–708; G. Evans and R. Anderson, “The Political Conditioning of Economic Perceptions,” Journal of Politics 68, no. 1 (2006): 194–207; G. Evans and M. Pickup, “Reversing the Causal Arrow: The Political Conditioning of Economic Perceptions in the 2000-2004 U.S. Presidential Election Cycle,” Journal of Politics 72, no. 4 (2010): 1236–51; Gerber and Huber, “Partisanship, Political Control, and Economic Assessments.”

23.

Lewis-Beck, “Does Economics Still Matter?”; Lewis-Beck et al., “Economics, Party, and the Vote.”

24.

The size of the effect of A2 can be termed here as the level of electoral accountability, or the strength of the economic vote in a special case where government performance is evaluated specifically on the matters of the economy. This article thus uses the terms of electoral accountability and, if relevant, economic voting to indicate the arrow A2, the linkage between government evaluation (or economic performance), and vote choice.

25.

Wlezien et al., “Economic Perceptions and Vote Choice.”

26.

Anderson et al., “Endogenous Economic Voting.”

27.

Evans and Anderson, “The Political Conditioning of Economic Perceptions”; Evans and Pickup, “Reversing the Causal Arrow.”

28.

W. Van der Brug, C. Van der Eijk, and M. Franklin, The Economy and the Vote: Economic Conditions and Elections in Fifteen Countries (Cambridge: Cambridge University Press, 2007).

29.

Tillman, “Economic Judgments, Party Choice.”

30.

Kayser and Wlezien, “Performance Pressure.”

31.

Ibid., 376.

32.

To address the independent role of the real economy in isolation from that of subjective perception in predicting voting decision, I included annual percent change in GDP (one-year lagged) into my models. The data was taken from the World Bank database.

33.

P. C. Ordeshook, “Institutions and Incentives,” Journal of Democracy6 (1995): 46–60; P. Reddaway, “Instability and Fragmentation,” Journal of Democracy 5 (1994): 13–19; S. White, R. Rose, and I. McAllister, How Russia Votes (Chatham, NJ: Chatham House, 1997).

34.

Brader and Tucker, “Pathways to Partisanship”; G. Evans and S. Whitefield, “The Structuring of Political Cleavages in Post-communist Societies: The Case of the Czech Republic and Slovakia,” Political Studies 46, no. 1 (1998): 115–39; H. Kitschelt, et al. Post-communist Party Systems: Competition, Representation, and Inter-party Cooperation (Cambridge: Cambridge University Press, 1999); I. McAllister and S. White, “Political Parties and Democratic Consolidation in Post-communist Societies,” Party Politics 13, no. 2 (2007): 197–216.

35.

T. Brader and J. A. Tucker, “Following the Party’s Lead: Party Cues, Policy Opinion, and the Power of Partisanship in Three Multiparty Systems,” Comparative Politics 44, no. 4 (2012): 403–20; T. Brader, J. A. Tucker, and D. Duell, “Which Parties Can Lead Opinion? Experimental Evidence on Partisan Cue Taking in Multiparty Democracies,” Comparative Political Studies 46, no. 11 (2013): 1–33; H. E. Hale, Why Not Parties in Russia? Democracy, Federalism, and the State (New York: Cambridge University Press, 2007); G. Pop-Eleches and J. A. Tucker, “Communism’s Shadow: Postcommunist Legacies, Values, and Behavior,” Comparative Politics 43, no. 4 (2011): 379–408.

36.

M. Harper, “Economic Voting in Postcommunist Eastern Europe,” Comparative Political Studies 33, no. 9 (2000): 1191–1227; A. C. Pacek, “Macroeconomic Conditions and Electoral Politics in East Central Europe,” American Journal of Political Science 38, no. 3 (1994): 723–44; D. V. Powers and J. H. Cox, “Echoes from the Past: The Relationship between Satisfaction with Economic Reforms and Voting Behavior in Poland,” American Political Science Review 91, no. 3 (1997): 617–33.

37.

R. M. Duch, “A Developmental Model of Heterogeneous Economic Voting in New Democracies,” American Political Science Review 95, no. 4 (2002): 895–910.

38.

J. Fidrmuc, “Economics of Voting in Post-communist Countries,” Electoral Studies 19, no. 2/3 (2000): 199–217; Tucker, Regional Economic Voting.

39.

J. Zielinski, K. M. Slomczynski, and G. Shabad, “Electoral Control in New Democracies: The Perverse Incentives of Fluid Party Systems,” World Politics 57, no. 3 (2005): 365–95.

40.

M. Bernhard and E. Karakoç, “Moving West or Going South? Economic Transformation and Institutionalization in Postcommunist Party Systems,” Comparative Politics 44, no. 1 (2011): 1–20.

41.

D. Bell, “The Resumption of History in the New Century,” in The End of Ideology, rev. ed., ed. D. Bell (Cambridge: Cambridge University Press, 2000); O. Knutsen, Class Voting in Western Europe: A Comparative Longitudinal Study (Lanham, MD: Lexington, 2006); Walczak et al., “Long- and Short-Term Determinants of Party Preferences.”

42.

S. M. Lipset and S. Rokkan, Party Systems and Voter Alignments: Cross-national Perspectives (New York: Free Press, 1967).

43.

Ibid.

44.

A. Abramowitz and S. Webster, “The Rise of Negative Partisanship and the Nationalization of U.S. Elections in the 21st Century,” Electoral Studies 41 (2016): 12–22; S. Iyengar, G. Sood, and Y. Lelkes, “Affect, Not Ideology: A Social Identity Perspective on Polarization,” Public Opinion Quarterly 76, no. 3 (2012): 405–31; L. Mason, “‘I Disrespectfully Agree’: The Differential Effects of Partisan Sorting on Social and Issue Polarization,” American Journal of Political Science 59, no. 1 (2015): 128–45.

45.

K. Arceneaux, “Can Partisan Cues Diminish Democratic Accountability?,” Political Behavior 30 (2008): 139–60; J. Huber, G. Kernell, and E. Leoni, “Institutional Context, Cognitive Resources and Party Attachments across Democracies,” Political Analysis 13, no. 4 (2005): 365–86; H. Lavine and T. Gschwend, “Issues, Party and Character: The Moderating Role of Ideological Thinking on Candidate Evaluation,” British Journal of Political Science 37 (2006): 139–63; Shively, “The Development of Party Identification among Adults.”

46.

Fidrmuc, “Economics of Voting in Post-communist Countries,” 200.

47.

R. Dix, “Cleavage Structures and Party Systems in Latin America,” Comparative Politics 22, no. 1 (1989): 23–37; S. P. Huntington, Political Order in Changing Societies (New Haven, CT: Yale University Press, 1968); R. Rose and W. Mishler, “A Supply-Demand Model of Party-System Institutionalization: The Russian Case,” Party Politics 16, no. 6 (2010): 801–21.

48.

Franklin et al., “Electoral Change”; P. Mair, “The Freezing Hypothesis: An Evaluation,” in Party Systems and Voter Alignments Revisited, ed. L. Karvonen and S. Kuhnle (New York: Routledge, 2001); M. Tavits, Post-communist Democracies and Party Organization (New York: Cambridge University Press, 2013).

49.

I decided not to use the first and fourth waves of the CSES as the question regarding the general performance of the government was not asked; they only surveyed how the respondents evaluate their country’s economy over the past twelve months. Although the economy is one of the important criteria with which the government is judged, I find the general performance question that makes direct reference to the government more relevant to the research question of this study. The full list of the CSES electoral studies used in the individual-level analysis is found in .

50.

I exclude those postcommunist countries from both levels of analyses that have not been consistently classified as “Free” for the period analyzed here by the Freedom House, namely, Albania, Belarus, Kyrgyzstan, Russia, and Ukraine, because comparing these countries to established democracies might be irrelevant. In addition, the 1997 Lithuanian study is also dropped from the individual-level analysis as the then incumbent party, the Democratic Labor Party of Lithuania, did not put up any candidate for the 1997 presidential elections, which makes voting for the incumbent infeasible.

51.

I. Alcañiz and T. Hellwig, “Who’s to Blame? The Distribution of Responsibility in Developing Democracies,” British Journal of Political Science 41 (2010): 389–411; S. Rabe-Hesketh and A. Skrondal, Multilevel and longitudinal Modeling Using Stata: Volume I (College Station, TX: Stata Press, 2012).

52.

One might suggest to employ three-level modeling taking individual respondents as a first level, countries as second, and survey modules as third because of the structure of the CSES data. However, the fact that this analysis only takes two modules renders three-level modeling infeasible. To control for the possible effects of different modules, I include a dummy variable indicative of the Module 3 respondents.

53.

I use cmp in Stata which is designed to fit a large family of multi-equation and multi-level estimators (D. Roodman, “Estimating Fully Observed Recursive Mixed-Process Models with cmp,” Stata Journal 11, no. 2 [2011]: 160), fulfilling the requirements of the tests here to take into account both endogeneity among variables and multi-level nature of the data.

54.

A list of the incumbent parties in all country-elections is presented in .

55.

In case of multiple partisanship, it is questionable into which partisan group, incumbent or opposition, those who claim partisanship towards both incumbent and opposition parties may be categorized. Both can be theorized. Since those with multiple partisanship have alternatives to which their vote choice can be switched when their incumbent evaluation is not positive, they can be assumed to perform similarly as opposition partisans. Contrarily, as they anyway have attachments towards incumbent parties, they could behave similarly as incumbent partisans. As I find the latter more convincing, I regard them as incumbent partisans for this analysis. My robustness tests with the former categorization produce more or less the same results not altering any major findings of the analysis.

56.

Following Braumoeller’s (“Hypothesis Testing and Multiplicative Interaction Terms,” International Organization 58, no. 4 [2004]: 807–20) advice that “in any interaction of k independent variables, a full set of $\sum_{k = 1}^{n} (\begin{matrix} n \\ k \end{matrix})$ coefficients must be estimated” to avoid biased estimation (811), all factorial variables of the composite variables are tested, though their effects are not reported here if not relevant.

57.

The literature of the “clarity of responsibility” (B. G. Powell and G. D. Whitten, “A Cross-national Analysis of Economic Voting: Taking Account of the Political Context,” American Journal of Political Science 37, no. 2 [1993]: 391–414) has claimed that the evaluation–vote linkage should weaken when voters find it difficult who is accountable for policy outcomes, such as under minority or coalition governments. As this study is cross-national, it may be desirable to take into account such varying levels of clarity between country-years. I, however, did not test the clarity effects in my models because, first, the regressions already employ multiple-way interactions, ending up with too many combinations of the composite variables to be tested (fifteen combinations in the case of the regional-comparison model), thus adding one more composite variable, the clarity one, complicates the regression results too excessively by generating thirty-one combinations, most of which are irrelevant to the analysis, and second, even my own preliminary test for the separate effect of the clarity (using various indicators of the government types in accordance with the previous studies) did not produce any significant or correctly signed coefficients for my sample, meaning that the clarity thesis may not be applicable to this particular set of times and countries.

58.

For every value of government evaluation, the predicted probabilities of voting for incumbent parties are calculated and averaged from all combinations of the values of the other variables in the model.

59.

To see how broadly my theories can be empirically cast, as they speak to a general relationship regarding voting behavior at least at the individual level, I expanded my sample to all democracies in the CSES data and ran the same model with the following twenty-three cases added: Brazil (2002, 2006, 2010), Chile (2005), Spain (2004, 2008), Greece (2009), Israel (2006), Japan (2004, 2007), South Korea (2004, 2008), Mexico (2003, 2006, 2009), Philippine (2004, 2010), Portugal (2002, 2005, 2009), Turkey (2011), Uruguay (2009), and South Africa (2009). The model produced more or less the same results: the interactive effect of incumbent partisanship and incumbent evaluation on incumbent vote ( $β_{22}$ ) is −0.105*** and that of opposition partisanship ( $β_{23}$ ) is −0.080***. This shows that the restrictive effect of partisanship on the evaluation–vote linkage is a more generally applicable relationship beyond the Western and postcommunist regions.

60.

I employ feasible generalized least squares (FGLS) estimation method using xtgls command in Stata.

61.

Kayser and Wlezien, “Performance Pressure”; Van der Brug et al., The Economy and the Vote.

62.

For the aggregate-level analysis, three election cases are added, Ireland (2011), New Zealand (2011), and Poland (2011), which were within the time frame of this analysis but dropped from the individual-level analysis as their CSES surveys were a part of the fourth wave making it impossible to construct the consistent variable for incumbent evaluation as noted above. As the macroeconomic data for those cases are available, I included them in the aggregate-level models. The number of the cases also decreases by one in comparison to the individual-level sample in which there are two separate CSES election studies for Germany in 2002 as two different survey methods, mail-back and telephone, were used. As for the aggregated partisanship, I averaged the results of the two studies.

63.

A. Bell and K. Jones, “Explaining Fixed Effects: Random Effects Modelling of Time-Series Cross-sectional and Panel Data,” Political Science Research and Methods 3, no. 1 (2015): 133–53; T. Clark and D. Linzer, “Should I Use Fixed or Random Effects?,” Political Science Research and Methods 3, no. 2 (2015): 399–408.

64.

Given that the president, as the head of government, is held more responsible for policy outcomes over the legislature in presidential systems, I use presidential elections for France, Russia, and the United States.

65.

I include vote shares for incumbent parties in the previous elections (at t – 1) to control for auto-correlation.

66.

F. Grotz and T. Weber, “Party Systems and Government Stability in Central and Eastern Europe,” World Politics 64, no. 4 (2012): 699–740; T. Hellwig and D. Samuels, “Electoral Accountability and the Variety of Democratic Regimes,” British Journal of Political Science 38 (2008): 65–90; D. Jones, “Partisan Polarization and Congressional Accountability in House Elections,” American Journal of Political Science 54, no. 2 (2010): 323–37; A. Roberts, “Hyperaccountability: Economic Voting in Central and Eastern Europe,” Electoral Studies 27 (2008): 533–46.

67.

L. M. Bartels, “Electoral Continuity and Change, 1868–1996,” Electoral Studies 17, no. 3 (1998): 301–26; N. Lupu and S. Stokes, “Democracy, Interrupted: Regime Change and Partisanship in Twentieth-Century Argentina,” Electoral Studies 29, no. 1 (2010): 91–104.

68.

M. Pederson, “Changing Patterns of Electoral Volatility in European Party Systems, 1948-1977,” in Western European Party Systems: Continuity and Change, ed. H. Daalder and P. Mair (Beverly Hills, CA: Sage, 1983), 29–66.

69.

Anderson, “The End of Economic Voting?”; M. S. Lewis-Beck and M. Stegmaier, “Economic Determinants of Electoral Outcomes,” Annual Review of Political Science 3 (2000): 183–219; Van der Brug et al., The Economy and the Vote.

70.

Fidrmuc, “Economics of Voting in Post-communist Countries”; Roberts, “Hyperaccountability”; Tucker, Regional Economic Voting.

71.

The data are taken from the World Bank database (accessed on June 20, 2015).

72.

E. K. Scheuch, “Theoretical Implications of Comparative Survey Research: Why the Wheel of Cross-cultural Methodology Keeps on Being Reinvented,” International Sociology 4, no. 2 (1989): 155.

73.

Tillman, “Economic Judgments, Party Choice.”

74.

Kayser and Wlezien, “Performance Pressure.”

Dong-Joon Jung is a senior researcher at the Institute for Peace and Unification Studies (IPUS) at Seoul National University, South Korea. He received his PhD in political science from the University of Florida in 2015. With a regional focus on post-communist Europe, his research interests center upon democratization and regime transitions, comparative voting behavior, parties and elections, civil society, political attitudes and gender.