Does (dis)liking the governing parties correlate with democratic (dis)satisfaction?

Abstract

Lodge and Taber claim that likeability heuristics influence political evaluations. We shed new light on this theory by examining whether (dis)liking the governing parties correlates with democratic (dis)satisfaction. We run hierarchical linear models with data from Modules 2, 3, 4 and 5 of the Comparative Study of Electoral Systems. We show that individuals who like the governing parties are more satisfied with democracy, even when controlling for winner/loser status and ideological congruence. We also show that political sophistication strengthens the correlation between (dis)liking the governing parties and (dis)satisfaction with democracy. This study provides new insights into the connection between likeability heuristics and political evaluations.

Introduction

In The Rationalizing Voter, Lodge and Taber (2013) argue that likes and dislikes serve as powerful heuristics guiding political evaluations (see also Capelos, 2010; Harsgor and Nevitte, 2022; Sanchez and Dunning, 2021; Sniderman et al., 1991; Uslaner, 2015). Although highly influential within the political behavior literature (Arceneaux and Vander Wielen, 2017; Guess and Coppock, 2020; Jerit and Zhao, 2020; Kiil, 2025; Kim S-Y et al., 2010; Kraft et al., 2015; Pérez, 2016; Redlawsk and Pierce, 2018; Roseman et al., 2020), scholars have not yet fully examined Lodge and Taber’s theory in relation to a central attitude in the field: democratic (dis)satisfaction (Linde and Ekman, 2003; Lipset, 1959). From this perspective, this study has two primary—and complementary—goals: first, it examines whether likes and dislikes towards the governing parties shape democratic (dis)satisfaction. ¹ Second, it explores the moderating role of political sophistication (Lodge and Taber, 2013). Through these analyses, this study not only provides greater context to the heightened affective polarization (Heltzel and Laurin, 2020) and eroding democratic legitimacy (Angiolillo et al., 2024; Dalton, 2004; Linde and Ekman, 2003; Lipset, 1959), but offers new knowledge on the factors motivating democratic (dis)satisfaction (Linde and Ekman, 2003; Lipset, 1959).

This study addresses its two research objectives using cross-national data from Modules 2, 3, 4, and 5 of the Comparative Study of Electoral Systems (CSES). ² It reveals two important patterns. First, individuals who like the governing parties express more satisfaction with democracy, even after controlling for winner/loser status (Blais and Gélineau, 2007) (see Appendix A for additional references) and ideological congruence (Kim M, 2009) as well as accounting for potential reverse causation through panel data (Kartal, 2024) (see Appendix A for additional references). Second, consistent with Jerit and Davies (2018), Lodge and Taber (2005, 2013), McGraw (2018) and Vitriol et al. (2023), political sophistication strengthens the link between (dis)liking and (dis)satisfaction.

From this perspective, our study contributes to the literature in five meaningful ways. First, it offers new insights into citizens’ tendency to act as pragmatic cognitive misers (Popkin, 1991; Stroh, 1995). Second, it extends Erhardt (2023) and Ridge (2022) by using data from 34 countries and a measure that captures (dis)like towards all coalition parties. ³ Third, it advances knowledge on important correlates of democratic (dis)satisfaction, such as populist radical right voting (Feitosa et al., 2025; Goovaerts et al., 2020) (but see Dassonneville and McAllister, 2023), the dissemination of false information (Humprecht, 2025) and the support for democratic institutions (Claassen and Magalhães, 2022) and principles (Graham and Svolik, 2020). Fourth, this study highlights the negative effects of populist parties’ attacks on governing parties (Heinisch, 2003; Immerzeel and Pickup, 2015), especially their potential contribution to the erosion of democratic legitimacy (Linde and Ekman, 2003; Lipset, 1959). Fifth, it broadens our understanding of motivated reasoning and partisan bias beyond the traditional focus on prior beliefs (Bayes and Druckman, 2021) and partisan identities (Greene, 1999).

Our study proceeds as follows: first, we review the reasons for a connection between (dis)liking the governing parties and (dis)satisfaction with democracy. Subsequently, we consider the potential moderating effect of political sophistication. We then discuss the data, variables and methods used in the analysis. We continue with a presentation of the results and conclude with a summary of our main findings and a discussion of avenues for future research.

Likeability heuristics and political evaluations

As outlined earlier, Lodge and Taber (2013) provide strong theoretical grounds for a link between (dis)liking the governing parties and democratic (dis)satisfaction. A central claim in Lodge and Taber’s work is that individuals’ likes and dislikes—even when incidental—powerfully shape how they evaluate political objects (see also Damasio, 1999). The authors identify two distinct yet mutually reinforcing processes through which this influence unfolds. The first, affect transfer, denotes a simple, immediate and spontaneous impact of likes and dislikes on evaluative judgments (Jennings et al., 2024; Lodge and Taber, 2013; Schwarz and Clore, 1988). ⁴ The second, affect contagion, captures a more gradual, indirect and mediated pathway through which likes and dislikes shape political evaluations (Cassino et al., 2007; Clore and Isbell, 2001). ⁵

Lodge and Taber (2013) therefore provide a well-grounded framework for understanding how feelings towards governing parties may shape evaluations of democracy: through affect transfer and affect contagion, likes and dislikes towards such parties may color judgments of the democratic system. Lodge and Taber’s account of likes and dislikes—particularly their contribution to political evaluations—further converges with two major research traditions. First, it aligns with scholarship emphasizing the role of shortcuts—including likes and dislikes (Abelson et al., 1982; Capelos, 2010; Marcus, 1988; Schafer, 2000; Sniderman et al., 1991)—in opinion formation and decision-making (Cutler, 2002; Hobolt, 2007; McDermott, 2005). Second, it resonates with theories of motivated reasoning and partisan bias. Specifically, by influencing political evaluations, likes and dislikes may give rise to these two processes. ⁶

Even so, systematic empirical evidence on the relationship between (dis)liking and (dis)satisfaction remains limited. Although Erhardt (2023) and Ridge (2022) engage with this relationship, their contributions are nonetheless indirect. From this perspective, this study extends existing research by conducting a direct examination of the (dis)liking–(dis)satisfaction nexus. It further advances the scholarship on four key fronts. First, it draws on data from multiple countries rather than a single case. Second, it includes controls for winner/loser status, ideological congruence with governing parties and identification with these parties, as well as study-level confounders by means of study fixed effects. Third, following the literature on attitudes towards government (Dassonneville et al., 2021; Finke, 2009; Reher, 2018), it relies on a measure of (dis)like that accounts for all governing parties. Fourth, it examines the moderating role of political sophistication.

To be sure, our focus on political sophistication as a moderating factor is not arbitrary; it follows a well-established line of research. Lodge and Taber (2013) contend, for instance, that individuals who possess the cognitive capacity to reflect on their opinions generally become even less receptive to disconfirming information (see also Lodge and Taber, 2005; McGraw, 2018; Valli et al., 2023; Vitriol et al. (2023); but see Tappin et al., 2020, 2021). Similarly, Kahan (2013) affirms that individuals who are better equipped to engage in deliberate, effortful reasoning are particularly susceptible to motivated biases. Jerit and Barabas (2012) posit, as well, that those operating in information-rich environments exhibit a greater tendency to absorb congenial content while resisting opposing claims (see also Nyhan and Reifler, 2010; but see Hart et al., 2011; Jerit and Barabas, 2012; Luskin, 1990; Ottati and Isbell, 1996; Prior et al., 2015).

Political sophistication moderates the relationship between (dis)liking and (dis)satisfaction through at least three mechanisms. First, political sophistication enhances individuals’ capacity to produce counter-arguments when exposed to uncongenial information (Taber et al., 2009). Second, political sophistication exposes individuals to large volumes of information, which would demand substantial time and cognitive effort for conscious deliberation (Lodge and Taber, 2013; see also Cutler, 2002). Third, and of particular relevance, political sophistication facilitates affect transfer and affect contagion by strengthening the link between likes and dislikes and the political objects they target (Lodge and Taber, 2005).

Thus, we hypothesize that:

• Hypothesis 1. (Dis)liking the governing parties correlates with democratic (dis)satisfaction.

• Hypothesis 2. Political sophistication strengthens the correlation between (dis)liking the governing parties and democratic (dis)satisfaction.

Data and variables

We assess Hypotheses 1 and 2 by means of data from Modules 2, 3, 4 and 5 of the CSES. Although the cross-sectional nature of the data precludes causal inference, it still provides insights into descriptive patterns, which constitute a critical step in knowledge building (Gerring, 2004; Lupu and Warner, 2022). It also enables a real-world assessment of the link between (dis)liking and (dis)satisfaction (Coulombe et al., 2024), without relying on artificial, researcher-driven and/or context-specific interventions (Abelson et al., 1982; Gadarian and Lau, 2011). Furthermore, by covering 34 countries with a Polity IV score of 6 or higher (Farrer and Zingher, 2019; Hernández et al., 2021; Nemčok and Wass, 2021; Ridge, 2022), the CSES data enable a systematic, large-scale examination of the relationship between (dis)liking and (dis)satisfaction.

Independent variable

We measure (dis)liking the governing parties—our main independent variable—by means of party like and dislike questions (Garretson, 2018; Kim S-Y et al., 2010; Lodge and Taber, 2013) (see Appendix B for a discussion on implicit measures). ⁷ To account for parties’ different weights in a governing coalition (Fortunato et al., 2021; Plescia and Kritzinger, 2018), we follow Blais and Bodet (2006) (see Appendix A for additional references) and multiply party (dis)likes by the number of portfolios that each party holds. ⁸ Although a different approach would also take into account the relative importance of each portfolio, it is unclear how this could be calculated in an objective, non-discretionary way. Furthermore, research highlights that weighted portfolio shares closely resemble their original, unweighted values (Warwick and Druckman, 2001). To enable comparisons with other independent variables, we standardize the weighted values to a mean of zero and a standard deviation of one (see the distribution of the resulting (dis)liking variable in Figure C1 of Appendix C; see also Table C1 for summary statistics).

Dependent variable

We assess democratic (dis)satisfaction—our dependent variable—by means of the question: ‘On the whole, are you very satisfied, fairly satisfied, not very satisfied, or not at all satisfied with the way democracy works in [COUNTRY]?’ The maximum value (4) corresponds to respondents who report being very satisfied, and the minimum value (1) corresponds to those who report being not at all satisfied (see Figure C2 of Appendix C for the distribution of responses to this question; see Appendix A for works using the same question or similar versions).

Moderating variable

We operationalize political sophistication—our moderating variable—using an index combining political interest, education and attention to political matters. ⁹ Political interest assesses an internal motivation to learn about politics. As Luskin (1990) explains, individuals with a stronger interest in politics are more likely to notice the political information they encounter and to engage with it more thoughtfully. Education gauges exposure to political information, as longer schooling tends to result in greater overall exposure to political content (Luskin, 1990). It also reflects the development of cognitive skills that facilitate the comprehension and retention of political information (Carpini and Keeter, 1996) (see Appendix A for additional references). Like education, attention to political matters relates to the acquisition of political information. We nevertheless include it in the index, as it more directly captures that dimension (see Gallina (2023) for a similar approach). Principal component analyses further indicate that political interest, education and attention to political matters all contribute to the political sophistication index (see Appendix B for information on the operationalization and coding of each item as well as on the creation of the political sophistication index).

Control variables

We control for six potential correlates of both the main independent and dependent variables: winner/loser status, ideological congruence with the governing parties, identification with these parties, gender, education and age. ¹⁰ In line with Anderson and Guillory (1997), Dahlberg and Linde (2017), Esaiasson (2011), Kern and Kölln (2022), Nadeau et al. (2023), Plescia (2019) and Singh (2014), we measure winner/loser status by means of a vote choice question. Consistent with Anderson and Guillory (1997), Esaiasson (2011), Kern and Kölln (2022), Nadeau et al. (2023) and Stiers et al. (2018), ‘2’ denotes individuals who voted for a party that entered government, ‘1’ denotes abstention and ‘0’ denotes voting for a party that did not enter government (the reference group). ¹¹ We compute ideological congruence by calculating the distance between individuals’ ideological self-placement and the average left–right position of the governing parties (Best and Seyis, 2021; Kim M, 2009; Mayne and Hakhverdian, 2017). Following Blais and Bodet (2006) and Ferland (2021), and to ensure consistency with the (dis)liking measure, we weight party positions by the number of portfolios each party holds. As in Aldrich et al. (2020) and Ridge (2022), we operationalize identification with governing parties using two party closeness questions: ‘Do you usually think of yourself as close to any particular party? Which party do you feel closest to?’, where ‘2’ refers to respondents who report feeling close to a governing party, ‘1’ refers to those who report no party closeness and ‘0’ refers to those who report feeling close to a party that is not in government (the reference group) (see Lambert et al., 1986; Thomson, 2017; Thomson and Brandenburg, 2019; Weisberg, 2002). We code respondents’ gender as a binary variable, where ‘0’ corresponds to men and ‘1’ to women. We identify respondents’ age by means of a continuous variable.

Models

To examine the correlation between (dis)liking and (dis)satisfaction, we rely on hierarchical linear models (HLMs) with study fixed effects. ¹² Through this approach, we account for the nested structure of the data (Fullerton et al., 2010) and control for any unmeasured study-level factor that could potentially bias estimates (Allison, 2009). By using linear models, we further provide coefficients that can be meaningfully compared across models (Allison, 1999; Karlson et al., 2012), are more directly interpretable (Bergan et al., 2022; Huang, 2022; Nadeau et al., 2023) and do not vary with other variables’ values (Berry et al., 2010). Nonetheless, in a series of robustness checks, we depart from this baseline modeling strategy by estimating an ordinal logit model, incorporating countries as a third level of analysis or as fixed effects, and allowing the effect of (dis)liking to vary through a random slope. As shown below, the results remain substantively unchanged.

To further evaluate the robustness of the results, we estimate six supplementary models. The first restricts the data to CSES studies featuring a midpoint option (‘neither satisfied nor dissatisfied’). The second focuses on CSES studies conducted after government formation, as identified by ParlGov and complementary sources like IPU Parline. The third incorporates an interaction term between (dis)liking and interview date. The fourth features a trichotomous measure of (dis)liking, where ‘1’ refers to those with a score between 0 and 0.49, ‘0’ refers to those with a score of 0.5 (the reference category) and ‘2’ refers to those with a score between 0.51 and 1. ¹³ The fifth makes use of a generalized additive mixed effect estimator, allowing the connection between (dis)liking and (dis)satisfaction to take any functional form. The sixth relies on panel data from the 2015 Canadian federal elections, the 2012 French presidential elections and the 2011 Spanish federal elections. Since these elections resulted in power shifting from the governing to the former opposition parties, the influence of (dis)satisfaction (measured after the elections) on (dis)liking (measured before the elections) should be minimal or even negative (Baccini and Urpelainen, 2014; Conyon and He, 2014; Eveland et al., 2005; Kartal, 2024; Lehoucq and Pérez-Liñán, 2014; Lenz, 2009; Steinberg and Malhotra, 2014; Vliegenthart and Lefevere, 2018). As expected, the correlation between (dis)liking and (dis)satisfaction stands up to these tests. It also strengthens—rather than weakens—over time and is not primarily driven by an effect of disliking.

After we examine the main hypothesis—(dis)liking correlates with (dis)satisfaction—we turn to the follow-up question of whether political sophistication moderates this association. We investigate this aspect through conventional moderation tests: specifically, we regress (dis)satisfaction on (dis)liking, political sophistication and their interaction. For consistency with the other tests, we rely on a hierarchical linear model. Yet we also estimate a hierarchical ordinal logit model and present the results as a robustness check. This additional test yields consistent results.

Results

Before delving into the regression results, we ran descriptive analyses in search of evidence in favor of or against our expectations. Although descriptive analyses do not account for potential confounders, they can nonetheless illuminate patterns that might otherwise be overlooked by researchers (Clayton, 2014). They do so without relying on the assumptions underlying inferential models (Angrist and Pischke, 2009). For these reasons, researchers often begin with descriptive techniques before turning to regression analysis (Haden, 2019).

Figure D1 of Appendix D reports the connection between (dis)liking and (dis)satisfaction. To allow the dataset to ‘speak’ for itself, the figure displays a locally weighted scatterplot smoothing (lowess) curve. Two elements stand out. First, in line with expectations, individuals who like the governing parties are more satisfied with democracy. Second, the functional form accounting for the correlation between (dis)liking and (dis)satisfaction is almost perfectly linear, suggesting that the regression estimations can be based on a linear function. Figures D4 and D5 of Appendix D further report the correlation between (dis)liking and (dis)satisfaction among the low- and high-sophistication groups. As expected, these correlations are consistently lower among the former than the latter: 0.246 versus 0.297 for the minimum and maximum sophistication values, and 0.390 versus 0.429 for the bottom and top sophistication quintiles, respectively.

Figure 1 presents the regression results. They indicate that greater liking of the governing parties correlates with higher satisfaction with democracy. Importantly, these findings remain robust not only across ordinal logit regressions (see Figure 1) but different model specifications (see Tables E1–E4 and Figure E1 of Appendix E)—specifically, analyses including a middle category (‘neither satisfied nor dissatisfied’), only post-government formation studies, a random slope for (dis)liking the governing parties, a third level (country), country fixed effects and a time lapse between (dis)liking and (dis)satisfaction. Additional tests reveal that the coefficient for (dis)liking strengthens as the distance between the election and interview date increases, and that liking and disliking have equivalent effects on democratic (dis)satisfaction (see Appendix F for additional country-level analyses and moderation tests across different socio-economic groups).

Figure 1.

Average marginal effect (AME) of (dis)liking the governing parties on democratic (dis)satisfaction. (a) Hierarchical linear model. (b) Hierarchical ordinal logit model.

The previous results strongly suggest a correlation between (dis)liking and (dis)satisfaction. Figure 2 and Table G1 of Appendix G turn to the hypothesized moderation effect of political sophistication. They show that the (dis)liking–(dis)satisfaction nexus becomes stronger at higher levels of political sophistication. In other words, politically sophisticated individuals express greater democratic satisfaction—when they like the governing parties—than less-sophisticated ones (see Jerit and Davies (2018); Lodge and Taber (2005, 2013); McGraw (2018); Vitriol et al. (2023) for similar arguments; see also Appendix H on additional tests with the CSES political information questions).

Figure 2.

Average marginal effect (AME) of (dis)liking the governing parties on democratic (dis)satisfaction among those with the maximum and minimum values of political sophistication. (a) Hierarchical linear model. (b) Hierarchical ordinal logit model.

Conclusion

This study sought to extend the insights of Lodge and Taber (2013) by examining whether likes and dislikes also correlate with democratic (dis)satisfaction. Consistent with their framework, we showed that the more an individual likes the governing parties, the more they are satisfied with democracy, even when controlling for winner/loser status and ideological congruence. We further showed that the more politically sophisticated rely more on their (dis)like of the governing parties when judging the functioning of democracy than the less sophisticated.

It is unclear whether a correlation between (dis)liking and (dis)satisfaction is normatively good or bad. Existing research on party cues suggests that heuristics can be helpful when accurate, but harmful when they are not (Dancey and Sheagley, 2013). Following this logic, it may be that (dis)liking the governing parties leads to evaluations of democracy that are inconsistent with individuals’ views on key elements of this system, such as their representation in political decisions. Future research should examine this possibility.

Future research should also examine whether there is contextual variation in the relationship between (dis)liking and (dis)satisfaction. Although we explore this element in supplementary tests, including the moderating role of two factors (government–opposition affective polarization and aggregate political sophistication), future research should further examine this aspect of the relationship. We also leave for future research the study of ways to reduce the tendency—particularly among political sophisticates—to rely on affective considerations when evaluating democracy. For instance, would these individuals rely less on this heuristic when provided with incentives to think more carefully? What form might such incentives take? A third and final aspect left for future examinations concerns the extent to which Lodge and Taber’s theory would also apply to other kinds of political evaluations. Based on the observed findings, their theory seems to represent a powerful tool to understand what elements individuals consider when making such evaluations.

Supplemental Material

sj-pdf-1-ips-10.1177_01925121261423135 – Supplemental material for Does (dis)liking the governing parties correlate with democratic (dis)satisfaction?

Supplemental material, sj-pdf-1-ips-10.1177_01925121261423135 for Does (dis)liking the governing parties correlate with democratic (dis)satisfaction? by Fernando Feitosa, Bjarn Eck, Jean-Benoît Pilet and Éric Bélanger in International Political Science Review

Supplemental Material

sj-zip-2-ips-10.1177_01925121261423135 – Supplemental material for Does (dis)liking the governing parties correlate with democratic (dis)satisfaction?

Supplemental material, sj-zip-2-ips-10.1177_01925121261423135 for Does (dis)liking the governing parties correlate with democratic (dis)satisfaction? by Fernando Feitosa, Bjarn Eck, Jean-Benoît Pilet and Éric Bélanger in International Political Science Review