Politicians’ Theories of Voting Behavior

Abstract

While political scientists regularly engage in spirited theoretical debates about elections and voting behavior, few have noticed that elected politicians also have theories of elections and voting. Here, we investigate politicians’ positions on eight central theoretical debates in the area of elections and voting behavior and compare politicians’ theories to those held by ordinary citizens. Using data from face-to-face interviews with nearly one thousand politicians in 11 countries, together with corresponding surveys of more than twelve thousand citizens, we show that politicians overwhelmingly hold thin, minimalist, “democratic realist” theories of voting, while citizens’ theories are more optimistic and policy oriented. Politicians’ theoretical tendencies—along with their theoretical misalignment from citizens—are remarkably consistent across countries. These theories are likely to have important consequences for how politicians campaign, communicate with the public, think about public policy, and represent their constituents.

“The Labor Party is not going to profit from having these proven unsuccessful people around who are frightened of their own shadow and won’t get out of bed in the morning unless they’ve had a focus group report to tell them which side of bed to get out.”

— Paul Keating, 2007

“I could stand in the middle of Fifth Avenue and shoot somebody, and I wouldn’t lose any voters.”

— Donald Trump, 2016

INTRODUCTION

Do voters select parties that will implement their desired policies, or are they largely concerned with seeing their political team win and the other team lose? When voters support a party, do they focus on the character and competence of the party leader, or are they primarily interested in the party’s policy commitments? Are voters prospective, oriented to the future, or are they retrospective and oriented to the past? These kinds of questions are central to political science research on elections and voting behavior. The theories that researchers have developed to answer them are among the most well-known and widely debated in political science.

Elected politicians figure prominently in these theories: their policy commitments, career aspirations, and campaign tactics are central to many political science accounts of how elections work. Yet, politicians also have their own beliefs about elections and voting. Sit with a politician as the room empties after a town hall meeting, or accompany a politician as they walk from door to door on the campaign trail, and you will soon discern the outlines of their theories of why citizens vote, how voters make their choices, and the forces that shape citizens’ political beliefs. When a politician complains that their party is too obsessed with focus groups, or brags that they could shoot a person in the street without electoral consequence, these comments tell us something not only about the politician’s personality and values but also offer clues about their working theories of elections and voting behavior.

These working theories have often gone unnoticed by political scientists, but there is good reason to expect that they matter a great deal for politics. Political science research on “lay theories” of politics, while limited, has consistently found that these theories are strongly related to political behavior and policy attitudes among both citizens and political elites (Kertzer and McGraw 2012; Rad and Ginges 2019); for instance, politicians who think of their constituents as policy oriented rather than identity oriented report spending more time on policy-related tasks (Lucas, Sheffer, and Loewen 2024), and politicians who think of constituents as oriented toward the long term are more likely to take a long-term perspective when facing temporal tradeoffs in policy choices (Sheffer, Loewen, and Lucas 2023). Moreover, recent work suggests that the more politicians believe voters will retroactively hold them accountable, the more time and effort they spend gathering public opinion information (Soontjens and Walgrave 2021). Hence, politicians’ implicit theories matter; they create “psychological worlds” (Dweck 2012, 39) that shape their expectations about others, and how they themselves behave.

We can gain an especially clear picture of politicians’ theories of elections and voting behavior by comparing politicians’ beliefs to those of ordinary citizens. Because citizens are unlikely to have reflected deeply on the forces that shape elections and voting behavior, their implicit theories are likely to be less well structured and reflect more top-of-mind assumptions. Comparing politicians’ theories to those of their constituents thus allows us to understand if politicians develop distinctive theories. Moreover, theoretical misalignment between politicians and citizens may have consequences of its own for elite-mass communication and even, in some cases, for political representation and citizens’ democratic satisfaction.

Here, we use data from face-to-face structured interviews with nearly one thousand elected national and regional politicians in 11 countries, along with surveys of over twelve thousand citizens, to provide a first-ever systematic analysis of politicians’ theories of elections and voting behavior. We find that elected politicians hold widely varying beliefs on central theoretical debates in political science: debates about retrospective versus prospective voting, policy-driven versus leader-driven electoral selection, voter knowledge versus ignorance, and more. To clarify these theories, we estimate the latent theoretical types that lay beneath politicians’ responses and find that nearly three quarters of politicians embrace a “thin” or “minimalist” theory of voting behavior, one that broadly resembles “democratic realism” (Achen and Bartels 2016). Comparing politicians’ theories to ordinary citizens, we find that politicians’ beliefs differ dramatically from those of the citizens they represent: in nearly every country we study, politicians are more likely than citizens to see voters as leader oriented rather than policy oriented, retrospective rather than prospective, egocentric rather than sociotropic, focused on single issues rather than multiple issues, concerned about the short term rather than the long term, and “blind” rather than “clear-eyed” in their retrospection. While nearly three quarters of politicians embrace a “thin” and realist theory of voting behavior, citizens are much more evenly divided between the realist perspective and an alternative theory in which voters are more policy oriented, knowledgeable, and engaged.

POLITICIANS’ THEORIES AND THEIR CONSEQUENCES

Theories of elections and voting are empirical accounts of election outcomes and the voting behavior that generates them. Theoretical beliefs are distinguished from other beliefs in being conceptual, explanatory, and predictive (Gelman and Legare 2011; Gopnik and Meltzoff 1998). Theories are conceptual in the sense that they provide concepts (e.g., “retrospective voting”) with which to organize the world into meaningful categories and explain empirical phenomena. Theories are explanatory in that they provide plausible causal accounts of events and outcomes; explanatory statements like, “the President lost because citizens were upset about the economy” imply underlying theories of voting behavior. Finally, theories are predictive in that they enable individuals to develop expectations about the consequences of their actions; statements like, “there is no way the party machine will allow him to become the Presidential nominee” are predictions grounded in implicit theories (Lucas, Sheffer, and Loewen 2024).

To make this more concrete, an example may be helpful. In spatial voting theory, each voter is typically assumed to hold a bundle of policy preferences that can be meaningfully summarized in some low-dimensional latent space (often characterized as a left–right spectrum); this bundle is called an “ideal point.” Political candidates and/or parties compete with one another by proposing their own bundles of policy promises; voters consider these promises and select the party or candidate whose proposed ideal point is closest to their own in latent space (Downs 1957). This theory provides a set of concepts (ideal points and spatial proximity) with which its users can provide explanations of electoral outcomes (“Party A had become too extreme, allowing Party B to build a new coalition of centrist and left-of-center voters”) and make related predictions about the future.

A starkly contrasting theory of elections and voting is Christopher Achen and Larry Bartels’s (2016) democratic realism. Synthesizing decades of political science research, together with their own original analysis, Achen and Bartels argue that voters are decidedly incapable of making choices based on calculations of spatial proximity. Instead, most voters make choices based on longstanding group identities and vague, short-term assessments of their well-being. Democratic realism not only provides theoretical concepts of its own (e.g., “blind retrospection”), but also offers very different explanations and predictions than those that arise from, among others, spatial voting theory.

These competing theories offer very different organizational, predictive, and explanatory perspectives on politics. Importantly, to the extent that these or other theories are held by politicians who are actively involved in politics, they are likely to generate widely varying behavior. In a study of Canadian local politicians, for example, Sheffer, Loewen, and Lucas (2023) found that politicians who believe voters focus on the short term are significantly more likely to opt for short-term rather than long-term solutions when facing temporal policy tradeoffs (Jacobs 2011). Politicians’ theories of voters have also been found to shape their responsiveness to public opinion (Soontjens 2022) and their choices about how they allocate their available working time, with “democratic realist” politicians spending less time on policy-related activity and more time on communication with constituents (Lucas, Sheffer, and Loewen 2024). These theories thus appear to have important consequences for how politicians choose to do their jobs. ¹

These consequences are likely to extend beyond how politicians reason about policy or how they choose to spend their time. In some circumstances, politicians’ theories may generate a form of self-fulfilling prophecy, inadvertently creating the electorate that politicians’ theoretical beliefs lead them to expect. For example, politicians who believe that voters are short-sighted and retrospective may endorse excessive preelection spending, exacerbating the patterns documented in research on electoral business cycles (Alesina and Roubini 1992) and signaling to citizens that elections are indeed “about” short-term policy or material rewards. Similarly, politicians who believe that voters are oriented to identity-based appeals rather than policy-based appeals are likely to spend more time communicating with constituents about relevant in-groups and out-groups than about policy, heightening the salience of group identities in election campaigns and political debates. And if politicians see voters as personalistic and leader-focused, rather than focused on parties’ policy promises, they may find it more difficult to stand up to their own party leaders, even when those leaders take norm-violating action in power, believing that their own political survival is strongly tied to their leader’s success (Matovski 2021). More broadly, then, politicians’ theoretical beliefs about what voters want—demand for personalism, demand for identity-based appeals, demand for short-term policy solutions, and so on—are likely not only to shape politicians’ own behavior, but may also serve over the longer term to reorient voters’ behavior as well. For all of these reasons, understanding politicians’ theories of elections and voting behavior should be an important and even central component of our general understanding of political elites.

We can gain an especially clear understanding of politicians’ theories by comparing politicians’ views to those of ordinary citizens. As Joshua Kertzer (2022) has argued, comparing political elites to ordinary citizens is valuable for normative, theoretical, and methodological reasons. At a normative level, most theories of political representation assume that politicians resemble constituents in ways that allow politicians to respond to constituents’ needs and allow constituents to meaningfully assess their representatives’ performance (Mansbridge 2003; Pitkin 1967). Together with decades of research on policy representation (Miller and Stokes 1963; Soroka and Wlezien 2009), political scientists have also explored other ways in which politicians might be expected to resemble their constituents, including personality traits (Dynes, Hassell, and Miles 2022; Hanania 2017), reasoning and problem-solving (Sheffer et al. 2018), and values and norms such as altruism and cooperation (Enemark et al. 2016; LeVeck et al. 2014). Identifying the respects in which politicians do or do not resemble their constituents—and then assessing the normative significance of this alignment or misalignment—is a central feature of contemporary theories of political representation (Mansbridge 1999; Urbinati and Warren 2008).

Comparing political elites to ordinary citizens is also crucial for theories of elite political behavior and elite-mass divides. This is important not only for ongoing debates about what it is that distinguishes politicians’ attitudes, characteristics, or decision-making processes from those of ordinary citizens (Kertzer and Renshon 2022), but also for more specific theories of political communication and democratic satisfaction. When citizens say things like, “politicians think we’re stupid” or “politicians think we don’t pay attention to what they do,” researchers often interpret these statements as indicators of political disengagement or cynicism. But what if politicians are more likely than citizens to think that voters are uninformed and ignorant? What if they are more likely than citizens to think that voters pay no attention to politicians’ actions? These differences between politicians and citizens, if they exist, would reveal an important and overlooked individual-level predictor of miscommunication, misunderstanding, and even dissatisfaction among citizens with their representatives—which would require very different solutions than other sources of dissatisfaction, such as ideological misalignment or poor performance in office.

Alignment or misalignment between elites and citizens is also theoretically important because we have good reason to suspect that elites do differ from ordinary citizens in their theories of elections and voting behavior. Political elites differ from other citizens not only in their demographic characteristics, such as age, wealth, and education (Carnes and Lupu 2023), but also in their personality traits (Hanania 2017), partisanship and political engagement (Enders 2021), and in the way they approach relevant decision-making processes, such as bargaining (Sheffer, Loewen, and Lucas 2023). Relatedly, politicians have distinctive opportunities to interact not only with voters but also with other politicians, including more experienced elites and campaign strategists who may socialize them into particular theories of “how things really work” in politics (Esaiasson and Holmberg 2017; Fenno 1977). ² These compositional, social, and cultural factors give us good reason to expect politicians’ beliefs about elections and voting to differ from those of ordinary citizens.

Finally, elite-mass comparisons have methodological implications. Studies of political elites—especially active, elected politicians in major national or regional executives and legislatures—are costly and time-consuming. In some domains, differences between political elites and the mass public (or between “top” politicians and more accessible elites, such as municipal politicians), are relatively small (Sheffer et al. 2018; Teele, Kalla, and Rosenbluth 2018). Understanding these differences allows researchers to pursue less costly research strategies, where appropriate, while still illuminating important features of elite political behavior (Kertzer and Renshon 2022).

Importantly, these comparisons are valuable even if citizens have reflected much less deeply than politicians on elections or voting behavior. Because citizens’ theories are likely, on average, to be drawn from culturally accessible and less well-structured narratives about politics, they provide us with a baseline against which to judge the distinctiveness of politicians’ theories. As long as citizens understand the questions being asked of them—and we provide evidence that the overwhelming majority of citizens do offer coherent responses to our questions—then comparing politicians’ theories to those of citizens offers valuable insight into how politicians see politics. This is true even if, as we expect, citizens’ theories are likely to be less stable and less predictive of behavior than those of political elites.

In sum, past research in political science, together with a larger interdisciplinary research tradition, offers good reason to expect that politicians’ theories of elections and voting behavior have important consequences for politicians’ behavior. These theories, however, have thus far gone largely unnoticed in political science research. ³ Our purpose in this article is thus to provide a comprehensive and comparative overview of politicians’ theories of elections and voting behavior.

MEASURING THEORIES OF ELECTIONS AND VOTING

To measure politicians’ theories of elections and voting behavior, we developed eight novel survey questions, each of which captures an enduring theoretical debate in political science. To select these debates, we focused on four criteria. First, we chose to focus on elections and voting behavior because we expect politicians to have developed theoretical beliefs in this area; this contrasts with other theoretical debates in political science (such as theories of the policy process or executive-bureaucratic relations) in which politicians may have had less opportunity to develop theories. Second, we focus on enduring theoretical debates in the elections and voting behavior field—debates that appear regularly in handbooks, textbooks, and synthetic reviews. Third, because our research is comparative and exploratory, we sought to cover a wide variety of theoretical debates, rather than focusing on repeated measures of a smaller number of debates. Finally, we focus on debates about individual voting behavior, rather than theories of system-level responsiveness or representation, such as theories of thermostatic responsiveness or issue evolution (Soroka and Wlezien 2009). While these macro debates are important, and political elites may well have theoretical beliefs about them, we begin by focusing our attention on a group about whom politicians are likely to have invested a great deal of thought: individual voters.

Based on these criteria, we selected eight debates to include in our interviews with politicians and citizens. The first of these is policy versus identity voting. As we noted earlier, political scientists in the spatial voting tradition argue that voters rely on their policy preferences to select their preferred candidates (Downs 1957; Jessee 2012; Schonfeld and Winter-Levy 2021). More generally, many theories of policy responsiveness assume that citizens’ policy preferences influence government policy in part through an electoral selection mechanism (Ansolabehere, Snyder, and Stewart 2001). However, an equally longstanding tradition rejects the notion that citizens even have coherent bundles of policy preferences with which to make their voting decisions (Converse 2006; Kinder and Kalmoe 2017), arguing that these choices are instead driven by factors such as citizens’ longstanding group identities—especially partisanship (Achen and Bartels 2016; Mason 2018). ⁴ This remains an area of spirited debate.

The second theoretical debate we selected concerns voters’ short-term versus long-term orientations. Inter-temporal choices are at the heart of policymaking, and it is commonly argued that policies tend to be biased toward the short term, in part because representatives have electoral incentives to cater to an impatient public (Ashworth 2012; Jacobs 2011). Research in psychology and economics emphasizes people’s tendency to be myopic in their preferences; citizens tend to be short-sighted and focused on the near rather than far future (Streich and Levy 2007; Urminsky and Zauberman 2015). However, empirical studies that corroborate this idea of myopic citizens in the context of elections and voting is more scattered in its conclusions. Healy and Malhotra (2009) do find that voters, in the context of policies dealing with natural disasters, support immediate relief aid rather than future disaster prevention, which suggests that voters tend to be averse to short-run costs that are connected to long term responsible policymaking. Jacobs and Matthews (2012; 2017), in contrast, show that voters are myopic in favoring secure short-term policy benefits, but emphasize that this does not imply that voters are fundamentally short-sighted. Voters are not impatient, they argue, but focus more on the short term simply because they are uncertain about the future. The character of voters’ short-term or long-term orientations thus remains an active scholarly debate.

Third, we ask if our respondents see voters as knowledgeable or ignorant. Empirical scholarship has long debated the degree of citizens’ policy-specific knowledge (Gilens 2001), general political knowledge (Delli Carpini and Keeter 1996), or political sophistication (Luskin 1987). While there is a broad consensus that political knowledge is associated with positive outcomes such as civic participation, correct voting (Lau and Redlawsk 1997; Lupia 2006), and political activism (Verba, Schlozman, and Brady 1995), the level of knowledge that citizens bring to their voting choices remains an area of debate (Achen and Bartels 2016; Fowler 2020).

Fourth, we measure theories of single-issue versus multiple-issue voting. Since Converse (2006) first proposed the idea of “issue publics”—voters who pay close attention to particular issues and who vote on the basis of parties’ stances on those issues—political scientists have debated whether such voters actually exist. As we already noted, many political scientists have suggested that voters’ policy beliefs are simply too weak and unstable to genuinely shape their choices (Achen and Bartels 2016; Cohen 2003; Kinder and Kalmoe 2017), and even those who do believe policy attitudes are important for voting tend to assume that bundles of issue positions, rather than single issues, are what matters (Fowler et al. 2023; Jessee 2012). Even so, a new analysis by Ryan and Ehlinger (2023) used a novel survey question and “bespoke” conjoint experiments to show that a substantial fraction of the American public does appear to belong to issue publics. This new approach is likely to provoke considerable new research—and debate—about the presence or absence of genuine single-issue voters.

Fifth, we explore the debate between those who see voters as motivated by political leaders versus those who see voters as focused on parties and their substantive ideas. There is a well-documented long-term process of personalization in democratic politics, wherein leaders’ personal authority becomes increasingly significant amidst weakening political parties (Rahat and Kenig 2018). In contrast, others follow classic research in spatial voting theory (Lau and Redlawsk 1997) by providing evidence for voters’ attention and responsiveness to changes in parties’ policy positions and ideology (Klüver and Spoon 2016; Seeberg, Slothuus, and Stubager 2017; Serra 2010). The relative importance of leadership competence (vs. policy and ideas) in vote choice is a longstanding focus in electoral research (Lanz 2020; Petrocik 1996), including work on both presidential candidates (Miller, Wattenberg, and Malanchuk 1986) and party leaders (Garzia 2011; Valgarðsson et al. 2021). These studies have recently been extended to voter support for political leaders who violate democratic principles (Carey et al. 2022; Frederiksen 2022) and the impact of competence on voting preferences (Green and Jennings 2017). Although some work suggests that the importance of leaders’ competence for vote choice has increased in recent years, its influence relative to parties’ substantive ideas remains a subject of active debate.

Our sixth debate is retrospective versus prospective voting. Classical theories of democratic representation view voters as future-oriented individuals who are driven largely by policy expectations (Downs 1957). Under this “promissory” model (Mansbridge 2003), voters make choices based on the match between their own policy preferences and the policies that candidates and parties offer (Naurin and Thomson 2020). In contrast, voting based on already implemented policies is considered retrospective, and a distinguished theoretical literature argues that voters’ decision-making is based largely on evaluations of representatives’ past behavior (Ferejohn 1986; Fiorina 1981). While a great deal of evidence indicates that citizens consider information on past performance when making their electoral choices (Healy and Malhotra 2013), prospective theory continues to receive considerable attention (Fowler et al. 2023; Jessee 2012). ⁵

Seventh, we measure egocentric versus sociotropic theory. An important question about citizens’ assessment of their incumbents’ performance is whether voters are egocentric in their evaluations—so-called “pocketbook” voting—or sociotropic, assessing the overall state of the national economy or other broad features (Healy, Persson, and Snowberg 2017; Lewis-Beck and Lockerbie 1989; Lockerbie 2006). Early rational choice models (Downs 1957) implied that voters would be egocentric, focusing on personal well-being, but considerable research has found that many voters instead respond to the state of the national economy and the incumbent government’s performance on the national economy (Clarke et al. 2004; Kinder and Kiewiet 1979; 1981). Others have reinforced this view with a more general argument that voters look beyond their own situation when casting their vote, acting with “sociotropic” rather than “egocentric” retrospection (Fiorina 1978; MacKuen 1983). Still, recent research has questioned the sociotropic consensus (De Benedictis-Kessner and Hankinson 2019; Healy, Persson, and Snowberg 2017), and it is also possible that retrospective voters evaluate the state of the nation and their own well-being—a distinction that is methodologically challenging to disentangle (Feldman 1982).

Finally, we explore the theoretical debate about citizens’ competence to assess their elected representatives’ performance. In classical retrospective voting theory, citizens hold their elected representatives accountable for their actions by considering indicators of their well-being (whether egocentric or sociotropic) during the full course of a government’s time in office (Fiorina 1981; Key 1966). This “clear-eyed” retrospection—holding governments accountable for what they can control, but ignoring changes over which governments have no control—incentivizes politicians to anticipate their constituents’ preferences and communicate the reasons for their actions to citizens (Mansbridge 2003). However, retrospective voting can secure this representational connection only if voters’ assessments are genuinely linked to politicians’ performance, and a prominent tradition of political science research has argued that voters’ assessments are in fact based on considerations that have nothing to do with politicians’ actions, such as the outcome of college football games (Healy and Malhotra 2009), local shark attacks (Achen and Bartels 2016), and extremely short-term economic fluctuations (Achen and Bartels 2016). These findings have prompted new studies that seek to question the “irrelevant events” results or argue that such events in fact provide voters with valuable information (Ashworth 2012; Ashworth, Bueno De Mesquita, and Friedenberg 2018; Fowler and Hall 2018).

Having selected these theoretical debates, we developed questions that describe each debate in accessible language. We provide the full wording for each of our questions in Table 1. In each question, we identify each side of the debate and ask respondents to position themselves within the debate on a 0–10 scale, with each pole appropriately labeled. Further, we field-tested all of these questions in surveys of local politicians in Canada, the United States, and Belgium. In each case, question response patterns and open-ended follow-up questions confirmed that politicians understood the questions, felt comfortable placing themselves in the theoretical debates, and even, in many cases, enjoyed the opportunity to express their views. ⁶ These questions are designed to be accessible to politicians and citizens alike; past elite-citizen comparisons suggest that politicians and citizens tend to respond to survey questions and prompts in similar ways, even on quite technical and specialized tasks (Kertzer 2022).

Table 1. Overview of Question Wording and Short Labels

Two additional features of these questions are worth emphasising. First, the order of the questions in the table carries no implied ranking—we consider all eight questions equally important, and the order of presentation of the items was randomized for both politicians and citizens. Second, our setup—with distinct questions for each theoretical debate—allows for but does not require that respondents’ positions on the theoretical debates are strongly related to one another. Among political scientists, we know that some combinations of theoretical positions are more common, and even perhaps more logically coherent, than others. However, research on implicit theories outside political science has demonstrated that lay theories are much more flexible than those developed by scientific professionals (Gelman and Legare 2011; Rad and Ginges 2018), and our pilot studies indicated that respondents might combine their theoretical positions in a wide variety of ways. Our questions allow for many possible theoretical positions not only in terms of the respondent’s answer to each theoretical item but also in terms of their positions across the eight theoretical debates.

DATA: THEORIES OF ELECTIONS AND VOTING IN 11 COUNTRIES

We examine politicians’ and citizens’ theories of elections and voting behavior in 11 countries: Australia, Belgium (Flanders), Canada, Czechia, Denmark, Germany, Israel, the Netherlands, Portugal, Sweden, and Switzerland. In each country, our questions were part of surveys fielded in the framework of the POLPOP project. ⁷ While these countries are similar to one another in being established Western democracies, they are quite diverse in terms of electoral systems, including majoritarian as well as proportional systems, large and small district sizes, strong and weak party systems, hybrid systems, and so on. These systemic differences necessarily influence why and how voters in those systems make their decisions. At a more individual level, these 11 countries are also diverse in terms of politicians’ lived experiences: the role of political parties in politicians’ careers widely varies, as do the lengths of their careers, the amount of turnover expected at each election, the size and characteristics of the constituencies they represent, the amount of staff support they receive, the media and how they cover politics and politicians, and so on. In the present article, we focus primarily on describing and comparing politicians’ theories, with the institutional and other country-level variation allowing us to check whether these differences hold across contexts.

To study politicians’ theories of elections and voting, we draw on extensive face-to-face surveys collected from 982 elected national and regional politicians between March 2022 and March 2023 (see Table 2)—an unprecedented dataset of elected political elites (Kertzer and Renshon 2022). Moreover, our sample of participating politicians is broadly representative of the full population in terms of gender, seniority, and ideological position (for more information on the sample, see Section 1 of the Supplementary Material). While response rates vary substantially across countries, the total number of completed surveys is exceptionally high for research with active members of parliament. In most countries, all national members of parliament were the target population, and in federal countries like Belgium and Canada, provincial or state parliamentarians were also asked to participate. In Israel, Sweden, and Australia, an election was called during the fieldwork period, and our target population thus included politicians who were not reelected as well as reelected and newly elected members of parliament. Politicians were asked to participate by local researchers, first via email and then, if contact details were publicly available, also via telephone.

Table 2. Data Collection: Fieldwork Periods and Response Rates

Concretely, a 30-minute Qualtrics-programmed survey was put to politicians by local researchers in each of the participating countries. Politicians always completed the survey in the presence of a researcher—who was either physically present or present in an online meeting (see Section 1 of the Supplementary Material for more information). This way, we ensured that politicians themselves and not their staffers completed the questionnaire, and we could respond to clarification questions as the politicians progressed through the survey. Importantly, however, while the interviews were conducted in face-to-face settings, politicians completed the survey portion of the interview, which we use here, using identical questions and the identical survey platform (Qualtrics) as citizens, and researchers could not see politicians’ responses as they completed the survey. ⁸

Next, we fielded an online population survey in March 2022 in each country to compare politicians’ theories of voting behavior and elections with those of citizens. In collaboration with Dynata, around two thousand citizens of voting age were targeted in each country from existing online panels, with recruitment quotas for age and gender (crossed), and education level. ⁹ To adjust for remaining imbalances, we computed post-estimation raking weights using age, gender, education, and party choice in the most recent national election. ¹⁰ Due to the modular structure of the public opinion survey, half of the respondents in the public opinion survey were randomly assigned to complete our questions on theories of elections and voting behavior; hence, we have responses from about one thousand citizens in each country (see Table 2). ¹¹ In the questionnaire, citizens were shown the same eight statements on elections and voting behavior and they too were asked to indicate their position on each 11-point scale. The phrasing for these questions was identical to the politician survey, and here, too, the item order was randomized.

POLITICIANS’ THEORETICAL BELIEFS

We begin with Figure 1, which summarizes the distribution of responses to our eight theory questions among politicians (in purple) along with citizens (in green), with pooled responses in panel A and country-specific results in panel B. Several important results are immediately visible. First, and most obviously, responses on all of the theory questions vary—for all eight questions, responses range widely across the available response options. The theoretical items we have measured are indeed debates, with a substantial proportion of respondents on each side of every question.

Figure 1. Theory Questions: Distribution of Politician and Citizen Responses

Note: Summary of the distribution of citizen responses (in green) and politician responses (in purple) to eight questions about elections and voting behavior. Pooled responses in panel A and country-specific responses in panel B. Columns are distinct questions (see Table 1 for full wording), and rows in panel B are countries. Response options range from 0 to 10.

A second important finding in Figure 1 is the similarity of the citizen and politician distributions across countries. In general, both the politician and citizen distributions look quite similar within each question as we scan from top to bottom in each column. This visual pattern is strongly confirmed in statistical tests; in the Supplementary Material, we show that in just two cases (of twenty two) is more than 10% of the variation in theoretical positions explained by cross-country rather than within-country variance. ¹² Despite considerable institutional and political-cultural variation across our case countries, the distribution of theoretical beliefs among both politicians and citizens is strikingly similar.

This cross-national similarity contrasts starkly with the third and most important finding in Figure 1: clear differences on most questions between the politician and citizen distributions. In the first column (unfair vs. fair blame), for example, the politicians’ distribution is shifted leftward and the citizens’ distribution is shifted rightward in all countries, suggesting that politicians tend to be more likely than citizens to see voters as “blind” rather than “clear-eyed” when making retrospective judgments about government performance. Similarly, in the far-right column (short-term vs. long-term focus), politicians once again skew left and citizens skew right. In this case, it appears that politicians are more likely than citizens to think voters focus on short-term rather than long-term considerations.

To formalize this comparison, Figure 2 summarizes estimates of expected differences between politicians and citizens on each question. In the top panel, each coefficient is drawn from a separate OLS model, regressing survey responses for each item on a politician/citizen indicator variable along with country fixed effects. In the figure’s remaining panels, we provide country-specific coefficients. Full tables for these models are available in the Supplementary Material.

Figure 2. Differences between Politicians and Citizens

Note: Summary of average difference between politicians and citizens for each item: black coefficients are statistically significant (p < 0.05), gray coefficients are not. Top panel provides overall differences from models that include country fixed effects. Bottom panels provide country-specific differences, by question. Full model tables are available in the Supplementary Material.

The coefficients in Figure 2 confirm that there are substantively large differences between citizens’ and politicians’ theories of elections and voting—differences that are generally consistent across countries. In the top panel, the first two coefficients reveal that politicians are more likely than citizens to think of voters as leader-focused rather than ideas-focused and to think of voters as egocentric rather than sociotropic. In both cases, the differences are substantively important, approaching an expected within-country difference of one point on a 0–10 scale. The smaller panels illustrate that these findings are consistent in direction in 10 of 11 countries for leadership vs. ideas and in all 11 countries for sociotropic vs. egocentric voting. ¹³

The next two coefficients in the top panel are not statistically significant. Politicians are no more likely than citizens to think of voters as identity oriented rather than policy oriented, nor are politicians more likely than citizens to think of voters as ignorant rather than knowledgeable. In both cases, the country-by-country breakdowns in the bottom panels indicate that these pooled null findings are not merely the result of country-level variation that is “canceled out” in a pooled model: the policy vs. identity relationship is null in eight of eleven countries and the knowledge vs. ignorance relationship is null in seven of eleven countries. These null findings are theoretically interesting because they suggest that politicians do not simply take what we might think of as more “cynical” theoretical positions than citizens across the board. The null findings also help to confirm that citizens are not more inclined than politicians to merely provide socially desirable responses: if the citizen responses were more contaminated by social desirability, we would expect this to be especially visible in the “knowledge vs. ignorance” question. ¹⁴ To be sure, plenty of politicians and citizens believe that voters are not especially knowledgeable in their voting decisions, but this position is no stronger, on average, among politicians than citizens. It is equally striking, in an environment of strong elite polarization and debates about “identity politics” in many democracies, that politicians are no more likely than citizens to think of voters as motivated primarily by group identities rather than policy commitments.

The remaining coefficients in the top panel of Figure 2 are the questions for which politicians tend to select lower values than citizens. For prospective and retrospective voting, the difference is relatively modest (about 0.5 points on the 10-point scale), with politicians having a slightly higher overall tendency to hold prospective theories. Notice, however, that this difference is statistically significant in just four countries. The three remaining questions are much stronger and more consistent: politicians are substantially more likely than citizens to think that voters unfairly blame elected representatives for events that are outside the government’s control; more likely to think of voters as short term rather than long term in their focus; and more likely to think voters focus on single issues rather than many issues when voting. In all three cases, these differences are substantively large—well over one point on the 0–10 scale—and, as the country-specific breakdowns reveal, remarkably consistent in direction and significance across countries.

Overall, then, we find that politicians differ quite profoundly from citizens in their theoretical beliefs about elections and voting behavior. These politician–citizen differences are much more consistent in direction, statistical significance, and magnitude than the cross-national differences. While theoretical beliefs vary widely among both politicians and citizens—in all eight cases, the theoretical debates we have identified are indeed debates, with many citizens and politicians on both sides of each debate—we see remarkably similar general tendencies across countries, despite substantial differences in electoral institutions, party systems, and political cultures.

FROM BELIEFS TO THEORIES: POLITICIANS’ THEORETICAL TYPES

At a glance, theoretical tendencies in Figure 1 and Figure 2 appear to hang together in coherent bundles: politicians tend to be more likely than citizens to think of voters as leader oriented, egocentric, and oriented toward short-term and single-issue considerations. Borrowing from Achen and Bartels, we might say that politicians appear to incline more strongly toward “democratic realism,” whereas citizens incline toward what we might call a “democratic optimist” theory of elections and voting behavior. In other words, the differences between politicians and citizens may be differences not only in beliefs about specific theoretical debates, but may also cohere into more differences in deeper and more general theories of voting behavior.

To explore this possibility, we used a latent class analysis (LCA) to organize politicians and citizens into more general latent classes on the basis of their responses to each of the eight theory questions (Linzer and Lewis 2011). Our goal in this analysis was to inductively identify the latent “theories” of elections and voting beneath responses to the individual theory items. We thus began by simplifying each question into three categories: a position on one side of each theoretical debate (e.g., sociotropic voting), a position in the exact center of the 0–10 response scale, and a position on the other side of the theoretical debate (e.g., egocentric voting). We then used these simplified theoretical positions in a LCA, fitting solutions ranging from 2 to 20 classes and recording class membership values and fit statistics for each solution. We provide additional detail on our class selection criteria and fit statistics, as well as robustness tests using alternative coding approaches and clustering solutions, in Section 6.1 of the Supplementary Material.

Our analysis indicated that a four-class solution struck an attractive balance between substantive interpretability and statistical fit. We summarize this four-class solution in Figure 3. In the top panels, we report the proportion of citizens (left) and politicians (right) who belong to each of the four classes. In the remaining panels, we provide the full distribution of responses to each question, organized by class membership. These distributions allow us to interpret the results of the LCA and help to justify the labels we have applied to each of the four classes.

Figure 3. Politician and Citizen Membership in Four Latent Theory Types

Note: Summary of latent class analysis describing politicians’ and citizens’ membership in four latent theory types. Top panel summarizes percentage of citizens (left) and politicians (right) belonging to each class. To aid in interpretation, bottom panels summarize the distribution of responses to each theory question among members of each latent class.

To interpret the distributions in the bottom of Figure 3, notice the general tendency in responses across the first row: based on the visible peaks in the distributions, these respondents tend to think of voters as fair in their retrospective assessments, policy oriented, prospective, sociotropic, multiple-issue-focused, interested in policy rather than political leaders, knowledgeable, and oriented to the long term. These respondents are democratic optimists, expressing a confident view of voters as policy oriented, knowledgeable, prospective decision makers. More than a third of our citizen respondents belong to this category, while far fewer politicians—just 16%—belong to this latent class.

The second latent class contrasts starkly with the first: individuals in this category tend to see voters as unfair in their blame, identity oriented, retrospective, egocentric, single-issue-focused, leader-driven, ignorant, and short-termist. These views largely correspond to what Achen and Bartel’s describe as “democratic realism,” where voters are seen as blindly retrospective, group oriented, and generally rather ignorant about politics. Politicians are much more likely to be democratic realists than democratic optimists—nearly three quarters of the politicians in our sample (73%) are democratic realists. Among non-elites, in contrast, we see an even distribution across the democratic optimist and democratic realist groups—about a third belong to each class.

The two remaining classes in the LCA, while interesting, are of less substantive importance. The third class captures respondents who tend to choose the middle value or very moderate values across the theory questions. While these respondents do have views on some questions, they are clearly uncertain in their theoretical beliefs, and we therefore describe them as the “undecided” theorists. This group is small, but by no means insignificant, among both citizens (17%) and politicians (10%).

Finally, a small but discernible fraction of citizens appear to choose higher values (between 6 and 10) across all eight issue items (11% of citizens, extremely few politicians). These respondents may be especially susceptible to acquiescence bias, choosing the second theoretical position in each question, but the most likely explanation is that these respondents are simply inattentive and move through the questions too quickly. ¹⁵ We thus label this group “Inattentives.” Notably, almost no politicians fall into this final class.

Taken together, the findings in Figure 3 indicate that politicians and citizens have starkly different theories. While citizens are quite evenly divided in their theories between democratic optimists and democratic realists (with the final third falling into the undecided or acquiescence camps), politicians are overwhelmingly democratic realist in their orientation. These differences are substantively large and statistically significant in every country in our study. ¹⁶

While these general politician–citizen differences are consistent across countries, the proportion of politicians who are democratic realists does vary. Figure 4, which summarizes politicians’ latent class membership by country (focusing on the two most theoretically important classes), confirms that a majority of politicians are democratic realists in each country. However, the figure also reveals striking variation across countries. In some countries, more than four in five politicians are democratic realists (such as Czechia, Canada, and Israel), whereas other countries have a substantial minority of democratic optimists among elected representatives (such as Switzerland, Australia, Sweden, and Denmark).

Figure 4. Politicians’ LCA Types, by Country

Note: Breakdown of Democratic Realism and Democratic Optimism types by country.

The patterns in Figure 4 are not intuitive, with institutionally and culturally similar countries (such as Canada and Australia) at opposite ends of the spectrum and very different countries (such as Belgium and Israel) closely resembling one another. However, the results in Figure 4 are remarkably consistent, in that in each of our 11 country cases, there is a clear majority of politicians who we can identify as democratic realists.

DISCUSSION AND CONCLUSION

This article has provided what is, to the best of our knowledge, a first-ever analysis of politicians’ theories of voting behavior. Drawing on data from face-to-face structured interviews with nearly one thousand politicians, we found that elected politicians hold widely varying views on central debates in elections and voting but tend, on average, to think that voters are unfair in their retrospective assessments of politicians’ performance, identity oriented rather than policy oriented, retrospective, egocentric, single-issue-focused, leader oriented, relatively uninformed, and oriented to the short term. In each of the 11 countries we study, we found that a majority of politicians belong to a latent class that we call “democratic realism”—a thin, minimalist, relatively pessimistic view of voters’ capacities.

Beyond these general findings, many particulars are also notable. For instance, while politicians across our 11 countries tend to believe that voters place more emphasis on the qualities of party leaders than on those parties’ principles, politicians do not differ from citizens in their views about voters’ orientation to identity versus policy. This reflects a political elite that deviates from ordinary citizens in their theories primarily in the weight it gives to personalistic considerations. Politicians may be motivated to adopt this view because it makes their personal “brands” more consequential for their own success or their party’s fortunes (and for some politicians, such as those elected in single-member districts, this may be a natural conclusion). For others, however, it may be an expression of a (potentially misguided) belief that voters have a strong attachment to leaders, a phenomenon that is closely associated with the weakening of party systems (Rahat and Kenig 2018) and has more recently been argued to be a facilitating factor in processes of democratic backsliding (Matovski 2021).

Politicians also differ strongly from citizens in their beliefs about the prevalence of single-issue voters. Some politicians may be motivated to adopt this belief if they are themselves focused on a single major priority as legislators, or if their party is a distinct issue-owner. Whatever the individual motivations, politicians with single-issue theories of voting may be more inclined to develop (or perhaps more cautious about resisting) single-issue and niche parties, including radical right populist parties in Western democracy, who have gained electorally from focusing on the single issue of immigration (Dennison 2020; Mudde 1999).

More broadly, politicians’ theories of voter demand for single-issue focus, personalism, short-term policy, or other representational behavior and policy outputs, are important factors to consider when evaluating representation gaps, elite political behavior, and concrete policy outcomes in future research. That politicians’ views differ so strongly from those of citizens could also have implications for existing theories of non-elite political behavior, and in particular for models of vote choice and policy responsiveness. Such models often make similar assumptions on citizens and elites—for example, that they are both myopic (e.g., in models of the electoral business cycle) or are similarly interested/disinterested in fulfilling policy goals (e.g., in models of spatial voting). If politicians and voters have divergent views, as we document here, then there is value in reexamining these models and whether their predictions hold in light of updated assumptions. We see this as a priority for future work.

More generally, we hope that our findings will spark a new interest in elite theories of politics and their consequences. We see numerous opportunities to deepen and clarify our findings. For example, while we found that politicians are more likely than citizens to cluster into a “democratic realist” theoretical perspective, the results in Figure 1 also demonstrate that politicians are quite variable in their theoretical beliefs. Future research should explore this variation in more detail, seeking to understand how politicians’ individual characteristics (their ideological positions, their personality types, sociodemographic backgrounds, and leadership positions) and career experiences (the parties into which they were recruited, the length of their careers, their electoral history) relate to their theoretical beliefs. Related work could explore how these theories develop throughout a politician’s career, along with the kinds of experiences (e.g., electoral victory, electoral defeat, prominent elections in other jurisdictions) that shape this development. Going beyond individual politicians, the cross-national differences we document (see Figure 4) suggest that there is also promise in the institutional and structural factors that affect how elected officials in different polities develop their views. Emerging methodological developments in LCA, enabling computationally efficient multilevel LCA with covariates (see Di Mari et al. 2023; Lyrvall et al. 2024), offer a particularly promising path forward for exploring heterogeneity in politicians’ theories.

Future studies should also explore the implications of politicians’ theories for their behavior as representatives. Evidence from past research suggests that politicians’ theoretical beliefs are importantly related to how they think about public policy (Sheffer et al. 2023). This work could be extended to studies of politicians’ communication strategies, policy prioritization, risk-taking behavior, campaign tactics, and their cooperation with other politicians. It could also be extended to important behaviors among citizens, such as shifts in turnout (Kostelka and Blais 2021) and citizens’ more general “participation repertoires” (Oser 2022). Much of this work could be observational, connecting politicians’ survey responses to observed behavior. To enable more precise causal inference, however, implicit theories could also be induced in experimental settings; researchers in other disciplines have found that implicit theories can be experimentally induced even in instances when individuals hold strong beliefs, and these experiments would be valuable for measuring the consequences of politicians’ theories for behavioral tasks (Dweck 2012). Panel studies measuring how politicians’ theories develop throughout their careers in response to socialization and accumulated experience, election outcomes, and changes to patterns of voting behavior, will also help to clarify the causal mechanisms that shape politicians’ theories.

Finally, we see considerable potential for studies of politicians’ theories in other domains of politics. As we noted earlier, we expect that all democratically elected politicians possess quite well-developed theories of voting behavior. But politicians may have other theories that are also consequential for their actions. For instance, politicians’ theories of the policy process—how issues arise on the public agenda, how decision makers allocate attention to problems, the role of the public service, and so on—are also likely to be important for politicians’ engagement in the policy process (Hall 1993; Stone 1989). Politicians’ other theories—ranging from implicit theories of the economy (Rubin 2003) to theories of the causes and consequences of protest activity—are equally worthy of attention. Understanding the “psychological worlds” that these politicians inhabit will, we hope, ultimately clarify the concrete worlds of political participation, representation, and public policy that their theories help to create.