1 Introduction

The relationship between public opinion and government behavior is of vital importance for democracies. A common conception of democracy holds that public opinion should guide governments in their policy formulation. For example, Robert Dahl (1971, p. 1) states that “a key characteristic of a democracy is the continuing responsiveness of the government to the preferences of its citizens, considered as political equals”. Political scientists report that this normative premise is at least partially satisfied in empirical reality—research agrees that governments are indeed responsive (Burstein 2003; Wlezien and Soroka 2007), although inequalities along characteristics such as income persist (Elkjær and Klitgaard 2021; Elsässer and Schäfer 2023). That is, what policies governments actually implement is influenced by public opinion.

This literature review engages with the theory and empirical research on the effects of public opinion on political phenomena from voting behavior to policymaking. It approaches the topic from the perspective of Comparative Political Economy (CPE) research on electoral politics and policy formulation (CPE research hereafter). It is widespread in CPE research to characterize public opinion as an exogenous force that subsequently shapes politics and policies. That is, citizens first form their opinion and then decide who to vote for, which in turn influences government formation and ultimately policies. I call this the exogenous preferences perspective. For example, Philipp Rehm argues that labor market risk exposure leads people to demand stronger social policies, which then translates into actual social policies via the voting mechanism (Rehm 2009, 2011, 2016, 2020).

In this review, I juxtapose the theoretical conception of CPE research with an alternative theory from political sociology and communication, which argues that public opinion is more open to interpretation and less centered around actual policies than assumed by CPE. Citizens support abstract ideals rather than specific policy preferences and they rely on simplifying heuristics, one of which is that people adapt their attitudes according to the communication of political elites. I call this the endogenous preferences perspective.

The goal of this literature review is to outline what CPE research can learn from the endogenous preferences perspective. I will argue that CPE has much to gain from incorporating insights from the contradicting theory. First, citizens are rarely guided by specific policy preferences—such as support for specific welfare programs—in their political behavior. As most individuals lack the required sophistication and knowledge, they rather rely on more abstract core values, such as egalitarianism. Second, political elites such as parties influence public opinion because citizens rely on elite cues to form their opinion. This alternative perspective suggests that the relationship between public opinion and downstream phenomena such as politics and policies is less deterministic and more open to interpretation than assumed by CPE research. While these points regard CPE research in specific, this review has implications for political science research in general, as the assumption of public opinion assuming an exogenous causal role is widespread and reaches beyond the confines of CPE.

It is important to note that CPE research is aware of the shortcomings of the exogenous preferences perspective. For example, Silja Häusermann and Herbert Kitschelt explore the transformation of left parties in a forthcoming volume (Häusermann and Kitschelt in press). In line with the dominant CPE approach, they stress the importance of public opinion and socio-economic structure, for example when they explain the electoral demise of social democratic parties via the declining share of industrial workers and the associated preference shift. However, they also highlight that parties do not simply respond to voters’ demands in the narrow sense of policy preferences. Parties are rather forced to engage in programmatic campaigns that appeal to fundamental predispositions of voters. This incorporates insights from the endogenous preferences perspective, which is what I recommend as a way forward.

The remainder of this article will proceed as follows. The next section explains the dominant paradigm in CPE research, which I call the exogenous preferences perspective. It develops a stylized theoretical model to illustrate the theory, demonstrates that it is actually influential in CPE, and reviews empirical research on it. The third section introduces the contradicting theoretical approach—the endogenous preferences perspective—and reviews empirical research on it. The fourth section discusses where the two perspectives differ and subsequently outlines what CPE research can learn from the opposing theory. The fifth section develops the fundamentals of a theory on the relationship between public opinion and policymaking and the final section concludes with a discussion on the implications for specific CPE literatures.

2 The exogenous preferences perspective

Researchers from Comparative Political Economy (CPE) and adjacent fields assign public opinion a crucial role in the determination of various political phenomena such as voting behavior, party positions, government selection, and policymaking. I will argue in this section that many contemporary contributions share a common theoretical view that links public opinion to all these phenomena. I first develop a stylized theoretical model to render this theoretical thought visible. The section subsequently illustrates how widespread the worldview is in CPE research.

2.1 Theoretical model

Figure 1 presents a stylized theoretical model that links public opinion to voting behavior, government selection, and policymaking. The purpose of the model is to clarify how large parts of contemporary CPE theory think about public opinion and its implications. The crucial point is that public opinion is an exogenous causal factor—that is, public opinion emerges without inputs from the political process itself and then goes on to shape voting behavior, government composition, and policies. The model focuses on the essentials and is therefore oversimplified—I do not claim it to be a perfect representation of the discussed research. One of these simplifications is that the model equates public opinion solely with policy preferences, which are normative beliefs about the design of specific government policies (Aalberg 2003, ch. 1).

Fig. 1
figure 1

Stylized model of the exogenous preferences perspective. (Source: Author’s illustration)

In more detail, the theoretical model starts with socio-economic conditions. CPE primarily focuses on objective material conditions—that is, income and wealth—but especially recent contributions also focus on conditions only partially related to material distributions, such as education, age, and social class.

People derive their policy preferences from the socio-economic structure and their position in it. CPE primarily expects that people rely on their economic self-interest (Meltzer and Richard 1981; Romer 1975), supporting policies that grant them more material resources (e.g., Ahrens 2022a; Rueda and Stegmueller 2019). But the research field has been broadening its theoretical scope, including expectations that individuals also form their preferences based on sociotropic motivations such as a wish for fairness and altruism (Ahrens 2019; Cavaillé 2023; Dimick et al. 2017, 2018).

Thereafter, policy preferences influence voting behavior. People are assumed to be guided by instrumental rationality. They aim to vote the party into power that offers the optimal policy bundle, which assumes that voters are (at least minimally) informed about the policy platforms of parties and past government performance (Downs 1957; Rueda and Stegmueller 2019). First, people prospectively vote for the political party that promises to enact policies most in line with their preferences (Downs 1957; Rueda and Stegmueller 2019). As CPE stresses material conditions, this entails that the economically disadvantaged should vote for parties promising greater state intervention and redistribution (Emmenegger et al. 2015; Rueda and Stegmueller 2018b, 2019). Second, voters retrospectively base their vote on an assessment of whether the previous government parties performed well (Duch and Stevenson 2008). The literature primarily focuses on assessments of past economic circumstances such as unemployment or inflation (Duch and Stevenson 2008; Lewis-Beck and Stegmaier 2013). But retrospective performance voting also includes evaluations of previous policy outputs. The expectation is that recently governing parties should be punished (rewarded) when they implemented policies contradicting (coinciding) with voters’ policy preferences (Lewis-Beck and Stegmaier 2013; Pierson 1994, 1996; Tilley et al. 2018).

Next, the aggregation of individual votes influences the government composition. Left parties with a progressive policy agenda will govern with higher probability when the electorate favors more redistribution and economic protection. This mechanism generally occurs in all democratic countries, but its importance depends on existing institutions. The policy preferences of the electorate may induce more votes for a particular type of party—such as a right-leaning party opposing redistribution—but who ends up in government depends on electoral rules and the existing party system (e.g., Iversen and Soskice 2006). However, there generally is a relationship between policy preferences and the composition of government.

Finally, the government composition influences actual policies. More left-leaning governments should implement more left-leaning policies, and vice versa. In other words, it truly makes a difference who governs (Hibbs 1977; Korpi 1983; Schmidt 1996), which is more contested than one may assume.Footnote 1 Overall, the model expects a causal relationship between public policy preferences and policymaking for two reasons. First, policy preferences influence who governs, which thereafter influences policies because parties are both willing and able to implement promised policy programs. Second, parties seek re-election, which incentivizes them to follow public opinion once in office (Adams et al. 2004; Stimson et al. 1995).

The exogenous preferences perspective rests on two axiomatic foundations. First, voters’ policy preferences are viewed as an exogenous force that causally affects political behavior, politics, and policies. Exogeneity entails that preferences are independent from the communication and policy positions of political elites, such as parties. Put differently, voters know what they want and assert their will via the voting mechanism. Second, politics and policies ultimately depend on the socio-economic structure. The reason is that peoples’ preferences depend on their social and economic position in their country. The model therefore takes a deeply structuralist perspective.

2.2 Examples from the literature

This section shows that the exogenous preferences model continues to be influential in CPE. It is important to note that few contemporary contributions are as simplistic as Fig. 1. For example, research shows public officials to misperceive the preferences of their voters, thus weakening the link between preferences and policymaking (Sevenans et al. 2023; Walgrave et al. 2023, 2024). However, I argue that many contributions continue to expect that—under specific conditions—public opinion is sufficiently exogenous and causally affects downstream political phenomena. I demonstrate this by reviewing influential CPE contributions.

The constrained partisanship model by Beramendi et al. (2015) is a first case in point. It addresses the constraints parties face in advanced capitalist countries. Reportedly, parties carefully align their policy platforms with the prevailing preferences of the electorate while simultaneously managing the limitations imposed by a shifting economic structure. Choosing a strategy has become increasingly difficult for parties because voters’ preferences have become increasingly multidimensional while forces such as globalization concurrently constrain the available policy menu.

Work on the political implications of economic risk argues that workers facing more severe risks—such as a higher probability of losing employment because of technological change—shape both government composition and policies. Philipp Rehm (2009, 2011) argues that people experiencing risk demand social protection via the welfare state. As public opinion influences policies, realized welfare policies depend on the distribution of risks within countries (Rehm 2011, 2016, 2020). Furthermore, Thomas Kurer and his coauthors theorize that labor market risk threatens the social status of workers, to which they respond via activating traditionalist predispositions to uphold their status. The result is stronger electoral success of radical right parties (Abou-Chadi and Kurer 2021; Kurer 2020; Schöll and Kurer 2024).

David Rueda and Daniel Stegmueller (2019) argue that the political behavior of individuals depends on their support for redistributive policies. This demand for redistribution depends on both their current and expected income relative to others in society. Citizens who support more redistribution vote for left-leaning parties with higher probability because these parties promise to implement social policies and progressive taxation (see also Rueda and Stegmueller 115,116,a, b).

The still influential insider-outsider theory by David Rueda (2005, 2006) holds that—in advanced capitalist democracies—labor is divided into insiders in stable employment and outsiders on the insecure labor market fringes, such as in fixed-term employment or agency work. Insiders and outsiders have contrasting demands for social protection due to their varying economic circumstances, which creates a conundrum for social democratic parties, as it is difficult to satisfy both camps with their policies. Put between a rock and a hard place, social democratic parties serve the interests of insiders and uphold—or even strengthen—job protection measures while neglecting policies benefitting outsiders, such as unemployment assistance (Rueda 2005, 2006, 2014).

To be sure, I expect that few authors would fully subscribe to the model displayed in Fig. 1. It represents a simplified theoretical conception that allows researchers to focus on variables they deem important. For example, it is obvious to everyone that many determinants of voting behavior or policymaking are left out of the model. Nevertheless, this review shows that many CPE contributions hold that the exogenous preferences perspective offers a suitable theory to explain a meaningful proportion of politics and policymaking.

2.3 Empirical evidence

The exogenous preferences perspective receives support from a mountain of empirical evidence. All nodes and connections displayed in Fig. 1 were tested in empirical studies. While there are doubts about several of the connections, many studies present evidence supporting the model. For example, people’s position in the socio-economic structure predicts their policy preferences (Ahrens 2022a; Alesina and La Ferrara 2005; Margalit 2013, 2019; Rueda and Stegmueller 2019) and policy preferences predict voting behavior (Lewis-Beck and Nadeau 2011; Quinlan and Okolikj 2020; Rueda and Stegmueller 2019). Furthermore, many studies suggest that who governs really makes a difference for policies (Bandau and Ahrens 2020; Cusack and Beramendi 2006; Garritzmann and Seng 2020, 2023; Haffert 2021; Schmitt 2016).

Studies on individual connections are important, but this section will focus on the overarching expectation—a relationship between policy preferences and policymaking—to fence in the review. It would offer great support for the model if it could be shown that the policy preferences of the public are related to the policies that governments implement. So, what does the evidence say?

The responsiveness literature shows that voters’ policy preferences covary with the policy preferences of parties (Adams et al. 2004, 2006, 2009; Adams and Ezrow 2009; Hobolt and Klemmemsen 2005). When people take a more left-leaning stance, parties in general—or at least the particular parties these individuals vote for—also endorse more left-leaning policies. Further research also confirms a relationship between public policy preferences and the policies governments implement (e.g., Carrubba 2001; Hakhverdian 2012; Stimson et al. 1995; Page and Shapiro 1983; see the reviews by Burstein 2003 and Wlezien and Soroka 2007). The thermostatic model garnered considerable empirical support, which suggests that policymakers adapt policies in reaction to changes in public opinion over time (Wlezien 1995; Wlezien and Soroka 2007).

Recent responsiveness research has largely focused on democratic inequalities, showing that the preferences of advantaged voters with higher income and education predict policy outputs much better than preferences of the disadvantaged. Beginning with studies on the US (Bartels 2008; Gilens 2005, 2012; Gilens and Page 2014) and echoed by follow-up studies on European countries (Elkjær 2020; Elsässer et al. 2018, 2021; Hobolt and Klemmemsen 2005; Pinggera 2021; Schakel et al. 2020; Schakel 2021; Schakel and Burgoon 2022), this research points out democratic inequalities. Nevertheless, it shows that public policy preferences correlate with policy outputs for parts of the population (see the reviews by Elkjær and Klitgaard 2021 and Elsässer and Schäfer 2023).

A shortcoming of the responsiveness literature is that it is largely based on analyses of observational data with no causal identification strategies even though it explicitly assumes public opinion to adopt a causal role. Studies reliably establish correlation between public opinion and party behavior, but it is far from clear whether this correlation results from a causal relationship. It cannot be ruled out (yet) that the correlation emerges because there is a shared determinant of policy preferences and policy outputs. Another possibility is reverse causality—that is, policies and party positions influencing citizen preferences.

To tackle these endogeneity concerns, several studies use methods with more careful causal effect identification. First, researchers utilized instrumental variable strategies to eliminate endogeneity. Using political institutions and voter demographics as instruments, their results suggest a causal impact of public opinion on policies and party preferences (Breznau 2017; Carrubba 2001; Hill and Hurley 1999). A second set of studies uses time series divergences to show that public opinion shifts often predict subsequent policy changes but not the other way around (Hakhverdian 2012; Page and Shapiro 1983; Stimson et al. 1995). This supports the argument that public opinion causally affects policymaking. The findings underscore that, overall, there is considerable support for the exogenous preferences model.

3 The endogenous preferences perspective

Drawing from research in political communication, psychology, and sociology, this section presents an alternative theoretical conception of public opinion and voting behavior, which I will refer to as the endogenous preferences perspective. Although it does not theorize on the downstream phenomena—government selection and policies—the perspective stands in obvious contrast to the exogenous preferences perspective. However, the goal of this section is not to prove the exogenous preferences perspective wrong but to explore arguments that—as I will argue—have potential to enrich CPE research. As above, this section develops a stylized theoretical model and then presents empirical evidence to support it. Thereafter, it juxtaposes the theory to the exogenous preferences perspective.

3.1 A multidimensional conceptualization of public opinion

I draw from the literature on political belief systems to develop an alternative theoretical conception that considers the endogeneity of public opinion. This literature relies on a different conceptualization of public opinion, and it is important to describe it first. It departs from the finding that the assumption of sophisticated citizens with well-formulated policy preferences is mostly unreasonable. People voice an opinion on particular policy proposals when you ask them to, but few have substantiated knowledge about objective socio-economic realities and the functioning of policies (Goren 2013; Jensen and Zohlnhöfer 2020; Zaller 1992; Zaller and Feldman 1992). It is therefore essential to introduce elements of public opinion that are less specific than policy preferences and so do not require high political sophistication and up-to-date knowledge about policies and the world.

Rather than only including policy preferences, the endogenous preferences perspective uses a multidimensional conception of public opinion. It expects that individuals have hierarchically organized belief systems comprising several subcomponents (Aalberg 2003; Feldman 1988; Goren 2001, 2013; Rokeach 1973). I distinguish between three concepts: core beliefs, perceptions, and policy preferences.Footnote 2 Preferences follow the same definition as before—they are normative beliefs on how specific government policies should be designed (Aalberg 2003, ch. 1). Perceptions are beliefs about empirical conditions. These can be both descriptive (e.g., how much unemployment is there in my area?) and normative (e.g., how unfair is the income distribution?).Footnote 3 But the most important concept is core beliefs.

Core beliefs are deeply held, normative assessments about actors, behaviors, and desirable empirical conditions. It is beyond the scope of this article to give a detailed overview, but researchers focus on different core beliefs including values such as economic individualism and policy principles such as limited government (Aalberg 2003; Feldman 1988; Goren 2001, 2013; Rokeach 1973). Another deeply held belief which is influential in politics is partisan attachment—that is, an emotional attachment of an individual to a particular party (Campbell et al. 1960; Niemi and Jennings 1991).

In contrast to both preferences and perceptions, core beliefs are abstract rather than specific. Take the example of the policy principle limited government (Goren 2013), which holds that the government should take a restrained role and intervene little in the economy. This principle implies that the government should redistribute less and that larger inequality is acceptable, but the exact policy implications require interpretation. Therefore, support for the principle does not require detailed knowledge about the world and policies. The same can be said about partisan identification. It constitutes a rather abstract attachment to a party that—in itself—lacks detailed assessments about policies and the world.

Core beliefs are the most important element of public opinion because they serve as heuristics that allow voters to form their more specific beliefs and behavior, such as preferences and who to vote for. Political belief systems are organized hierarchically, with core beliefs at the top. Core beliefs serve as simplifying decision makers that influence how people perceive the world and what policies they prefer (Feldman 1988; Goren 2001, 2013; Rokeach 1973). Several studies show that core beliefs are stable predictors of more specific preferences, but not the other way around, implying that core beliefs assume the top role in peoples’ belief systems (Feldman 1988; Goren 2001; Hurwitz and Peffley 1987; Jacoby 2006; Peffley and Hurwitz 1985, 1993). Furthermore, core beliefs shape perceptions of the world, such as on the fairness of income distributions (Aalberg 2003). Lastly, core beliefs remain relatively stable over a lifetime (Feldman 1988; Goren 2005; Niemi and Jennings 1991; Peffley and Hurwitz 1993), which contrasts with the instability of preferences (Zaller 1992; Zaller and Feldman 1992).

3.2 Theoretical model

This section develops a theoretical model of the endogenous preferences perspective. It must be noted that the model displayed in Fig. 2 ultimately explains individuals’ voting behavior and therefore has a more limited scope than the exogenous preferences model in Fig. 1, which further includes government selection and policies. The reason is that the relevant theoretical insights only concern the topic of preference formation and voting.

Fig. 2
figure 2

Stylized model of the endogenous preferences perspective. (Source: Author’s illustration)

The model in Fig. 2 is based on the multidimensional conceptualization of public opinion and hence considers that people seldom have well-formulated policy preferences—they rather rely on deeply held core beliefs as simplifying heuristics to form their preferences and to decide who to vote for. Figure 2 thereby deviates from the exogenous preferences model as it no longer characterizes policy preferences as an exogenous force that shapes the political process. Policy preferences are still important because—as before—they influence voting behavior. However, voting behavior depends more on core beliefs such as partisan identification or economic individualism. Furthermore, Fig. 2 also specifies that people adapt their preferences and perceptions to the cues of political elites as a further simplifying heuristic. Especially the latter factor breaks with the CPE approach because it grants politicians—and therefore parties—an influence over public opinion, thereby making public opinion endogenous.

In more detail, the endogenous preferences perspective argues that voting behavior is a result of core beliefs as well as perceptions and preferences. Individuals generally vote for the party that best matches their core beliefs, perceptions, and preferences. This is most obvious for the core belief partisan identification, as an emotional attachment to a party serves as a reliable heuristic in the voting booth (Campbell et al. 1960). But the further elements of public opinion also matter for voting behavior. For example, left parties attract voters who hold more egalitarian core beliefs in the abstract, support stronger welfare policies specifically, and perceive the wealth distribution as more unfair.

Even though preferences and perceptions can matter, core beliefs are paramount. First, voters may reliably use their core beliefs to guide their voting behavior as core beliefs are abstract, implying that no fine-grained knowledge about parties and proposed policies is necessary (Goren 2013). In contrast, voting based on preferences and perceptions requires detailed knowledge that only the more sophisticated possess (Goren 2013; Zaller 1992; Zaller and Feldman 1992). Second, preferences and perceptions are endogenous to core beliefs—that is, these core beliefs shape what policies voters prefer and how they perceive the world in the first place (Bartels 2002). For example, voters deduce their support for the welfare state from their egalitarian values (Feldman 1988).

While core beliefs are vital, Fig. 2 also expects political elites such as parties to influence political attitudes. This may seem minor but it grants political elites influence over who people vote for. The reason is that political elites can influence how people perceive the world and what policies they favor, which (partially) influences their voting decision. For example, we will see in the empirical evidence section that voters readily align their policy preferences to the positions taken by the political party they voted for (e.g., Slothuus 2010; Slothuus and Bisgaard 2021a).

Why are people influenced by political elites? The reason is that—just as core beliefs—elite communication serves as a heuristic that simplifies political behavior (Goren 2005; Zaller 1992; Zaller and Feldman 1992). As discussed, most people lack the political sophistication necessary for deliberate policy voting. People possess limited information about political issues and often care little about them. As a resolve, people adapt the opinions offered by political elites. The motivation for doing so is twofold (Slothuus and Bisgaard 2021a, p. 897). First, elite opinions simplify the world—once you trust a specific opinion source, it becomes unnecessary to form an independent opinion as you can simply adopt their thinking. Second, elites can become so integral to the self that rejecting their opinion requires rejection of the self, implying that people adopt foreign thought to avoid inner inconsistency. This becomes relevant when people form a strong partisan identification (Campbell et al. 1960).

Zaller (1992) offered perhaps the most nihilistic view of preferences (see also Zaller and Feldman 1992). He theorized that people rarely reveal real preferences in surveys but rather answer questions based on random draws from a basket of currently accessible considerations. People have ideological predispositions (in my terminology: core beliefs), but what considerations are available largely depends on elite rhetoric—that is, the opinions people gather from actors such as political parties via exposure to media. As elite rhetoric shifts to the right (left), people’s baskets will be populated by more right-leaning (left-leaning) considerations, resulting in more right-leaning survey responses.

Zaller’s theory—as well as further political science contributions—help to understand under what conditions people rely on their own thinking to develop preferences and perceptions rather than adapting them from elites, such as parties (Zaller 1992; Zaller and Feldman 1992). First, people with more political sophistication are in a better position to assess policies, implying that they rely less on the input from political elites (Bullock 2011; Goren 2013). Second, people are less likely to adopt positions that contradict their core beliefs, especially as these core beliefs become stronger as well as when the beliefs more relevant for a policy under discussion (Bullock 2011; Cavaillé and Neundorf 2023; Slothuus and Bisgaard 2021b). Third, especially those with a strong partisan identification adapt their preferences to that of their party, implying that weaker or non-existent partisan identification is associated with more independent political thought (Slothuus 2010; Slothuus and Bisgaard 126,127,a, b).

The importance of core beliefs in the model begs the question of how they are formed. The theory displayed in Fig. 2 suggests that they primarily develop during the impressionable years and are therefore based on peoples’ socio-economic origin—that is, their childhood, adolescence, and early adulthood (Jennings et al. 2009; Niemi and Jennings 1991; O’Grady 2019; Rico and Jennings 2016). People adopt their beliefs from their surroundings via a mix of social transfers and independent deductions based on shared circumstances.Footnote 4 In this way, the socio-economic structure experienced during the impressionable years plays a central role in determining people’s political attitudes and—ultimately—voting behavior. For example, young people living in poorer areas with poorer parents who experience more economic hardship tend to develop stronger economic egalitarianism and therefore are more likely to vote for left parties.

3.3 Empirical evidence

The previous section presented a theoretical vision of what public opinion is, how it originates, and how it influences people’s voting behavior. This has obvious implications for government selection and policymaking, even though these phenomena are not included in the model. This section reviews empirical evidence underscoring the endogenous preferences perspective, with a focus on the effect of elite cues on public opinion.

First, empirical evidence suggests that both core beliefs and policy preferences are related to voting behavior. It will be unsurprising for political scientists that partisan identification—a core belief—exerts a strong influence on voting behavior (Bartels 2000; Bonneau and Cann 2015; Campbell et al. 1960; Dalton 2020). The probability that those with a partisan identification vote for their respective party and not for another party is much higher. Goren (2013) shows that further core beliefs also matter. He finds that the vote choices of US citizens can be meaningfully explained by three deeply held policy principles: limited government, traditional morality, and military strength.

The evidence also suggests that policy preferences matter. Several studies find that the probability of those who support more redistribution to vote for left-leaning parties is higher (Gingrich 2014; Quinlan and Okolikj 2020; Rueda and Stegmueller 115,116,a, b, 2019). As the model in Fig. 2 suggests, however, policy preferences are endogenous to core beliefs (Feldman 1988; Goren 2001, 2013; Rokeach 1973). Goren (2013) finds that only the most sophisticated US citizens adapt their vote choice to their preferences once these core beliefs are accounted for.

The remainder of this section focuses on studies assessing the influence of partisan cues and partisan identification on policy preferences. Empirical research offers strong evidence for this relationship. When parties change their policy positions or their policies, especially those with a partisan identification follow suit and adapt their preferences. A first set of studies uses experimental designs to gauge the impact of partisan cues on policy preferences. Broockman and Butler (2017) had US legislators send official communications with randomized policy sentiment to their constituents, who subsequently endorsed these positions even when they were unsubstantiated and clashed with recipients’ prior beliefs. Barber and Pope (2019) show that US republicans endorsed liberal—and therefore ideologically contradicting—positions championed by Donald Trump. Bullock (2011) discovers that people adapt their healthcare spending preferences to positions taken by partisan actors even when these are at odds with their ideology. Nelson’s (2004) review also shows that partisan actors can manipulate the importance attributed to topics and policies.

A second set of studies uses observational data. In his study of the transformation of the British welfare state, O’Grady (2022) demonstrates the increasing hostility of Labour voters towards the welfare state following the ideological realignment of the Labour party under Tony Blair. Several studies use natural experiments based on panel surveys where a party suddenly reversed their policy stance between two panel waves (Nemčok et al. 2023; Slothuus 2010; Slothuus and Bisgaard 126,127,a, b). The results generally show that—in reaction to the policy switch—party supporters suddenly change their policy preferences and align them with the position of their party. For example, supporters of the Danish Liberals and Danish People’s Party—but not the supporters of other parties—became more hostile towards the unemployment insurance after their parties unexpectedly changed their policy stance overnight (Slothuus and Bisgaard 2021a).

Public opinion depends on elite cues, but people do not blindly follow their party. Especially those who are informed about an issue (and therefore do not require “ideological assistance”) and those incentivized to oppose a party position due to their core beliefs or material interests stick with their prior convictions (Bullock 2011; Cavaillé and Neundorf 2023; Slothuus 2010; Slothuus and Bisgaard 2021b).

Overall, these studies offer convincing evidence on the partial endogeneity of policy preferences and downstream phenomena, such as voting behavior. These studies are convincing because they mostly rely on statistical designs that credibly identify causal effects. They use a mix of lab, survey, and natural experiments from different countries and points in time. Evidence in the social sciences can hardly become more convincing.

4 What can CPE learn from the alternative theory?

This section asks what CPE researchers can learn from the endogenous preferences perspective. It first illustrates the essentials points where the two theoretical models diverge. The section subsequently carves out opportunities to advance CPE theories inspired by elements of the endogenous preferences perspective.

4.1 Comparison of the two perspectives

The crucial difference is that the exogenous preferences perspective depicts policy preferences—and thereby socio-economic structure—as the deciding factor over the political sphere. Parties are utterly exposed to voters’ policy preferences, to which they have to adjust by skillfully addressing support coalitions of diverse voters with shared interests. The endogenous preferences perspective, in contrast, holds that core beliefs are paramount, and that parties have sway over policy preferences and the perceived importance of issues. The task of parties rather becomes to skillfully develop political communication that convinces voter segments of the benefits of voting for said party. To be successful, parties cannot fundamentally contradict the core beliefs and preferences of their voters. At the same time, they can innovate and convince voters that—against the backdrop of current socio-political circumstances and developments—their vision will translate voters’ beliefs into action.

I will use the Robin Hood Paradox as an example to illustrate the difference. It poses the question of how it is possible that disposable income inequality is on the rise in democracies although a majority of citizens should enforce inequality reduction motivated by both material self-interest and distributive justice concerns (Ahrens 2022b; Dimick et al. 2017; Meltzer and Richard 1981).

In line with the exogenous preferences model, Meltzer and Richard (1981) theorize that relatively poor voters support more redistribution compared to the relatively rich because the relatively poor can expect material benefits (see also Romer 1975). The implication is that higher income inequality—implying greater benefits of redistribution for the relatively poor—should result in greater demand for redistribution. The voting mechanism translates this demand into parties taking a more pro-redistribution stance, pro-redistribution parties winning elections, and governments redistributing more via social transfers and taxation. Contemporary CPE theory also suggests that the relationship between inequality and demand for redistribution may result from social affinity or fairness considerations (Ahrens 2022b; Dimick et al. 2017, 2018).

In contrast, the endogenous preferences perspective is more open and hinges on the communication strategies of parties. To be sure, the electoral viability of party strategies depends on socio-economic realities (rising inequality) and public opinion (such as egalitarian values), with more egalitarian voters tending to support parties that favor more redistribution. However, whether redistribution will actually be implemented depends on whether (left) parties endorse redistribution and whether their framing efforts resonate with voters. Furthermore, the communication strategies of political adversaries and their persuasiveness matters. Overall, socio-economic structure and public opinion matters, but it is open to interpretation.

The endogenous preferences perspective can be used for potential explanations of the absence of Robin Hood policies. Right-leaning parties—in partnership with the business lobby—have shown to be successful in convincing voters of the perils of redistributive policies such as wealth and inheritance taxes (Emmenegger and Marx 2019; Hilmar and Sachweh 2022). Concurrently, left parties may not focus their strategies and political communication on fiscal redistribution via taxation because they lack policy knowledge and prematurely shy away from the political headwind that right parties will generate (Elsässer et al. 2023; Fastenrath et al. 2022; Fastenrath and Marx 2023). In summary, there might be a potential for redistribution, but right parties effectively combat it and left parties fail to politicize the issue.

4.2 Lessons for CPE research

The theoretical contradictions are substantial, and one could assume that the best course of action is to favor one model over the other based on the available empirical evidence. However, it is important to remember that theoretical models constitute idealistic conceptions, which necessarily simplify the political sphere. The theoretical contradictions are less consequential than one may assume because both models contain valuable truths about the world.

In this section, I argue that CPE research employing the exogenous preferences perspective can learn from the endogenous preferences perspective. I discuss in what areas CPE can incorporate insights from the endogenous preferences perspective, focusing on the topics of political attitudes, supply and demand, theoretical determinism, and stability and change.

4.2.1 Political attitudes

CPE research could devote greater attention to core beliefs such as egalitarianism rather than specific policy preferences, for example support for unemployment insurance. The endogenous preferences perspective suggests that more abstract beliefs guide political behavior and the way people make sense of the political world. In contrast, policy preferences are flimsy, easily manipulated, and only shape voting behavior among the most sophisticated. In many cases, people cannot be expected to articulate substantiated policy preferences because they lack knowledge about socio-economic realities (Engelhardt and Wagener 2018; Fernández-Albertos and Kuo 2018) and the functioning of policies (Geiger 2018; Jensen and Wenzelburger 2021).

4.2.2 Political supply and demand

A core lesson is that the political supply side is important even for analyses that focus purely on the demand side. The mechanism that links public opinion to voting behavior, government selection, and policymaking hinges on how and to what extent particular issues are politicized by political actors in a polity. Even when there is latent support for political decisions—such as reducing income inequality—this support may not be funneled into the political process if parties fail to focus their communication and policy strategies on this issue. Across countries and years, it is not guaranteed that issues receive public attention and that all policy opportunities are put on the table.

4.2.3 Theoretical determinism

The endogenous preferences perspective suggests that politics and policymaking are less predetermined than assumed by the exogenous preferences perspective. The exogenous preferences perspective leaves little room for interpretation, as the preferences that shape political behavior, politics, and policies are deeply rooted in the socio-economic structure. Therefore, the structure decides over political phenomena, at least in the long term. The endogenous preferences perspective rather recommends to perceive politics and policymaking as open, as parties and governments have considerable leeway in their strategies and behavior. Public opinion and structural factors surely are important, but they are not binding. The element of public opinion that shapes political outcomes to the most extent—core beliefs—is too vague to narrowly define a set path.

4.2.4 Stability and change

The endogenous preferences perspective suggests that politics can both be more stable and unstable compared to the exogenous preferences perspective. The exogenous preferences perspective expects stability because the socio-economic structure—and thereby preferences—change gradually. The endogenous preferences perspective, however, suggests that there might be even more stability because the element of public opinion that matters most—core beliefs—changes only gradually over time (Inglehart 1977, 2008). Turning to change, both perspectives expect that fundamental shifts are induced by shocks, such as financial crises (Limberg 2019, 2020). The exogenous preferences perspective cites changing policy preferences, but this approach underestimates the possibility for change because it does not consider that shocks offer parties an opportunity to assert a new interpretation of the right path forward. That is, shocks allow parties to shape how voters evaluate the socio-economic reality and what solutions they favor. Effective political entrepreneurship can therefore enable more fundamental change after shocks than expected by the exogenous preferences perspective.

5 A theory of public opinion as an elastic corridor

The pressing question remains of how public opinion shapes politics and policies. This section develops the fundamentals of a theory that answers this question. It incorporates the insights discussed above and formulates how and to what extent public opinion matters. The theory draws from Busemeyer et al. (2020), Grossman and Guinaudeau (2021), and the party politics literature on issue competition and agenda-setting (e.g., Dennison and Kriesi 2023; Green-Pedersen 2019; Green-Pedersen and Mortensen 2010). I argue that public opinion can be conceived of as an elastic corridor that constrains the strategies and behavior of political parties (which reverberates the “snakes in a tunnel” metaphor from Grossman and Guinaudeau 2021). Public opinion shapes downstream political phenomena while remaining malleable and leaving wiggle room for parties.

Public opinion influences political behavior—and therefore politics and policies—because people (partially) base their voting behavior on their policy-related attitudes, in particular their core beliefs such as egalitarianism and to a lesser extent their policy preferences. Voters electorally reward (punish) parties that offer a political vision and policies that coincide (contrast) with their beliefs and preferences. However, especially core beliefs are abstract and require interpretation to translate them into specific policies. This is what parties do—they develop communication strategies and policy packages and try to convince voters that these coincide with their beliefs.

Due to this inaccuracy, public opinion can be characterized as a corridor that defines an opportunity space for parties. Parties are incentivized to stay within the corridor, as they can expect electoral rewards for doing so. But the corridor can be wide, and parties may shape public opinion within its walls to attract as many voters as possible. Parties fight over what issues are important by making them salient, the best course of action regarding policies, and also perceptions of the world, such as candidate competence or fairness evaluations. Since public opinion offers a corridor, there is leeway.

The relevance of public opinion therefore hinges on the position and width of the corridor. These depend on several factors. The position of the corridor obviously depends on the current attitudes of the public. It will include more left-leaning positions when public opinion currently leans more to the left. But the exact position differs between parties as they target different voters. For example, left parties are incentivized to adopt a more egalitarian stance because their voter base is more egalitarian (Adams and Ezrow 2009; Hibbs 1977; Schmidt 1996).

The width of the corridor depends on the saliency and complexity of an issue. The corridor is narrower and constrains parties more when an issue is salient and perceived as important (Busemeyer et al. 2020; Green-Pedersen 2019). When a particular issue such as unemployment is currently salient, parties are incentivized to follow the attitudes of the public more closely. In contrast, parties have more leeway when issues are hidden from the public eye (Culpepper 2010), which weakens the relationship between public opinion and downstream phenomena. Similarly, the corridor becomes narrower when an issue is less complex, which implies that voters are actually able to reason about an issue and form independent preferences. For example, Culpepper (2010) outlines that the regulation of corporate takeovers is largely developed independently from public opinion, as voters are unable or unwilling to engage critically with this complex issue.

Especially the latter point suggests that the corridor depends more on voters with high political sophistication because they are in a better position to reason about issues. Put differently, the opportunity space they create is narrower because they are less hindered by the complexity of the political world. The opinions of more sophisticated—and thereby better educated and wealthier—voters should therefore have a stronger impact on politics and policies.

The public opinion corridor constrains parties, but it remains elastic—it is possible for parties to fundamentally change how the public thinks about an issue and thereby shape the corridor. This is possible when parties are able to assert a new interpretation of a problem and the necessary solutions for it. Shaping the corridor does not require that parties change the core beliefs of voters. Voters rely on a mix of different beliefs and preferences that are not fully consistent (Converse 2006). Parties can therefore establish a new interpretation of an issue that speaks to different beliefs held by the voters. Such fundamental change should be especially likely after exogenous shocks.

6 Implications for specific literature strands

I close the review with a discussion on three literature strands in contemporary CPE research: unequal representation, policy preferences, and economic voting. I outline the implications of the endogenous preferences perspective for these literatures and derive recommendations regarding established theories and future research avenues. The goal is not to criticize these literatures but rather to identify how they may be developed in the future. I focus on these particular literatures because they coincide with my research interests and expertise, and I expect that the lessons from the endogenous preferences perspective extend beyond these.

First, research on unequal representation has discovered that there is a larger congruence between public opinion and policy outputs for richer and better educated citizens. Researchers argued that mechanisms such as lobbying, disparities in political participation, politicians’ misperceptions of voter preferences, and inequalities in descriptive representation may drive this inequality. However, the exact mechanisms remain an open question (Elkjær and Klitgaard 2021; Elsässer and Schäfer 2023).

The elastic corridor theory proposes another mechanism that could drive empirical patterns. Unequal representation may be a result of the stronger political sophistication that the educated and wealthy tend to have. The theory argues that parties and governments have a stronger incentive to respond to more sophisticated voters because they not only rely on abstract core beliefs but also base their voting behavior on more specific policy preferences. Moreover, the politically sophisticated have a better understanding of what policies the government implements, for example because they follow the news. Due to retrospective voting, governments have a stronger incentive to implement policies that are congruent with the preferences of elite citizens.

Second, there is a large literature on policy preferences in CPE that focuses on inequality-related attitudes, above all preferences regarding redistribution and the welfare state (e.g., Alesina and La Ferrara 2005; Cavaillé 2023; Rehm 2009, 2011; Rueda and Stegmueller 2019). The analysis of policy preferences is conducted because preferences both normatively should and empirically do influence phenomena such as politics and policies according to CPE.

The endogenous preferences perspective suggests that public opinion research could devote greater attention to core beliefs such as policy principles (Goren 2013). It argues that policy preferences only shape politics and policies among the most sophisticated. Most people rely on their core beliefs—as well as cues from political elites—in their political behavior. These elements of public opinion have been largely neglected in CPE research despite their importance. For example, it is intriguing what core beliefs people rely on when they deal with inequality-related issues as well as how these beliefs have developed over time and whether they are malleable.

Third, the economic voting literature expects that people rely on retrospective voting. Most research focuses on retrospective economic evaluations, arguing and finding that the probability of voters continuing their support for incumbent parties increases when the economy is running well (Duch and Stevenson 2008; Lewis-Beck and Stegmaier 2013). However, the theory also expects that the past policy performance of governments should guide retrospective voting behavior. The expectations that welfare cutbacks and tax hikes should be punished at the subsequent election are especially common in CPE research (e.g., Pierson 1994, 1996).

The endogenous preferences perspective suggests that policy-based retrospective voting may have limited relevance because of partisan bias. That is, retrospective policy evaluations are at least partly endogenous to peoples’ partisan attachment and further core beliefs. People who vote for a specific party because of their partisan identification and, for example, egalitarian beliefs also evaluate the policies implemented by said party more favorably. What is more, the endogenous preferences perspective highlights that people seldom have well-grounded policy preferences and that their attention to politics is limited, which removes a core requirement for retrospective voting. The endogenous preferences perspective therefore offers an explanation for why empirical research often fails to find policy-based retrospective voting (e.g., Ahrens and Bandau 2023; Giger and Nelson 2011).Footnote 5