Solving the (False) Dilemma: An Ecological Approach to the Study of Opinion Constraint

Abstract

Since Converse’s paper, opinion constraint has been defined as the degree to which voters hold ideologically consistent opinions across different issues. Yet, scholars have found that opinions departing from the liberal/conservative categories constitute alternative ways of organizing political preferences. This suggests a methodological dilemma: how can we assess the consistency of opinions based on empirical, rather than theoretically predefined, criteria? This article proposes measuring constraint as the extent to which citizens’ policy preferences resemble those of their most preferred political parties (a top-down approach). To do so, it relies on data from the 2019 European Election Studies and the 2019 Chapel Hill Expert Survey. Analyses show that a top-down measure of opinion constraint correlates weakly with pre-existing measures of this concept (discriminant validation). Findings also suggest that well-established hypotheses about the predictors and effects of constraint are confirmed when using the top-down measure (nomological validation).

Keywords

opinion constraint issue positions public opinion political parties European countries

Introduction

Converse (1964) claimed that ideologically sophisticated voters should exhibit an ideological, coherent, constrained belief system.¹ The belief system is ideological because it relies on abstract concepts, coherent because attitudes are bound together, and constrained because it is anchored to pre-existing individual predispositions (i.e. ideological preferences). In this early conceptualization of opinion constraint,² that is, the capacity to rely on a consistent set of ideas, sophisticated voters are expected to build their attitudes along a single dimension defined by the liberal-conservative continuum. However, Kinder (1982) argued that the role of ideology in structuring voters’ attitudes is negligible, which suggests that the public is mostly ‘innocent of ideology’. In this regard, research has demonstrated that people act and think in many different ways (Conover and Feldman, 1984; Treier and Hillygus, 2009), often diverging from a single and a priori-defined ideological dimension. Indeed, it has been found that heterogeneity is systematic in the electorate and that opinions not aligning with the mainstream liberal-conservative structure often represent alternative, rather than unsophisticated ways of organizing political preferences (Baldassarri and Goldberg, 2014).

Against this backdrop, it seems particularly important to rethink the measure of opinion constraint in the political science literature. Traditional indicators systematically underestimate the political skills of those who deviate from a one-dimensional ideological continuum. However, there is large evidence that the public does not conform to a single dimension to organize their political thoughts (Van der Brug and Van Spanje, 2009). In addition, past techniques also consider individuals in a ‘vacuum’, as they do not consider the context in which voters have to form their opinions. Research has instead stressed that political parties play a central role in guiding public opinion formation. The primary ingredient of public opinion is political predispositions (Kinder, 1998), but ‘the transition from individual predispositions to political choices does not necessarily come naturally’ (Leeper and Slothuus, 2014: 132). The political contexts, and in particular political parties, have the crucial role of activating predispositions and facilitating their application to political decisions. In doing so, political parties not only provide the alternatives among which voters are called to choose, but also provide cues about how those alternatives should be understood.

In light of the apparent lack of one-dimensional ideological thinking in the electorate, and in light of the mobilizing role of parties, an indicator of opinion constraint defining what is ‘(un)constrained’ solely based on the adherence to the liberal-conservative continuum, and without accounting for the role of political parties, fails to comprehensively grasp the concept under study. In updating the measure of opinion constraint, the aim is to do the following: (1) be inclusive of those opinions that depart from the traditional ideological continuum and (2) create a context-related index, where parties’ structure of opinions and their influence on voters are taken into account.

To do this, I avoid predefined ideological standards and consider instead parties as empirical criteria to assess opinion constraint as they structure the ‘ecology’ in which voters organize their political attitudes. I argue that political parties are a valid benchmark to understand the consistency of voters’ opinions and propose a new empirical strategy to build the index of opinion constraint. To do that, I rely on data from the 2019 European Election Studies (EES; Schmitt et al., 2022) and the 2019 Chapel Hill Expert Survey (CHES, Bakker et al., 2020). The index is validated through a three-step procedure (Adcock and Collier, 2001). Findings suggest that a top-down measure of opinion constraint correlates only weakly with pre-existing measures of this concept, providing evidence for discriminant validation. Regression models also confirm that well-established hypotheses about the predictors and effects of constraints when using the top-down measure (nomological validation). In particular, gender, age, education, occupation, interest, and knowledge are significant predictors of constraint. In terms of outcomes, those who hold constrained opinions are more likely to vote for the party they feel close to, less likely to engage in vote switching between elections and have a better understanding of where parties are on the left/right axis.

A New Definition of Opinion Constraint: The Mobilizing Role of Parties

As stated by the ecological theory of ideological consistency (Sniderman, 2017), political parties are the lenses through which voters make sense of the political world, as they provide shortcuts, build the debate, construct the discourse, and offer cues (Sniderman, 2017; Zaller, 1992). To develop a coherent system of ideas, voters do not need to – as maintained by traditional approaches to the study of opinion constraint – master abstract political concepts, because parties do this for them. Indeed, research on opinion formation demonstrates that the way in which the public forms (or changes) policy preferences is greatly influenced by how political parties present and interpret issues, the so-called party framing effects (Slothuus, 2010). This literature clarifies that voters do not form opinions solely based on their own judgement, but taking into account the cues they receive from political elites (Petersen et al., 2010). In the real world, party competition on policy issues exposes the public to multiple issue frames. However, there is large evidence that voters are inclined to follow the issue frame promoted by the party they feel attached to (Seeberg et al., 2017; Slothuus, 2010; Slothuus and Bisgaard, 2021; Slothuus and De Vreese, 2010), which means that they are likely to form a policy preference consistent with the issue frame of their closest party. Therefore, the influence of parties on public opinion formation happens through the mechanism of partisanship (Sniderman, 2017). Since parties play a key role in public opinion formation, it is reasonable to expect that citizens develop a coherent set of ideas when they are familiar with (and conform to) the discourse of parties.³ In particular, to account for the individual selective exposure to partisan-coherent issue frames, this article uses the most preferred party as benchmark to evaluate voters’ consistency. Narrowing down the focus to the consistency with the most preferred party permits to cancel out the noise generated by crosscutting policy frames and focus specifically on the partisan-motivated reasoning. This brings the advantage of working with a measure that truly resembles real world conditions, in which voters do not process all cues that are available in the party system, but rather shape (or, in more rare occasions, update to respond to changing party positions, see Slothuus and Bisgaard, 2021) their opinions by looking at the stances of their most preferred party.

In this vein, opinion constraint is understood as the extent to which voters conform to the political preferences (i.e. issue configurations) of the party they most like. Since this operational strategy implies benchmarking individual opinion consistency to the characteristics of higher-level entities (i.e. political parties), I refer to this as a top-down index of constraint. Technically speaking, analyses in this article first calculate policy issue structure based on parties’ and voters’ positions, respectively. Then, they compute the distance between parties’ and voters’ structure of policy opinions. The more individuals deviate from their most preferred party’s positions on issues, the more they are considered unconstrained (see the methodological section and the Online Appendix for further details).

Origins of Opinion Constraint and Its Empirical Indicators

Converse (1964: 207) defined opinion constraint as ‘the success we would have in predicting, given initial knowledge that an individual holds a specified attitude, that he holds certain further ideas and attitudes’.⁴ In other words, he expected that a liberal position on one issue would predict a liberal position on another issue, and vice versa. This empirical strategy was based on two techniques. First, Converse categorized respondents on the basis of the ideological categories utilized to evaluate candidates (i.e. their level of conceptualization). This method has been criticized for not considering the different communicative and verbal skills among individuals (Marcus et al., 1974) as well as for evaluating individuals’ sophistication of political ideas mostly on a discretional and arbitrary basis. It is likely that the level of conceptualization measure employed in the years following Converse’s paper (Nie et al., 1976; Pierce, 1970) was soon discarded by scholars for these reasons. The second technique correlates attitudes on domestic and international issues, assuming that people rely on the same liberal/conservative continuum to evaluate issues on different areas. Yet, this is not how voters’ minds generally work. As Jackson and Marcus (1975) showed, different individuals might have diametrically opposed ideas and be both perfectly constrained since they respond to different ideological dimensions. The technique based on correlations has the major problem of relying on an a priori definition of the connection between issues.

Moreover, scholars in previous years have pointed out that the use of correlations is an inappropriate measure of opinion constraint. For instance, Weissberg (1976) claimed that correlations do not grasp the levels of attitude consistency as Converse assumed. In the hypothetical case in which all respondents hold conservative positions on issue x and liberal positions on issue y, the measure of covariation will be counterintuitively high since the ordering of respondents is the same on both issues. Similarly, Barton and Parsons (1977) maintained that correlation coefficients are affected not only by the structuring of beliefs, but also by the heterogeneity of populations and that, for this reason, results might be misleading. They indeed showed that ‘if one of the populations compared is essentially homogeneous, the correlation coefficients between the different attitudes in that population will approach zero, even though individuals hold highly consistent or predictable attitudes’ (Barton and Parsons, 1977: 164).

However, some studies have remained coherent with this line of research, carrying on the study of issue attitude consistency using correlations (Baldassarri and Gelman, 2008; Enders, 2019; Tomz and Sniderman, 2005). This branch of the literature has stressed the importance of relying on a multi-dimensional, or ‘hierarchical’, model of ideological constraint (Hurwitz and Peffley, 1987). These studies often combine a horizontal and a vertical measure of opinion constraint, where vertical constraint is understood as the consistency between the ideological orientations (e.g. left/right positions) and issue positions, while horizontal constraint is the consistency among issue positions (Enders, 2019; Federico and Hunt, 2013). In the ‘partisans without constraint’ thesis, those have been defined as issue alignment, that is, correlations between pairs of issues; and ideological alignment, that is, correlation between attitudinal items and the ideological identification (Baldassarri and Gelman, 2008; Kozlowski and Murphy, 2021). These two dimensions of constraint are understood to be related (Judd and Krosnick, 1989), ‘in the sense that horizontal constraint can be thought of as a consequence of the higher-order vertical linkage between issue positions and a central ideological self-placement’ (Federico and Hunt, 2013: 95). Nevertheless, other studies have preferred to focus exclusively on the horizontal measure constraint and have proposed different statistical techniques to evaluate the consistency and stability of issue positions. Among others, Cronbach’s alpha (Bartle, 2000; Sturgis et al., 2005; Valentino et al., 2002), latent class analysis (Linzer, 2006), multi-trait multi-method confirmatory factor analysis (Freeze and Montgomery, 2015), and correlations between factor scores (Ansolabehere et al., 2008).

One Concept, Different Approaches: Towards the Solution of the Dilemma

In the 1970s and 1980s, research investigating opinion constraint moved towards an individual-based approach (Jackson and Marcus, 1975; Marcus et al., 1974) in an attempt to define the ‘personal’ ideological space. These studies clarified two crucial points. First, that constraint cannot be assessed by the degree of correlation between various issue positions since a person can respond liberally on one issue and conservatively on another just because such issue organization is coherent with his or her rationale. Second, that constraint cannot be assessed by correlating issue positions to the ideological outlook because different issues might evoke different ideological positions from the same individual (Jackson and Marcus, 1975). Therefore, a reliable measure of opinion constraint should evaluate the consistency of attitudes on the basis of the general ideological dimensions that a person employs. For example, Jackson and Marcus (1975) proposed to use the ability of an individual to relate consistently a set of specific political issues to their general ideological dimensions (see also Marcus et al., 1974). In their survey, respondents need to evaluate the similarity between 15 political terms and then they are asked to relate each issue to one of the 15 general political concepts on a scale. However, this method has the shortcoming of not being parsimonious (for 15 political concepts, respondents have to make 105 pairs’ judgements). In addition, it does not consider the actual positions of voters on issues, which remains of foremost importance if we want to know how attitudes are structured.

A few years later, this approach was expanded by another strand of research under a more psychological perspective, through the so-called schema theory. Schema theorists have tried to map the structure of voters’ opinions, relying on a broader definition of constraint than the one in Converse. Indeed, as Conover and Feldman (1984: 121) pointed out, ‘people organize their political worlds in richer and more diverse ways than implied by the traditional approaches to mass belief systems’, suggesting that scholars should go beyond the traditional liberal-conservative axis of interpretation. Yet, these studies have been severely criticized for vague measurements, lack of conceptual clarity, and for the limited theory on which they lie (Kuklinski et al., 1991). One of the main limitations depicted by Kuklinski and colleagues is that these studies neglect the social context in which individuals live. As they point out, ‘We are still social beings, and theories of social cognition must eventually take account of the fact’ (Kuklinski et al., 1991: 1346). Because of their limitations, these studies have fallen into disuse. During the nineties, scholarly attention shifted to the study of political knowledge (Delli Carpini and Keeter, 1996) and political awareness (Zaller, 1992), with analyses looking specifically at the level of opinion constraint in the electorate having become rare.

In an effort to overcome a priori expectations in the measurements of constraint, recent works have studied the spatial organization of voters’ ideas. These studies aim at taking ‘Converse’s idea of functional interdependence between issues one step forward’ by conceptualizing ‘a belief system as a network of interconnected political beliefs’ (Baldassarri and Goldberg, 2014: 54). This approach treats voters’ attitudes as an extended system of ideas and not just as connections among pairs of issue positions. By means of relational class analysis, Baldassarri and Goldberg (2014) are able to categorize citizens based on their type of spatial organization. They find evidence of multiple and heterogeneous belief systems, deriving especially from differences in individuals’ social identities. Similarly, Boutyline and Vaisey (2017: 6) ‘interpret a set of survey responses as an empirical manifestation of the belief network, where the belief items are nodes and the associations between the beliefs are weighted ties’. They show that, contrary to Baldassarri and Goldberg’s findings, heterogeneity is the exception and not the rule.

Other branches of study have recently introduced another key element to the study of attitudes’ consistency: the role of elites and parties. Scholars have found that clearer clues (in terms of polarized options or the use of brand names) can support citizens in forming more constrained opinions (Levendusky, 2009; Tomz and Sniderman, 2005). Research focusing on public opinion formation has also stressed that parties are able to activate political predispositions and facilitate their application to political decisions mainly through the partisan-motivated reasoning mechanism (Leeper and Slothuus, 2014). Focusing on consistent political thinking, Sniderman (2017) has developed an ecological theory of ideological consistency, arguing that parties help voters develop their ideological outlook on politics. Contrary to what Converse (1964) maintained, citizens are not required to organize and bring meaning to the political arena and ideological labels, since parties, which have incentives to build policy reputations to ‘sell’ their platform and attract votes, do so for them. In this vein, Sniderman (2017: 51) argued that ‘the right place to start to understand political reasoning is not the psychology of citizens but the constraints of politics – especially the institution of electoral competition through the medium of political parties’.

This article builds upon the ecological theory of ideological consistency and argues that the measurement dilemma (Marcus et al., 1974) derived by the necessity to evaluate what is coherent without imposing a priori-defined standards is in fact a false dilemma. Voters are indeed embedded in a context where political parties as the main actors play the important role of labelling, advertising, and bringing meaning to the political debate. Against this background, opinion constraint should be understood and studied as an attribute of a specific context. Therefore, by focusing on the resemblance between citizens’ and parties’ configurations of policy preferences, it is possible to overcome the long-lasting measurement dilemma and investigate opinion constraint under a new perspective. The next section discusses the creation and testing of a new indicator of opinion constraint in detail.

A New Empirical Strategy: The Top-Down Approach

Starting from the consideration that mass opinions should be studied within the context in which they are embedded, the measure I develop follows a top-down approach. This approach is already familiar to the political science literature. Studies investigating the congruence between the masses and elites have relied on a similar rationale and similar variables (issue positions or ideological placement of both parties/candidates and voters) through different theoretical lenses, that is, correct voting (Lau et al., 2014), representation gap (Dalton, 2020), issue proximity (De Sio and Franklin, 2012), and quality of representation (André and Depauw, 2017). The top-down approach for the study of opinion constraint differs mostly in its purpose: measuring the voters’ system of policy ideas and relating it to parties’ system of ideas. Following Sniderman (2017) and the ecological theory of ideological consistency, this study focuses on the input that voters are provided with in their thinking process (Dalton, 2020), that is, parties’ discourse, starting from the consideration that parties are the key actors that provide voters with cues, labels, and meaning. In particular, as citizens follow the issue framing of the party they feel attached to (Slothuus, 2010; Slothuus and De Vreese, 2010), opinion constraint is measured as the degree to which voters’ system of ideas resemble those of their most liked party.

To create the constraint index, I rely on data from the 2019 EES and the 2019 CHES, which permits having the same issues for both parties and voters (question wording for all the employed variables can be found in the Online Appendix). In particular, the policy issues at stake include the following: state regulation and control of the economy, redistribution of wealth, same-sex marriage, civil liberties, immigration, and the environment. This selection of issues permits covering all the main policy issue dimensions that the literature about political space and issue structure have detected (e.g. Van der Brug and Van Spanje, 2009): the economic dimension (state regulation and control of the economy, redistribution of wealth), the ‘old’ cultural dimension (i.e. lifestyle issues, here same-sex marriage), the ‘new’ cultural issue (civil liberties, immigration), and the environmental issue. For the analyses, I import data from CHES into the EES data set by assigning each individual the issue position of the party they prefer most (for a step-by-step description of how data are imported and how the index is created, see Figure A1 in the Online Appendix). Such preference is calculated on the basis of a respondent’s propensity to vote (PTV). PTVs are variables indicating the electoral preferences (utilities) for all the parties, as respondents are asked to specify the probability of voting for each of them from 0 to 10 (Van der Eijk et al., 2006). These variables permit overcoming the limitation of binary survey questions, which impose upon respondents the choice of a single party option over all the others. Indeed, they are defined as ‘non-ipsative measures’ (Van der Eijk, 2017), providing insight about individuals’ degrees of preferences towards different parties. Moreover, PTVs have been demonstrated to be more strongly related to actual electoral behaviour than other measures of party choice (Van der Eijk et al., 2006).

For those who have a most preferred party (i.e. they attribute the highest utility to just one party), it suffices to identify the party that received the highest PTV and attribute to an individual voter the positions of that party. The majority of respondents in the EES sample (19,324) fall within this first case. However, in the data set, 6785 respondents have attributed their highest utility to more than one party. In this case, the most preferred party has been selected on the basis of the results of a regression model (for simplicity, this analysis has been run in stacked data form,⁵ which permits considering the effects of the independent variables on all parties at once). The most preferred party has been predicted by the ideological proximity (distance between the perceived parties’ positions on the left/right axis and left/right self-placement) and voter–party affinity⁶ based on socio-economic variables (gender, age, education, rural vs urban cleavage, religious affiliation, church attendance, objective and subjective social class, marital status, and citizenry). The highest predicted PTV resulting from this model has been selected and the related party has been categorized as the most preferred party for those respondents who did not have a single highest utility score. As a final step, data have been imported from the 2019 CHES and all respondents have then been matched with the policy positions of the party they most like based on their highest PTV and their highest predicted PTV. The outcome of this process is a ‘respondent*variables’ data set reporting the issue positions of the respondent and of the party that is most preferred by the respondent.

The first step of the analysis consists of estimating the policy issue structure of parties. To do this, six ordinary least squares (OLS) regression models with robust standard errors per country have been estimated using parties’ issue positions from the 2019 CHES (i.e. parties’ positions assigned at the individual level on the basis of the highest PTV). In each model, one issue position is regressed on all the other issues, under the assumption that in a belief system all the ideas are bound together (Converse, 1964). Predicted values are saved for each of the six models, representing the predicted position on issue x given the positions on all the other issues. The same procedure is repeated for issue positions from the 2019 EES, calculating the policy issue structure of voters. I then estimated the absolute distance between parties’ and voters’ predicted values for each issue. To create the index of constraint, the row mean among all six distances were computed. The mean has also been weighted for the salience⁷ at the party level of each issue (see Figure A1 in the Online Appendix for the details about how the weighted mean is computed). This follows the idea that an individual that is not constrained on an issue deemed less important by their most preferred party should not be penalized as much as an individual that is not constrained on an issue that is salient in the discourse of their most preferred party. The final index is recoded and reversed so as to vary between 0 and 10, where 0 means minimum constraint while 10 means maximum constraint. The distribution of the final index of constraint is presented in Figure 1.

Figure 1.

Distribution of the Index of Top-Down Constraint.

Validity Tests

As Adcock and Collier (2001) suggested, political researchers willing to test whether their indicators capture the concept they want to measure can go through different validity steps. First, the content validation, which responds to the question:

does a given indicator adequately capture the full content of the systematized concept? (Adcock and Collier, 2001: 538).

This underlines the need to evaluate whether the indicator contains some inappropriate elements or excludes some necessary elements. As mentioned before, the literature investigating opinion constraint has been crowded and scholars have usually relied on two sorts of variables to measure constraint: individual policy positions and ideological placements. The top-down approach measure of constraint instead uses policy positions of both parties and voters. Since it builds on the ecological theory of consistency, it excludes one element (i.e. ideological orientation), which is redundant in this theoretical framework (voters give meaning to the ideological labels because of parties) and it considers new actors, that is, the parties. By incorporating this new element, the measure is able to consider individuals as embedded in their political contexts, overcoming the shortcomings of previous research focusing on voters’ structure of ideas (Kuklinski et al., 1991). However, this first ‘conceptual’ step of validation is incomplete if considered alone. The next two validity steps can give more insights about how the top-down measure adapts to the concept.

The convergent or discriminant validation asks whether

the scores produced by alternative indicators of a given systematized concept [are] empirically associated and thus convergent (Adcock and Collier, 2001: 540).

For this step, other indices of constraint that have been used in the literature are built and correlations with the top-down measure are checked. A strong correlation would signal that the indicators are measuring the same concept (convergence), while weaker associations would indicate a discriminant validation (Adcock and Collier, 2001). As mentioned above and contrary to the rationale in the literature, the top-down measure does not assume issue positions connect to a single predefined ideological standard, but rather relies on a combination of both parties’ and voters’ issue positions. For this reason, it is reasonable to expect the correlations not to be strong and to point towards a discriminant validation of the top-down index.

To test this, I select six indicators of opinion constraint that have been predominantly used in the recent literature. More specifically, I rely on the three-dimensional measure developed by Federico and Hunt (2013) and replicated by Enders (2019). First, a dimension of vertical constraint computed as the average absolute difference between one’s self-placement on the ideological self-identification scale and their self-placement on the issue attitudes scales. Second, another dimension of vertical constraint that corresponds to the number of instances in which individuals placed themselves on the same side of the issue scale as they placed themselves on the ideological self-identification scale (see also Weissberg, 1976). Third, a dimension of horizontal constraint, that is, the average standard deviation of each individual’s issue attitudes across all the issues (see also Barton and Parsons, 1977). Furthermore, in the literature, other efforts to validate opinion constraint indicators (Tomz and Sniderman, 2005) have relied on five indices: ‘(1) the proportion of citizens who took consistently liberal or consistently conservative positions on all issues (Weissberg, 1976); (2) the proportion who located themselves at exactly the same point on the liberal conservative scale for all issues; (3) the proportion who expressed nearly identical views – all within one point of each other – on all issues; (4) the standard deviations of response scores across a set of issues (Barton and Parsons, 1977); and (5) the absolute distance between positions on issues’ (5).

Indices (1) and (4) are already covered by Federico and Hunt’s (2013) indicators (calculated at the individual rather than at the group level⁸). To those, indicators (2), (3) and (5) have been added as individual measures. To ensure comparability, all indicators are recoded between 0 and 10. Further details about how these indicators have been built can be found in the Online Appendix (Table A1).

Table A2 in the Online Appendix shows the summary statistics of the different indicators of constraint. The top-down constraint has the highest mean (7.27), followed by distance between ideology and issue placement (7.14) and issue distance (6.74). The horizontal constraint has a lower mean (5.94), as well as the other measures of vertical constraint, that is, the number of consistent placement (4.37). This preliminary analysis confirms that constraint tends to be higher than suggested by past techniques when individuals are benchmarked to political parties. In other words, the top-down measure represents a more inclusive measure of constraint. The two other dichotomous variables are the strictest in determining the constraint level of voters: only 1.5% (377 respondents) of the population is deemed constrained according to the exact same position measure and around 2% (611 respondents) when it comes to the almost identical position measure. In addition, to gain insight into the profile of the most constrained group of people, Table A3 in the Online Appendix reports the composition (percentage points) of constrained voters in terms of socio-demographic variables and political interest across all different indicators of constraint. In the analysis, only those with values equal to or higher than 8 for the continuous variables and those with values of 10 for the dichotomous variables are included. The table shows that, across all indicators, constrained people tend to be highly interested in politics and highly educated. They are also predominantly men, as well as between the ages of 25 and 55. This is in line with previous findings about the profile of so-called sophisticated voters, understood as those who have a better understanding of politics (Delli Carpini and Keeter, 1996).

Coming to the convergent/discriminant validation, the results of the correlation analysis between top-down constraint and the other indicators of constraint are shown in Table 1. The measure of horizontal constraint is very highly correlated with issue distance (0.96) and with the indicator of vertical constraint built on the ideology–issue distance (0.60), although only weakly correlated with the other vertical constraint measures (0.19). The two indicators of vertical constraint are related to each other (0.44), but the one based on ideology–issue distance has higher connection with issue distance (0.57). The indicator capturing the placement on the exact same point of the scale on all issues (number (2) in the list above) does not correlate with any of the other variables, while the other similar indicators, that is, holding almost identical positions across issues, is moderately connected to issue distance and horizontal constraint (0.32 in both cases). As for the top-down constraint, correlations are in general quite weak: the highest coefficient is 0.20, which indicates the correlation between the top-down index and the ideology–issue distance (vertical constraint). Correlations are low between top-down constraint and horizontal constraint (0.11), as well as between top-down constraint and the other indicators of vertical constraint (0.10). They are close to 0 between top-down constraint and the measures retrieved from Tomz and Sniderman (2005). On one hand, horizontal constraint, vertical constraint (especially the one built on distances), and issue distance seem to converge towards the same concept (convergent validity); on the other hand, the weak correlations between the top-down index and all the other indicators provide evidence for discriminant validation.

Table 1.

Pearson’s’ Correlations Between the Top-Down Constraint Index and the Other Individual-Level Indicators of Constraint (Correlations Significant at p Value < 0.05).

	Top-down constraint	Horizontal constraint (SD)	Vertical constraint (ideology–issue distance)	Vertical constraint (N of consistent placements)	Issue distance	Almost identical positions on issues	Exact same positions on issues
Indicators from Federico and Hunt (2013)
Top-down constraint	1
Horizontal constraint (SD)	0.111	1
Vertical constraint (ideology–issue distance)	0.203	0.602	1
Vertical constraint (N of consistent placements)	0.103	0.193	0.441	1
Indicators from Tomz and Sniderman (2005)
Issue distance (5)	0.078	0.963	0.574	0.218	1
Almost identical positions on issues (3)	−0.008	0.328	0.198	0.057	0.324	1
Exact same positions on issues (2)	0.031	0.036	−0.015	0.054	0.020	−0.019	1

The third and last step consists in the nomological/construct validation, which asks:

in a domain of research in which a given causal hypothesis is reasonably well established, [. . .] is this hypothesis again confirmed when the cases are scored with the proposed indicator for a systematized concept that is one of the variables in the hypothesis? (Adcock and Collier, 2001: 542).

For this validation, it is useful to consider constraint both as a dependent and an independent variable, since in the literature it has been studied from both angles. When considering opinion constraint as a dependent variable, Luskin’s (1990) seminal paper comes to mind. He proposed an equation to study the predictors of sophistication,⁹ accounting for age, education, intelligence, occupation (personal and parental), political interest, and exposure to media. Building on this, I run a multilevel regression model with random intercepts per country to check whether these variables confirm to be predictors of constraint. Relying on the available information in the 2019 EES (see the Online Appendix for question wording), I regress the top-down constraint on age, education, occupation, political interest, and campaign attention, plus gender as control variables. Moreover, I include in the model an indicator of political knowledge. Previous research has indeed demonstrated that high levels of political information are associated with more structured attitudes (Federico and Hunt, 2013; Lupton et al., 2015; Zaller, 1992). As shown in Table 2, all variables proposed by Luskin (1990) exert a significant effect on the index of opinion constraint, with the exception of gender. In particular, age, education, and political interest affect constraint positively. The same applies to political knowledge. As for campaign attention, the effect is significant but negative. This does not differ much from Luskin’s results, who found no significant direct effect of media exposure on sophistication. In this specific case, due to data limitations, campaign attention to the 2019 European campaign has been used as proxy of media exposure, which might not exactly capture the concept of media exposure in general. This finding suggests that more attention during the 2019 European campaign has resulted in less constraint. Moreover, occupation does have an effect on constraint: those working in private industry and public and private services show to have higher constraint compared to those working in agriculture (reference category). Overall, the results suggest that well-established hypotheses about the predictors of opinion constraint are confirmed when using the top-down constraint index.

Table 2.

Model of the Predictors of Top-Down Constraint: Coefficients of Multilevel Regression Model.

	Top-down constraint
Gender (female)	0.016
Gender (female)	(0.019)
Std. coefficient	0.006
Age	0.007***
Age	(0.001)
Std. coefficient	0.082
Education	0.130***
Education	(0.017)
Std. coefficient	0.056
Interest	0.049***
Interest	(0.013)
Std. coefficient	0.03
Campaign attention	−0.006*
Campaign attention	(0.003)
Std. coefficient	−0.013
Knowledge	0.067***
Knowledge	(0.013)
Std. coefficient	0.039
State industry	0.06
State industry	(0.072)
Std. coefficient	0.011
Private industry	0.174***
Private industry	(0.065)
Std. coefficient	0.057
Public services	0.159**
Public services	(0.065)
Std. coefficient	0.05
Private services	0.123*
Private services	(0.065)
Std. coefficient	0.039
Other	0.099
Other	(0.066)
Std. coefficient	0.027
Constant	6.320***
Constant	(0.138)
Country variance	0.55
Country variance	(0.074)
Log likelihood	−27,575.822
AIC	55,179.64
BIC	55,288.06
Observations	17,057
Number of groups	28

Standard errors in parentheses.

AIC: Akaike information criterion; BIC: Bayesian information criterion.

p < 0.1; **p < 0.05; ***p < 0.01.

Turning to the outcomes of opinion constraint, the literature has often associated constrained belief systems with stable opinions (Ansolabehere et al., 2008; Bartle, 2000). Due to EES data’s cross-sectional design, variations in voters’ policy opinions over time are not observable, but it is possible to study whether there is stability in electoral behaviour. Building on the literature investigating the effects of political sophistication (of which constraint is considered to be one of the components, see, for example, Luskin, 1987), we know that high levels of political expertise are associated with lower likelihood of switching (Dassonneville, 2012; Dassonneville, 2014). Since it is not possible to distinguish between campaign and inter-election volatility in this context, the variable for switching is simply indicated by the match between the two vote variables included in the 2019 EES, that is, the party voted for at the last general elections (recall-question) and the party voted for at the 2019 European elections. I expect that those with higher constraint are less likely to switch between elections. In addition, the relationship between sophistication and understanding parties’ positions has been studied as well (Gerber et al., 2015). In this regard, I expect more constrained voters to place parties correctly on the left/right scale. This has been calculated as perceptual agreement, that is, ‘absolute distances between the respondent’s placement of each of his or her society’s political parties on a 10-point ideological scale and the mean placement of those parties by the remaining respondents from that society’ (Gordon and Segura, 1997: 133). Furthermore, sophisticated voters are said to be able to rely more effectively on their political preferences when casting a ballot. In particular, political sophistication ‘affects the way in which attitudes guide political behaviour’ (de Vries et al., 2011: 3), since political sophisticates have more accessible information on which rely for their decisions. In this regard, I expect those with a more constrained system of ideas to be more likely to link their votes to their political preferences, especially with respect to their party identification. Thus, I regress the match between the party one feels close to, and the party voted for at the 2019 European elections on the top-down constraint. Overall, these analyses tackle three different aspects of constraint’s outcomes: the stability of choices, the actual understanding of the political system and the link between preferences and behaviour. Three regression models have been run: a multilevel model with random intercepts at the country level for the perceptual agreement and two binary logistic multilevel models for the other variables. Moreover, in the models I control for socio-demographic characteristics (gender, age, and education) and other variables that have been used in the literature as indicators of political sophistication, that is, political interest and political knowledge (de Vries et al., 2011; Lachat, 2007).

Results for the three models are presented in Table 3. In all three models, the top-down measure of constraint has a significant effect in the expected direction: a negative effect on vote switching (Model 2) between elections, and a positive effect on the match between party identification and vote (Model 1) and perceptual agreement (Model 3). In the first model, the party ID/vote match is also affected by age and political interest, while in the case of switching, I found socio-demographic variables (gender and age) and political interest have an effect when controlling for top-down constraint. The perceptual agreement model is affected by constraint, gender, age, education, and knowledge. Overall, findings indicate that those who hold high constraint are more likely to vote for the party they feel close to, less likely to switch between elections and they have a better understanding of where parties place on the left/right axis. For the two binary multilevel logistic regression models, the odds ratios confirm that there is a relationship between constraint and the match between party identification and vote switching. More specifically, an odds ratio of 1.06 for the party ID/vote match shows that there is a 6% increase of the probability to vote for the party one feels close to for a 1-unit increase of opinion constraint. The odds ratio of 0.96 for vote switching indicates that there is a 4% decrease of the probability to switch for a 1-unit increase of constraint. To visualize the relationships between top-down constraint and the three dependent variables, Figure 2 shows the model-based predictions for this set of regression models. With regard to party ID/vote match and perceptual agreement, the effects displayed in the graphs are significant (confidence intervals at the extremes of the constraint scale never overlap). With regard to switching, confidence intervals overlap very slightly (they collide over only 0.2 points on a probability scale from 0 to 1) and we can therefore conclude that all effects are confirmed to be mostly significant.

Table 3.

Model of the Effects of Top-Down Constraint: Coefficients of Two Binary Multilevel Logistic Regression Models and a Multilevel Regression Model.

	Model 1: match party ID and vote	Model 2: switching	Model 3: perceptual agreement
Gender	−0.018	−0.095***	−0.101***
Gender	(0.042)	(0.03)	(0.015)
Std. coefficient	−0.19	−0.096	−0.045
Age	0.0208***	0.005***	0.004***
Age	(0.001)	(0.000)	(0.000)
Std. coefficient	0.746	0.187	0.072
Education	−0.04	−0.027	0.215***
Education	(0.038)	(0.026)	(0.014)
Std. coefficient	−0.05	−0.31	0.112
Interest	0.225***	0.037**	−0.003
Interest	(0.027)	(0.018)	(0.009)
Std. coefficient	0.382	0.064	−0.002
Knowledge	−0.015	0.019	0.033***
Knowledge	(0.028)	(0.02)	(0.001)
Std. coefficient	−0.02	0.029	0.023
Top-down constraint	0.058***	−0.03**	0.077***
Top-down constraint	(0.016)	(0.012)	(0.006)
Std. coefficient	0.171	−0.083	0.094
Constant	−1.21***	0.359**	6.517***
Constant	(0.217)	(0.155)	(0.098)
Country variance	0.36	0.19	0.13
Country variance	(0.100)	(0.053)	(0.037)
Log likelihood	−6819.9923	−13,019.686	−27,814.515
AIC	13,655.98	26,055.37	55,647.03
BIC	13,714.91	26,118.43	55,717.64
Observations	11,676	19,583	18,874
Number of groups	28	28	28

Standard errors in parentheses.

AIC: Akaike information criterion; BIC: Bayesian information criterion.

p < 0.05; ***p < 0.01.

Figure 2.

Predicted Probabilities of Party ID/Vote (Based on Model 1) and Switching (Based on Model 2) at Different Levels of Top-Down Constraint and Linear Predictions of Agreement (Based on Model 3) at Different Levels of Top-Down Constraint.

Altogether, all well-established hypotheses from the political sophistication literature seem to be confirmed when the top-down constraint is used. This offers evidence for the nomological validation.

Discussion

The aim of this article is to test a new index of opinion constraint that can overcome the problems related to a one-dimensional ideologically defined measure of opinions’ consistency. By relying on the ecological theory of ideological consistency, which suggests that parties provide voters with ideological meanings, voters’ constraint is understood on the basis of parties’ configurations of policy preferences (Sniderman, 2017). Against this background, a new index of constraint is built using a top-down approach and defining opinion constraint as the extent to which voters conform to the political discourse of the party they most like. The resulting top-down index has shown to correlate weakly with other pre-existing indicators of opinion constraint that rely on internal consistency across issue positions or between ideological orientations and issue positions. The finding suggests that discriminant validation occurs, confirming that the top-down constraint grasps a different concept than the past techniques. Analyses have also tested well-established hypotheses on this index, considering the top-down constraint both as a dependent and an independent variable. In terms of predictors of constraint, gender, age, education, interest, knowledge, and occupation have confirmed to have an impact on constraint (campaign attention does too, but with a negative sign). In line with previous research, top-down constraint is found to have an effect on vote switching, knowledge of parties’ placement and match between party identification and the actual vote. This has confirmed a nomological validation of this measure.

This article started from the consideration that the traditional approach to studying opinion constraint did not adapt to the real world since it considered organization of ideas that do not fall within theoretical expectations as non-ideological (Converse, 1964). The aim of this article was to solve this measurement dilemma and to respond to the question: how can we study voters’ consistency of opinions while accounting for the fact that people mostly do not rely on a single ideological dimension to build their opinions? This is not a negligible question, as revising the measure of constraint has fundamental implications in the field of electoral and public opinion studies. First, this permits considering the heterogeneity of the electorate. Although it has been clear for a long time that ideology could not satisfactorily predict voters’ preferences, studies have not adapted their expectations. This means that the minimalist interpretation of the capacity of the electorate (e.g. Converse, 1964) might have forgotten the most important point: that voters do not necessarily need to be one-dimensional or rely on uniform ideological criteria to be politically competent. Not by chance, when the focus moves away from measuring consistency with specific ideological dimensions, the proportion of those who are constrained clearly increases (see Table A2 in the Online Appendix). This can open the door to new interpretations of the capacity of the electorate, and perhaps to the fact that voters can understand politics more than what we are used to think.

In addition to the heterogeneity of the electorate, there is a second important implication that should be discussed. As the top-down measure of constraint includes political parties in its computation, it provides an indicator that is context-sensitive. While existing research has acknowledged that consistency comes from the discourse of the elites (Converse, 1964), it has not taken this into account in its operationalization. We know, however, that the meaning of political labels and ideologies can dramatically vary across contexts (Piurko et al., 2011) and that discussion over policy issue preferences takes different forms (in terms of characteristics of the discourse and salience) according to the historical period and the political history of a given country. Issue preferences therefore emerge from the interaction between individuals and their political environment (Druckman and Lupia, 2000), which is unique in space and time. Not accounting for this means not only neglecting the importance of the electoral and party context, but also most importantly assuming that being constrained means the same thing across different countries. Ultimately, relying on a measure that is both more inclusive and context-sensitive provides a tool through which we can respect both individual and contextual heterogeneity at the same time.

The top-down approach permits overcoming the criticisms of past measures while also offering an operationalization that is easily applicable to different contexts, since it relies on variables generally available in electoral and expert surveys and it is applicable to any party system. One of the main advantages of this index is that it measures voters’ opinion consistency by accounting for multiple aspects of the electoral environment cueing their opinions: how parties frame policy issues, how and if voters listen to them, whether certain issues are more salient than others in parties’ discourse, if parties modify over time their issue framing and whether voters follow them. The top-down index indeed permits to rely on a measure that resembles real-world conditions, where people are not in a vacuum but respond and adapt to the characteristics of their political context that can change over time. Analyses investigating the relationship between elites and citizens (to cite but a few, studies of electoral mobilization, party rhetoric, issue ownership, representation gap, campaign effects and voting behaviour) can benefit from such a context-sensitive measure. In particular, this index is useful not only to study people’s opinion consistency per se, but also to investigate its relationship with voters’ attitudes vis-à-vis the political system. As the opinion formation literature points out, voters’ receptivity of party messages is often conditional to whether parties are perceived as trustworthy actors (Slothuus and Bisgaard, 2021). In this sense, top-down constraint can be related to dimensions that were not specifically tackled by traditional measures, for example, satisfaction with the electoral menu or with democracy in general, trust in political institutions and actors, and external political efficacy.

Moreover, the use of this measure can be expanded to the broader literature investigating Politics and International Relations. By aggregation at the country level, the top-down constraint gives a measure of how well in a given society voters relate their issue positions to their most preferred party, which we could define as voters’ goodness of fit to the party system. As parties effectively shape public opinion especially when they are trusted (Slothuus and Bisgaard, 2021), the aggregate measures of top-down constraint can be used as a proxy of voters’ trust in political actors or, in broader terms, voters’ level of satisfaction with the electoral menu and can be studied in connection with different macro-level variables to explore several research questions. For example, do democracies with higher levels of constraint show higher electoral participation and involvement of people in the deliberation process? What are the contextual characteristics (effective number of parties, electoral rules, etc.) that favour higher levels of constraint in a society? Are highly constrained societies more ideologically polarized? In constrained systems, are parties’ policy platforms more distinguishable? Do constrained societies show higher or lower levels of political violence, for example, do politicians rely on negative rhetoric? and do they make use of uncivil language or false information? The top-down approach therefore allows for analyses that are able to account for the different characteristics of the voter–party relationship, but also enlarges the spectrum of the variables to which opinion constraint can relate, expanding its relevance beyond its traditional narrow scope.

Yet, it is also important to consider that this research comes with a number of limitations. First, it does not consider other actors that could influence voters in forming their political opinions. As individuals are embedded in a specific social and political context, voters can indeed receive cues from a variety of actors, for example, unions, churches, social media, social groups, family, and so on. In addition, it does not include political elites, actors that have been considered as baseline to evaluate voters’ consistency of opinions in past research (Converse, 1964). However, the lack of sufficiently extended policy issue data in the European context that could be comparable between candidates and voters makes political parties the most reasonable choice in terms of data availability. After all, parties not only provide voters with the tools to understand the political world, but also represent the political offer that people deal with when casting a vote (Sniderman, 2017). Furthermore, this article does not consider alternative party-level data: a comparison with manifesto data could make sure that an expert-based measure of party positions is not affected by ideological bias (Curini, 2010), which might be reflected in an excess of consistency of certain parties compared to others. Yet, recent research has shown that expert survey data and manifestos can be equally trustworthy and that expert surveys can sometimes provide even better estimates than manifestos when it comes to party positions on different topics (Ecker et al., 2022). Moreover, data from Chapel Hill guarantee the comparability of data on parties’ and voters’ issue positions (in terms of both question wording and response category). Such correspondence is unparalleled by any other party-level data sets. One downside, however, is that analyses in this article are limited to six policy issues, which are those that are comparable between the 2019 EES and the 2019 CHES. In addition, analyses focussed on a cross-country validation at the expenses of longitudinal analyses. It is also important to note that the creation of the top-down constraint follows the assumption that voters have a most preferred party. In other words, it does not consider the hypotheses of ‘partisan dealignment’, that is, the decline of the sense of partisanship in many countries (Dalton et al., 2000). This limitation can be justified by the fact that voters in a political system eventually come to the ballot and they have to make a decision, i.e. they have to develop a preference by taking into consideration the range of party options they are offered. Finally, such analyses did not test the effect of contextual variations in terms of party systems. Indeed, further research could focus on the impact of party competition patterns, such as polarization, the deinstitutionalization of the traditional party system as regards the rise of challenger parties, the salience of anti-elite rhetoric, and other characteristics on opinion constraint.

Supplemental Material

sj-docx-1-psx-10.1177_00323217221147774 – Supplemental material for Solving the (False) Dilemma: An Ecological Approach to the Study of Opinion Constraint

Supplemental material, sj-docx-1-psx-10.1177_00323217221147774 for Solving the (False) Dilemma: An Ecological Approach to the Study of Opinion Constraint by Marta Gallina in Political Studies

Footnotes

Acknowledgements

The author would like to thank Pierre Baudewyns (Catholic University of Louvain), Ruth Dassonneville (University of Montréal), Stefano Camatarri (Autonomous University of Barcelona), Lewis Luartz (Riverside University), and the anonymous reviewers for their valuable comments.

Author’s note

Marta Gallina is now affiliated with Department of Political Science and Public Law, Autonomous University of Barcelona, Bellaterra, Spain.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

Marta Gallina

Supplemental Information

Additional Supplementary Information may be found with the online version of this article.

Notes

Author Biography

Marta Gallina received her PhD in Political Science from the Catholic University of Louvain (Belgium). She has been Lecturer at the Catholic University of Lille (France) and post-doc researcher at the Catholic University of Louvain (Belgium) and at Waseda University (Japan). Currently, she is Juan de la Cierva Postdoctoral Fellow at the Autonomous University of Barcelona (Spain). Her research interests regard the study of political behaviour, issue dimensionality, opinion consistency and Voting Advice Applications. Her work appeared in scientific journals, such as Acta Politica, Swiss Political Science Review, Regional and Federal Studies, and Environmental Politics.

References

Adcock

Collier

(2001) Measurement Validity: A Shared Standard for Qualitative and Quantitative Research. American Political Science Review 95 (3): 529–546.

André

Depauw

(2017) The Quality of Representation and Satisfaction with Democracy: The Consequences of Citizen-Elite Policy and Process Congruence. Political Behavior 39 (2): 377–397.

Ansolabehere

Rodden

Snyder

(2008) The Strength of Issues: Using Multiple Measures to Gauge Preference Stability, Ideological Constraint, and Issue Voting. American Political Science Review 102 (2): 215–232.

Bakker

Hooghe

Jolly

, et al. (2020) 2019 Chapel Hill Expert Survey (Version 2019.1). Chapel Hill, NC: University of North Carolina, Chapel Hill.

Baldassarri

Gelman

(2008) Partisans without Constraint: Political Polarization and Trends in American Public Opinion. American Journal of Sociology 114 (2): 408–446.

Baldassarri

Goldberg

(2014) Neither Ideologues nor Agnostics: Alternative Voters’ Belief System in an Age of Partisan Politics. American Journal of Sociology 120 (1): 45–95.

Bartle

(2000) Political Awareness, Opinion Constraint and the Stability of Ideological Positions. Political Studies 48 (3): 467–484.

Barton

Parsons

(1977) Measuring Belief System Structure. Public Opinion Quarterly 41 (2): 159.

Boutyline

Vaisey

(2017) Belief Network Analysis: A Relational Approach to Understanding the Structure of Attitudes. American Journal of Sociology 122 (5): 1371–1447.

10.

Conover

Feldman

(1984) How People Organize the Political World: A Schematic Model. American Journal of Political Science 28 (1): 95–126.

11.

Converse

(1964) The Nature of Belief Systems in Mass Publics. Critical Review 18 (1–3): 1–74.

12.

Coppock

Green

(2022) Do Belief Systems Exhibit Dynamic Constraint? The Journal of Politics 84 (2): 725–738.

13.

Curini

(2010) Experts’ Political Preferences and Their Impact on Ideological Bias: An Unfolding Analysis Based on a Benoit-Laver Expert Survey. Party Politics 16 (3): 299–321.

14.

Dalton

(2020) The Representation Gap and Political Sophistication: A Contrarian Perspective. Comparative Political Studies 54: 889–917.

15.

Dalton

McAllister

Wattenberg

(2000) The Consequences of Partisan Dealignment. In: Dalton

Wattenberg

(eds) Parties without Partisans. Oxford: Oxford University Press, pp.37–62.

16.

Dassonneville

(2012) Electoral Volatility, Political Sophistication, Trust and Efficacy: A Study on Changes in Voter Preferences during the Belgian Regional Elections of 2009. Acta Politica 47 (1): 18–41.

17.

Dassonneville

(2014) Political Sophistication and Vote Intention Switching: The Timing of Electoral Volatility in the 2009 German Election Campaign. German Politics 23 (3): 174–195.

18.

De Sio

Franklin

(2012) Strategic Incentives, Issue Proximity and Party Support in Europe. West European Politics 35 (6): 1363–1385.

19.

de Vries

van der Brug

van Egmond

, et al. (2011) Individual and Contextual Variation in EU Issue Voting: The Role of Political Information. Electoral Studies 30 (1): 16–28.

20.

Delli Carpini

Keeter

(1996) What Americans Know about Politics and Why It Matters. Yale University Press.

21.

Druckman

Lupia

(2000) Preference Formation. Annual Review of Political Science 3 (1): 1–24.

22.

Ecker

Jenny

Müller

, et al. (2022) How and Why Party Position Estimates from Manifestos, Expert, and Party Elite Surveys Diverge: A Comparative Analysis of the ‘Left–Right’ and the ‘European Integration’ Dimensions. Party Politics 28 (3): 528–540.

23.

Enders

(2019) Conspiratorial Thinking and Political Constraint. Public Opinion Quarterly 83 (3): 510–533.

24.

Federico

Hunt

(2013) Political Information, Political Involvement, and Reliance on Ideology in Political Evaluation. Political Behavior 35 (1): 89–112.

25.

Freeze

Montgomery

(2015) Static Stability and Evolving Constraint: Preference Stability and Ideological Structure in the Mass Public. American Politics Research 44 (3): 415–447.

26.

Gerber

Nicolet

Sciarini

(2015) Voters Are Not Fools, or Are They? Party Profile, Individual Sophistication and Party Choice. European Political Science Review 7 (1): 145–165.

27.

Gordon

Segura

(1997) Cross–National Variation in the Political Sophistication of Individuals: Capability or Choice? The Journal of Politics 59 (1): 126–147.

28.

Hurwitz

Peffley

(1987) A Hierarchical Model of Attitude Constraint. American Journal of Political Science 29 (4): 1100–1120.

29.

Jackson

Marcus

(1975) Political Competence and Ideological Constraint. Social Science Research 4 (2): 93–111.

30.

Judd

Krosnick

(1989) The Structural Bases of Consistency among Political Attitudes: Effects of Expertise and Attitude Importance. In: Greenwald

Pratkanis

Breckler

(eds) Attitude Structure and Function. Hillsdale, NJ: Erlbaum, pp.99–128.

31.

Kinder

(1982) Enough Already about Ideology: The Many Bases of American Public Opinion. Paper Prepared for Delivery at the 1982 Annual Meeting of the American Political Science Association, Denver, CO, 1–4 September.

32.

Kinder

(1998) Opinion and Action in the Realm of Politics. In: Gilbert

Fiske

Lindzey

(eds) The Handbook of Social Psychology. New York: McGraw-Hill, pp.778–867.

33.

Kozlowski

Murphy

(2021) Issue Alignment and Partisanship in the American Public: Revisiting the ‘partisans without Constraint’ Thesis. Social Science Research 94: 102498.

34.

Kuklinski

Luskin

Bolland

(1991) Where Is the Schema? Going Beyond the ‘S’ Word in Political Psychology. American Political Science Review 85 (4): 1341–1380.

35.

Lachat

(2007) A Heterogeneous Electorate. Political Sophistication, Predisposition Strength and the Voting Decision Process. Baden-Baden: Nomos.

36.

Lau

Patel

Fahmy

, et al. (2014) Correct Voting Across Thirty-Three Democracies: A Preliminary Analysis. British Journal of Political Science 44 (2): 239–259.

37.

Leeper

Slothuus

(2014) Political Parties, Motivated Reasoning, and Public Opinion Formation. Political Psychology 35: 129–156.

38.

Levendusky

(2009) Clearer Cues, More Consistent Voters: A Benefit of Elite Polarization. Political Behaviour 32 (1): 111–131.

39.

Linzer

(2006) Ideological Constraint and the Quality of Governance in New and Established Democracies. Available at: https://citeseerx.ist.psu.edu/document?repid=rep1&type=pdf&doi=f31613cf159eb948654b6b72da247d66fbcce30e (accessed 9 January 2023).

40.

Lupton

Myers

Thornton

(2015) Political Sophistication and the Dimensionality of Elite and Mass Attitudes, 1980−2004. The Journal of Politics 77 (2): 368–380.

41.

Luskin

(1987) Measuring Political Sophistication. American Journal of Political Science 31 (4): 856–899.

42.

Luskin

(1990) Explaining Political Sophistication. Political Behavior 12 (4): 331–361.

43.

Marcus

Tabb

Sullivan

(1974) The Application of Individual Differences Scaling to the Measurement of Political Ideologies. American Journal of Political Science 18 (2): 405–420.

44.

Nie

Verba

Petrocik

(1976) The Changing American Voter. Cambridge: Harvard University Press.

45.

Petersen

Slothuus

Togeby

(2010) Political Parties and Value Consistency in Public Opinion Formation. Public Opinion Quarterly 74 (3): 530–550.

46.

Pierce

(1970) Party Identification and the Changing Role of Ideology in American Politics. Midwest Journal of Political Science 14 (1): 25.

47.

Piurko

Schwartz

Davidov

(2011) Basic Personal Values and the Meaning of Left-Right Political Orientations in 20 Countries. Political Psychology 32 (4): 537–561.

48.

Schmitt

Hobolt

Brug

, et al. (2022) European Parliament Election Study 2019, Voter Study. GESIS, Cologne. ZA7581 Data file Version 2.0.1, https://doi.org/10.4232/1.13846

49.

Seeberg

Slothuus

Stubager

(2017) Do Voters Learn? Evidence that Voters Respond Accurately to Changes in Political Parties’ Policy Positions. West European Politics 40 (2): 336–356.

50.

Slothuus

(2010) When Can Political Parties Lead Public Opinion? Evidence from a Natural Experiment. Political Communication 27 (2): 158–177.

51.

Slothuus

Bisgaard

(2021) How Political Parties Shape Public Opinion in the Real World. American Journal of Political Science 65 (4): 896–911.

52.

Slothuus

De Vreese

(2010) Political Parties, Motivated Reasoning, and Issue Framing Effects. The Journal of Politics 72 (3): 630–645.

53.

Sniderman

(2017) The Democratic Faith: Essays on Democratic Citizenship. New Haven, CT: Yale University Press.

54.

Sturgis

Roberts

Allum

(2005) A Different Take on the Deliberative Poll: Information, Deliberation, and Attitude Constraint. Public Opinion Quarterly 69 (1): 30–65.

55.

Tomz

Sniderman

(2005) Brand names and the organization of mass belief systems. Available at: https://www.researchgate.net/profile/Paul-Sniderman/publication/228917249_Brand_Names_and_the_Organization_of_Mass_Belief_Systems/links/552441a30cf2b123c5173904/Brand-Names-and-the-Organization-of-Mass-Belief-Systems.pdf (accessed 9 January 2023).

56.

Treier

Hillygus

(2009) The Nature of Political Ideology in the Contemporary Electorate. Public Opinion Quarterly 73 (4): 679–703.

57.

Valentino

Traugott

Hutchings

(2002) Group Cues and Ideological Constraint: A Replication of Political Advertising Effects Studies in the Lab and in the Field. Political Communication 19 (1): 29–48.

58.

Van der Brug

Van Spanje

(2009) Immigration, Europe and the ‘New’ Cultural Dimension. European Journal of Political Research 48 (3): 309–334.

59.

Van der Eijk

(2017) Selecting the Dependent Variable in Electoral Studies: Choice or Preference? In: Fisher

Fieldhouse

Franklin

, et al. (eds) The Routledge Handbook of Elections, Voting Behavior and Public Opinion. Abingdon: Routledge, pp.445–457.

60.

Van der Eijk

Franklin

(1996) Choosing Europe? The European Electorate and National Politics in the Face of Union. Ann Arbor, MI: University of Michigan Press.

61.

Van der Eijk

van der Brug

Kroh

, et al. (2006) Rethinking the Dependent Variable in Voting Behavior: On the Measurement and Analysis of Electoral Utilities. Electoral Studies 25 (3): 424–447.

62.

Weissberg

(1976) Consensual Attitudes and Attitude Structure. Public Opinion Quarterly 40 (3): 349–359.

63.

Zaller

(1992) The Nature and Origins of Mass Opinion. The Nature and Origins of Mass Opinion. Cambridge: Cambridge University Press.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.04 MB