Daniela Beyer and Miriam Hänni
The public opinion – policy linkage has received scholarly attention for a long time. After all, this linkage is not only a key characteristic of democracy, but one of the most important aspects and quality criteria of a functioning representative democracy. Despite more than 50 years of political science research, there is still a lot of controversy about how the linkage between public opinion and policy actually works. Two related but distinct strands have formed in the literature – one focusing on responsiveness, the other on congruence. While both schools of thought are ultimately interested in the link between public opinion and representatives’ position or behavior they pursue two different strategies leading to confusion over the concepts and measurement in question. We provide a mutually exclusive conceptualization of congruence and responsiveness and structure the review of the extensive literature accordingly. In addition to providing greater theoretical coherence, our conceptualization fosters further development in the field by deliberately combining the two concepts with the research strands on public policy and representation. We conclude with a call for a more integrated research agenda and introduce a novel concept of ‘congruent responsiveness’.
A key characteristic of a democracy is the continuing responsiveness of the government to the preferences of its citizens
— Dahl, 1971, 1
The foundational idea of representative democracy is that elected political elites represent citizens by responding to their preferences and concerns. We do not need elections if not for giving citizens the chance to elect those representatives who are closest to their viewpoints. We do not need parties if not for interest aggregation and organization of the political process. Most importantly, we do not need representatives if not to ensure that public opinion is transmitted into public policy.
Even after 50 years of political science research on representation, there is still a lot of controversy about how citizens and their representatives are linked in terms of preferences, priorities, opinion, and policies. Over the years, two related but often distinct strands have formed in the literature — one focusing on congruence between citizens and representatives, the other on responsiveness of representatives to citizens. While both schools of thought are ultimately interested in the link between public opinion and representatives’ position or behavior in parliament, they pursue two different strategies. This divergence has often led to imprecision and confusion over the concepts in question and their measurement.
In this review we distinguish between congruence and responsiveness and divide the literature accordingly. We disentangle the two concepts empirically and conceptually and thereby offer a reference for scholars in the field. This endeavor is not only important for scholars of representation but also of public policy. We reconnect public policy ideas which deal with processes of policy-making and the evolution of laws with the voters as the backbone of representative democracy. Despite the significance of the public in public policy theories they rarely deal with public opinion explicitly.
We first discuss the concept of congruence that statically examines the overlap between citizens and their representatives’ ideologies, policy positions, or issue priorities. The literature does often not consider where this linkage comes from nor does it discuss a causal direction. We then focus on responsiveness, which we define as a dynamic and causal relationship between representatives and their constituents. Responsiveness requires that (shifting) constituent preferences change representatives’ preferences, behavior or, ultimately, policy outputs. Hence, responsiveness comes closer to the theoretical idea of representation while congruence is more successful in assessing whether the majority gets what it wants. As such, both concepts are crucial for the evaluation of representative democracy, both require scholarly attention and, while cautioning against unclear conceptualizations, stronger interactions between the research strands would be desirable.
We depart from the premise that congruence neither implies responsiveness nor responsiveness necessarily guarantees congruence. A perfect situation in a representative democracy, however, would see both fulfilled. In this context both congruence and responsiveness were high, corresponding in terms of direction, and related to majority will. In this ideal state, responsiveness is sufficient for high levels of congruence and congruence is necessary for responsiveness. This is not meant to say that we should not consider both concepts individually but that their combination defines the full extent of successful representation of citizens’ will.
Scholars of the two concepts address different research questions. When congruence is used as a dependent variable, we find a strong emphasis on the effect of electoral systems and parties/partisanship (e.g. Powell, 2009; Blais and Bodet, 2006). When acting as the independent variable, scholars often explain democratic satisfaction with the level of congruence (e.g. Mayne and Hakhverdian, 2017; Ezrow and Xezonakis, 2011). Since responsiveness, on the other hand, examines the more dynamic link between public opinion and policy, an overwhelming majority studies the effect of public opinion on policy outputs. Among others, scholars engage in conditional explanations (e.g. Wlezien and Soroka, 2012; Hobolt and Klemmensen, 2008), issues of causality (Hakhverdian, 2012), and differences between population subgroups (e.g. Grimes and Esaiasson, 2014). A smaller and more recent share of the literature uses responsiveness as an independent variable to predict democratic support (e.g. Esaiasson et al., 2017) or tries to explain varying levels of responsiveness with institutional designs or the composition of parliaments (e.g. Bird et al., 2011).
In the following, we review the literature out of which we build clear, mutually exclusive definitions and illustrate them graphically. This task entails a combined literature review of the two concepts, a discussion of their use, and an outlook to the relationship between public opinion and public policy research strands.
Both responsiveness and congruence are forms of (...) representation, but they capture different dimensions of democratic performance.
— Lax and Phillips, 2012, 148
Policy responsiveness is a goal of democratic government—that government action responds to the preferences of its citizens. It is conceptually distinct from “representation,” whereby government actions mirror the preferences of public opinion.
— Erikson, 2013, 1
Despite (or due to) the sheer amount of literature on congruence and responsiveness the two concepts are still not clear. While researchers increasingly differentiate between responsiveness to public opinion and their congruence (Wlezien, 2017b, 2), the definitions are often ambiguous and vary from study to study. In this section we provide a conceptualization that follows the current state of the art but provides clearly distinguishable definitions along the following dimensions:1 static vs. dynamic, overlap vs. causal relationship, representation of the majority will vs. responsive behavior. For this purpose, it is important to differentiate between the theoretical constructs and their measurement. We are aware that measurement difficulties at times prevent scholars from mutually exclusively delimiting and identifying the concepts empirically. Nonetheless, the precision of the theoretical constructs is decisive for further development of the literature.
In essence, the major theoretical difference between the two concepts lies in their dynamics (see for instance Weissberg’s (1976 distinction between majoritarian and covariational congruence, or Erikson’s (2013) definition of responsiveness). While congruence refers to the static accordance between citizens and elites at a given point in time, responsiveness suggests a dynamic relationship that relates preferences for policy changes (or public opinion change) to changes in policies. Responsiveness, thus, includes a causal element which is absent from the notion of congruence. For congruence, the movements of public opinion and policy/opinion are not necessarily related, but responsiveness includes the idea of responding elites. In empirical research on responsiveness, however, it has often been difficult to establish causality.2
Congruence and responsiveness can be further distinguished through how they emerge. Congruence results mostly from electoral competition and is part of the responsible party model where representatives are expected to transform citizens’ mandate into policies that match their preferences. Responsiveness, by contrast, happens due to rational anticipation of future elections when representatives try to please voters ahead of elections by enacting policies that are in line with their preferences (see also: Arnold and Franklin, 2012; Stimson et al., 1995). Wlezien and Soroka (2016) refer to this distinction as post-election and between-election representation.3 We now discuss each of the concepts in more detail.
The relationship between citizens’ ideologies, attitudes, preferences, and opinions and those of their elected representatives as well as policy outputs is covered in the concept of congruence. Ideologies in this context are usually expressed as left-right self-placement, attitudes define people’s standpoint on an issue, preferences imply a rank ordering, and opinions are the more abstract concept. Accounts on congruence use all four approaches to examine the overlap between citizens’ and (their) elites’ viewpoints. More recently, scholars started to analyze congruence between citizens’ viewpoints and actual policy outputs. The extent of overlap presents a key component of every representative democracy and influences citizens’ evaluation of the functioning of democracy.
Congruence can be conceptualized in three different ways, pictured according to Golder and Stramski’s (2010) approach in Figure 1 below. In the original framework, the axis defines left-right ideology, but it can also be applied to the comparison of other positions. The simplest one-to-one relationship compares the position of an individual citizen to his or her elected representative — i.e. dyadic representation. Such matching is most useful in political systems with a strong representational link between legislators and their constituents (i.e. in single member districts). The relationship is generalizable to other systems when a citizen’s position is compared to a particular party or government instead of an individual legislator. As a result of increasing interest in collective representation (e.g. Weissberg, 1978), the next step was a many-to-one relationship that compares priorities, positions or preferences of citizens to those of their representative (party/government). Finally, Golder and Stramski (2010) introduced the many-to-many relationship that match the distribution of citizens’ to the distribution of representatives’ preferences. Depending on the research question not all approaches are equally appropriate. They have to be applied carefully to the respective dependent variable.
While the uni-dimensional concept of ideological congruence was the dominant approach for a long time, more recently, the literature increasingly engages with congruence of policy preferences and priorities, and considers policy outputs. It often moves beyond the comparison of citizens’ and elites’ standpoints and applies it as a dependent variable to study the effect of electoral institutions, institutional contexts, and party strategies (see section 3.1).
In contrast to congruence, responsiveness typically analyzes the effect of public opinion or preferences on policy outputs. It introduces a dynamic component which is absent from the concept of congruence. When political actors are responsive to their citizens they react (respond) to opinion changes during the policy-making process. Responsiveness requires that if public opinion moves in a certain direction, policy-makers adapt policy outputs in the same direction (e.g. Eulau and Karps, 1977; Erikson, 2013). As a result of the new policy, citizens readjust their preferences about policy change — i.e. the gap between preferred and actual policy lessens. Policy responsiveness thus requires an action on behalf of the political actors and at least conceptually implies a causal relation between public opinion and public policy. In empirical applications, however, it is often difficult to establish if it is really public opinion affecting policy change or if a third factor influences both simultaneously.4
Figure 1: Different Approaches to Congruence
Policy responsiveness can be specified in two ways. On the one hand (1), policy responsiveness is defined as a correlation between prior public opinion and policy outputs. In this view, responsiveness is achieved if public preferences at time t affect policy outputs at time t+1. This also includes the case when a new government changes policies in response to pre-existing public opinion — i.e. a change in public opinion is not always necessary. In contrast to many empirical applications, however, we oppose the definition of responsiveness as a mere correlation between opinion and policy at the same point in time. We subsume a simple correlation without a dynamic temporal relationship under congruence and not responsiveness because it only measures how opinion and policy overlap at a given point in time and not if elites respond to public opinion (see section 2.1.3). On the other hand (2), responsiveness can be measured as first-differences in public opinion and policy outputs. In this case, policy responsiveness is achieved if changes in public opinion lead to changes in policy in the same direction. While most studies apply the first approach (1) (correlation with lagged public opinion) we believe that the second approach (2) corresponds better to the theoretical concept of responsiveness as outlined by Pitkin (1967) or Dahl (1971). Regressing changes on changes comes closer to the idea of political elites responding to their citizens’ changing opinions.
The discussion of congruence and responsiveness underlines that while the two concepts are closely related, they are conceptually and empirically distinct. Responsiveness does not necessarily lead to congruence but congruence can be a result of responsiveness. When the ideal of representative democracy is fully achieved, we should see both fulfilled (Lax and Phillips, 2012, 148). In the absence of this ideal, however, congruence and responsiveness may be at odds. Especially when conceptualized as a correlation, policy can move in the same direction as public opinion (responsiveness) without being in line with majority will (congruence) (Canes-Wrone, 2015, 148). More specifically, responsiveness would already be achieved when citizens’ shift in a more liberal (conservative) direction which results in more liberal (conservative) policies. This shift does not result in a congruent decision, however, when policies remain more conservative (liberal) than the individual or mean/median/majority prefers (see also Wlezien (2017b)). The ultimate distinction is whether the majority gets what it wants (Wlezien, 2017b, 2).
Therefore, one could argue that congruence is more important for representative democracy than responsiveness, although it measures the much simpler concept. Responsiveness tries to draw the link between preferences and outputs, whereas congruence only recently moved into a similar direction. One of the main reasons is that assessing the public’s preferred level of policy is hardly possible due to a lack of adequate individual level data. In many policy areas, it is difficult to determine what and how much the public wants (Wlezien, 2017b, 562).
Figure 2: Congruence vs. Responsiveness
We graphically distinguish the concepts in Figure 2. Congruence, the upper part of the figure, defines the overlap between citizens and elites’ ideologies, policy positions or priorities, or policy outputs, at a given point in time. Policy responsiveness, on the other hand, pictured in the lower part of the figure, examines (the change in) public opinion and the resulting adaption of public policy. As a result, it can move on different scales (e.g. public preferences influencing policy outputs) and not necessarily in full accordance. For policy change to be characterized as responsive, it needs to move in the same/preferred direction, but not parallel.
The difference between the two concepts can be formalized in the following way, where P refers to policy and P* to public preferences: (adapted from: Soroka and Wlezien (2010, 36) and Wlezien (2017b, 2-3))
The first line displays policy responsiveness. On the left, responsiveness is conceptualized as the effect of past preferences on policy change or implementation, respectively. Scholars in this tradition conceptualize responsiveness as a correlation. On the right, we include a change-oriented design of responsiveness. Here, changes in policy outputs are a function of changes in public preferences. The second line represents the concept of congruence. We understand congruence as a continuum. If there is perfect congruence we observe a slope of B = 1 and a constant of a = 0, i.e. a congruent decision. This implies that the level of preferred and actual policy is identical. If the slope of B 6=1 one can interpret it as a certain degree of congruence. A continuous understanding of congruence is relevant, for instance, in studies assessing which electoral systems achieve more or less congruence.
The difficulties of previous studies with the two concepts illustrate the need for clearer conceptualizations. The colloquial use of language often complicates distinguishing the two concepts and contributes to the ambiguities observed in the literature. Responsiveness for example is often explained as congruent changes of public opinion and public policy, which, however, must not be confused with the concept of congruence. The confusion that the lax use of language causes, can be nicely illustrated in path breaking Page and Shapiro (1983, 177) who conclude that: "there has been a great deal of congruence between changes in policy and changes in opinion (...)". While the authors make use of the word "congruence" they do not refer to congruence in public opinion and public policy but to changes in the same direction - i.e. responsiveness.
More problematic is the resulting confusion of concepts such as in Brettschneider (1996, 293), where congruence is used as a defining criteria of responsiveness: "policy responsiveness is defined as the congruence of collective attitudes towards political issues with the legislative behavior of representatives, parties, and the government.” Here, the confusion does not stem from the choice of words, but from the intermingling in the definition.
Another type of conceptual confusion occurs when scholars talk about one concept but measure the other. Sometimes this mismatch occurs without clarifying the measurement concept, in other cases the difference between concept and measurement is made explicit. An example for the latter is Canes-Wrone’s and Shotts’ (2004) article in which the authors talk about presidential responsiveness while measuring congruence between public opinion and the respective president’s proposed budgetary authority.
Powell pursues a different approach to circumvent the issue of measuring responsiveness and combining it with congruence. He defines responsiveness as the replacement or continuation of incumbent policy-makers in a way that is responsive to citizens’ votes (Powell, 2000, 15-16). The underlying argument is that we only know about citizens’ preferences through what they reveal in their vote choice (Powell, 2000, 15). Congruence, on the other hand, is conceptualized and measured as ideological congruence, which Powell labels representational congruence (Powell, 2000, 16 & 159ff).
The literature is full of examples to underline such conceptual imprecision, which emphasizes the importance of more careful conceptualizations in future research. This call for clearer distinctions does not imply that the concepts should be studied separately. Quite to the contrary, for a comprehensive view on representation a combination can be useful. Some early examples of this include Achen (1978) and Brettschneider (1996). The combination became more pronounced and gained momentum in recent years (e.g. Lax and Phillips, 2009, 2012; Matsusaka et al., 2010), probably also because the two literatures study different dependent variables. Those studying representation of elected officials often focus more congruence, those interested in policy decisions on responsiveness.
In addition to the conceptual difficulties, the operationalization and measurement of congruence and responsiveness presents major challenges (see e.g. Powell, 2000). Both approaches relate data on public input to some form of output:
Whether we consider congruence or responsiveness, the data for the left hand side of the equation, the public input, is the same. Survey data plays the most important role — for public opinion, preferences, priorities and ideologies. Some authors use survey questions on specific policies to examine whether citizens want more or less of a certain policy (e.g. Page and Shapiro, 1983) or whether they support a given policy proposal (e.g. Lax and Phillips, 2009, 2012). An alternative approach that has mostly been introduced by Wlezien and Soroka (e.g. 2012) is to use citizens’ spending preferences. Identifying adequate survey data on policy preferences is often very difficult and involves unresolved challenges about how to measure what the public actually wants (Wlezien, 2017b).
Particularly popular among scholars interested in comparisons across time or space, citizens’ left-right self-placement is sometimes used as a proxy for public preferences. While it may at times be the only available source (e.g. Hakhverdian, 2012) it is certainly not ideal to measure responsiveness as it does not cover opinion or preferences, but broad ideological stances. It is, therefore, usually better suited to measure congruence.
For the earlier stages in the decision-making process, scholars use priority or most-important problem questions (MIP) to assess issue priorities of the population (e.g. Spoon and Klüver, 2014; Bevan and Jennings, 2014; Jones and Baumgartner, 2004; Jones et al., 2009; Flavin and Franko, 2016). In the case of policy positions and preferences the literature deals with distributions whereas for priorities scholars apply more macro-oriented, aggregate measures such as percentages.
What mostly differentiates congruence and responsiveness in terms of measurement is the elite’s side (the right hand side of the equation) and especially the lag structure.
The measurement of congruence depends on the definition of who is congruent to whom. In a one-to-one relationship, a small absolute distance between a citizen and a representative implies high congruence. For the more common many-to-one relationship, scholars are either interested in the absolute distance between the median citizen or citizens’ aggregated priority position and the representative (party/government), in the average absolute difference, or in the absolute distance relative to the dispersion of citizens’ preferences (Golder and Stramski, 2010, 92-93). More recently, research increasingly studies preferences of particular subgroups in the society (e.g. Bartels, 2008; Gilens, 2011, 2012; Giger et al., 2012; Branham et al., 2017). In Golder and Stramski’s (2010) proposed many-to-many relationship, the authors compare the distribution of citizen and representative preferences. Focusing on ideological congruence on a 0-10 left-right scale, probability distribution functions show the overlap of preference distributions, the area between the cumulative distribution functions are the actual measure of congruence. This approach shows that while the probability distributions’ location can be identical, differing shapes (e.g. representatives’ uniform vs. citizens’ normal distributions of preferences) can lead to very different levels of overlap (Golder and Stramski, 2010, 97).
Citizens positions are combined with elites’ positions by assessing either citizens’ perceptions of parties’ (representatives/governments) positions on the same scale (e.g. Golder and Stramski, 2010), experts’ or elites’ assessments of the latter (e.g. Powell, 2006), or data from comparative data collection efforts like the comparative manifestos project or the comparative agendas project (e.g. McDonald et al., 2004). Studies on congruent issue priorities or policy congruence are more likely to rely on such comparative data collections, which code manifestos or documents from the legislative process (questions, speeches, bills, laws) into policy areas. They enable a comparable overview over the policy attention of parties, governments, and legislators.
The measurement of policy responsiveness depends on whether it is conceptualized as a correlation or first difference. The first requires the comparison of levels with a time lag, the second a comparison of first differences. Both approaches, however, predominantly rely on two types of sources. They use survey data to assess public input and derive information about legislators’ or governments’ behavior in office (through laws, speeches, roll-calls) from official documents.
In addition to the above presented survey approaches, the concept of policy mood enjoys high popularity among scholars in the U.S. as a measure of public opinion. The concept and measurement was first introduced by Stimson (1991). It only recently travelled beyond the U.S. when Guinaudeau and Schnatterer (2017) applied it to the European Union. Public mood is defined as follows and measured through a series of repeated survey questions:
Mood is the major dimension underlying expressed preferences over policy alternatives in the survey research record. It is properly interpreted as left versus right- more specifically, as global preferences for a larger, more active federal government as opposed to a smaller, more passive one across the sphere of all domestic policy controversies. Thus our public opinion measure represents the public’s sense of whether the political "temperature" is too hot or too cold, whether government is too active or not active enough (Stimson et al., 1995, 548).
With regard to the dependent variable, the right hand side of the equation, three main strategies have emerged. Some scholars analyze actual policy outputs by examining the introduction of laws or the voting behavior of individual Members of Parliament (MPs) (e.g. Page and Shapiro, 1983; Lax and Phillips, 2012). Others focus on the symbolic behavior and analyze government speeches or parliamentary questions (e.g. Hakhverdian, 2012). A third tradition studies legislative output on a more global level by analyzing the liberalism or conservatism of a particular parliament by applying policy rating scales5 to roll-call votes (e.g. Stimson et al., 1995). Only recently, a fourth tradition has developed which departs from studying the link between citizens and the political elite through survey data and some political output and resorts to experimental designs. Most popular are field experiments which study the reaction of MPs to letters or emails by (mostly fake) constituents (e.g. Butler and Nickerson, 2011; Butler and Broockman, 2011; Butler et al., 2012; Butler, 2014), but also survey experiments with political elites gain importance (e.g. Butler, 2014; Öhberg and Naurin, 2016; Arnesen and Peters, 2017).
The two seminal works on congruence and responsiveness — Miller and Stokes (1963) and Page and Shapiro (1983) — both pursued a similar goal: testing the ties between representatives and their citizens. Yet, while the underlying idea is similar, the authors follow different strategies. Should there just be congruence (no matter how it evolves) or should representatives take a more active role and respond to their constituents’ opinion changes?6
Searching for Congruence or Responsiveness respectively in publication titles listed in the "Web of Science"7, lets us reliably portray the literature’s evolution in the field. In total, 346 congruence- and 374 responsiveness-related scholarly books and articles have been published since the 1900s (annual data is available since the 1970s). This shows the importance of the two concepts.
As Figure 3 depicts, publication numbers and thus scholarly attention are rising for both concepts since the 1990s. Congruence experienced initial peaks around 2005, responsiveness some years later. Since 2010, publications on responsiveness outnumber those on congruence. Despite the seemingly higher scholarly attention for responsiveness, the line for congruence is smoother. Overall, the attention for both concepts clearly increased over time. Representation, which can be seen as the overlying concept — experienced an even stronger but parallel increase in scholarly attention and consequently publication numbers. Publications dealing with both concepts (such as Lax and Phillips (2009, 2012)), however, are rare. There are no findings in the Web of Science before the 1990s and only 45 publications in total. Only the last 5 years marked a sharp increase which underlines our claim for a better conceptualization and subsequent combination of the two research strands.
We now move into the individual literature reviews for both congruence and responsiveness. Many excellent reviews on congruence and responsiveness cover the vast literature in the field (e.g. Esaiasson and Wlezien, 2017; Manza and Cook, 2002a; Wlezien, 2017b; Canes-Wrone, 2015; Burstein, 2003, 2010; Shapiro, 2011; Wlezien and Soroka, 2016). In contrast to most other reviews, however, we pay attention to include not only studies from the U.S., but also from other world regions when appropriate.
Figure 3: The Evolution of the Literature
Citizens’ perceptions of representation is often "shaped by the degree of congruence between their own views and those of political elites” (Greene and Reher, 2017, 1). The degree of overlap that congruence measures can take very different forms. Existing studies therefore have to be distinguished along two lines. First, it matters who is congruent to whom as elaborated in the measurement section. Secondly, and this is what structures the following depiction of the state of the art, we have to distinguish ideological congruence that often focuses on electoral consequences, attentional congruence that occurs at the earliest stage of the policy-making process and analyzes shared priorities, and policy congruence defining the actual similarity in various policy fields.
Miller and Stokes’ seminal article on “Constituency Influence in Congress” (1963) is a good example for how congruence entered the research on representation. Analyzing interviews with incumbents, non-incumbent opponents, and a sample of constituents in the context of the 1958 congressional election as well as roll-call votes, the authors provide an account of the ties between representatives and their constituents. They find that representatives’ policy views and their perceptions of constituents’ attitudes successfully predict roll call behavior (Miller and Stokes, 1963, 51).
Descriptive studies on congruence show that mass-elite congruence varies across time and space (Lupu et al., 2017; Miller and Stokes, 1963, 282), but that even in relatively recent democracies in Latin America and Eastern Europe considerable levels of congruence emerge (Lupu et al., 2017, 282). Yet, the heterogeneity between countries is striking. Among Latin American democracies for instance countries like Uruguay and Chile show much higher levels of congruence than Bolivia and Ecuador at the other end of the scale (Luna and Zechmeister, 2005, 413). Scholars explain the variation with different political institutions and electoral systems, party differences and political majorities, voters’ characteristics and other context factors. Congruence is thereby often understood as a "quantified measure of the level of political representation" (Luna and Zechmeister, 2005, 413).
Starting in the 1960s studies on ideological congruence — namely the one-dimensional left-right proximity of citizens and their representatives — dominated the first generation of congruence studies (e.g. Miller and Stokes, 1963). The contradicting findings of this generation spurred two debates. The so called ’ideological congruence controversy’ (Powell, 2009) arose in response to diverging results about electoral systems’ effects. Some scholars found that proportional representation was particularly conducive to ideological mass-elite congruence (Huber and Powell, 1994; Powell, 2009) whereas others reported no differences between electoral systems (Blais and Bodet, 2006; Golder and Lloyd, 2014; Golder and Stramski, 2010; Lupu et al., 2017).
Powell (2009) steps into the ’ideological congruence controversy’ arguing that the different findings in congruence studies stem not from measurement but time period differences. He claims that the relationship between electoral rules and ideological congruence is robust to different measurements, but effects have declined in recent years as a consequence of parties’ convergence towards the center in first-past-the post systems.
Golder and Stramski (2010) who introduce the many-to-many conceptualization, in turn, provide evidence that legislatures in countries with proportional representation are more likely to be congruent with voters’ ideological preferences than countries with majoritarian electoral systems (Golder and Stramski, 2010, 104). Congruence between citizens and their governments, however, does not differ between electoral systems. This challenges the prevailing idea of a trade-off between majoritarian systems with high accountability and proportional systems with an advantage in terms of ideological congruence.
Rohrschneider and Whitefield (2012) also examine the impact of electoral institutions on congruence, finding that it is conditional on voters’ characteristics and the level of partisanship within the electorate. In contexts of highly partisan electorates, majoritarian institutions increase ideological distances between parties and voters since voters are spread across the whole ideological spectrum while parties converge towards the center. De-aligned voters, by contrast, also tend to converge more towards the center. As a result the distances remain smaller. Majoritarian systems are thus more congruent when partisanship is weaker, the opposite is true for proportional systems.
Besides the examination of the level of congruence, the concept has been used as an explanatory variable. Studying citizens’ satisfaction with democracy. Ezrow and Xezonakis (2011) find in a cross-national analysis of 12 countries between 1976 and 2003 that when parties’ policy choices are more congruent to the mean voter’s ideological position overall citizen satisfaction with democracy increases. In one of the most recent accounts, ideological congruence is approached from a different angle. Mayne and Hakhverdian (2017) are interested in the effect of sociotropic and egocentric judgments of congruence on democratic satisfaction. The authors show that only egocentric congruence boosts satisfaction with democracy. These findings clearly indicate that conclusions for the functioning of representation, citizens’ evaluation thereof, and electoral consequences heavily depend on how congruence is measured.
Wlezien (2017a) recently links this research to electoral success of U.S. presidents. Studying the costs of ruling, he shows that U.S. presidents tend to lose more votes the longer they are in power. He associates the costs of ruling with an increasing misrepresentation of voters.
Jones and Baumgartner (2004) were the first to define congruence as shared priorities between representatives and the public. They compare Congress’ attention allocation, measured with topic coded congressional hearings and laws, with citizens’ issue priorities as observable in MIP survey responses. Finding an “impressive congruence between the priorities of the public and the priorities of Congress across time” (Jones and Baumgartner, 2004, 1) and also evidence for congruence between citizens’ priorities and law-making, the authors conclude that the public is “seriously involved in the agenda-setting process, not an ignored bystander” (Jones and Baumgartner, 2004, 20). The only difference is that the public “lumps” concerns while congress “splits” issues, dealing with a broader variety of policies.
In a more recent account, Reher (2015) combines voter surveys from the European Election Study with candidate survey data to examine whether congruence in priorities has a similar effect on democratic satisfaction as ideological congruence. The results indicate that the linkage is the same and that the effect increases with democratic experience.
With a focus on the proclaimed democratic deficit in the European Union, Beyer (2017) finds that European Council Conclusions are just as congruent to citizens’ concerns as the individual member states who move between 30 and 60 percent of MIP-relevant agenda shares.
Flavin and Franko’s (2016) focus on unequal representation during the agenda setting phase adds to the literature on inequality (see below) by showing that differences are already present at earlier stages of the policy-making process.
Compared to studies of ideology or policy positions studies on priorities often take a more aggregated view on representation and compare percentages in the population and the elite. They seldom make references to priorities of the median voter or take the priority distribution directly into account. Of course, this is also due to unresolved methodological challenges when using priority data.
Although the literature on attentional congruence offers the advantage of distinguishing between important and less important issues when measuring the quality of representation, scholars have focused more on congruence in specific policy areas.
Monroe (1998) provides one of the first and most detailed over time comparisons of policy congruence. Comparing public opinion and public policy between 1960-1979 and 1980-1993 in the U.S., he finds that although policy outputs are consistent with policy preferences in 55 percent of the cases, a decline of eight percent occurred towards the end of the period. Issues that are more salient to the public are more likely to be congruent.
More recent scholarship focuses on explanations for and consequences of policy congruence. Louwerse (2012), for example, examines the extent of parties which fulfill their collective mandate with the responsible party model. The comparison of a majoritarian and a consensus democracy, the UK and the Netherlands, yields similar, reasonable levels of congruence. Levels of mandate fulfillment are higher in absolute terms for majoritarian systems and in proportional terms for proportional systems. His findings again go back to and should be connected more closely with the ’ideological congruence controversy’.
De Sio and Franklin (2012) developed the issue yield model that focuses on parties’ campaign issue choices. According to these authors, parties act strategically to minimize electoral risk, using issue congruence to make themselves more attractive to voters.
In the U.S. another strand of literature has evolved around the question how voter characteristics affect levels of congruence. Bartels (2008) and Gilens (2005, 2011, 2012) argue that the rich are better represented than the poor. They show that the size of the gap between voters’ preferences and representatives’ policy choices depends on citizens’ economic status. Flavin (2012) is one of the more recent examples in this tradition. He still finds that economically well situated constituents are advantaged in the political process. Policy making is more congruent with the preferences of high income than low income citizens in the U.S.. This is the case both in state policies in general as well as in social issues like the death penalty and abortion. Flavin and Franko (2016) furthermore find that this pattern already evolves at the agenda-setting stage examining differences in issue priorities between rich and poor citizens and bill introduction. Studying unequal representation comparatively, Giger et al. (2012) come to similar conclusions regarding the inferior representation of the preferences of the less economically advantaged citizen although they reveal important cross-national differences. Bhatti and Erikson (2011), on the other hand, contradict these studies and report no significant differences in congruence to more affluent voters. Similarly, Branham et al. (2017) show that policy outcomes are only slightly more favorable towards the rich, also because the rich and middle income groups often agree.
Beyond the focus on inequality, others focus on particular policy fields to draw a more detailed picture of congruence: Arnold et al. (2012) explain domestic parties’ positions on European integration preferences with voters’ preferences as well as party and electoral characteristics. Their results indicate that parties are in fact responsive to voters’ preferences in the realm of European integration. Immigration policies and the proclaimed gap to more restrictive public preferences has also received attention, particularly resulting from the recent increase in salience among the publics in Europe. Morales et al. (2015), for example, have comparatively examined the effect of politicization on congruence in this policy field. They show that the salience and intensity of the public debate as measured in extensive media coverage are conducive to policy congruence. The authors report cross-national variation both regarding the existence of the gap itself and the change in immigration policies as well as with respect to the elements of politicization (Morales et al., 2015, 1509ff.). Likewise, Norrander and Wilcox (1999) find for abortion policies that grass-root activism and public opinion match and are reflected in state policy.
Vasilopoulou and Gattermann (2013) apply the question of “matching policy preferences” to the case of the European Union, finding that the level of congruence between MEPs and their voters varies across issues. It depends on the party family, the frequency of contacts, and MEPs seniority. Euroscepticism, electoral systems, and EU membership length also play a role.
In the first comparative study to include multiple policy fields, Stecker and Tausendpfund (2016) not only analyze the similarities and differences of policy congruence between 45,000 citizens and 31 governments in 15 Western and Central Eastern European countries on six issue dimensions, but also focus on the consequences for satisfaction with democracy. Incongruence between citizens’ and governments’ views are associated with lower satisfaction with democracy. Also including ideological congruence, Stecker and Tausendpfund (2016, 506) show that citizens are not only interested in congruence on the left-right axis, but also on other issues, most importantly in the area of redistribution and European integration. However, ideological congruence remains the most important factor for satisfaction with democracy. Political interest acts as a moderator: the more politically interested citizens are, the more they suffer from ’policy deviations’ (Stecker and Tausendpfund, 2016, 506). The authors’ argument that they bring a "multidimensional perspective to the study of policy congruence between citizens and governments in a time when the preference structure of parties and citizens has become increasingly complex" gives reason for the continuing academic interest on a topic that has its roots in the middle of the last century.
The question whether political actors adhere to normative ideals and are responsive to their constituents when enacting policies have occupied political scientists for decades (e.g. Dahl, 1971). However, few believe that politicians are responsive due to a normative ideal. Rather, they act in line with citizens’ wishes because it raises politician’s re-election chances when they are responsive to citizens (Stimson et al., 1994).
Empirically, many studies which analyzed politicians’ responsiveness to public opinion report high overall levels of responsiveness (but see: Jacobs and Shapiro, 2000). We first discuss aggregate studies of public opinion and policy before shifting to research on individual policy fields. Most of the initial work has focused on the United States and compared voters’ preferences on the national, district, or state level with the behavior of political actors or policy outputs. They typically focus on how changes in public opinion lead to changes in public policy — ideally with time-series data, but to satisfy our definition at a minimum by measuring public opinion at time t and examining its effect on policy output at time t+1.
Page and Shapiro (1983) are among the first to apply this approach which has since developed into the state of the art in responsiveness research. The authors examine public opinion in the United States with survey data from the 1930s to the late 1970s and its relation to actual policy outputs (studying various policy fields and including time lags). In a nutshell, they study if policy moves in the same direction as public opinion. Due to the temporal ordering of policy they are able to show that opinion causes policy more frequently than vice-versa (Page and Shapiro, 1983, 185 & 189).
Also using time series data but studying the public opinion — policy link on an even more aggregate level, Stimson (1991) and Stimson et al. (1995) show that policy outputs are mostly in line with the public mood. They argue that public opinion is often vague and uncrystallized but that politicians can derive information about public opinion from a broad public mood which moves in either a more liberal or conservative direction. Analyzing the relation between this public mood and legislation introduced by the U.S. Congress, the president, and the Supreme Court they find that policy generally reacts well to changes in public mood (Stimson et al., 1995, 557, see also Erikson et al. 2002, Stimson 1991). Studying the second dimension of public mood Nicholson-Crotty et al. (2009) analyze if U.S. federal criminal justice policy responds to public mood for more or less punitive policies. They find that political actors respond to preferences about more or less punitive policies when enacting criminal justice legislation and not to preferences about spending.
Going one step further Wlezien and Soroka contributed in important ways to the development of the dynamic model of representation. With their seminal work on the thermostatic model of representation they teach us that responsiveness of political actors to public opinion and responsiveness of the public to policy changes go hand in hand. Political actors react to changes in public opinion, which leads to a change in the public’s policy demands. If policy moves in the desired direction the demand for change declines. (among others: Wlezien, 1995, 2004; Soroka and Wlezien, 2010).8
Since then many studies have confirmed these findings. Developing more finegrained theoretical models of responsiveness, they have increasingly turned towards conditional explanations of responsiveness (e.g. Wlezien and Soroka, 2012; CanesWrone et al., 2001; Cohen, 1997), variance in responsiveness towards specific subgroups (e.g. Rigby and Wright, 2011; Druckman and Jacobs, 2011; Gilens and Page, 2014; Jacobs and Page, 2005; Wlezien and Soroka, 2011)9, responsiveness in specific policy fields (see section 3.2.2), or experimental approaches (e.g. Butler and Nickerson, 2011; Butler and Broockman, 2011; Butler, 2014). Studies on population subgroups focus for instance on inequality in representation across income groups (some on a more aggregate level, others regarding specific policy areas). We learn that policy makers are more responsive to the rich than the poor, organized business interests, and experts (Rigby and Wright, 2011; Druckman and Jacobs, 2011; Gilens and Page, 2014; Jacobs and Page, 2005). Wlezien and Soroka (2011) provide contradictory evidence. They show that policy preferences do not differ between groups in many areas (but see Page et al. (2013) for an opposing argument). Thus, even if inequality in representation exists, it may not affect policy in most fields. Experimental approaches provide for the first time truly causal evidence of responsiveness. Butler and Nickerson (2011) for instance explicitly study how information about voters’ preferences affects MPs vote choice on a specific spending proposal. They show that MPs who were randomly selected to receive survey information about their district’s preferences, where more likely to vote in line with constituents’ opinion than those who did not receive such information. Other experimental work is focused less directly on policy responsiveness10, but still presents important evidence about relationships between citizens and legislators. It shows that responsiveness depends on voter and MP characteristics such as race or socio-economic background (e.g. Butler and Broockman, 2011; Butler, 2014).
Moving towards a more comparative approach researchers have studied responsiveness in the US not only on the national, but also on the subnational level. These studies faced huge challenges in terms of data availability in the past (Shapiro, 2011). Nevertheless, by now there is little disagreement that public opinion also matters for policy making in the states. Introducing an aggregate measure of state ideology that became similarly popular as public mood Erikson et al. (1993) showed that variation in state policies can be explained by the ideological orientation of state publics.
The U.S. political system is of course extraordinary in that it encourages a particularly strong link between representatives and represented. One of the first studies outside the U.S. was conducted by Brettschneider (1996) who studied responsiveness (and congruence) in Germany. While he conceptualizes the concepts ambiguously (see above) he distinguishes them more clearly in the empirical analysis and provides evidence for both, responsiveness and congruence, in Germany. Since then studies on responsiveness have gained foothold in Europe as well: Hakhverdian (2012) follows Stimson et al. (1995) and uses time-series data to study how public opinion and policy outputs relate to each other. He addresses the challenges associated with time-series data by isolating left-right positions from annual budget speeches delivered by the Chancellor of the Exchequor to the House of Commons. Applying a similar research design as Page and Shapiro (1983) or Stimson et al. (1995) he finds strong support for the responsiveness hypotheses: public opinion change often leads to public policy changes (as expressed in speeches) in the subsequent year. Likewise, Wlezien and Soroka (2012) who study public spending and spending preferences across 17 OECD countries find that preferences for spending influence actual government spending significantly. They qualify their statement with the finding that the effect is conditioned by a country’s institutional design.
Driven by the importance of parties in European democracies a growing literature analyses how parties — not governments or legislatures — respond to changes in citizen preferences or priorities. Using manifesto data and voters’ left-right placements, these studies find that parties respond to shifts within the electorate if voters shift away from the parties’ policy positions (Adams et al., 2004). The effect appears to be more pronounced for center and right parties (Adams and Somer-Topcu, 2009), who primarily respond to shifts from the mean voter (Ezrow et al., 2011), whereas leftist parties’ policy positions are more stable (Adams and Somer-Topcu, 2009), and only affected by shifts from their core supporters (Ezrow et al., 2011). Butler et al. (2016) discuss a different pathway of party responsiveness. Linking the literature on European and U.S. systems they evaluate how individual MPs take up constituency preferences to lobby for policy changes within their own party. While MPs cannot as easily respond to voter preferences in strong party systems as in the U.S. they use voter cues to change the policy position within their own party.
Overall, the literature on responsiveness finds strong evidence that government, parliaments and individual political actors are responsive to public opinion — be it in the U.S. or beyond. Despite some differences between political systems, most studies conclude that public opinion influences public policy strongly.
More recently, scholars of responsiveness have followed the example of scholars on issue congruence and assess the consequences of policy responsiveness for democratic satisfaction or the acceptability of political decisions (e.g. Esaiasson et al., 2017; Esaiasson and Wlezien, 2017; Arnesen and Peters, 2017). Others examine under which circumstances citizens value and perceive responsiveness (Rosset et al., 2017; Bowler, 2017), or responsiveness to particular subgroups in society (Grimes and Esaiasson, 2014).
A sizable number of studies within and beyond the United States study responsiveness on a more disaggregated level and focus on responsiveness in specific policy domains. Compared to the more generalized or global studies using for example public mood or ideology scores these studies have the advantage of providing direct evidence of how public opinion in a certain policy field affects policy output in this field. This makes the proposed causal relationship more explicit, and uncovers differences in responsiveness between policy domains (Wlezien and Soroka, 2016). Unless they combine very diverse sets of issues, however, they come at the cost of a narrower picture of how policy works in a given context. Aggregate measures allow researchers to study the link between public opinion and policy in the absence of issue-specific public policy data (Burstein, 2010).
The field of foreign policy has attracted particular attention. One of the most studied questions is if public opinion influences defense spending (Manza and Cook, 2002b). Most of these studies find robust and quite sizable effects of public opinion on spending for security issues (e.g. Wlezien, 1996; Bartels, 1991; Hartley and Russett, 1992). For the time period 1965-1990 Hartley and Russett (1992), for instance, find that public opinion significantly influences governments’ military spending level. Bartels (1991) even predicts the costs of public opinion: he estimates that the strong demand for higher defense spending increased the budget by 17 billion US Dollars, accounting for almost 10 percent of the total defense budget of 1982 (Bartels, 1991, 464).
Also beyond the U.S. defense spending is one of the most popular issues for the study of responsiveness. Comparing the effect of public opinion on changes in defense budgets in five countries (United States, Great Britain, France, Germany, and Sweden) Eichenberg and Stoll (2003) find clear evidence that public policy responds to public opinion in all countries but Sweden.11 These findings are corroborated for Great Britain by Soroka and Wlezien (2005).
Outside the field of foreign policy fewer studies analyze responsiveness in specific policy fields, probably also due to a lack of time-series public opinion data in other areas. An exception are Lax and Phillips (2009) who study the effect of public opinion on gay rights for public policy for both congruence and responsiveness on the state level- although as pointed out above, their operationalization of responsiveness lacks the temporal dimension that we deem necessary. In terms of responsiveness they find for all policies that more favorable public opinion increases the likelihood that a particular policy is adopted within a state. However, in terms of congruence the results are less encouraging: despite high responsiveness only 50 per cent of the policies are congruent with public opinion. They extend their study in a subsequent contribution where 39 policies in eight policy fields are analyzed. They find a positive effect of public opinion on public policy in all but four cases (significant for half of the policies), but again less optimistic results in terms of congruence (Lax and Phillips, 2012).
Other examples from the U.S. include studies on the effect of public opinion on abortion (Norrander and Wilcox, 1999), child labour, capital punishment and women rights (Mooney and Lee, 2000; Erikson, 1976). Many U.S. studies on specific policy domains are actually conducted on the state- or county-level. They often introduce institutional features such as direct democracy as conditional explanations of varying state-level responsiveness (see also Shapiro, 2011). While there is disagreement about the importance of institutional features, few question the importance of public opinion for state-level policy making (among many others Gerber, 1999; Matsusaka, 2008; Lupia et al., 2010; Percival et al., 2009).
Also comparative studies have extended their focus. Comparing the UK and the U.S., Bevan and Jennings (2014) for instance find that legislative and executive agendas are responsive to citizens’ priorities on a broad number of topics (such as education, environment, law). Others show that political actors generally respond to public opinion in such diverse fields as health (Soroka and Wlezien, 2005), same-sex marriage, EU membership, and economy (in the UK) (Hanretty et al., 2017), and social policy spending (Brooks and Manza, 2006). By contrast, Hobolt and Klemmensen (2008) report a more nuanced picture of responsiveness. They study responsiveness in six policy fields (defense, law and order, public health, housing, education, and social services) by comparing public priorities (measured through most important problem questions) to governments’ policy priorities as expressed in speeches. They further analyze changes in public expenditure within each field as a proxy for policy behavior and find that responsiveness is not constant across time and space, but depends on the institutional and strategic context (Hobolt and Klemmensen, 2008, 332).
Despite strong empirical evidence for responsiveness some remain skeptical and doubt that public opinion influences policy outputs in important ways. The debate evolves around two main topics. (1) Some question if the public even has a coherent "opinion" to which politics could be responsive. (2) Others, while not doubting a correlation between public opinion and public policy, challenge the direction of the relationship: they argue that politicians and policy outputs influence the public’s view on policies rather than vice-versa.
The problem of Inconsistent Public Opinion
A first strand of literature which is critical of the link between public opinion and public policy questions whether the public has consistent and meaningful views which political actors can follow. Most famously Converse (1964) showed in the 1960s that survey respondents change their responses when they are interviewed repeatedly. He interpreted this phenomenon as "non-attitudes". If this is true it poses a great challenge to theories of responsiveness. How can political actors be considered responsive to public opinion, when public opinion does not exist (e.g. Ferejohn and Kuklinski, 1990, 3)? Consequently, some argue that we can only expect policy responsiveness on issues which the public is greatly interested in over a prolonged period of time (e.g. Burstein, 2003, 30). Continuing and constant responsiveness, by contrast, would exceed the cognitive capacities of the public and political actors.
The Problem of Reversed Causality12
The discussion evolving around the issue of reversed causality can be illustrated with Cohen (1997). Cohen discusses the dilemma of U.S. presidents who are caught between expectations of policy leadership and demands for responsiveness to public opinion. He argues that presidents should not only be required to respond to public opinion but that they also need to provide policy leadership (Cohen, 1997, 1). Cohen’s core argument is that presidents are only responsive to the public if it does not interfere with their policy agenda (Cohen, 1997, 32). Otherwise, presidents attempt to influence public opinion in a way favorable to their policy agenda through public speeches and other ways of public outreach. Similarly, Jacobs and Shapiro (2000) claim that political actors use polls and mass communication to influence public opinion and manipulate citizens to support or at least accept policy proposals that they would otherwise reject. Rather than actually changing public opinion (but see Page and Shapiro (2010, ch. 8)), presidents typically try to influence the salience of a topic and thus the problem awareness of the public. Presidents then appear responsive despite having "created" public opinion themselves by raising public’s awareness to a particular issue (Beckett, 1999).
This somber view of policy responsiveness is contested by those who study both directions of causality within the same framework and use time-series data to control the temporal order. Already Page and Shapiro (1983) show that the effect of public policy on public opinion is weaker and less consistent than the effect of opinion on policy making. In a methodologically more advanced study Hakhverdian (2012, 1398-1399) finds three decades later that at least in the UK, policy does not affect public opinion in any statistically meaningful way.
By contrast, using an instrumental variable approach to solve some of the difficulties associated with measuring the effect of policy on public opinion13, Gabel and Scheve (2007) find that at least for the issue of European integration elites are able to influence public opinion - namely, more negative elite messages about European integration significantly reduce public support for Europe.
These conflicting findings are reconciled, for instance, by Steenbergen et al. (2007) who find that political elites both, shape and respond to public opinion on the issue of European integration. In a similar way, Wlezien’s thermostatic model also implies that public policy, both, shapes and responds to public opinion (Wlezien, 1995).
Given this review, we conclude with a discussion of what we see as promising opportunities for future research in terms of concepts and empirics. We discuss the marginal role of public opinion in public policy research, how different visions of democracy affect levels of congruence and responsiveness, and finally propose a novel combined concept of congruence and responsiveness.
The concepts of congruence and representation deal with the relationship between citizens’ and elites’ preferences and priorities and public policy-making. As such, it would be natural to find the link between public opinion and public policy also in public policy theories. Yet, although questions of congruence and responsiveness have attracted great attention among scholars of representation they have remained surprisingly absent from public policy research. Among the major public policy theories, we exemplary present the Punctuated Equilibrium Theory, Multiple Streams Theory, and the Advocacy Coalition Framework. Public opinion is often not directly integrated in tests of common public policy theories. They rarely consider which role their constructs play for representation.
The Punctuated Equilibrium Theory (PET) (Baumgartner and Jones, 1993), for example, is a theory of agenda-setting in the context of attention scarcity. Policy change on constrained agendas only happens when a certain threshold is reached. In this case, there is overreaction defined as punctuations. Venue shopping (overcoming the institutional limitations) or changing policy frames are major factors enabling a topic to make it to the agenda. Public opinion changes could affect policy frames. PET theory, however, does not explicitly incorporate the public and instead focuses more on external shocks and interest groups. With the side statement "public opinion reacts to public policy more than it causes it" Baumgartner and Jones (1993, 247) dismiss the role of public opinion for their theory of the agenda-setting process. Only very recently this appears to change. While measuring responsiveness indirectly due to data limitations, Fagan et al. (2017) for instance show that the extensiveness of punctuations differs in response to varying democratic responsiveness.
The Multiple Streams (MS) Theory (Kingdon, 1984) explains policy change and agenda setting with a process-oriented approach. In Kingdon’s framework, agenda setting works through three independent but interdependent concepts that jointly create windows of opportunity for policy change: the problem stream, policy stream, and political stream. These streams run parallel most of the time. Policy change occurs when a window of opportunity opens which leads the streams to cross (Béland and Howlett, 2016). Kingdon (1984) discusses public opinion in the context of the media and its agenda setting function. Public opinion thus potentially plays a role in the problem stream, which covers perceptions of public problems. According to the MS theory public problems reach awareness due to crisis or feedback effects that raise public attention. Nevertheless, in empirical applications of the theory public opinion plays a subordinate role.
Finally, the Advocacy Coalition Framework (ACF) (Jenkins-Smith and Sabatier, 1993) was developed as a theory of the policy process which explains policy change through policy learning. Policy is introduced in subsystems (issue specific networks) where coalitions compete over influence in policy making. Most of the time policy change happens incrementally through policy learning. Coalitions learn through observing the effects of policies, comparing them to their secondary beliefs and updating them when needed. In times of crisis external shocks may lead to subsystem instability and provoke change. While public opinion is one of several external factors that may explain policy change (especially change in core beliefs) it has only recently been added as a factor that may play an internal role in policy change (Jones and Jenkins-Smith, 2009; Weible et al., 2011).
This short overview of public opinion in public policy theories underlines that the two research fields have developed mostly independently from each other since the 1980s. We believe that much could be learned if scholars of representation and public policy joined forces to fully understand the policy process in modern democracies. Scholars of representation probably overestimate the role of public opinion in policy-making and neglect other pressures on legislators such as attention scarcity, the influence of advocacy coalitions or interest groups. Scholars of public policy by contrast have not yet incorporated the causal effect of public opinion in their models. Public policy theories do not seem to acknowledge the role of representation. Nor do they question how external shocks work and that they might only gain elites’ attention through public opinion. The integration of representation and public policy theories into a common framework may therefore provide a fruitful endeavor for future research. Recent developments in the field might be interpreted as a sign that this movement is already starting (e.g. Bevan and John, 2016; Jones and Baumgartner, 2004; Fagan et al., 2017).
As outlined in our conceptualization above, one could argue that responsiveness comes closer to the theoretical idea of representation while congruence is more successful in assessing whether the majority gets what it wants. Yet, the literature on the respective concepts introduces a puzzle: despite the generally agreed upon claim that congruence measures reflection of the majority’s will, we learn that proportional systems are better (or at least as good) in enabling congruence between citizens and their representatives (see section 3.1.1).14
Surprisingly, the literature on responsiveness seldom deals with the effects of electoral systems on the level of responsiveness. The only exception are accounts on parties by Adams et al. (2006) or political contestation by Hobolt and Klemmensen (2005; 2008) and Ezrow (2007) who study responsiveness in the context of party competition. Also, Wlezien and Soroka (2016) offer a theoretical discussion of this point. All other articles that study policy responsiveness and electoral system related concepts in fact measure congruence (e.g. Kang and Powell, 2010). What we learn from Adams et al. (2006) is that niche parties do not follow public opinion shifts in the way mainstream parties do and if they do moderate their position they are not electorally rewarded. Since niche parties as a phenomenon of proportional systems and mainstream parties seem to be more responsive, the overall responsiveness of parties might be higher in majoritarian systems. Hobolt and Klemmemsen (2005) analyze the effects of responsiveness for political contestation finding that public opinion drives policy due to the threat of electoral sanctions which are more likely in proportional systems. Measuring executive policy priorities in opening speeches, the authors argue that more party competition and vulnerability of the government make the executives in proportional systems more responsive to public opinion changes. In their subsequent article Hobolt and Klemmensen (2008, 313) claim that "directly elected executives are more responsive to public priorities than indirectly elected executives" but that "executives in plurality systems are less responsive to the public’s priorities than executives in proportional systems". They find that rhetorical responsiveness (speeches) is highest in Denmark and the U.S .and lowest in the UK, but that American presidents show the highest level of effective responsiveness (budgets). Finally, Wlezien and Soroka 2016 argue that majoritarian systems are better in achieving responsiveness to public opinion change between elections, due to (1) their single-party governments and (2) higher electoral incentives. Proportional systems, by contrast, are better in achieving congruence via elections than majoritarian systems. This argument is corroborated by Ezrow (2007). He finds that the effect of changing variance in voter’s policy preferences is stronger in majoritarian than proportional systems.
Considering the divergent views in the literature it remains up to future research to resolve the question of whether electoral systems have a similar effect on responsiveness as on congruence.
The previous sections taught us that “congruence can be low even if responsiveness is relatively high” (Canes-Wrone, 2015, 148) but equally that congruence can be high while responsiveness is low. As such, the two concepts measure two distinct mechanisms: responsiveness being closer to the idea of representation, congruence measuring the output of representation, no matter how it evolved. Both concepts are thus important aspects of representation. Citizens expect their representatives to be responsive, but what they actually perceive is congruence. Methodologically, both concepts have advantages and disadvantages. While responsiveness does not require measuring the public’s and policy-makers’ preferences on the same scale (Canes-Wrone, 2015, 150) it suffers from difficulties of empirically identifying the causal relationship. Congruence in turn allows us to establish if majority will and their representatives or policy outputs are in line, but the right survey questions are often difficult to come by.
This review illustrates that both concepts influence each other and at times require combined analysis in order to understand the fully understand representation. Policy-makers’ perceptions of the current level of congruence affect their responsiveness. If responsiveness moves policy to citizens’ preferred point, the output leads to congruence. We argue that also the strong linkage of the two concepts requires further attention. Only the combination of both concepts gives us a complete picture of the functioning of representation. Moving beyond the existing literature, there are two possible ways to include both concepts into a combined analysis. One is to twist the idea of responsiveness in a way that responsiveness is only deemed to be fulfilled if it is in line (congruent) with majority opinion, thus leading to congruence. We call this mechanism congruent responsiveness.
Figure 4: The Ideal Case of Congruent Responsiveness
Figure 4 illustrates the idea of congruent responsiveness. Congruent responsiveness refers to all cases in which responsiveness leads to a congruent situation, thus a policy adaption that follows public opinion change from time t to t+1 in a way that then matches the majority’s will at time t+1. This is illustrated by the box at time 2 containing public opinion and public policy at the same level. At time 1 public opinion and public policy may or may not be in sync (in figure 4 they are not). Congruent responsiveness occurs if policy change leads to an overlap of public opinion and public policy at time 2. We do not imply that studies of congruence and responsiveness are not meaningful in their own right. Rather, we want to propose an additional concept that allows researchers to combine the strength of both ideals and might further our understanding of representation. We are aware that the application of this concept is subject to huge data challenges.
A two-step process is the second option for integrating the two concepts better.15 The first step analyzes how representatives respond to citizens’ demands. The second step focuses on the consequences: namely on how responsiveness affects the overlap in public preferences and policy outputs (congruence).
[The concepts of congruence and responsiveness are] slippery to conceptualize and difficult to observe. The last half-century of theoretical and empirical research in political science has taught that such essential concepts as citizens’ preferences, political influence, and policy consequences are fraught with exquisitely complex problems for analysis.
— Powell, 2000, 15-16
Slippery concepts and difficulties associated with their operationalizations and measurement are related. The occurrence of both issues is even more problematic. Trying to circumvent one or the other issue, scholars oftentimes tweak the concepts and their empirical application. As a result, we see a multitude of different approaches to a topic that has severe consequences for representative democracies. Not achieving congruence or responsiveness can lead to political disenchantment and the rise of populist parties who precisely promise to respond to all those demands that the incumbent political elites do not seem to hear. With this review, we provide a clear-cut, mutually exclusive definition of congruence and responsiveness and consider recent developments within the respective literatures. An important aim of the article was to establish new conceptual boundaries where necessary. While emphasizing both concepts’ importance as currencies of representation, we show that congruence and responsiveness are two sides of the same coin that require clear conceptualizations and measurement strategies but are jointly required for the full picture of the functioning of representation in any modern democracy.
This literature review provides what we believe is necessary for scholarship to move forward: a clear distinction of the concepts, graphical and formal clarification, and an accordingly structured literature review on the most important contributions in the field. We not only cover the path breaking U.S. literature, but also include contributions from other world regions which often qualify the relationship and introduce conditional explanations. The combined review which takes the specificities and commonalities of both concepts into account paves the way for more integrated research designs.
Our final discussion section develops on this vision. We suggest to strengthen the link between scholars of public policy and public opinion for a more complete understanding of the public policy process and stress the need to study congruence and responsiveness in a common framework. Responsiveness is not sufficiently informative about the quality of representation without knowing whether the majority gets the policies it wants. Congruence in turn is an incomplete evaluation criterion for the representational process as it may have little to do with legislators’ actions in the policy process. Combining both allows a comprehensive study of the quality of representation. Especially in recent years the combination of both concepts has gained attraction and we hope that we encourage many to follow.
1 Wlezien (2017b) also contributed to a better conceptualization, focusing more on congruence than responsiveness.
2 Research on congruence and responsiveness analyzes citizens’ and elites’ priorities, attitudes, preferences and public opinion. We are aware of the conceptual differences between these terms and acknowledge their respective importance. When we compare and summarize multiple contributions at the same time, however, it is not always possible to keep the distinctions for linguistic reasons. We try to be as precise as possible, but sometimes choose one of the concepts over the other to improve reader friendliness. To some extent the terms are thus used interchangeably although they have of course slightly different meanings.
3 We elaborate on the effect of electoral systems in section 4.2.
4 While responsiveness in the sense of responding to the public’s wishes implies causality from a conceptual point of view - actors consider public opinion and respond to it when enacting policy, many researchers acknowledge that it is difficult to demonstrate empirical causality through regression type analyses (Wlezien and Soroka, 2016; Wlezien, 2017b). More recent experimental approaches also put the empirical focus on causality instead of correlations (Butler and Nickerson, 2011).
5 Groups such as Americans for Democratic Action (ADA), Americans for Constitutional Action (ACA), or the American Conservative Union (ACU) produce scales which intent to measure the policy behavior of members of the U.S. congress.
6 This review is concerned with the relationship between public opinion and public policy or elected representatives and the public, respectively. We acknowledge that there is a similar debate about responsiveness of court decision making to public opinion which lies beyond the scope of this review (e.g. Brace and Boyea, 2008; Giles et al., 2008)
7 and (Policy OR Representation OR "Public Opinion") as modifiers to exclude irrelevant literature
8 A somewhat related literature studies when public policies produce changes in public opinion (Soss and Schram, 2007).
9 Not all of these studies fulfill the temporal ordering that we deem necessary. However, as long as they conceptually measure responsiveness we still discuss them here.
10 This research may be subsumed under Eulau and Karps’ 1977 concept of service and Essaiasson and Wlezien’s discussion about the importance of listening and explaining policy choices to voters.
11 They explain the null finding for Sweden by a lack of variance: Sweden’s defense budget and public opinion remains almost stable over the period of investigation.
12 There is an extensive literature on framing (e.g. Chong and Druckman, 2007), priming (e.g. Ewoldsen et al., 2007), and cueing (e.g. Gilens and Murakawa, 2002) studying how political actors influence public opinion which is directly related to the topic discussed here (see also Zaller, 1992). As the focus is on responsiveness and congruence rather than non-responsiveness, however, the discussion is deliberately restricted to few influential contributions in the field of responsiveness.
13 For instance, time series data cannot solve the endogeneity problem of public opinion and public policy. If elites correctly anticipate public opinion they may be responsive despite enacting policies prior to the measurement of public opinion.
14 However, work by Warwick (2016; 2010) and Lupu and Pontusson (2011) may shed doubt on the mesaurement approaches of some of these studies. Warwick, for instance, criticize that representation of the median voter is much weaker than often assumed.
15 Wlezien (2017b) makes a similar point in his discussion on congruence and responsiveness.