Does public policy reflect the wishes of the public? To what degree? Under what conditions? These are important questions for political science and more broadly, even for those who debate the value of such representation (see, e.g., Mansbridge Reference Mansbridge2003; Sabl Reference Sabl2015), some of whom advance alternative yardsticks (Saward Reference Saward2019; Disch Reference Disch2021; see also Pitkin Reference Pitkin1967), including moving “beyond responsive representation” itself (Bagg Reference Bagg2024; also see Shapiro Reference Shapiro2016; Rehfeld Reference Rehfeld2018). Not surprisingly, given its clear and widespread scholarly (and political) importance, there now is a massive amount of research on the representation of public opinion in policymaking.
Most of the existing work focuses on positions. A common approach is to compare the policy opinions of people in districts and those of their representatives, or else the latter’s roll call votes. This was first employed in research on dyadic representation that began with Miller and Stokes’s (Reference Miller and Stokes1963) examination of members of Congress, and continues today, both in the US and abroad. Here, scholars focus on the association between aggregate district opinion and the actions of their representatives, and there is evidence of a connection that varies across issues.Footnote 1 Another common approach compares the positions of the broader public with those of institutional actors, e.g., political parties and governments. Beginning with Weissberg’s (Reference Weissberg1978) analysis of Congress, this traditionally is referred to as collective representation, research on which has exploded.Footnote 2 Golder and Stramski (Reference Golder and Stramski2010) provide a nice review and assessment of different approaches in this tradition, focusing on legislatures.Footnote 3 This research also relates to that on surrogate representation (Mansbridge Reference Mansbridge2003).
A newer, strain of research examines collective representation of priorities. Here, scholars assess the correspondence between what the public considers to be important, typically “problems,” and the attention government actors, e.g., legislatures, pay to different policy areas (Hobolt and Klemmensen Reference Hobolt and Klemmensen2005; Jones et al. Reference Jones, Larsen-Price and Wilkerson2009; Bevan and Jennings Reference Bevan and Jennings2014; Traber et al. Reference Traber, Hanni, Giger and Breunig2022). This is a growth industry and for good reason given that agenda-setting matters for policymaking, i.e., getting on the agenda is a necessary condition for policy representation.
The research demonstrates a lot of positive associations between the public and politicians’ positions and priorities.Footnote 4 It is important to be sure but offers only limited information about policy representation, since positions, priorities, and policies are different things. Positions may correlate with policy, but they are not one in the same, and while agenda representation may be necessary for policymaking, it is not sufficient; indeed, it may be that we observe representation of public priorities and misrepresentation of public preferences for policies in the government outputs that result.
There is an increasing amount of work on policy, however. Virtually all of this research examines correlations between opinion and policy decisions, e.g., usually laws. There is evidence from research across countries (e.g., Brooks and Manza Reference Brooks and Manza2007; Wlezien and Soroka Reference Wlezien and Soroka2012), sub-nationally across units such as the U.S. states and also cities and school districts (e.g., Erikson et al. Reference Erikson, Wright and McIver1993; Tausanovitch and Warshaw Reference Tausanovitch and Warshaw2014; Lax and Phillips Reference Lax and Phillips2012; Berkman and Plutzer Reference Berkman and Plutzer2005), and within national and subnational units over time (e.g., Eichenberg and Stoll Reference Eichenberg and Stoll2003; Soroka and Wlezien Reference Soroka and Wlezien2010; Caughey and Warshaw Reference Caughey and Warshaw2022; McGann et al. Reference McGann, Dellepiane-Avellaneda and Bartle2023).Footnote 5 Here, scholars again have found positive associations, though ones that vary with differences in issues and institutions.
Some may worry that policymakers do not really represent preferences in policy, and that those preferences simply reflect what governments do (Achen and Bartels Reference Achen and Bartels2016). Policymakers (and policies) are not randomly assigned, of course, and we know that people in different states and countries, like different districts, are not the same, so there is reason to suppose that the former to some extent reflect the latter.Footnote 6 That said, findings of dynamic effects are especially noteworthy – indeed, more causally compelling – as changes in policy temporally follow preferences. Although policymakers may induce opinion change in order to justify policy actions, there is little evidence that future changes in policy predict current changes in preferences, an important finding that points to other, seemingly exogenous causes (see Soroka and Wlezien Reference Soroka and Wlezien2010). There also is growing evidence of “thermostatic” public response, where policy feeds back negatively on preferences for more policy, in direct contrast with what we would expect with opinion leadership, namely, positive feedback (e.g., Wlezien Reference Wlezien1995; Erikson et al. Reference Erikson, MacKuen and Stimson2002; Kustov Reference Kustov2025). The patterns also are strikingly similar across subgroups, including income, education, and party identification (Soroka and Wlezien 2010; Reference Soroka and Wlezien2022), which is as we might expect given “parallel publics” (Page and Shapiro Reference Page and Shapiro1992).Footnote 7 Policy representation is a real thing; indeed, policymakers appear to be responsive.
While this conclusion may come as little surprise to some readers, it does directly challenge recent essays on representative democracy, perhaps most notably Achen and Bartels (Reference Achen and Bartels2016). Those scholars highlight the limitations of the public to monitor elected officials and hold them accountable, providing little information or incentive to represent in policy once in office. Other earlier research sets what seems a higher bar, spelling out the series of links that are required in the chain of representation from public interests to policy outputs to actual outcomes (Fung Reference Fung, Rein, Moran and Goodin2006). Given that breaks in any of the links introduce a democratic “deficit,” and the seemingly fatal one(s) that Achen and Bartels proffer, it may be surprising that we observe any representation at all.
This essay traces theoretical conditions for effective policy representation, especially the responsiveness of policymakers to public preferences. The examination focuses on the “input” and “output” sides of the equation, those relating to public demand and policy supply. Of the many links in the representational chain, these are of special importance. Consider that public demand provides the information policymakers can represent as well as the incentive to respond; without these, there is no basis for doing so.Footnote 8 While necessary, the demand for responsive policy is not sufficient, as effective supply depends on policymakers having the information, interest, and capacity to respond to the public’s preferences.
In what follows, I dig deeply into the conditions for effective demand and supply and the existing research addressing the degrees to which they are satisfied. The analysis makes clear that responsiveness requires a great deal of both the represented and the representatives (and scholars too), that these requirements are not easily met, and yet that they sometimes are, helping to explain the findings of previous research I briefly described above. It also highlights aspects about which we still know relatively little, much of which relates to policy supply, an unfortunate imbalance that may be understandable given the critical importance of public demand. The concluding section contemplates some of the leading avenues for future research on these underpinnings as well as the nature of representation and its political consequences. The portrait I offer in what follows is a very broad one, leaving some features incompletely covered or else placed in the online appendices.
On policy representation and responsiveness
It is useful at the outset to be absolutely clear about our object of interest – policy representation and responsiveness – and so I provide a basic formalization to accompany the text that follows. If there is representation, actual government outputs would be positively related to the public’s preferences, as in the following function:
$P = f\left\{ {P*} \right\},$
where P designates policy and P* the public’s preferred level of policy, both on the same scale. The focus is on preferences and not interests since the latter are difficult to conceive and actually measure, even as some scholars (e.g., Fung Reference Fung, Rein, Moran and Goodin2006) prioritize them and others attempt to impute them (also see the discussion in Soroka and Wlezien Reference Soroka and Wlezien2010).Footnote 9 Research typically concentrates on the median citizen, which is of special importance in democratic theory (Hotelling Reference Hotelling1929; Black Reference Black1948; Downs Reference Downs1957; Buchanan and Tollison Reference Buchanan and Tollison1984), though policymakers may be more responsive to others, as we address later in the essay.Footnote 10 We have representation if the government provides a lot of policy when the median voter wants a lot, i.e., when P* is high. This would not mean that the person gets exactly what she wants, i.e., that there is a match between P* and P, what we usually refer to as “congruence.”
We can formally depict this congruence between policy and the average citizen, borrowing from Achen (Reference Achen1978):
$P = a + B{\rm{ }}P* + \varepsilon, $
where the stochastic term ϵ reflects the other influences on policy besides public preferences. The units of analysis here can be spatial, such as countries, states, and cities, or policy types, and they can be temporal as well, where representation is dynamic, usually designated with a subscripted t. For expository purposes, and to begin with, let us consider preferences and policy across different country units at a particular point in time. If there is congruence, the coefficient (B) for opinion would be a perfect “1.0” and the intercept (a) would equal “0”; if a is non-zero, policy is shifted to the left or right of P*, depending on the intercept’s sign.
Now, if B is greater than 0 and less than 1, there still would be an association between P* and P but policy would not exactly match opinion, and is what some scholars refer to as “responsiveness” (Lax and Phillips Reference Lax and Phillips2012; also see Achen Reference Achen1978).Footnote 11 The gradient of policy would not be sufficiently steep, however. Of course, the relationship could be too steep, where B>1, e.g., where policy is too liberal in liberal countries and too conservative in conservative ones. There could be bias as well, indicated by the intercept in equation 2. But, these parameters only take us so far, as we also are interested in the fit of the equation, i.e., how well P and P* line up, which increases as the influence of other factors (ϵ) decreases. With perfect representation, B would equal 1, a would equal 0, and the variance of ϵ would be 0. Of course, a match between policy and the preference of the median person may still leave many people dissatisfied with the result, particularly the greater the variance of public preferences, i.e., where some favor a large government role and others a very small one. This obviously can matter a lot for politics and the integrity of political systems (also see Golder and Stramski Reference Golder and Stramski2010).
If there is representation, it can come about in two well-known ways. The first is indirect, where representatives and the represented simply want the same things, as might result if we select people who reflect our views, i.e., where the two are aligned. But what happens when the representatives and the represented want different things? Do politicians do what “they” want? Or what “we” want? This leads to a consideration of the second, more direct way to representation – genuine responsiveness. Although scholars often use the word to designate any positive association between preferences and policy, many conceive of it as more active. Indeed, some see responsiveness as explicitly dynamic, where policymakers follow the public or subsets of the public when the preferences of the latter change, i.e., when there is an opinion “shock” and the government responds (Russo Reference Russo, Cotta and Russo2020; Soroka and Wlezien Reference Soroka and Wlezien2010; Caughey and Warshaw Reference Caughey and Warshaw2022). This responsiveness occurs in between elections.
Dynamic responsiveness
There is dynamic responsiveness when public preferences shift in a more liberal direction and policymakers provide more liberal policy, i.e., policy adaptation.Footnote 12 This still would not mean that the public literally is directing policymakers, as it may be that the public serves mostly to limit their discretion, what Key (Reference Key1961) referred to as “dikes” constraining politicians’ actions. Importantly, responsiveness does not require that legislators actually change their policy positions, just that the policy agenda changes and the legislative body goes along.Footnote 13 In practice, using observational data, we only behold “statistical” responsiveness. We just cannot be sure based even on a robust, systematic relationship between policy and opinion over time that politicians are literally responding to the public. It may be that both the public and policymakers respond independently to the same things, which we cannot easily test.Footnote 14 It alternatively may be that policymakers effectively lead the public instead of following. Disentangling the direction(s) of causality is not easy, as we cannot experimentally manipulate public opinion, and it is not clear that we should even if we could. Dynamic approaches do offer insight, however, as we can see how preferences and policy influence each other over time, as discussed.Footnote 15 Based on that research, there appears to be responsiveness, even as we cannot be sure that there is an exact match between preferences and the magnitude of policy change.
Let us contemplate what such responsiveness requires and the degree to which the conditions apply given previous research. What follows focuses on the represented and their representatives, which can be thought of as part of two broad “sides” of the connection – one relating to inputs, or policy demand, and the other to outputs, or policy supply.
The “input” side
It almost goes without saying that for there to be effective responsiveness, people must have preferences for policy, without which policymakers could not respond. As Achen (Reference Achen1975) observed, “Democratic theory loses its starting point.” Preferences also must add up sensibly, as public choice theorists have underscored – see Riker’s (Reference Riker1983) essay on the subject that raises certain complications, ones that are taken up in Mackie’s (Reference Mackie2003) re-assessment. Riker stressed the possibility that preferences at the individual level do not necessarily produce a rational public preference, i.e., there can be a “cycle,” where a majority prefers option a to option b and option b to c, and yet prefers option c to a.Footnote 16 While an important theoretical challenge to policy responsiveness to public preferences, Mackie (Reference Mackie2003; Reference Mackie, Novak and Elster2014) highlights research showing that cycling is not very common and, where it does exist, tends to be confined to a subset of similar policies. So, do people have preferences?
On preferences
In his early statement on public opinion and democracy, Key (Reference Key1961) held that people did have policy “attitudes,” even as these were isolated, often thin, and sometimes weakly held. The body of research that has followed largely supports this view, even as it reveals considerable variation across issues and individuals (see Erikson and Tedin Reference Erikson and Tedin2019). People are more likely to have preferences on some issues than others, and some people are more likely to have preferences than others. This is highly intuitive.
Also early on, Converse (Reference Converse and Apter1964) highlighted “nonattitudes” in his classic statement on belief systems. He found high levels of instability in survey responses among most, but not all respondents, and inferred an absence of true preferences. That is, most people responded in what appeared to be random ways, what we might expect based on early critiques of the public (Lippmann Reference Lippmann1922). Achen’s (Reference Achen1975) reanalysis led to a more charitable conclusion, attributing response instability to question wording that led people with true attitudes to respond differently at different points in time. While Smith (Reference Smith, Turner and Martin1985) revisited the nonattitudes thesis and sided more with Converse, Zaller and Feldman (Reference Zaller and Feldman1992) proffered a middle ground, focusing on the importance of “considerations” when respondents answer survey questions. In effect, rather than simply retrieving a mental summary statistic from long-term memory, people sample from a distribution of relevant considerations, which produces response instability, where people provide one response at one point in time and another response at another. This is more common for some issues and individuals than others, echoing a recurring theme in the literature.
All of this said, what does it mean to say that people have policy preferences? Consider that we have focused on the representation of P*, the public’s preferred level of policy. Do people really have such specific preferences, even in very broad areas like defense or health care? We know that survey organizations rarely, if ever, ask respondents about how much policy they want, e.g., how much government spending. Survey organizations still do regularly ask about support for specific policies, including some with which people may have direct experience, e.g., whether the minimum wage in the U.S. should be $12.00. This may elicit something close to a preferred policy, but notice that responses at best tell us something about people’s support for that particular amount, not what they want. For instance, people may support raising the minimum wage to $12.00 but prefer it be $15.00, or some other amount, even including $11.00 or $10.00. There may be majorities in support of a range of specific policies in most areas.
This is not just angels on heads of pins; to assess policy congruence, after all, we need to be able to match up policy with what the public actually wants on the same scale, which is not possible using responses to our hypothetical minimum wage item (also see Wlezien Reference Wlezien2017a). The public’s preferred policy is even less clear when questions ask about more general policy actions, e.g., “raising the minimum wage.” Much the same is true for bills under consideration, which may include a variety of specific policies that most people just do not know, as appeared to be the case with the Affordable Care Act, i.e., Obamacare (Mettler Reference Mettler2011). Research by Hill and Huber (Reference Hill and Huber2019) further implies that responses to questions about particular bills – ones on which legislators voted – are not very revealing about true support for the bills. A growing body of research nevertheless assesses the relationship between public support for specific policies and government adoption of those policies, what Weissberg (Reference Weissberg1976) refers to as “majoritarian congruence” (see, e.g., Monroe Reference Monroe1979; Reference Monroe1998; Lax and Philips Reference Lax and Phillips2012; Rasmussen et al. Reference Rasmussen, Reher and Toshkov2019).Footnote 17 While such examinations cannot reveal whether the public is getting the policies it wants, they may provide evidence about whether the public is getting certain policies it prefers to the status quo.Footnote 18 The possibility seems to provide the basis for the policy “satisfaction” that Bartels (Reference Bartels, Lupu and Pontusson2023) highlights in his essay on political inequality.Footnote 19
That survey organizations do not ask people how much policy they want seems understandable, at least in most policy domains. Consider how you would respond if asked about how much the government should spend on national defense. Most of us would have a hard time even for this particularly important and salient area. These also are (very) general preferences, and providing more specific guidance, such as how to spend funds on different defense programs, is almost inconceivable for all but small numbers of experts. That said, people do appear to have quite specific preferences in some areas, such as abortion, including the exact circumstances under which it should be legal (Hout et al. Reference Hout, Perrett and Cowan2022). They not only express preferences when asked, they answer consistently, which is what we would expect if responses reflect true preferences.
The meaning of responses is less clear in other areas where survey organizations ask support for specific policies, e.g., whether “capital gains from selling stocks and other assets should be taxed at the same rate as other income.” People may express support or opposition when asked, but are these responses informative about preferences and, if so, to what degree? Part of the problem is that survey organizations usually have not included a “don’t know” (DK) response option, leading people to express a preference even when they don’t have one, guessing randomly or else based on heuristics.Footnote 20 Based on almost all of the survey research, it just is not clear how much policy people want and whether they really support the many specific policies about which they are asked. Importantly, very recent research (Fowler and Howell Reference Fowler and Howell2023) elicits people’s preferred levels of policy in selected areas and some other work attempts to directly assess congruence, matching preferences and policy on the same scale (Simonovits and Payson Reference Simonovits and Payson2023). This is encouraging, and I look forward to developments in this space.Footnote 21
Even where we can capture the public’s policy preferences in some way, it is important whether they are exogenous to government action, a common criterion of genuine responsiveness. If policies positively influence public preferences, after all, any claims of opinions “causing” those outputs are difficult to countenance; at the very least, they are compromised. This is well-known and taken seriously by some scholars, particularly those evaluating the dynamic relationships between the two, as we have already seen (e.g., Soroka and Wlezien Reference Soroka and Wlezien2010). That said, preferences do not emerge in a vacuum, and there is little gainsaying “the open interplay of opinion and policy” that V.O. Key (Reference Key1961) recognized decades ago. Competing politicians attempt to influence what we care about and want, and the public then gets to decide, what some have referred to as the “court of public opinion.” I find this a useful metaphor, as it parallels what businesses face in the marketplace. They try to gain our custom but are not always successful, and even where they are, it may not last as competitors enter the marketplace and compete, and our preferences evolve.
Consider the iPhone. Before it was on offer, consumers did not have it on their shopping lists, e.g., “milk, bread, eggs, an iPhone.” But Apple thought we might like it, and attempted to determine what specifically we would prefer as they produced the original model, and then as they updated. Politicians might do much the same, though motivated by votes, a form of what Mansbridge (Reference Mansbridge2003) referred to as “anticipatory representation,” though they also may be interested in doing something else, at least as much as they can given electoral repercussions. That “something else” could relate to politicians’ own preferences, which may reflect what they think is best for the country, or other interests that we will consider later, when turning to the output side. Whatever the reasons, the potential for such disjunctures places a premium on the public to notice what governments do, and to reward and punish based on their actions, as we will see.
Although survey organizations rarely ask about people’s specific preferred levels of policy – their absolute preferences – they often ask about their relative preferences. These questions relate to changes in policy, whether the government should do more or less. A common battery taps support for spending, and the wording from the General Social Survey is standard:
“Do you think we are spending too little, too much, or about the right amount on [national defense]?”
Responses to this type of question clearly are basic and may better reflect people’s preferences – that they have a sense of whether the government should do more or less (Stimson Reference Stimson1991; Soroka and Wlezien Reference Soroka and Wlezien2010). They also may offer information about the public’s preferred levels of policy, at least where relative preferences are thermostatic (see Wlezien Reference Wlezien1995). The public’s relative preferences (R) would equal the difference between the preferred policy temperature (P*) and the actual policy temperature (P). If the average person expresses support for more policy, therefore, the underlying preferred level of policy would appear to be greater than the actual policy.
This is the inference that Bartels (Reference Bartels2015) draws. While seemingly reasonable, it is (highly) questionable, and for a number of reasons (see Soroka and Wlezien Reference Soroka and Wlezien2010; Wlezien Reference Wlezien2017a). To begin, while there is evidence of thermostatic public response, relative preferences in all domains are not thermostatic. Even where they are, it is not clear what the neutral point is. Part of the problem is that responses differ depending on question wording, e.g., “welfare” vs “the poor,” where sizable majorities favor spending less on the former and more on the latter (Wlezien Reference Wlezien2017a). Even where question wording does not make a difference to survey responses, we do not know what the public has in mind in each policy area and whether this matches what researchers have in mind when measuring policy.Footnote 22 Perhaps most importantly of all, responses for many questions, e.g. those about spending and whether the government should “do more,” register unconstrained preferences, which often are more supportive of spending increases than those that require people to pay for such increases with spending cuts, increased taxes, or debt (Hansen Reference Hansen1998; Kolln and Wlezien Reference Kolln and Wlezien2024).Footnote 23 For the most part, then, we have only a very general idea of what people want, and policymakers may have even less of an understanding.
Issues and moods
Do people really have preferences in different policy areas? Previous research shows that public preferences across domains ebb and flow together over time, what Stimson (Reference Stimson1991) famously referred as policy “mood.” Although preferences in many areas are related, some quite closely, they are not reducible to a single mood, or two moods for that matter (Wlezien Reference Wlezien2004; Soroka and Wlezien Reference Soroka and Wlezien2010). Issues matter and their salience/importance is a defining feature. People care more about some issues than others. They place a greater weight on them. Their preferences on those issues are – at least might be – more consequential. News reporting is likely to pay more attention to them.Footnote 24 Politicians too. The apparent role of abortion in the 2022 U.S. Congressional elections underscores the point.Footnote 25 This is not to say that most people have meaningful, fully separable preferences on all or most policies, even despite the tendency for many to treat survey responses in different areas in this way.
Even where people have preferences in different policy areas, to which ones does the public want elected officials to pay attention? This is asking about priorities, the degree to which people want government to act on their preferences. To capture public priorities, scholars typically rely on percentages of people saying that different issues are the most important problems (MIPs) facing the nation. Although these provide limited information about the importance of policy areas (see Wlezien Reference Wlezien2005), they are revealing about what areas the public thinks most deserve attention, i.e., important problems. They do not make clear how much attention the public wants in the different policy domains, however.Footnote 26 The main problem is that responses do not directly map onto the policy agenda (Wlezien Reference Wlezien, Rohrschneider and Thomassen2020). For example, where 45% of the public say the economy is the MIP does not mean that the public wants policymakers to spend 45% of their policymaking time on the economy. Likewise, where zero people think transportation is the MIP does not mean that the public wants policymakers to spend literally no time on transportation. Given this, it is not clear what correlations between those measures and policy behavior, such as numbers of speeches, bill introductions, and hearings, tell us. Positive relationships point to representation of public priorities in policymaking but leave a lot to be desired from the point of view of assessing both responsiveness and congruence.Footnote 27
On information
While critical to responsiveness, people holding preferences is not enough. They also need to have information about government policy actions. This makes it possible to not only evaluate politicians based on their positions but guide them as well, particularly when instructing politicians to change policy. Consider that in order to send a meaningful signal about increasing or decreasing the government’s role in an area, i.e., to express clear relative preferences, people need to be able to compare what they want to the policy status quo.Footnote 28 This was the motivation for the development of the thermostatic model, which has spawned a now-large literature showing that people do appear to have some information, even if thin and quite general (see, e.g., Wlezien Reference Wlezien1995; Reference Wlezien2017a; Erikson et al. Reference Erikson, MacKuen and Stimson2002; Jennings Reference Jennings2009; Soroka and Wlezien Reference Soroka and Wlezien2010; Pacheco Reference Pacheco2013; van Hauwaert Reference van Hauwaert2023). There again is considerable variation across issues, much less so across individuals, including income, education, and party identification (Soroka and Wlezien Reference Soroka and Wlezien2010), though the last of these may matter some depending on party control of the White House (Branham Reference Branham2018).Footnote 29 That said, people have limited information about government policies even in the most salient domains, and there are many areas where but a small segment of the population knows what is going on (Mettler Reference Mettler2011).Footnote 30 While the public does not have much information about government policies, it does have some – far from perfect to be sure yet greater than nothing. This is important, as it provides a basis for effectively guiding elected officials and holding them accountable, at least in a general way on certain salient issues and/or more broadly across various domains (Wlezien Reference Wlezien1995; Erikson et al. Reference Erikson, MacKuen and Stimson2002; van Hauwaert et al. Reference van Hauwaertn.d.).
Preferences, information, and political judgments
To incentivize elected officials to pay attention to – and reflect – public opinion when making policy, there also needs to be some accountability. For this, people need to use both their preferences and information about government actions when making political judgments.Footnote 31 One expression involves approval of incumbents, where voters punish parties (and/or politicians) in government for pushing policy too far or not far enough. There is evidence of such effects in U.S. presidential and congressional elections, where governing parties pay a price for policy extremity (Wlezien Reference Wlezien2017b; Grossmann and Wlezien Reference Grossmann and Wlezien2024).Footnote 32 The tendency may hold in other countries, and help explain electoral patterns such as the cost-of-ruling, where vote shares for parties in government tend to decline the longer they are in power.
While elections may mostly be referendum judgments on sitting governments, they also involve choice between alternative parties and candidates, where expectations regarding the latter also may matter. We expect policy preferences to be most influential on vote decisions when politicians take different positions. Where there is not a real difference, preferences on particular issues still may influence voters’ evaluations of the different actors, but there is a limited basis for choice. This underscores the importance of information about both the actions of the sitting government and those competing for office. That there is evidence of voters taking into account the proximity of competing policy positions when choosing (e.g., Enelow and Hinich Reference Enelow and Hinich1984; Jessee Reference Jessee2012; Fowler Reference Fowler2020) is noteworthy. Although others argue that proximity matters less than directionality (especially Rabinowitz and Elaine MacDonald Reference Rabinowitz and Elaine MacDonald1989; MacDonald et al. Reference MacDonald, Listhaug and Rabinowitz1991), both may be operating (see Iversen Reference Iversen1994; also see Adams et al. Reference Adams, Samuel Merrill and Grofman2005). Prediction models that incorporate choice also outperform those based solely on referendum judgments (see, e.g., Erikson and Wlezien Reference Erikson and Wlezien2012).
The “vote” equation
Although a reliable connection between policy preferences, government actions, and political judgments may help to motivate politicians, it may not be necessary. The threat of such effects may be enough (see Caughey and Warshaw Reference Caughey and Warshaw2022), which may help explain representation where research shows limited accountability on Election Day, such for U.S. state legislative offices (Rogers Reference Rogers2023).Footnote 33 Of course, it may be that there is not much electoral threat and people just do not (always) want delegates when it comes to policy, particularly issues on which they do not have strong opinions, and so the representation of preferences does not impact their judgments of politicians (McCrone and Kuklinski Reference McCrone and Kuklinski1979; also see Carman Reference Carman2007). Electoral reward and punishment thus provide indications of the importance of different issues to the public, what often are referred to as “salience” weights. That said, the vote is a very coarse instrument of accountability, one that can encompass a broad range of considerations, which I will summarize shortly.Footnote 34 This is not to say that incentives are unimportant for policy responsiveness – without them, it depends entirely on the dispositions of elected officials – but to highlight the possibility that they may not be particularly strong.
The foregoing considers consequences of policy for elections in the form of referendum judgments, and also the choice on offer; where competitors for office advance different policies, people can choose between them based on their preferences and the weights they attach to those issues. Importantly, doing this does not necessarily produce majoritarian policies, as the distribution of salience weights across policy domains is critical, keeping in mind that these can only sum to 100% for each individual.Footnote 35 While many of us care about a variety of issues, and place small (or moderate) weights on each of them, some put all of their weight on one issue. These “single issue” voters often have extreme preferences, the intensity of support for which can lead to “frustrated majorities,” in the words of Seth Hill (Reference Hill2022). Even if people have clear policy preferences and vote on that basis, therefore, we do not expect to see majoritarian outputs, at least not always. This might help explain differential representation of groups in policymaking (also see Hutchings Reference Hutchings2003), and also the emergence of populist parties (and candidates), which appears to have roots in political misrepresentation itself (Sorace Reference Sorace2023).
While policy matters for voters, it is not all that matters. There are lots of things on the right-hand side of the vote equation, many of which may have little to do with policy. We already have considered certain “outcomes,” for which there is a lot of good evidence of effects in voting behavior research, particularly regarding the economy, something that is influenced by policy but not completely under government control. Other outcomes can matter (Green and Jennings Reference Green and Jennings2017), and this may be particularly true where they are salient and associated with policy but easier to observe (Buchanan Reference Buchanan2023).Footnote 36 The degree to which government decisions, implemented policy outputs, and outcomes are connected varies substantially across domains. Some, like immigration, may be mostly under the direct control of elected officials, but others, like housing, much less so (also see Hartland Reference Hartland2023).
Not all conditions matter to voters, however. For example, despite some suggestions to the contrary (Achen and Bartels Reference Achen and Bartels2016), shark attacks do not impact U.S. presidential elections (Fowler and Hall Reference Fowler and Hall2018), and it is not clear that the many other seemingly “irrelevant events,” like floods, droughts, and college football outcomes, do either.Footnote 37 Even where factors substantially outside the control of presidents, like economic growth, impact those elections, it is not that events at the very end of the cycle are decisive (see Wlezien Reference Wlezien2015). There is a lot more to the story of policy and outcomes, of course, which I further explore in Appendix A.
The list of factors that influence party and candidate evaluations includes characteristics of candidates (and/or party leaders) and voters’ social group memberships and identification (see, e.g., Converse Reference Converse and Apter1964; Miller et al. Reference Miller, Wlezien and Hildreth1991; Greene Reference Greene2004; Lewis-Beck et al. Reference Lewis-Beck, Jacoby, Norpoth and Weissberg2008; Kane et al. Reference Kane, Mason and Wronski2021; but also see Clifford et al. Reference Clifford, Simas and Suh2025). Although the latter have featured in the earliest research on voting, their relevance in the literature has ebbed and flowed over time, and surged of late, particularly with the growing salience of attitudes toward political parties (especially see Huber Reference Huber2022). Recent history highlights that democratic principles themselves may factor into voters’ decisions, even if only to a limited degree (Jacob Reference Jacob2025).Footnote 38
Of course, some of these variables may be associated with policy, such as group memberships and identification, and partisan affiliation itself, which is not randomly assigned, after all (also see Huddy Reference Huddy2018). And affective party polarization may owe more to policy considerations than anything else (Orr et al. Reference Orr, Fowler and Huber2023). Objective outcomes that matter also may be associated with government policy decisions and outputs, though to a variable degree, and here perceptions may be decisive. Even attitudes toward “democracy” are not unrelated to policy winning and losing (Anderson et al. Reference Anderson, Blais, Bowler, Donovan and Listhaug2005). But these and other non-policy variables are not entirely about the policy actions political actors support and undertake.
The fundamental point of the foregoing discussion is that people’s political judgments are not solely – or even primarily – determined by policy considerations. These non-policy aspects of political judgments diminish the weight of policy in those evaluations, which has implications for elected officials, at least those motivated by reelection: the less policy drives voters, the less the public’s policy preferences should drive these politicians. Moreover, the complexity of the equation increases the demands on policymakers to both notice what matters and to what degree. Given these demands, it is hard to picture politicians not relying on heuristics to take action. This is even true when focusing solely on the role of “policy,” as there are multiple domains, the public salience weights on which need to be estimated, relative to each other (and the many non-policy factors), and the public’s policy preferences do too. Getting all of this right is not easy for analysts armed with complete datasets, statistical prowess, and time; it may be harder still for politicians, particularly when the salience of issues is changing, e.g., immigration. Then again, they seem to be quite good, on average, at evaluating political conditions and navigating them, even despite research showing that they often misperceive public priorities and preferences expressed in surveys. Reelection rates are impressive, after all.
The bottom line to the research on political judgments is that there is public demand for policy responsiveness but it is limited, focused on certain highly salient issues but otherwise very general, the latter of which seems much in line with Stimson’s (Reference Stimson1991) view of public opinion. There thus is a correspondingly limited incentive for politicians to attempt to respond even where the public has policy preferences and information about policy actions, though it still may be enough to motivate politicians to some degree. The actual supply of policy is further limited.
The “output” side
I have outlined the basics of the input side to effective policy responsiveness, which constitute critical, necessary conditions, seriously limiting of possibilities for effective representation. This parallels some of the claims of other scholars, including Achen and Bartels (Reference Achen and Bartels2016), though a review of the full literature indicates that there is at least some effective demand – the public has preferences and information about policy, and these factor into political judgments, including election outcomes themselves. This is of fundamental importance for policy responsiveness, as demand is a necessary precursor, and may help make more understandable what we do observe. Even where that demand exists, however, it is not sufficient for policy supply, and we now want to consider under what conditions elected officials follow our lead. One possibility is that those in government are like us, possibly because of advantageous selection; if so, there is reason to suppose they also behave like us as conditions change. If this is not true, however, and there is good reason to think that they are not a representative sample (see, e.g., Butler Reference Butler2014), we need them to respond to us.
On perceptions
For there to be such responsiveness, the most basic requirement is that elected officials must notice public opinion – both priorities and preferences. This is well-known in the literature on political representation, dating back to the earliest formulations and tests of the dyadic relationship between representatives of legislative districts and their constituents (Miller and Stokes Reference Miller and Stokes1963; Hedlund and Friesema Reference Hedlund and Paul Friesema1972). Although there is reason to suppose that representatives do pay attention, often actively, research has produced mixed evidence – findings of accurate perceptions on some issues (under some circumstances) and inaccuracies elsewhere (also see Erikson and Tedin Reference Erikson and Tedin2019). More recent research highlights errors in estimating percentages – and majorities – of the public in support of different policies (Broockman and Skovron Reference Broockman and Skovron2018; Hertel-Fernandez et al. Reference Hertel-Fernandez, Mildenberger and Stokes2019; Pereira Reference Pereira2021; Walgrave et al. Reference Walgrave, Jansen, Sevanans, Soontjens, Pilet, Brack, Varone, Helfer, Vliegenthart, van der Meer, Breunig, Bailer, Sheffer and Loewen2023a). This may come as little surprise given that many politicians may not rely heavily on polling data to assess public opinion, and those who do may lack detailed information. Various other issues complicate measurement and assessment here, including survey error (see Zelizer Reference Zelizer2025), and whether politicians get opinions right or wrong may tell us little about perceptions of the public’s preferred levels of policy, as we have seen. (Given the importance and evidence of dynamic responsiveness, what may be most crucial is that politicians notice preference change). Regardless, for misperceptions to preclude effective policy responsiveness depends on how politicians’ estimates of opinion aggregate up into party and government views, which presumably are what matter for policymaking.
Any perceptual disjunctures may be especially concerning where people have meaningful policy preferences, but how they impact policy responsiveness is not at all clear. Misperceptions evidently are not a major problem where policymakers do appear to respond to public preferences, but they may help explain the variation across issues that we have discussed. Some scholars have noted that politicians’ errors may reflect the salience of issues to voters, where we observe a closer match on issues about which voters care (Walgrave et al. Reference Walgrave, Jansen, Sevanans, Soontjens, Pilet, Brack, Varone, Helfer, Vliegenthart, van der Meer, Breunig, Bailer, Sheffer and Loewen2023a; also see Soroka and Wlezien Reference Soroka and Wlezien2010; Lax and Phillips Reference Lax and Phillips2012). Such a pattern could emerge simply because of selection, where voters elect like-minded people whose actions reflect the preferences (and salience weights) of the average voter.Footnote 39 Salience is not the only factor that may structure politicians’ (mis)perceptions, of course, and the others may be more problematic, as they can compromise effective responsiveness when the public really wants it (see Jones and Baumgartner Reference Jones and Baumgartner2005; Baumgartner and Jones Reference Baumgartner and Jones2015). Political institutions themselves may matter (see Pereira Reference Pereira2021). All of this said, any consequences of the apparent misperceptions for policymaking remain to be seen.
On motivation
Even armed with accurate information about what the public wants, politicians still may not follow preferences, as they need to be motivated to represent. That is, they need to care – intrinsically or extrinsically – about representing us. Part of the representational story may be selection, though keep in mind research demonstrating that the officials we choose are different from us – older, more educated, more affluent, more male, and possibly more conservative than the average person/voter (Butler Reference Butler2014; Carnes and Lupu Reference Carnes and Lupu2023; Gerring et al. Reference Gerring, Jerzak and Öncel2024). (For more on “who gets represented,” which has been the subject of a very large and rapidly expanding corpus of research, see Appendix B). To the extent elected officials are not like us, incentives are of special importance, particularly effects of policy actions on voters’ political judgments, at least for those motivated by reelection. This encourages representatives to take actions that align with pivotal voters, though other political participation also may be impactful, both to correct perceptual errors and provide motivation.Footnote 40
That policy appears to influence election outcomes thus is of real importance.Footnote 41 To the extent that there are incentives and these motivate governments, they are in large part future-oriented, about what voters will do in an election to come, per Mansbridge (Reference Mansbridge2003) anticipatory representation. Erikson et al. (Reference Erikson, MacKuen and Stimson2002) refer to this as “rational” anticipation in recognition of the fact that responsive officials need to forecast what voters will reward and punish at the polls when making policy (or doing other things). Predicting the future is never easy, but may be especially difficult for governing politicians given the complexity of the vote equation and the very general nature of the public’s preferences in many domains. That said, we know a lot about what happens on Election Day, and approval ratings are revealing in advance (Erikson and Wlezien 2012; Lewis-Beck and Stegmaier Reference Lewis-Beck and Stegmaier2013).
The “policy” equation
Just as political judgments, including votes, do not only reflect policy considerations, policy actions do not only respond to the public. This is well known, and Kingdon (Reference Kingdon1973) provided a useful basic model early on, in which public salience plays a useful conditioning role. Where the public cares, elected officials tend to care more about public opinion; where it does not, there is (more) space for other actors.Footnote 42 Chief among these are political parties and interest groups, both of which have featured a lot of scholarship on policy in the U.S. and elsewhere. Political party control is known to matter (see, e.g., Wlezien Reference Wlezien1996; Erikson et al. Reference Erikson, MacKuen and Stimson2002; Wlezien Reference Wlezien2004; Soroka and Wlezien Reference Soroka and Wlezien2010; Bevan and Greene Reference Bevan and Greene2018), leading governments to push policy off to the “left” or “right,” including on salient issues, the effects of which depend in part on political institutions and polarization. Such differences are exactly what we would expect based on views of parties as competing coalitions of interests, following Aldrich (Reference Aldrich1995), and primary elections in the U.S. might be reinforcing here (Cohen et al. Reference Cohen, Karol, Noel and Zaller2008), perhaps especially with the increase in safe seats and gerrymandering. Interest groups also are important players unto themselves, especially on low salience issues (see Culpepper Reference Culpepper2012).Footnote 43 Of special concern, at least since Olson (Reference Olson1965), is the influence of small, narrowly-focused groups that may lead to policies that reflect their special interests (also see Schattschneider Reference Schattschneider1960). Part of this relates to the role of money in politics, particularly in the U.S., the influence of which appears to be complex (see, e.g., Hall and Wayman Reference Hall and Wayman1990; Ansolabehere et al. Reference Ansolabehere, de Figueiredo and Snyder2003; Hertel-Fernandez Reference Hertel-Fernandez2019).
Since Truman (Reference Truman1951) articulated the “pluralist” model of politics, interest groups have been seen as critical actors in the representational process. To simplify his view, the center of mass among competing groups drives government decision-making, and this helps produce policies that approximate the preferences of the average person. Olson (Reference Olson1965) directly challenged that view, highlighting that it is easier for smaller groups to form, leading to correspondingly biased policy actions, though he also made clear that there are ways for larger, more “public” groups to organize, and they have. Perhaps not surprisingly, some recent research sees interest groups as often reflecting public opinion and aiding its representation, much as Truman had theorized (especially see Rasmussen and Willems Reference Rasmussen, Willems, Harris, Bitonti, Fleisher and Skorkjaer Binderkrantz2021). Other research on policy decision-making finds either minimal interest group participation and effects (Burstein Reference Burstein2014) or limited, status quo enforcing ones (Baumgartner et al. Reference Baumgartner, Berry, Hojnacki, Kimball and Leech2009). In the “quiet” politics associated with low salience domains, which are quite common, there is reason to suppose that special interests can have out-sized influence (Culpepper Reference Culpepper2012; Reference Culpepper2021). Of course, it may be that decisionmakers also rely substantially on their own “more informed” views (Lucas et al. Reference Lucas2025), and this might help explain the alignment(s) observed in studies of unequal representation (also see Appendix B).
On implementation
Most research on policy representation has focused on government decisions, typically laws but other actions as well, including executive directives and court rulings (see Jessee et al. Reference Jessee, Malhotra and Sen2024), not on the implemented outputs that follow (see Buchanan et al. Reference Buchanan, Dias, Wlezien and Rudolph2022). Thus, even when the necessary conditions obtain and we observe that policy decisions reflect public preferences, this does not mean that they guide what governments actually do. The connection between decisions and outputs is closer in some areas than others, but laws often are fairly general, requiring executive officials to take decisions, which may make government outputs less representative of public preferences. Indeed, this is precisely where much of the research on interest groups is focused, even in high salience domains, as what happens at the implementation phase often is quiet. Some research on appropriations decisions and the spending that results documents a surprising disjuncture between the two, which has consequences for assessments of policy responsiveness (Wlezien and Soroka Reference Wlezien and Soroka2003; also see Angell Reference Angell2025).
In his APSA presidential address that argued for the importance of public administration to theories of politics and democracy, Matthew Holden (Reference Holden2000) highlighted the “public” aspect of that side of government. This is promising and there also are signs of growing scholarly interest in what governments do (especially see Bertelli Reference Bertelli2021), but the role of the public is mostly to be determined. Many may wonder how much guidance the public can and does provide to those charged with implementing policy decisions.Footnote 44 That point in the policymaking process also may be more susceptible to less representative forces, including special interests, as has been well documented for decades (see, e.g., Hamm Reference Hamm1984). The public still may have a role to play there, if only because of the possibility that they will hold officials accountable for undesirable outcomes (also see Miller Reference Miller2013).
On capacity
Importantly, politicians also need to be able to represent. Just because they accurately assess public opinion and want to represent does not mean that they can. The budget constraint is a classic limit to government action, as elected officials can only do so much, i.e., they have to pick and choose (Soroka and Wlezien Reference Soroka and Wlezien2004; Reference Soroka and Wlezien2005; Wlezien and Soroka Reference Wlezien and Soroka2012; Ezrow et al. Reference Ezrow, Hellwig and Fenzl2020).Footnote 45 But agendas are limited too, if less strictly (Baumgartner and Jones Reference Baumgartner and Jones1993; Jones and Baumgartner Reference Jones and Baumgartner2005); governments cannot do everything, even where those in charge have the desire to act, which clearly complicates policy responsiveness. Political institutions matter as well, both electoral and governing ones, as we will see.
Beyond the basic model
The foregoing provides a basic characterization of the necessary conditions for effective policy responsiveness relating specifically to demand and supply, and the degree to which they are met. Meeting all of the conditions we have set out for the two “sides” of the does not ensure effective representation, i.e., while necessary, they are not jointly sufficient, as there is more to the story relating to policy inputs and outputs. Most importantly, there is variation in context, particularly that associated with political institutions. Their impact relates directly to both the input and output sides, so I could have reflected them in the foregoing discussion, but chose to separate them to make clear what a context-free “model” looks like and to provide institutions the full attention they deserve.
Institutions and variation in public perceptions of policy
Institutions impact policy effects on the public’s policy preferences and political judgments in various ways. Federalism is one feature that has received increasing attention from researchers of political behavior and public opinion. Traditionally, scholars highlighted the benefits of subnational government(s) for citizen engagement in more local issues and the ease of understanding and monitoring government decisions (e.g., Mill 1861; de Toqueville Reference de Toqueville1835; also see Porter Reference Porter1977; Winthrop Reference Winthrop1976), which can aid the representation of regional differences in public opinion (Acton Reference Acton and Rufus Fears1985; Elazar Reference Elazar1987). More recent work points to complications federalism introduces for accountability, particularly when governments at different levels, e.g., the nation and the state, are involved in policymaking. There, federalism lowers the “clarity of responsibility” (Hobolt et al. Reference Hobolt, Tilley and Banducci2013), potentially confusing the attribution of responsibility for outputs (and outcomes) (Cutler Reference Cutler2010; Wlezien and Soroka Reference Wlezien, Soroka, Enns and Wlezien2011), even as citizens are able to partly navigate their way (e.g., Anderson Reference Anderson2006; Arceneaux Reference Arceneaux2005; Niemi et al. Reference Niemi, Stanley and Vogel1995). That is, citizens may have a harder time determining which governments are responsible for policy actions, and to what degree. This is an obvious problem for voters, as they may incorrectly assign credit and blame, reducing accountability for actions and consequences at each level of government.
This vertical “mixing” of governments in policy areas can compromise public preference signals themselves, e.g., people adjust preferences thermostatically for more government action at one level of government based on actions taken at another (see Wlezien and Soroka Reference Wlezien and Soroka2012a). Federalism thus may have the consequence of distorting policy responsiveness to (true) public preferences even where officials are motivated to represent us and respond to our expressed preferences. Other institutions may weaken the public’s connection with policymakers, of course, but federalism may be of special importance.Footnote 46 Representation at different levels within federal systems also matters, and research there is on the rise (e.g., Berkman and Plutzer Reference Berkman and Plutzer2005; Caughey and Warshaw Reference Caughey and Warshaw2022; Rogers Reference Rogers2023; Tausanovitch and Washaw Reference Tausanovitch and Warshaw2014).
Institutions and variation in government responsiveness
Political institutions influence the government response to public preferences even on salient issues where politicians are aware of preferences, want to represent them, and are not constrained. It is well known that political systems vary across space, especially countries. The primary ways that they differ relate to electoral and governing institutions, the latter of which we already considered to some degree in the foregoing discussion of federalism.Footnote 47 Whether and how other institutional differences matter for the policy behavior of elected officials has been the subject of much more research, particularly as regards electoral institutions.
Electoral institutions
Beginning with Lijphart (Reference Lijphart1984), scholars have concentrated on the representational consequences of electoral systems. The primary focus here is on differences between plurality and proportional elections, where coalitions that often result in the latter are seen as better representing the median voter than single-party governments in the former. To Powell (Reference Powell2000), the proportional “vision” tends to produce more centrist governments because the coalitions effectively average across governing partners. The theoretical expectations are strong, and Powell also found supporting evidence. Others have not, however, most notably Blais and Bodet (2006), who observed that proportional representation (PR) leads (seemingly causally) to more parties that are less centrist to begin with, and so the resulting coalition governments are not more representative than those in majoritarian systems. More recent research explicitly theorizes and analyses the influence of party polarization, which is positively influenced by PR but also reflects other factors (Powell Reference Powell, Dalton and Anderson2011). This may help explain the differences in the results of different studies focused on different time periods with different levels of party polarization.
This research focuses on positions, not policy, and the implications apply mostly to the immediate aftermath of an election. For example, following Powell, the coalition governments that form after elections are (more) representative of the electoral center on Election Day. But, what do we expect as policy preferences change in between elections? Consider a similar “shock” to preferences in proportional and majoritarian systems. How do governments in these two systems respond? Our expectations might be quite different to what we would predict based on Powell’s research, at least for policy. Wlezien and Soroka (Reference Wlezien and Soroka2012) argue that coalition governments should be less responsive, increasingly so as the size of the coalition increases, and analysis of budgetary policy is confirming (also see Soroka and Wlezien Reference Soroka and Wlezien2015). There are various reasons to expect this difference, including constraints on change in between elections that coalition agreements impose, even as they provide a basis for action after elections (Bergman et al. Reference Bergman, Angelova, Back and Muller2024; Kluver et al. Reference Kluver, Back and Krauss2023; Muller and Strom 2008; Peterson and de Ridder Reference Peterson and de Ridder1986).Footnote 48 Less obvious but perhaps just as important is that parties within the governing coalition may react differently to the opinion shock (Calvo et al. Reference Calvo, Hellwig, Chang, Schofield, Caballero and Kselman2013).
The main point is that electoral institutions, seemingly together with party systems, may condition government responsiveness to the public’s policy preferences in complex ways. The structures that may enhance representation after elections may dampen it in between. Even where the input conditions and other output conditions are met, therefore, electoral institutions may impede the representation of public opinion, regardless of the systems we inhabit.
Government institutions
Government institutions also may impact responsiveness. Here, there is considerably less research, though it is well known that the division of policymaking power can matter. Some aspects are straightforward, such as the independent judiciary, especially the Supreme Court in the U.S., which can review and declare unconstitutional actions taken by the president and Congress. This can lead to countermajoritarian decisions, even if less often than news coverage might lead us to think (Artabe and Badas Reference Artabe and Badas2023), as can other institutional features in the U.S., and elsewhere.Footnote 49
Division between legislative (and executive) powers can impact responsiveness as well. Of special significance may be the horizontal separation between executive and legislative branches in presidential systems, especially Madisonian ones, which provide balance. This sharing of powers obviously makes it harder to undertake policy, even where both institutions want to do something, as it necessitates coordination across branches.Footnote 50 The same is not true in parliamentary systems, at least not to the same degree, as taking action is easier, other things being equal, e.g., electoral institutions. But note that ease of action does not imply greater responsiveness, as it also allows governments to take action that is not supported by the public (see Fagan et al. Reference Fagan, Jones and Wlezien2017).
In this context, it might be useful to think of parliamentary systems as affording executive discretion, which has been shown to reduce responsiveness. There is indirect evidence of such a tendency from research on policy change in the agendas subfield, both across states within the U.S. owing to gubernatorial power (Breunig and Koski Reference Breunig and Koski2009) and across countries associated with authoritarianism (Baumgartner et al. Reference Baumgartner, Carammia, Epp, Noble, Rey and Murat Yildirim2017). That research finds that the concentration of executive power at both levels is associated with more leptokurtic distributions of budgetary policy change, i.e., those where there are both more incremental innovations and larger ones than expected based on a normal distribution. This is not what we would observe were there high levels of responsiveness, at least assuming that the distribution of public demands is itself normal, for which there is strong theory and some empirical support (see Jones and Baumgartner Reference Jones and Baumgartner2005; Jones et al. Reference Jones, Larsen-Price and Wilkerson2009). These results are indirect, and only suggestive, so it is important that there is more direct evidence from work assessing the relationship to changing preferences, where horizontally-balanced systems appear to be more responsive (Wlezien and Soroka Reference Wlezien and Soroka2012). In theory, this is driven by “error correction,” where separate institutional actors can effectively check and balance, producing anticipation and adjustment. Early research on public opinion and spending for defense in the U.S. is illustrative, as it finds that the president’s proposals follow the public over time and that the Congress alters those requests to bring the final budget more “in line with” preferences (Wlezien Reference Wlezien1996). Also keep in mind that even divided government, where different parties control the executive and legislative branches, can produce policies the public supports (see Mayhew Reference Mayhew2005; Grossmann and Wlezien Reference Grossmann and Wlezien2024). That said, the tendency to do so may depend – and seemingly does – on the degree of polarization and effective balance, the existence of which is not hard-wired (Hughes and Carlson Reference Hughes and Carlson2015).Footnote 51
The findings relating to government institutions contradict with those for electoral institutions, as they imply that institutional “friction” is beneficial for responsiveness, not only detrimental. Recent research that explores the differences confirms both effects, pointing to the possibility that the friction associated with coalition governments has a very different effect to that associated with executive-legislative balance (Maedgen and Wlezien Reference Maedgen and Wlezien2024).Footnote 52 Put differently, while veto players all represent barriers to policy change (Tsebelis Reference Tsebelis2002), their influences on the association between the policy actions we observe and public opinion appear to differ. The particular mechanisms at work here remain to be seen, but it seems important research to do, promising to shed light on the conditions under which the status quo bias that veto points engender hurts or helps responsiveness.Footnote 53
The media
While explicitly political institutions are of fundamental importance to the functioning of representative democracies, the “mass media” are as well. Their role in providing information to the public has long been recognized (Dunaway and Graber Reference Dunaway and Graber2022), and is of (obvious) importance for effective accountability, which requires that people know what government is doing. That information also helps produce meaningful public preferences signals, the raw material of much policy responsiveness, particularly support for policy change.Footnote 54 These expectations reflect a long scholarly tradition in the social sciences (Deutsch Reference Deutsch1966; Easton Reference Easton1965; see Soroka and Wlezien Reference Soroka and Wlezien2022). That press freedom appears to matter for public responsiveness to policy actions thus should come as little surprise, as it impacts the information that people have (Hiaeschutter et al. 2021).Footnote 55 This also has consequences for policy responsiveness through the public preference signals that influence what governments do (Soroka and Wlezien Reference Soroka and Wlezien2010).Footnote 56
The media may impact what policymakers do more directly. News coverage may influence the information politicians have about public preferences (Walgrave et al. Reference Walgrave and Soontjens2023b), which can impact policy responsiveness, though this remains to be seen. Besides informing them, the media also may encourage responsiveness by its sheer presence as a potential monitoring agent, introducing possible risks to shirking behavior (Auslen Reference Auslen2024). Media presence may attract different kinds of politicians, ones who are more inclined to pay attention to public opinion, i.e., self-selection. Then again, media coverage may reflect the influence of other actors, and this may impact what policymakers do independently of public opinion (John Reference John2006). There is much work to be done in this space.
On institutions and representation
The preceding discussion highlights what are some institutional influences on policy responsiveness, mostly direct. While there are some additional limits on the public demand for policy, there are greater obstacles to policy supply even where public preferences exist and are transmitted to policymakers, and they have incentives to respond, are interested in doing so, and have the policy space. (Political polarization also is relevant, as we have already seen in our discussion, and I consider further in Appendix C).
Some concluding thoughts
Effective policy responsiveness demands a lot of the public, their representatives, and political institutions themselves. Given the requirements, that we observe as much as we do maybe should surprise, not the fact that the public often does not get the policies it wants. There is substantive policy representation, after all, and there also appears to be responsiveness, even as assessment of the exact congruence between public preferences and policy change remains elusive. While the practice of representative democracy may be a long way from the ideal some imagine, it often works better than the caricature offered by some scholars (e.g., Achen and Bartels Reference Achen and Bartels2016). Policy responsiveness is not the norm to be sure, primarily evident in highly salient domains, mostly at a general level. And, it may have declined over time (Monroe Reference Monroe1979; Reference Monroe1998; Jacobs and Shapiro Reference Jacobs and Shapiro2000; also see Barabas Reference Barabas2016), perhaps owing to some of the issues addressed in this essay, especially those relating to the output side.
Part of the difficulty for politicians (and scholars) is accurately mapping public inputs – what people care about and want. How much is about policy? What policy areas do people care about? What policies do they want? With what trade-offs? How much is about outcomes? Other things? We have surprisingly little information about this complex, multidimensional and politically important space, though people do appear to have preferences in some policy areas and some general idea of what politicians are doing and offering. There also appear to be electoral effects of policy behavior, which provide incentives to elected officials (and their competitors), even as they are very coarse mechanisms given how much gets packed into the vote, at least some of which is not policy related.
Then there is the difficulty of delivering outputs – government decisions and actions. Do politicians know what the public wants? Do they have effective control over the policies people prefer? What about outcomes? There is variation in the communication of inputs and the responsiveness of outputs, partly reflecting political institutions. Even where they are motivated to represent public preferences, it is just not easy for politicians to determine what exactly voters want, get it all passed, and then implement it. We thus should expect only some policy responsiveness, maybe not that much, and not because the public does not have preferences, care, and/or is completely inattentive to and informed about politics and policymaking. The output “side” matters, seemingly a lot more than the balance of research attention implies, which has favored public inputs.Footnote 57 This might help account for apparent representational disjunctures on certain noneconomic issues, e.g., immigration, which may have aided the rise of populist parties in various countries.
Most of the previous research focuses, sometimes only implicitly, on the representation of the median voter, but that may not be policymakers’ focus. This may not be a (major) concern where preferences across groups (and individuals) do not differ, as policy would end up in the same place regardless of who is decisive (Soroka and Wlezien Reference Soroka and Wlezien2008; Enns and Wlezien Reference Enns and Wlezien2011; Enns Reference Enns2015; Branham et al. Reference Branham, Soroka and Wlezien2017; also see Elkjær and Klitgaard Reference Elkjær and Klitgaard2024). Where preferences differ, however, officials have to choose, creating winners and losers. This highlights how inequality is inherent, really unavoidable, in representative democracies, as we all cannot get the policies we want, and winning and losing matter more the greater the differences in opinion, and electoral choice (also see Anderson et al. Reference Anderson, Blais, Bowler, Donovan and Listhaug2005). As one might expect, there appears to be some inequality in policy representation, just not as much as some surmise, and perhaps even less unequal responsiveness (also see Appendix B). Despite an emerging consensus about the facts of unequal representation in this unusually active research area, the sources of those patterns remain to be seen.
There is research to do on basic aspects of both sides of the representational equation, which I have tried to highlight in the essay.
The input side has been an active research area, and additional work is underway there, particularly on the measurement of the public’s preferred policies, allowing a more direct match with policy outputs themselves. There is more to do on that front but also on public priorities, the salience weights of different domains, which may be even more difficult to register than preferences (see Buchanan et al. Reference Buchanan, Dias, Wlezien and Rudolph2022). The information the public has about outputs and outcomes has received some scholarly attention in the U.S. and other countries, but the specifics are not clear, and additional research would benefit our understanding. More importantly, there is only a small amount of research on the electoral connection, mostly isolated to the U.S. and cast in a very general way, i.e., broad aggregates of policy (and opinion).
The output side has received less attention, but research is on the rise, particularly regarding whether policymakers notice what the public wants. There is more to do there, especially given developments in the measurement of public preferences, though even more on the impact of perceptions on policy responsiveness, about which we know little. The conditioning role of capacity constraints has received sporadic scholarly attention, and research on implemented government outputs is spottier still, perhaps not surprisingly given that it is far down the representational chain and difficult to study systematically (but see Bertelli Reference Bertelli2021). The latter remains an especially important limitation to research on policy representation and responsiveness. Given that the public’s preferences for policy often are means to the ends of real-world outcomes, understanding political “control” of the latter seems a worthy subject of scholarly research.
There is more to representation than policy adaptation and its effects, of course, some of which is policy-related (Esaiasson et al. Reference Esaiasson, Gilljam, Persson, Esaiasson and Narud2013). One involves “listening” to the public’s policy opinions. Do government officials allow space for public input? Recognize their priorities and preferences? Another involves “explaining” decisions. Do officials acknowledge that they are not doing what the public seems to want? Do they offer reasons for taking a different course? The public may demand and expect such responsiveness, and some scholars have started in on testing (see Esaiasson and Wlezien Reference Esaiasson and Wlezien2017). It is not the same as giving the public policies it wants, but it may be better than nothing for those who did not get the policies they prefer, and it is something that policymakers can do, keeping in mind the difficulty of controlling decisions and implementing them. To the extent that these non-adaptive actions satisfy voters, they may limit policy responsiveness, though literature strongly suggests that voters often want more. It just is not always easy to clearly register those demands and then have everything else that is required for policy responsiveness across various domains fall into place.
I see the foregoing as providing the basis for a more realistic “model” of representative democracy, one informed by the range of research on public opinion, voting, political institutions, policy, and political theory itself. From this perspective, the functioning of system is far from what some of us consider ideal, yet is much as we should expect given the demands it places on the public and policymakers. The theoretical conditions for effective responsiveness are clear, as are some of the empirical foundations, but much research remains. I hope this essay provides some useful guideposts.
Acknowledgments
Previous versions of this essay were presented at Reichman University in October, 2022, Tel Aviv University in December, 2022, and Ben Gurion University of the Negev, June, 2025 It was supported by a grant from the Institute for Humane Studies, and weaves together and extends themes developed in my previous research on the subject, much of which is with Stuart Soroka, to whom I am most grateful for our ongoing collaboration and discussions too. The essay also has benefited tremendously from the input of many others, including Wouter van Atteveldt, Luca Bernardi, Ian Budge, Peter Enns, Peter Esaiasson, Lawrence Ezrow, Gauthier Fally, Steve Fisher, Anthony Fowler, John Gerring, Liran Harsgor, Bryan Jones, Yael Kaplan, Young Seok Kim, Alexander Kustov, Liron Lavi, Peter Loewen, Jack Lucas, Jane Mansbridge, Naama Rivlin-Angert, Robert Rohrschneider, Shaul Shenhav, Zeynep Somer-Topcu, Miriam Sorace, Logan Strother, Israel Waismel-Manor, Kurt Weyland, and Fabio Wolkenstein, but especially Amnon Cavari, Mads Elkjaer, Katie Keil, Jack Maedgen, Jennifer Oser, Michal Shamir, Ranaan Sulitzeanu-Kenan, Stefaan Walgrave, the editor at JPP, Josh Ryan, and the four reviewers, all anonymous throughout the editorial process. I’m sure I have missed some folks; if so, I truly am sorry, and please let me know!
Supplementary material
To view supplementary material for this article, please visit https://doi.org/10.1017/S0143814X26101019.
Data availability statement
This study does not employ statistical methods and no replication materials are available.
Open access
