The Expertise Paradox: How Policy Expertise Can Hinder Responsiveness

Abstract

We argue that policy expertise constrains the ability of politicians to act on voter preferences. Representatives with more knowledge and experience in a given domain have more confidence in their own issue-specific positions. Enhanced confidence, in turn, may lead politicians to discount opinions they disagree with, producing a distorted image of the electorate. Two experiments with Swedish politicians support this argument. First, officials are more likely to dismiss appeals from voters in their areas of expertise and less likely to accept that opposing views may represent the majority opinion. Consistent with the proposed mechanism, in a second experiment we show that inducing perceptions of expertise increases self-confidence. The results suggest that representatives with more specialized knowledge in a given area may be less capable of acting as delegates in that domain. The study provides a novel explanation for variations in policy responsiveness.

Theories of political representation posit that re-election-seeking officials have incentives to be responsive to voters (for example, Dahl 1973, Pitkin 1967). Consistent with this view, prior work shows that politicians update their positions based on public opinion polls (Butler and Nickerson 2011; Pereira 2020) and election results (Adams, Merrill, and Grofman 2005; Somer-Topcu 2009). However, there is ample variation in patterns of responsiveness.¹ Lax and Phillips (2012) show that only half the policies adopted in US state legislatures are aligned with majority preferences. In Europe, there is growing evidence of inequalities in representation across different subconstituencies (Giger, Rosset, and Bernauer 2012; Homola 2019; Pereira 2021).

Variations in responsiveness may partially result from different institutional incentives (Hobolt and Klemmensen 2008; Soroka and Wlezien 2010), contextual dynamics (Ezrow and Hellwig 2014; Ezrow, Hellwig, and Fenzl 2020; Giger and Klüver 2016), or representational roles (Dynes, Hassell, and Miles 2018; Harden 2013). But responsiveness also varies according to the policy issue. For instance, Miller and Stokes (1963) demonstrated that US public opinion played a central role in shaping civil rights policies but only a negligible role in foreign affairs. Why do politicians dismiss voters' opinions on certain policy issues, despite incentives to be responsive?

This puzzle may partly be explained by the process through which politicians build their image of the electorate. In this article, we argue that policy expertise may constrain representatives' ability to be responsive. During their time in office – or from prior professional experience – elected officials often develop expertise in specific policy areas. Parliamentary committees encourage this form of specialization (Strøm 1998), and policy expertise is required for efficient policy making. However, specialized knowledge can also influence how politicians respond to new information and make decisions.

Individuals with more expertise in a given area tend to have more confidence in their own beliefs in that domain (Dunning 2005; Fisher and Keil 2016; Tetlock 2005) and to be less open to different views (Ottati et al. 2015). Self-confidence and dogmatism may lead politicians with expertise in a specific domain to disregard opposing views on that issue. Yet policy responsiveness requires accurate beliefs about the electorate (Broockman and Skovron 2018; Butler and Dynes 2016). Therefore, legislators may paradoxically be less capable of channeling constituent preferences or acting as delegates in areas where they are more knowledgeable. This argument is referred to as the expertise paradox.

We provide empirical support for the expertise paradox in a pair of survey experiments with Swedish elected officials. The experiments were included in two waves of the Panel of Politicians, a large biannual survey of representatives from all levels of government in Sweden. In the first experiment (N = 1,669), we asked politicians to evaluate a hypothetical policy appeal from a group of voters. The voters' position on the issue was always at odds with the elicited preferences of the official to hold constant the propensity of officials to discount opinions they disagree with (Butler and Dynes 2016). Additionally, we experimentally manipulated whether the initiative was within a policy area where the politician had a higher or lower level of expertise. The results confirm that officials are more likely to disregard contrasting views in areas where they are more knowledgeable. Officials in the high-expertise condition were less likely to accept that the group of voters: (1) understood the complexity of the issue; (2) based their opinion on facts, and; (3) that their position represented the majority opinion. The findings are not explained by whether the policy appeal was aligned with the majority opinion, by heterogeneity across issue areas, or differences in issue salience. Subgroup analyses suggest that the main effects are more consistently driven by officials with college degrees – a formal source of expertise – rather than ‘passive expertise’ accumulated over time in office.

In the second study, we directly assess the causal mechanism implicit in the expertise paradox. Consistent with our argument, we find that politicians randomly induced to perceive themselves as experts were less likely to express doubts about their decisions in office. On the other hand, heightened perceptions of expertise were not meaningfully associated with dogmatism.

We did not directly test whether politicians are more or less responsive in their areas of expertise. Instead, the experiments focused on the way politicians develop perceptions of voter preferences and how expertise shapes this process. This intermediary step influences the ability of legislators to act on voters' preferences (Broockman and Skovron 2018; Pereira 2021).² If policy expertise changes how politicians interact with voters and develop perceptions of public opinion, it alters the conditions for policy responsiveness. Our empirical strategy builds upon previous efforts to study the underlying processes of mass-elite policy responsiveness (Butler and Dynes 2016; Harden 2013; Naurin and Öhberg 2018).

The findings have important implications for research on political representation and legislative politics. The expertise paradox provides a novel explanation for variation in policy responsiveness. If elected officials with specialized knowledge play a central role in drafting new legislation (Makse 2021), the ability of citizens to control public policy is constrained whenever policy experts disagree with the majority. The study reveals a trade-off between expertise and the representational roles adopted by legislators. Although voters consistently prefer representatives who follow public preferences (Carman 2007; Converse and Pierce 1979; Dassonneville et al. 2021), the expertise required for policy making limits politicians' ability to act as delegates. At the same time, expertise can shield politicians from caving in to public opinion. However, the expertise paradox can also create obstacles to policy implementation. Dismissing public preferences can lead to lower compliance or resistance if policy solutions fail to take into account social context or cultural norms (for example, Wilkinson and Fairhead 2017).

Expertise and Disagreement Discounting

Efficient policy making requires policy expertise. In some contexts, representatives can partially outsource this skill (Hertel-Fernandez 2019). However, only elected officials can introduce bills and shape the legislative process directly. Recent studies have revealed specific ways in which the legislators' individual expertise can shape the policy-making process. Policy innovations tend to be introduced by representatives with more specialized knowledge (Makse 2021; Miler 2017). Seminal theories of legislative organization also recognize the value of expertise in policy making. According to Krehbiel (1992), the U.S. Congress committee system is meant to provide the chamber with the necessary information and expertise for legislating. Similar institutions designed to promote a division of labour in the legislative branch are common in Europe (for example, Mattson and Strøm 1995; Strøm 1998). Committees promote specialization within relevant policy jurisdictions and provide valuable information to parliaments (Shepsle and Weingast 1994). Committee members act as ‘low-cost’ specialists who make the legislative activity more efficient (Gilligan and Krehbiel 1989). Hence, the institutional framework of most contemporary parliaments encourages legislators to develop specific policy expertise (Mattson and Strøm 1995).

However, expertise can simultaneously shape how legislators evaluate information and make decisions. Expertise produces self-assurance (Fisher and Keil 2016; Tetlock 2005). Individuals with more knowledge and experience in a given domain tend to have more confidence in their own beliefs within that field. However, this confidence is often unwarranted (Dunning 2005). Individuals who are induced to believe they are experts tend to overestimate the accuracy of their beliefs (Arkes et al. 1987; Trafimow and Sniezek 1994). Specialized knowledge can produce illusions of understanding, partly because experts are less willing to admit that they do not know something in their area of specialization (Backus and Little 2020; Bradley 1981). Individuals tend to be more accurate at assessing their own knowledge in an unfamiliar domain than in a familiar field such as their college major (Fisher and Keil 2016). At the same time, recent scholarship suggests that expertise stimulates close-minded cognition or dogmatism (Ottati et al. 2015). Accordingly, social norms entitle experts to adopt more dogmatic and forceful positions while encouraging laypeople to be more open-minded and accepting of criticism.

If legislators with specialized knowledge express similar forms of overconfidence and dogmatism, policy expertise can impair the ability of elected officials to carry out other tasks. Re-election-seeking officials have incentives to act on public preferences (Butler and Nickerson 2011; Pitkin 1967).³ However, in order to be responsive, politicians must account for the preferences of constituents with whom they agree and disagree. We argue that policy expertise constrains the legislators' ability to incorporate the preferences of constituents they disagree with. The overconfidence induced by specialized knowledge can lead representatives to disregard opposing views in their areas of expertise. As a result, policy experts may develop perceptions of public opinion that are disproportionately influenced by their own preferences (relative to non-experts). Therefore, we argue that specialized knowledge, although a central feature in the policy-making process, can limit the ability of legislators to voice public preferences – what we call the expertise paradox.⁴

Our argument builds upon recent scholarship suggesting that legislators systematically dismiss the preferences of the constituents with whom they disagree (Butler and Dynes 2016). The tendency of officials to discount contrasting views may result from motivated reasoning. When faced with facts or opinions that challenge pre-existing beliefs, individuals are more likely to dismiss or actively counterargue these arguments (Lodge and Taber 2013). The expertise paradox hypothesis suggests that the propensity to dismiss contrasting views should be more acute in domains where legislators have more expertise.

A possible interpretation of this argument is that expertise is ‘bad’ for political representation. We disagree with this interpretation. Specialized knowledge is key to developing good policy solutions and in several domains we should not expect the majority of voters to have an informed opinion. Managing a public health emergency such as the COVID-19 pandemic is a good example of a context in which decision-makers should listen first and foremost to experts. However, ignoring public preferences can pose challenges to policy implementation. For instance, containment measures in response to a pandemic that are insensitive to specific cultural norms can backfire, leading to lower compliance and resistance (Wilkinson and Fairhead 2017).

Testing the Expertise Paradox

We fielded two original survey experiments with Swedish elected officials to test the expertise paradox hypothesis. The experiments were embedded in separate waves of the Panel of Politicians, a biannual panel survey with national, regional, and local representatives. The survey was administered by the Laboratory of Opinion Research at the University of Gothenburg.⁵

The goal of the first study was to understand whether policy expertise constrained the legislators' ability to incorporate public opinion signals. Politicians evaluated a hypothetical policy appeal from a group of voters. Direct interactions with constituents occur regularly and influence the legislators' perceptions of public opinion (Broockman and Skovron 2018; Pereira 2021). We experimentally manipulated elements of the request – including the issue area – to isolate the effects of expertise. According to the expertise paradox hypothesis, legislators should be more likely to disregard the opinion of voters in their areas of expertise. We designed a second study to directly assess our proposed causal mechanism. Study 2 tested whether heightened perceptions of policy expertise increased officials' self-confidence and dogmatism.

Both studies rely on subjective measures of expertise. The measures were built upon instruments validated in prior work on the topic (Ottati et al. 2015). Still, one concern with this strategy is that self-reports may simply reflect issue salience or self-confidence rather than specialized knowledge. We address this concern in two ways. In Study 1, the main results remained unchanged after accounting for variations in issue salience (Figure D5). Additionally, in Appendix E, we leverage the panel component of the survey to illustrate how the measure of expertise in Study 1 was rooted in genuine experiences. A subset of officials in Study 2 was asked to describe an instance when their expertise was useful in office (without referencing a specific issue area). Without prompting, nearly 40 per cent of respondents referenced the same issue area identified in Study 1 eight months earlier. This number grows to 80 per cent among officials who mentioned one of the five issue areas listed in Study 1. We interpret this result as reassuring. Self-reported measures of expertise in Study 1 do not simply capture abstract perceptions of expertise; they include concrete experiences of elected officials, either from their time in office or from formal training.

The Swedish Context

The Panel of Politicians is an online university-based panel that has been administered twice a year since 2011. The panel is composed of a diverse group of Swedish politicians. All major parties are proportionally represented in the sample save for the anti-establishment party, Sweden Democrats, which is somewhat underrepresented. Local councillors and mayors represent roughly three-quarters of the panel (74–5 per cent). The remaining sample is composed of politicians in regional and national offices, including the Riksdag.⁶ Participants are recruited via invitations in large surveys like the Comparative Candidates Survey and the Kommun- och landstingsfullmäktigeundersökningen, and through direct appeals on the websites of elected assemblies at all levels of government. Our first experiment was fielded in wave thirteen of the panel, administered between 10 October and 13 November 2019. A total of 1,861 elected officials participated in the survey (a 47 per cent response rate). The second experiment was integrated into wave fourteen and fielded between June 1 and July 6, 2020 (N = 1,348). Table B1 (Appendix B) summarizes the demographic characteristics of both elite samples.

Sweden's parliamentary system features strong and cohesive parties elected via proportional representation. Politicians are, first and foremost, party representatives. Yet, preferential voting creates incentives for legislators to cultivate a more personalized relationship with voters. Candidates on a party list can improve their ranking if they receive more than 5 per cent of the party votes in their constituency. Prior studies have found that although Swedish legislators are constrained by intra-party dynamics, they are committed to take constituency preferences into account (Naurin and Öhberg 2018) and to act on voters' signals (Butler, Naurin, and Öhberg 2017). Therefore, we expect the patterns uncovered here to generalize to most European countries that have (1) party-centric systems and (2) some incentives for officials to establish a personal connection with voters.

Party-centric systems can raise concerns about the substantive consequences of the expertise paradox. If each party collectively decides its policy positions, the individual biases of representatives with different areas of expertise may cancel out. However, at least in the Swedish context, legislators are particularly influential within their areas of expertise. As described in Table 1, according to the 2010 wave of the Swedish Parliamentary Survey, 53 per cent of Swedish Members of Parliament (MPs) reported having a very good chance of influencing the position of their party group within their area of expertise, compared to only 9 per cent outside their area of expertise. Hence, it is exactly in the areas in which legislators are more influential that their ability to incorporate constituent preferences might be constrained.

Table 1. Perceptions of within-party influence by area of expertise among Swedish MPs

Note: Each column represents the distribution of MPs' responses to the question: ‘How do you rate your ability to impact your party group's position on issues within/outside your own expertise?’.

Source: 2010 Swedish parliamentary survey.

Study 1: Expertise and Disagreement Discounting

The first study assesses whether policy expertise leads politicians to discount positions that go against their own. We asked elected officials to evaluate a hypothetical policy appeal made by a group of constituents. The content of the appeal was altered to isolate the effect of expertise on politicians' responses to the message. Two pre-treatment items provided the information needed to build the vignette: (1) measures of policy expertise and (2) the officials' preferences on the policies included in the vignette. Each item was described in turn, followed by the constituents' message and the outcome variables used to capture the response to the appeal.

Measuring Policy Expertise and Policy Preferences

Policy expertise results from years of accumulated knowledge and experience. The entire process through which individuals develop expertise cannot be credibly manipulated in a survey. Instead, the natural variation in expertise between respondents across five salient issue areas was leveraged: healthcare, education, immigration, social welfare, and housing. To measure policy expertise, we introduced the following item early in the survey:

Public officials have to deal with several different issues as part of their job, and it is impossible to be an expert in all of them. Below is a list of common issues that governments have to deal with. Please identify the areas in which you have (1) more and (2) less expertise on:

• • Healthcare

• • Education

• • Immigration

• • Social welfare

• • Housing

Officials reported the issue area in which they were (1) most and (2) least knowledgeable among the five domains listed.⁷ The distribution of higher- and lower-expertise issue areas was fairly uniform across respondents. While the modal area of expertise was education (30.5 per cent of respondents), the remaining four issue areas were selected by 11–24 per cent of public officials in the study. The same is true for the areas in which legislators have lower levels of expertise (Table B2). In Appendix E, we leverage data from an open question in Study 2 to show that this measure based on self-reports indeed captured the legislators' specialized knowledge and idiosyncratic experiences.

We used this information to randomly assign respondents to a high- or low-expertise condition. Later in the survey, those in the high-expertise condition received a message from a group of constituents asking the legislator to endorse a specific initiative in their field of expertise. Officials in the low-expertise condition received a policy appeal in the area in which they were least knowledgeable of the five domains provided.⁸

For each area of expertise, we identified one specific policy initiative. Table 2 lists the initiatives associated with each policy jurisdiction. We selected policies that were salient at the time of the study, based on pre-tests with data from the Swedish Parliamentary Survey and the Swedish National Election Studies.⁹ For each issue, elite and public support were sufficiently split so that any position on the issue could be seen as credible. For example, the healthcare initiative concerns whether private companies should be prevented from operating hospitals. At the beginning of the survey, we collected the official's positions on each of these initiatives. To avoid contamination, we included additional policy questions in this section of the survey, and the order of the items was randomized. Table 2 reports the share of legislators supporting each initiative. We used the officials' expressed preferences to ensure that the position of the group of constituents was constantly at odds with that of the public official. This way, we accounted for the tendency of legislators to disregard opinions they disagreed with (Butler and Dynes 2016). Holding policy disagreement constant in the vignette allowed us to isolate the effect of policy expertise.

Table 2. Issue areas, corresponding policy initiatives, and share of supporters in the Panel of Politicians

Policy Appeal and the Legislators' Response

Finally, we asked respondents to evaluate a putative message sent from a group of constituents.¹⁰ Legislators in the high-expertise condition received an appeal on the issue area in which they were most knowledgeable. In contrast, those in the low-expertise condition received a policy appeal in their least knowledgeable issue area. Box 1 provides an example of the message presented to public officials – in this instance, related to healthcare policy. As described above, the constituents' position on the policy was the opposite of the elicited preferences of the legislator to keep policy disagreement constant. Appendix A presents the vignettes for the remaining four policy issues.

Box 1. Example of constituents' appeal on one of the five issue areas: healthcare

Different groups of voters contact politicians with political propositions. Imagine the following: a group of voters is approaching you and wants you to [support/oppose] a proposal to ban companies from running hospitals. They believe that healthcare is facing major challenges. They see equality as a central issue. Their main argument is that a ban on companies operating hospitals makes it [easier/more difficult] for vulnerable patients to choose the healthcare they need.

We measured the legislators' reactions to the message by asking them to indicate their level of agreement with each of the following statements:

• • The group likely understands the complexities of the issue

• • The group likely based its opinion on facts

• • The group likely holds this position strongly

• • The group's opinion is the opinion of the majority of voters

The question wording comes from Butler and Dynes (2016). The first three statements capture the extent to which politicians recognized the validity of the arguments put forward by the constituents. The items provide specific rationales for politicians to discount the opinions conveyed in the message. While we could have directly asked legislators whether they would disregard the policy appeal, this approach would risk producing biased responses due to social desirability. The fourth statement captures the respondents' ability to acknowledge that opposing views may be shared by the majority of voters. The responses were recorded on seven-point Likert scales ranging from ‘completely disagree’ to ‘completely agree’. Each item was treated as a separate outcome variable. If the expertise paradox hypothesis is correct, legislators should be more likely to disagree with each of the statements when the message pertains to a high-expertise issue area.

To measure the causal effects of expertise on the legislators' responses to the policy appeal, we estimated linear regressions with covariate adjustment to improve precision (Gerber and Green 2012). The models account for the respondents' age, educational level, and party. However, the same substantive results are retrieved from bivariate regressions (see Figure D1 and Table C1 in Appendix D).

Results

Figure 1 presents the main results from Study 1. The y-axis lists the four statements used as outcome variables to assess the legislators' responses to the policy appeal. Each estimate represents the average difference in the level of agreement with the statements between officials in the high- v. low-expertise conditions. Therefore, the estimates capture the causal effects of policy expertise on the legislators' responses to the policy request. Negative values indicate less agreement with each of the statements.

Figure 1. The effects of expertise on legislators' ability to incorporate contrasting views.

Note: Points are estimates of the causal effect of policy expertise on legislators' assessments of voters' opinions. Horizontal narrow/wide lines are 95%/90% confidence intervals. Agreement with each statement listed on the y-axis corresponds to a distinct outcome variable. Estimates and standard errors are derived from linear models with covariate adjustment for party, age, and education. Table C1 reports the full model results.

The results are consistent with the expertise paradox hypothesis: policy expertise increases the propensity of public officials to discount contrasting views from constituents. On the one hand, there is no evidence that policy expertise shapes beliefs about the strength of voters' opinions. This item, along with the first two statements described on the y-axis, provided different justifications for respondents to dismiss the policy appeal. On the other hand, respondents in the high-expertise condition were significantly less likely to believe that the group of constituents understood the complexity of the issue and based their opinions on facts. Politicians are more likely to disregard opinions that go against their own in their areas of expertise. The statements included in the vignette provided concrete justification for this response. The effect sizes estimated are substantively meaningful. For instance, the top estimate in Figure 1 is −0.28 (standard error = 0.09). Since the mean level of agreement with the statement ‘voters understand the complexity of the issue’ is 2.8 (on a seven-point scale), the effect of policy expertise corresponds to a 10 per cent decrease in agreement with the statement. In turn, the estimated effect of expertise on agreement with the idea that the group of voters based their opinion on facts (mean = 2.78) is −0.35, which corresponds to a 12.6 per cent decrease relative to the mean value of the outcome.

Finally, officials in the high-expertise condition were less likely to accept that the position conveyed in the policy appeal reflected the view of the majority of voters (−0.17; S.E. = 0.08). Only 19.1 per cent of respondents (across conditions) accepted that the policy appeal represented the majority opinion. Since, by design, the position of the group of voters was always at odds with the preferences of the legislator, this result is not surprising. Prior studies have found that politicians systematically underestimate the level of support for policies they oppose (Converse and Pierce 1986; Pereira 2021). However, the patterns uncovered here suggest that elected officials are more likely to discount contrasting views in their domains of expertise.¹¹

Importantly, these patterns do not result from politicians in the high-expertise condition being more likely to receive a policy appeal that is counter to their own preferences (Butler and Dynes 2016). The vignettes were designed to hold policy disagreement constant. Also, the findings are not explained by the specific position held by the group of voters, either in favour or against the policy. The results remain unchanged after accounting for whether the policy appeal is aligned with the opinion of the majority of voters nationwide (Figure D2).

The study was not designed or powered to estimate effects by individual issue. Still, it is important to know whether the main findings were driven by any particular issue. This question is explored in three steps. First, the main models were re-estimated, excluding one issue at a time (Figure 2). If one particular issue was responsible for the patterns uncovered, meaningful differences relative to the estimates from the full sample should be observed. No evidence of differential effects by issue is found. Second, fixed effects by issue (Figure D3) were included. The main findings are robust to this specification, with the exception of the coefficient for ‘opinion aligned with themajority’, which is no longer distinguishable from zero. Both these analyses risk inducing post-treatment bias because the policy embedded in a given vignette is a function of treatment assignment (Montgomery, Nyhan, and Torres 2018). To avoid this concern, in Figure D4, we replicated the main models with fixed effects by pairing of higher/lower expertise domain selected pre-treatment by the subjects. This empirical strategy is consistent with the goal of leveraging differences across issues while accounting for any systematic variation across pairings. The same patterns are uncovered with this specification.¹² Together, the analyses suggest that systematic differences across issues or pairings of issues can only partially explain the tendency of legislators to discount contrasting views in their areas of expertise.

Figure 2. The effects of expertise on legislators' ability to incorporate contrasting views, excluding one policy issue in turn.

Note: Points are estimates of the causal effect of policy expertise on legislators' assessment of voters' opinions after dropping one policy issue. Horizontal narrow/wide lines are 95%/90% confidence intervals. Agreement with each statement listed on the y-axis corresponds to a distinct outcome variable. Estimates and clustered standard errors derived from linear models with covariate adjustment for party, age, and education.

The results could also be explained by differences in issue salience. Officials who have devoted more time to a given issue may care more about it. Hence, it is possible that legislators in the study were more likely to discount contrasting views in their areas of expertise not because of the overconfidence that comes with specialized knowledge, as hypothesized, but simply due to differences in issue salience. To account for this alternative mechanism, we replicated the main analysis controlling for how strongly respondents held their opinion on the issue in the vignette. The main effects remain unchanged after accounting for issue salience (Figure D5). This result suggests that the measures of expertise adopted in the study are not simply capturing issues that legislators deem most important to themselves or their constituents.

Finally, we assess how the list of issue areas included in the study may have affected the results. The experimental design required legislators to rank a limited set of policy areas in terms of expertise (see Table 2 above). These five domains are among the most relevant in local and regional governments from which the majority of officials in the sample were recruited (88–90 per cent, as described in Table B1). Still, this design constraint may potentially bias the results. In wave fourteen of the Panel of Politicians fielded seven months after Study 1, we provided respondents with a more extensive list of fourteen issues to address this concern.¹³ Two out of every three respondents recontacted (67 per cent) identified the same issue area in the original experiment and in the extended list of fourteen issues. More importantly, the same substantive results were uncovered when restricting the main analysis to this subset of respondents (see the full results in Figure D9). Hence, although this design constraint may have induced measurement error, it did not bias the results in a meaningful way.

Together, the findings suggest that policy expertise systematically constrains the legislators' ability to voice public preferences. Elected officials with more specialized knowledge in a given domain are more likely to overlook opinions that are contrary to their own. Experts are also more likely to believe that views opposed to their own do not represent the majority opinion. The following section explores how different sources of specialized knowledge can moderate the expertise paradox. The analysis provides a first test of the proposed causal mechanism.

The Role of Formal and Passive Expertise

Expertise in a given domain may come from the legislators' experience in office, prior professional experience, or formal education. Understanding the forms of knowledge that are more likely to result in the expertise paradox is important to shed light on its underlying mechanisms. In this section we explore how two sources of expertise moderate legislators’ tendency to discount contrasting views.

Formal expertise is a type of knowledge that results from an extended study of a particular topic (Palmiero et al. 2020). Formal education often takes place early in life and over a relatively short period of time. The knowledge acquired through a college degree can quickly be forgotten, giving rise to illusions of understanding and overconfidence (Fisher and Keil 2016). To measure formal expertise, we created a binary measure that takes a value of 1 if a legislator completed a college degree and 0 otherwise (50.1 per cent of officials in the survey had a college degree). Passive expertise, in turn, comes from ‘exposure through life experience and the position one occupies in a society or culture’ (Fisher and Keil 2016, 1,251). Officials with extended experience in office may feel more entitled to disregard contrasting views. Converse and Pierce (1986) found that French legislators with more years in office were more likely to misperceive constituency preferences. To capture passive expertise, we measured the respondents' time in office.¹⁴

According to our theoretical argument, the expertise paradox results from overconfidence and dogmatism. Both formal and passive expertise can produce overconfidence and less support for open-minded cognition. Hence, the main effects identified in the previous section are expected to be driven mainly by officials with either a college degree or more experience in office. To test these predictions, we reestimated the main model interacting with the treatment variable with each source of expertise, respectively.¹⁵

Figure 3 presents the results. Panel (a) describes the effects of expertise on disagreement for politicians with and without a college degree. The patterns suggest that legislators with a college degree were more likely to downplay the opinions of constituents they disagreed with in their area of expertise. For the three outcomes that provided results consistent with the expertise paradox in the main analysis, the coefficient for officials with a college degree is consistently larger (more negative). For instance, the effect of policy expertise on agreement with the statement that the message reflects the majority opinion is virtually zero among officials without a college degree (estimate = 0.02; std. error = 0.11), but negative and substantively large for respondents with a college degree (estimate = −0.36; std. error = 0.12). Officials with a college degree were, therefore, the significantly less likely to accept that opposing views in their area of expertise could represent the opinion of the majority (p-value for difference in means = 0.02).

Figure 3. The effects of expertise on the legislators' ability to incorporate contrasting views by form of expertise.

Note: Points are estimates of the causal effect of policy expertise on legislators' assessments of voters' opinion among officials with/without a college degree (panel a) and officials with above-/below-median experience in office (panel b). Horizontal narrow/wide bars are 95%/90% confidence intervals. Agreement with each statement listed on the y-axis corresponds to a distinct outcome variable. Estimates and standard errors derived from linear models with covariate adjustment for party and age. Full model results are in Tables C2 and C3.

In turn, Panel (b) describes how experience in office moderates the effects of expertise. The results provide less conclusive evidence. Representatives with above-median experience are slightly more likely to dismiss contrasting views, but the differences are small and indistinguishable from zero. The p-values for the difference in means in the first two outcomes are 0.51 and 0.42, respectively. The effects of expertise on the belief that contrasting views reflect the majority position (the two bottom estimates) are in the opposite direction. The estimate is distinguishable from zero only among politicians with less experience in office.

Together, the results suggest that the tendency of elected officials to discount contrasting opinions is more consistently driven by respondents with formal expertise. This result is in line with recent scholarship in cognitive psychology. Individuals with expertise that arise from an extended study of a particular topic are less like to exhibit overconfidence outside their fields of knowledge but more likely to overestimate their ability to explain their own areas of expertise (Fisher and Keil 2016). This overconfidence may result from ‘a failure to realize how much has been forgotten since they had maximum mastery of the topic’ (1260). Importantly, neither formal education nor experience in office is exogenous. Since treatment heterogeneity is not causally identified, the results should be interpreted as suggestive.

Study 2: Mechanism Test

In Study 1 we uncovered a relationship between expertise and how politicians respond to policy appeals from constituents. Officials with more knowledge of and experience in a given domain are more likely to discount opinions they disagree with. This relationship may be explained by a combination of overconfidence and dogmatism. Policy expertise can induce illusions of understanding, leading experts to overestimate the accuracy of their beliefs (Bradley 1981) and to engage in less open-minded cognition (Ottati et al. 2015). Consistent with this view, in the previous section, in the previous section we show that politicians with formal expertise are more likely to discount opposing opinions. In Study 2 we directly test the proposed mechanism.

The main goal of this study was to assess whether perceptions of policy expertise generate self-confidence and support for dogmatic attitudes. To induce perceptions of expertise, we randomly assigned officials into one of two groups. The first group was presented with the following expertise prime:

As a public official, you have to deal with a variety of policy issues. In some instances, this involves making decisions on issues you have extensive experience in. Please take two minutes to describe an instance when your specific policy expertise was important to solve an issue or push legislation forward.

Respondents in this condition reflected on a specific episode when their own expertise was important to solve an issue in office. This design builds upon prior experimental work using the controlled recollection of past events to induce specific beliefs and emotions (for example, Hadzic and Tavits 2019; Lerner et al. 2003). The goal of this intervention was to heighten respondents' self-perceptions of expertise. A second group received a placebo prime asking respondents to describe, in two minutes, how they first got interested in politics.¹⁶ We included a placebo to account for the possibility that the recollection task, per se, could influence our outcomes of interest while not affecting the respondents' perceptions of expertise.

After the recollection task, we asked both groups how much they agreed with the following statements: (1) ‘I often have doubts about my own decisions in office’ and (2) ‘It is a waste of time to pay attention to certain ideas.’ Respondents could record their responses on a seven-point scale from ‘Completely disagree’ to ‘Completely agree’. The first item measures the officials' degree of self-confidence. The second item captures their openness to different opinions (Ottati et al. 2015). Public officials who received the expertise prime were expected to be less likely to agree with the first statement and more likely to agree with the second.

This second experimentwas included in wave fourteen of the Panel of Politicians, in Sweden. See Table B1 (Appendix B) for sample descriptives. A total of 1,143 elected officials participated in this section of the survey. As in the first experiment, we rely on linear regressions with covariate adjustment to measure the causal effects of the expertise prime.

Figure 4 presents the main findings of Study 2. The estimates describe the causal effects of the expertise prime (relative to the placebo prime) in agreement with each of the statements described along the y-axis. The top estimate reveals that heightened perceptions of expertise are associated with more self-confidence. Public officials who received the expertise prime were less likely to express doubts about their decisions in office (−0.13; p-value = 0.08). This result is consistent with the argument that expertise is associated with more self-confidence. In turn, the effect of expertise on our measure of dogmatism (bottom estimate in Figure 4) is positive, but small and indistinguishable from zero (0.05; S.E. = 0.11). Representatives who received the expertise prime were not significantly more likely to agree with the statement that paying attention to certain ideas was a waste of time. This null effect may be explained, in part, by a relatively weak treatment or as a result of social desirability bias.

Figure 4. The effects of priming expertise on self-confidence and dogmatism.

Note: Points are estimates of the causal effect of expertise on self-confidence (top estimate) and close-minded cognition (bottom estimate). Horizontal narrow/wide bars are 95%/90% confidence intervals: estimates and standard errors derived from linear models with covariate adjustment for party, age, and education. Full model results can be found in Table C4.

Still, given the inherent challenges of effectively manipulating perceptions of expertise, we interpret these findings as reassuring. The results are consistent with the mechanism proposed to explain the expertise paradox. On average, politicians who received the expertise prime were less likely to express doubts about their own decisions and were (suggestively) less open-minded.

Discussion

Good policy outcomes require legislators to have expertise in different issue areas. Committee systems in most legislative branches encourage this form of specialization. At the same time, elections create incentives for responsiveness. In this article we argue that there might be a trade-off between electing policy experts and representatives who pursue policies that are consistent with voter preferences. This phenomenon is referred to as the expertise paradox.

Consistent with our argument, we show that policy expertise leads Swedish legislators to disregard the views of constituents with whom they disagree, regardless of the actual position expressed by voters. Experts are also less likely to believe that voters with preferences different from their own can reflect the majority's position. Hence, the specialized knowledge systematically constrains the legislators' ability to take public preferences into account when devising policies in their areas of expertise. In a second study we show that self-confidence induced by specialized knowledge can partially explain these patterns.

The patterns uncovered do not appear to be unique to a specific issue. Support for the expertise paradox was found across five salient policy jurisdictions: healthcare, education, immigration, housing, and social welfare. Yet, our findings also have important limitations. First, we held policy disagreement constant to ensure the experimental design was focused and well-powered. It is conceivable that the effects of expertise are less pronounced when politicians are asked to endorse policies they support. More research is needed to elucidate this question. That said, the expertise paradox is no less relevant if it is limited to instances in which officials are confronted with opposing views. Dealing with disagreement is an intrinsic element of the policy-making process. The ability of politicians to recognize the value of contrasting views is key to finding compromise solutions, which are often required to move policy forward (Gutmann and Thompson 2014)

Second, we propose that the expertise paradox results from a heightened sense of self-confidence, and close-minded cognition – processes that are associated with more specialized knowledge in a given domain. Study 2 establishes a relationship between expertise and self-confidence. Still, the mechanisms proposed are not exhaustive. For instance, it is possible that officials are more likely to discount public appeals in their areas of expertise because interest groups disproportionately target policy experts when trying to raise attention to an issue. This process could create case overload and lead legislators to become more sceptical about who is responsible for the appeal or the extent to which it is rooted in real concerns of the electorate.

Third, in Study 1, the officials encountered a group of constituents with a policy appeal, following similar protocols in other studies of responsiveness (for example, Butler and Dynes 2016; Öhberg and Naurin 2016). Constituency contacts are a common mechanism through which politicians receive public opinion signals and build their image of the majority. That said, responsiveness is, at least implicitly, about the actual position of the majority. The implications of this argument are potentially broader if politicians are also more likely to discount opinion polls in their areas of expertise. While this would be consistent with the expertise paradox, it remains an open question in this study.

Finally, the Panel of Politicians is composed mainly of local- and regional-level representatives. We speculate that the effects uncovered should be more prevalent among national-level legislators. On the one hand, recent scholarship has shown how local policy outcomes are responsive to mass preferences, refuting the traditional idea that local governments were less likely to act on voter preferences (de Benedictis-Kessner and Warshaw 2016; Tausanovitch and Warshaw 2014). On the other hand, national politicians have more resources to develop policy expertise. Perceptions of expertise that are better rooted in specialized knowledge are also more conducive to disagreement discounting. Future work is needed to elucidate this question.

The study contributes to scholarship on policy responsiveness, legislative politics, and elite behaviour. First, the expertise paradox provides a novel explanation for variation in mass-elite policy responsiveness. By restricting the ability of legislators to incorporate public signals, the expertise paradox implies that the prospects for constituency control are weaker when politicians disagree with constituents in their areas of expertise. Whether this implication is normatively positive or negative depends on one's understanding of the appropriate role of representatives. Sacrificing responsiveness over expertise can be seen as positive for representation (Bertsou and Caramani 2022). However, the expertise paradox can raise obstacles to policy implementation. If policy experts consistently disregard public opinion, constituents may be less willing to acquiesce. Policy solutions that are insensitive to social or cultural norms may lead to lower compliance or backlash (Wilkinson and Fairhead 2017).

The findings also shed light on the value of descriptive representation. In Study 1, politicians with college degrees were shown to be more likely to discount contrasting views in their areas of expertise. At the same time, recent scholarship shows that less privileged subconstituencies are less likely to convert their preferences into policy (Giger, Rosset, and Bernauer 2012; O'Grady 2019). Without questioning the importance of electing representatives with formal expertise, the results suggest that inequalities in policy responsiveness can be mitigated by electing a more diverse group of legislators.

The expertise paradox can also contribute to debates over parliamentary reform. Parliaments worldwide rely on their members’ specialized knowledge to find solutions for different policy issues. However, the results uncovered here suggest the existence of a trade-off between expertise and responsiveness. Reformers interested in curbing the systemic low levels of trust in the legislative branch (Hibbing and Theiss-Morse 1995; Leston-Bandeira 2012) may benefit from recognizing this trade-off and incorporate it into their efforts to strengthen the links between voters and their representatives without questioning the value of expertise.

Finally, the study joins recent scholarship on elite behaviour (for example, Butler and Dynes 2016; Esaiasson and Öhberg 2020; Pereira 2022; Sheffer et al. 2018). By focusing on the cognitive processes underlying the decisions of elected officials, these studies shed light on different questions left open in long-standing debates in the discipline. In this particular study, a natural next step involves exploring how elected officials can overcome the tendency to dismiss public opinion in their own areas of expertise.