David E. Broockman

1

How Do Citizens React When Politicians Support Policies They Oppose? Field Experiments with Elite Communication

David E. Broockman

Department of Political Science, University of California, Berkeley

210 Barrows Hall, Berkeley, CA 94720

Daniel M. Butler

Department of Political Science, Washington University in St. Louis

Campus Box 1063, One Brookings Drive, St. Louis, MO 63130-4899

Abstract

Politicians have been depicted as, alternatively, strongly constrained by public opinion, able to shape public opinion with persuasive appeals, or relativelyunconstrained by public opinion and able to shape it merely by announcing their positions. We shed light on how citizens react when politicians support policies they oppose with unique field experiments conducted in cooperation with sitting politicians.In the experiments, state legislators sent their constituents official communications with randomly assigned content. In some letters, the legislators took issue positions the receiving constituents had previously opposed, sometimes supported by extensive arguments but sometimes minimally justified. An ostensibly unrelated telephone survey found that citizens often adopted their representatives’ issue positions even when representatives offered little justification. Moreover, citizens did not evaluate their representatives more negatively when representatives took positions citizens opposed, even when minimal justification was offered. The findings suggest that citizens often afford elected officials significant truston policy matters.

“Your representative owes you, not his industry only, but his judgment; and he betrays, instead of serving you, if he sacrifices it to your opinion.” – Edmund Burke (1774)

Traditional conceptionsofrepresentative democracyoften envisionpublic opinion placing strong constraints on politicians’decisions.According to suchperspectives,electoral accountabilitycompelspoliticians to support policiesthatare broadly popular among citizens, lest they face electoral sanction (Downs 1957; Loewen and Rubenson 2010).Consistent with these theories, there is a strong relationship between public opinion and politicians’ policy positions(e.g., Brody and Page 1972).

Research on opinion leadership has long offered an alternative explanation forthis relationship:citizens mayadoptelites’ policy positionsas their own. Over the last several decades, evidencethat elites can influence public opinion in this manner has mounted(e.g., Abramowitz 1978; Gabel and Scheve 2007; Gooch 2014; Kahan 2012; Lenz 2009, 2012; Minozzi et al. 2014; Tesler 2014a). However,the nature of opinion leadership – and its implications for politicians’ accountability to public opinion –are increasingly contested.

On the one hand, classic theoreticalaccountscharacterizeopinion leadership as reflectingelite persuasion. These accountscontend thatelitesshape public opinion primarilyby highlighting how theirpolicy proposals are consistent with citizens’ preexisting values or will accomplish shared goals(e.g., Chong and Druckman 2007a, 2007b;Grose et al. 2014; Kinder and Sanders 1990; Ladd 2007; Jacoby 2000; Tesler 2014b;Zaller 1992).Such theories generally imply thatpoliticians’ accountabilityto public opinion remains relatively robust. If citizens reliably demand that politicians persuasively justify theirdecisions, politicians may tend to eschew taking positions they cannotpersuasively defend(e.g., Fenno 1978; Kingdon 1989).

However, adifferenttheoretical perspective suggests that, at least some of the time,opinion leadership is better characterized as a process ofposition adoption. Thisposition adoption perspective claims that citizensoften trust politicians’policy judgments over their own. Citizensthus are not expected toreliably evaluate their representativesmore negatively when learning their representatives support policies they oppose, and may instead adopt their representatives’ positions as their own–evenin the absence of persuasive justifications (e.g., Achen and Bartels 2006; Bartels 2005; Cohen 2003; Leeper 2013; Lenz 2009, 2012; Mackie and Cooper 1984; Zaller 2013).Such position adoption, to the extent it occurs, implies that citizens are inclined to trust politicians’judgment without demanding justifications. Consequently, public opinion may constrain politicians’ decisions far less.

The notion that such ‘position adoption’can describe public opinion even some of the time has been understandably controversial (for review, see Bullock 2011).On the one hand, a rich theoretical tradition in democratic theory suggests citizens may believe elites have superior policy information, expertise, or judgment (Burke 1774; see also Bianco 1994; Fox and Shotts 2009). Likewise, psychological studies suggest citizens may tend to defer to authority figures on even sensitive matters (e.g., Milgram 1974, Epilogue). However, this perspective sits at odds with most contemporary theories of democratic politics, which tend to envision democracy as the aggregation of citizens’ preferences (see Miller 1992 for review). Moreover, as we review below, existing empirical evidence in favor of this view remains potentially vulnerable to important critiques.

In this paper we examinethe empirical merit of thisposition adoptionhypothesis in a series of unique and difficult tests. Our evidencecomes from field experiments conducted in collaboration with sitting politicians in which we worked with them torandomly assign aspects of their official position-taking to voters.In the experiments, state legislators sent their constituents official communications with randomly assigned content. In letters to constituents in treatment groups, the legislatorsstaked out positions on salient and controversial issues these constituents had previously opposed.The policies in question were neither insignificant nor uncontroversial, and included decriminalizing marijuana, the minimum wage, and policies toward undocumented immigrants, among others.Further, legislators sometimes justified their positions on these issues with extensive arguments, but sometimes only minimally justified them. Ostensibly unrelatedtelephone surveysthen probed the effects of this position-taking on citizens’ evaluations of legislators and views on the issues.

We expected this contextto represent a relatively difficult test of theposition adoptionhypothesis: the politicians did not highlight any partisan similarity with voters, merely their status ascitizens’ elected officials – and, because the cooperating politicians were relatively unknown state legislators, just about the only thing many constituents in the treatment group knew about them was that they disagreed on a contentious issue.

Nevertheless, constituents’ reactions to these official legislative communications lend support to the position adoption perspective. When legislators sent their constituents a letter containingan issue position, these constituentswere significantly more likely to adopt this position. Moreover, constituents’ evaluationsof their legislatordid not significantlysour when thelettersstaked out positions the constituents had previously opposed, despite that constituents learned where their legislators stood.Finally,these patterns did not depend on whether legislators provided extensive justifications for their positions or largely omitted them: legislators appeared to move their constituents’ opinions towards their ownpositions just as successfully by essentially announcing their positions; likewise, citizens did not judge their legislators negatively even when their legislators clearly asserted a position with which the constituent had previously disagreed. (Nevertheless, citizens were equally likely to learn their legislators’ position.) There was no evidence that these patterns differed across issues.

These findings have important implications for understanding the nature ofpoliticians’ accountability to public opinion. To be sure, our evidence falls far short of providing a comprehensive account of democratic accountability: it does not suggest politicians canalwaysconvince their constituents of anything, nor thatthey can reliablyescape electoral punishment for everything.Indeed, we hope future research further explores the scope conditions on our findings, as they are sure to exist.However, our resultsdo provide unique evidence underscoring anotion deeply engrained in political thought: across a wide range of policy matters,citizens appear disposed to affordelected officialsa significant degree of trust.

How Do Citizens React When Politicians Support Policies They Oppose? Theoretical Perspectives And Existing Evidence

We focus on an empiricalquestion where three prevalent theoretical perspectives on public opinion and democratic accountabilityoffer sharply different predictions: how do citizens react when politicians support policies they oppose? To appreciate the differing answers to this question existing theoretical perspectives would offer, consider Table 1.

Table 1. How Do Citizens React When Politicians Support Policies They Oppose? Empirical Predictions of Competing Perspectives on Opinion Leadership

Empirical Question / Do leaders offer persuasive arguments appealing to citizens’ values? / Issue Voting Predictions / Elite Persuasion Predictions / Position Adoption Predictions
Do leaders lose favorability? / Yes / Yes / No / No
No / Yes / Yes / No
Do leaders change citizens’ issue opinions? / Yes / No / Yes / Yes
No / No / No / Yes

In its strictest form, the traditional view of democracy, which conceives of citizens as issue voters, suggests that politicians lose favor whenever they take policy positions their constituents oppose and cannot meaningfully influence citizens’ policy preferences. Indeed, in this view, just about the least politically advantageous thing a politician could do is announce to a constituent that they have a contrary position on an issue. The elite persuasion perspective would offer the same predictions if politicians did not offer persuasive argumentsjustifying their positions; but, if politicians can argue that their positions are consistent with citizens’ values and predispositions or frame them as such, they might successfully shape public opinion and avoid paying electoral costs for taking positions citizens once opposed. Finally, the position adoption perspective suggests that, ascitizens often trust legislators’ policy judgments, they will not react negatively when political leaders take positions they oppose and may evenadopt their positions, regardless of whether elites justify them.

No one thinks any of these perspectives describes how all voters think about all issues in all circumstances. Nevertheless, the possibility that ‘position adoption’can describe the politics of salient issues is itself controversial (seeBullock 2011for review). And, although a growing body of research is consistent with this notion,it contains potentiallyimportant ambiguities.[1]

One common empirical approach examines how citizens respond to elite position-taking and rhetoricin the real political world, but does not alter aspects of the elite communicationto which citizens are exposed (e.g., Abramowitz 1978; Lenz 2009, 2012; Zaller 1992).[2]While these studies find that citizens often adopt the views of their favored politicians, they leave thenature of this opinion change unclear because elite communication tends to contain a mix of source cues and persuasive messages. To appreciate this ambiguity,consider Abramowitz (1978)’s pathbreakingstudy of the Carter-Ford debates, which found that voters who viewed the debates tended to adopt their favored candidates’ position on unemployment insurance and did not alter their evaluation of their favored candidate based on their prior view on the issue. When Democrats adopted Carter’s position on unemployment insurance after the debate, were they swayed by Carter’s arguments, which presumably appealed to the liberal values to which Democrats tend to subscribe?Or,did Democrats change their opinions simply because they learned of Carter’s positions and trusted his judgment?[3] These processes imply quite different conclusions but are observationally equivalent in panel data often read as supporting the position adoption view (e.g., Lenz 2009, 2012).[4]The din of political disagreements tend to contain a mix of source cues and persuasive messages, thus leaving little direct evidence about how elites lead public opinion when they do.

Helping address this weakness, studies taking a second empirical approach have traded naturalism for greater control over the contents of elite communication. In particular, experiments in lab and survey settings have attempted to assess the processes that condition opinion leadership and electoral accountability to public opinionand have examined a broader set of issues (e.g., Bullock 2011; Cohen 2003; Peterson and Simonovits 2014).[5] However, while theoretically innovative, questions about external validity may remain(Findley et al. 2013; Grose 2014).For example, citizens aware they are being studied may feel compelled to misrepresent their true policy opinions and evaluations of politicians (Bullock et al. 2013;Orne 1962), but may react differently when a researcher is not watching over them at the time of exposure.

We contribute to this body of evidence with unique field experiments that combine the naturalism of field studiesand the control over message content previously only available in lab and survey settings. These experiments, described in the next section, were designed to test the predictions from Table 1:what happens when politicians take positions constituents oppose, and the extent to which extensive justificationscondition any impacts of such position-taking.

Political Context and Design Overview for Studies 1 and 2

We conducted our experiments in collaboration with eight Democratic state legislators from a Midwestern state.We conducted the first experiment in the summer of 2013 with one legislator and the second experiment in collaboration in the spring of 2014with seven legislators.

Table 2. Descriptive Statistics on Legislators and their Districts

Legislator / Terms of Service / 2012 Obama Vote Share / District Description
Study 1
A / 3 / 55% / Rural
Study 2
B / 1 / 90% / Urban
C / 1 / 65% / Urban/Suburban
D / 4 / 75% / Urban/Suburban
E / 2 / 80% / College Town
F / 1 / 50% / Rural
G / 1 / 65% / Urban/Suburban
H / 2 / 50% / Rural

The collaborating legislators all came from the same state, but represented diverse districts (see Table 2). Some of the legislators represented swing districts where Obama won the 2012 election by only a few percentage points, while others represented Democratic strongholds. The legislators were also split between urban, suburban, and rural areas.

Both experiments followed the same basic protocol (although there were important differences in their manipulations we describe later):

1. We worked with each legislator toidentify5-10 salient policy disputes in the legislature that year on which each legislator was actively supporting or opposing a side. We then conducted small statewide pilot surveys using Google Consumer Surveys(McDonald et al. 2012)to identify which four of the legislator’s positions were likely to have the least public support in their district.
2. We surveyed registered voters in each legislator’sdistrict about their positions on the legislator’s four issues and their approval of their legislator. TargetSmart Communications provided the voter lists and Winning Connections conducted the surveys.[6]Specific question wording and coding is given in the Supporting Information.
3. We retained all constituents in the sample who did not agree with the legislator’s position on at least one of these four issues. These votersrepresent the experiment’s sampling frame.These criteria excluded fewer than 5% of respondents.
4. These voters were then assigned to one of the treatments. These treatments differ by study and are described in greater detail below.The keyfeature of both studies is that legislators sent some, randomly-chosen constituents letters that containedthe legislators’ issue positions while others did not receive letters with the legislator’s issue positions.
5. Legislators sent the assigned letters from their legislative offices, on theirofficial letterhead, and in envelopes clearly marked as coming from the state capitol.
6. The week after the letters arrived,we conducted afollow-up survey to appraise the effect of the letters on constituents’ issue positions and evaluationsof the legislator.

This design presents several advantages over existing research. First, because werandomly altered aspects of the communication voters received, we are able to shed light on thetheoretical predictions described in Table 1. However, in contrast toresearch that alters simulated elite communication in the context of survey or lab environments, our field experimental design allows us to examine how citizens react to elites’ actual position-taking in the real world and when they are not aware they are being studied (e.g., Findley et al. 2013; Grose 2014; Loewen et al. 2010).Moreover, theletters also provided a uniquely strong and credible signal about legislators’ positions, as they came directly from the legislators themselves, not from a third party or a political opponent.These experiments thus provide a unique opportunity to examine how constituents react to legislators’ position-taking.

Study 1 – One Legislator: Issue Letter versus No Letter Control

Procedure

We designed experiment 1in collaboration with Legislator A(see Table 2) toexplore whetherpositions legislators take in mailings to constituents can affect constituents’ attitudes. We assigned the constituents in the study to either (1) a treatment group that received a letter in which Legislator A took a position the constituent did not agree with in the pre-survey (including those who did not have an opinion), or (2) a control group that received no letter at all.

The issues Legislator A discussed in his letters related to policies he was working on in the legislature and that he believed were important to his district. For example, Legislator A represents a rural district where mining for a particular mineral is an important industry. One of the issues dealt with whether local or state authorities should regulate mining of that mineral. The other three issues were also of relevance to the state at large: (a) should state monies be spent to expand school vouchers?, (b) should the state income tax be cut?, and (c) should school districts be allowed to raise property taxes?

We surveyed 1,210 voters in Legislator A’s district in May 2013and asked them about their positions on these issues in order to determine who would be included in the experiment’s sampling frame. Of these 1,210 voters, 64 (5%)already agreed with the legislator on all four issues and were removed from the sampling frame. The remaining 1,146 voters were randomly assigned with equal probability to either receive a letter or to not receive any letter, for a total of 573 voters in each group.

Legislator A then sent letters that contained his policy positions to the constituents in the treatment group. Each letter was personalized so that it presented the legislator’s position on up to two issues where constituents did not already agree with his position in the pre-survey.All letters began with a paragraph of biographical information about Legislator A, then up to two paragraphs with Legislator A’s position on the policy issues, followed by a closing paragraph. Supplementary Appendix B presents anonymized versions of these letters.

We also randomized which of the issues the letters discussed, allowing us to use both across-subject and within-subject randomly assigned variation for the opinion leadership analysis.[7]This necessitates two steps to assure unbiased inference (Gerber and Green 2012): we cluster standard errors at the respondent level and we condition on strata indicators for the number of issues on which each voter was persuadable.