Communication is central to political representation. Constituents frequently encounter messages from their elected officials through debates, advertisements, and campaign events. One prominent concern is that this provides politicians with an avenue to distort the electoral connection. By selectively targeting subsets of their constituents with messages that contain “crafted talk” or “tailored explanations,” politicians can potentially explain away poor performance or justify incongruent position taking that would otherwise reduce their support (e.g., Franz, Reference Franz and Ridout2013; Grose et al., Reference Grose, Malhotra and Van Houweling2015; Hersh and Schaffner, Reference Hersh and Schaffner2013; Jacobs and Shapiro, Reference Jacobs and Shapiro2000; Mayhew, 1974; McGraw et al., Reference McGraw, Timpone and Bruck1993).
We extend this line of research by examining the implications of one variety of elite messages—a politician’s justification for their stance on tax policy—for one notion of democratic accountability—the penalty candidates suffer for taking issue positions that differ from the preferences of their constituents. Drawing on political communication research, we develop a theory about the heterogeneous effects of justifications that employ either evidence or values based on whether or not constituents agree with the position a politician takes. We examine the effects of these justifications on public opinion using survey experiments and report three main findings.
First, while we observe heterogeneous responses to these justifications in a manner consistent with our theory, the magnitude of these effects is small and not substantively significant. Second, despite strong justification treatments and an environment that is otherwise favorable for elite influence on public opinion, we rule out large benefits for politicians to offer justifications relative to simply stating their stance on the issue (see also Broockman and Butler, Reference Broockman and Butler2017). Finally, we highlight a potential cost to selective messaging by showing that when these targeted messages reach unintended audiences they can backfire and reduce a candidate’s public support (see also Hersh and Schaffner, Reference Hersh and Schaffner2013). These results demonstrate limitations on the ability of politicians to generate additional leeway for incongruent position taking through messaging, at least on highly salient issues like the one considered here.
JUSTIFYING POLICY POSITIONS
When politicians stake out an issue position they are also afforded the opportunity to explain or justify their actions (e.g., Fenno, Reference Fenno1978, Kingdon, Reference Kingdon1973; Mayhew, 1974).Footnote 1 These justifications contain features such as compensatory rhetoric that provides constituents with a broader view of the legislator’s actions on similar issues (Grose et al., Reference Grose, Malhotra and Van Houweling2015), “crafted talk” to change constituent opinion on the policy under consideration (Jacobs and Shapiro, Reference Jacobs and Shapiro2000), or an explanation for holding a position that portrays the legislator as principled or reasonable to an audience that disagrees with their actions (Fenno, Reference Fenno1978; Mayhew, 1974).
Targeting justifications toward particular groups potentially allows politicians to reduce the costs of incongruent position taking. Using sources such as consumer data (Hersh, Reference Hersh2015; Hillygus and Shields, Reference Hillygus and Shields2008) or the letters they receive from constituents (Grose et al., Reference Grose, Malhotra and Van Houweling2015), elected officials can target messages based on group membership (e.g., gun ownership, religious affiliation; Schaffner and Hersh, Reference Hersh and Schaffner2013), geographic location (Franz, Reference Franz and Ridout2013), or the expressed policy preferences of the intended recipients (Grose et al., Reference Grose, Malhotra and Van Houweling2015). This approach allows politicians to tailor their message in order to portray their actions in the most favorable light for a particular group.
If they reach their intended recipients, these targeted messaging can improve the public’s view of a legislator (Hersh and Schaffner, Reference Hersh and Schaffner2013; McGraw et al., Reference McGraw, Timpone and Bruck1993). Closest to our approach, Grose et al. (Reference Grose, Malhotra and Van Houweling2015) use survey experiments to show that politicians can increase their support by emphasizing similar position taking on immigration issues to constituents that agree with their position and different, cross-cutting votes to those that disagree with their stance. While Grose et al. (Reference Grose, Malhotra and Van Houweling2015) identify the benefits of emphasizing information outside the scope of a particular issue to some constituents, it is unclear how this process proceeds on highly salient issues, where legislators must confine their justification to the issue at hand rather than changing the topic. In the next section, we draw on political communication research to establish the avenues of response open to legislators to justify their vote choice on a highly salient issue, income taxation of the wealthy.
Evidence and Values-based Arguments
Political communication scholars have identified several types of policy arguments employed in areas, like taxation, where both technical and symbolic considerations are applicable in a policy debate. One type—evidence-based arguments—make falsifiable predictions about the consequences of a policy and commonly refer to empirical evidence (Jerit, Reference Jerit2009; Lau et al., Reference Lau, Smith and Fiske1991). A second variety—values-based arguments—involve statements about the normative desirability (e.g., fairness) of a policy (Brewer, Reference Brewer2001; Nelson and Garst, Reference Nelson and Garst2005; Nelson et al., Reference Nelson, Clawson and Oxley1997).
Several recent studies compare the effectiveness of these types of arguments. Druckman and Bolsen (Reference Druckman and Bolsen2011) study the consequences of evidence and values-based arguments over new energy technologies and find that they are equally effective in shifting public support (see also Binning et al., Reference Binning, Brick, Cohen and Sherman2015). Marietta (Reference Marietta2012) compares these two types of arguments for issues such as gun control and also finds that neither type is innately more effective than the other in changing policy opinions, although they do have different consequences for evaluations of the speaker providing the argument (Marietta, Reference Marietta2009) and the strength of the attitudes that individuals eventually form (Marietta, Reference Marietta2008). We depart from this earlier research by focusing on the heterogeneous effects of values and evidence-based justifications for taking a particular policy stance on the evaluations a politician receives from different audiences.
Audience Opinion as a Moderator
We hypothesize that justifications employing either evidence or values will have heterogeneous effects based on the whether or not an audience agrees with the issue position taken by a politician. This claim builds on two distinct literatures in political psychology. First, prior research finds that evidence and values-based justifications place different information processing requirements on citizens. Evaluating the evidence that a legislator provides to buttress their position is cognitively intensive and requires an understanding of the nature of the prediction and its relevance to the current policy issue (e.g., Druckman and Bolsen, Reference Druckman and Bolsen2011). In contrast, values are more readily interpretable and allow individuals to quickly understand complex political settings (e.g., Goren, Reference Goren2012). This leads to the expectation that, relative to evidence-based justifications, values-based justifications will place an individual’s own position on the issue in sharper contrast with the politician they evaluate and lead elected officials to suffer more among those who disagree with their issue stance and gain additional support from those who agree with them.
Second, research on the psychology of self-affirmation (e.g., Binning et al., Reference Binning, Brick, Cohen and Sherman2015; Sherman and Cohen, Reference Sherman and Cohen2006; Steele, Reference Steele1988) stresses the critical role that maintaining a positive self-view has on attitudes in a variety of situations. For example, Minson and Monin (Reference Minson and Monin2012) find that survey respondents react in a more negative fashion to groups they disagree with when those out-groups emphasize their moral superiority. Such a dynamic is possible in the current setting because values-based justifications make normative assessments that might contradict a citizen’s own view that they have taken the “right” stance on the issue (see also Brewer, Reference Brewer2001; McGraw, Reference McGraw1998, 132).
Both of these account imply that evidence-based justifications have a smaller downside than values-based arguments as they indicate that the representative has simply come to a different understanding of the potential consequences of a proposal and do not highlight any values-based differences between the representative and their constituent. However, evidence-based justifications fail to affirm the values of those who support a legislator’s stance on an issue and so may increase support among this group to a smaller extent than a values-based justification. These accounts lead to two hypotheses as follows:
H1: Evidence-based justifications should generate increased support for a politician among individuals who disagree with a politician’s stance on an issue relative to values-based justifications.
H2: Values-based justifications should generate increased support among individuals who agree with a politician’s stance on an issue relative to evidence-based justifications.
RESEARCH DESIGN
We focus on the use of these types justifications in the debate over the appropriate tax rate for wealthy individuals. We have three reasons for this focus. First, the issue is highly salient and “easy” in that it has recently been at the center of political debate and contains symbolic features (Carmines and Stimson, Reference Carmines and Stimson1980). Assessing justification effects in this edge case—where individuals are likely to hold crystallized opinions—is an important step in determining the degree to which politicians can generate leeway using communication. Second, this policy highlights one area of disconnect between constituents and representatives. While 60% of the public supports increasing taxes on the wealthy, this policy has yet to gain traction in Congress (e.g., Pew, Reference Pew2012). Third, tax policy represents an area where both evidence- and values-based justifications are frequently offered by politicians in their public statements, ensuring our treatments maintain face validity (Gale and Slemrod, Reference Gale and Slemrod2001; Gordon and Miller, Reference Gordon and Miller2004).
We examine the effects of policy justifications using two survey experiments. The first experiment, referred to as Study 1 in what follows, was implemented on a nationally representative sample of respondents from GfK (formerly Knowledge Networks) through Time Sharing Experiments in the Social Sciences’ Short Studies Program. A follow-up, Study 2, included additional experimental treatments and used respondents recruited through Amazon’s Mechanical Turk.Footnote 2
Study 1 employed a 2×2 factorial design. Respondents read one of four candidate justifications for taking a particular position on the issue of tax policy for wealthy individuals. Half of the participants read about a candidate supporting a tax increase on the wealthy. The other half read about a candidate opposing the increase. Each candidate offered either an evidence-based or values-based justification for their position. The two arms of the experiment were independent, which created four conditions (i.e., Pro-Evidence, Pro-Values, Con-Evidence and Con-Values).
Each justification followed a similar format. First, the politician highlighted the particular dimension they would use to justify their policy stance. Next, they established their position and justified their stance. They then offered either evidence for the evidence-based justifications or an elaboration of the way in which values were applicable to the issue for the values-based justifications. Finally, the politician summarized their position by stating the policy was either the “right” (“wrong”) thing to do for values-based justifications or the “efficient” (“inefficient”) thing to do for evidence-based justifications.Footnote 3 After reading this justification, respondents evaluated the candidate on a 100-point support scale. Respondents provided their own tax policy opinion either before or after they read about the candidate (the order of the two questions was randomized).Footnote 4
Study 1 allows an initial examination of our theory on a nationally representative sample of respondents. However, length requirements prevented the inclusion of some follow-up questions to examine additional outcomes that justifications might affect.Footnote 5 Additionally, previous work in political communication (e.g., examinations of the effects of partisan cues on policy support, Nicholson Reference Nicholson2012) establishes that it is important to not only compare two different types of communications to each other, but to also compare these messages to a neutral condition in which no message is provided. To address these issues, we conducted a follow-up study on Amazon’s Mechanical Turk.
Study 2 contained six conditions. In each condition, respondents first offered their own opinion on the tax policy issue under consideration and then read about a member of Congress who took a position either in favor or in opposition to tax increases on the wealthy. The candidate then provided an evidence-based justification, a values-based justification or did not offer a justification for their position. After reading the justifications, respondents evaluated their likelihood of voting for the candidate in the next election on a 100-point scale and answered a battery of post-treatment questions.
While we focus on issue-based determinants of candidate support, an array of prior studies establish that candidate partisanship serves as a “pre-eminent electoral cue” (Basinger and Lavine Reference Basinger and Lavine2005, 171) and exerts a powerful influence on vote choice (e.g., Sniderman and Stiglitz, Reference Sniderman and Stiglitz2012). Given that our hypotheses concern the effect of justifications based on the issue alignment between a respondent and the politician they evaluated and that we lack clear expectations about how partisan cues would moderate the effect of justifications, we designed the study to ensure that co-partisanship was balanced across experimental conditions by randomizing the party of the politician that individuals evaluated (see Gerber et al., Reference Gerber, Huber, Doherty and Dowling2011 for another example of this approach).
RESULTS
Our hypotheses lead us to expect that the relative effectiveness of the evidence-based and values-based policy justifications will vary based on the alignment between the issue position advocated by the candidate and a respondent’s own issue position. In what follows, we refer to a respondent as congruent with a candidate if the two take the same tax policy position (e.g., if both the candidate and the respondent supported increasing taxes on the wealthy) and incongruent if they take opposing positions. Because our hypotheses are conditional on the congruence of the respondent and the evaluated candidate, we drop respondents who did not express an opinion about tax policy (i.e., respondents locating themselves at the middle of the scale or who did not answer the question at all) as well as those with missing data on the dependent variable.Footnote 6 Table 1 reports cell sizes for the analysis by experimental condition and issue alignment.
Table 1 Sample Sizes
Note: Entries are cell sizes by experimental condition and values of respondents attitudes about tax increases. The table excludes respondents with missing data on the dependent variable and/or the tax policy.
Table 2 displays the effect of these justifications on candidate support. The top panel of the table presents mean candidate support across the types of justifications separately for respondents taking congruent and incongruent policy positions relative to the candidate they evaluated. Comparing the values contained in each column of the table shows a consistent pattern across the two studies. Values-based justifications produced greater candidate support among respondents who were congruent with the candidate they evaluated. In contrast, evidence-based justifications performed better among incongruent respondents. While the signs of the between-justification comparisons (simple difference in means tests) reported in the bottom panel of Table 2 are consistent with our hypotheses, the magnitudes of these differences are small. The largest between-justification difference, for incongruent respondents in Study 2, is slightly less than four points on a 0–100 scale.Footnote 7
Table 2 Politician Support by Experimental Condition
Note: Entries are means of our dependent variable, scaled to 1100. Standard errors are in parentheses. Data from Study 1 are weighted using post-stratification weights provided by Knowledge Networks.
An additional research question is whether providing any justification is helpful in this setting, particularly given the limited difference in candidate evaluations between the two types of justifications. The bottom two rows of the table use data from the second study and display differences in politician support between each justification type and the condition in which the politician stated their position without offering a justification. These comparisons show that employing either type of justification failed to generate additional support beyond what a politician received when simply stating their position on the issue. Moreover, justifications appeared harmful when they reached a hostile audience. Candidates offering a values-based explanation for their position to incongruent respondents—a mis-targeted justification based on our theoretical expectations—received ratings that were over six points lower than offering no justification at all.
DISCUSSION
The above results indicate that politicians have a relatively limited ability to generate leeway for incongruent position taking on this issue using targeted justifications. A potential concern with these findings is that they reflect (1) weak justification treatments or (2) respondent inattention to the content of the justifications. We engage with each of these concerns in turn.
One concern is that these findings stem from weak justifications treatments. Several aspects of our treatments help to allay this concern. First, in a pre-test conducted on Mechanical Turk to select the justifications used in the experiment, the core features of the arguments employed here were largely evaluated as effective by respondents and exceeded the evaluations of frames characterized as “weak” in previous studies (e.g., Druckman et al., Reference Druckman, Peterson and Slothuus2013).Footnote 8 Second, the justifications in this experiment were lengthy and written with the intent of constituting a high-dose on public opinion, especially given the lack of other information about the elected official available to respondents. Merely on the basis of a “length equals strength” heuristic (e.g., Cobb and Kuklinski, Reference Cobb and Kuklinski1997, 95) these evaluations are more powerful than typical framing treatments that have produced effects on less salient issues. Finally, these justifications were closely modeled off of those regularly employed by politicians. To the extent these treatments are weak, they approximate the inability of politicians to develop particularly strong messages on this issue and capture respondent perceptions of the messages they would encounter in a policy debate.
A second concern is that respondents did not perceive the two types of justifications as sufficiently distinct, perhaps due to inattention to the content of the justifications. To address this concern, we asked respondents in Study 2 to evaluate the degree to which the argument they read relied on evidence. As intended, respondents clearly perceived the evidence-based justification as relying more on evidence than the values-based explanations (see Column 2 of Table B1 in online appendix).
CONCLUSION
Academics and pundits frequently raise concerns about the potential for targeted messaging to reduce political accountability (e.g., Franz, Reference Franz and Ridout2013; Grose et al., Reference Grose, Malhotra and Van Houweling2015; Hersh and Schaffner, Reference Hersh and Schaffner2013; Jacobs and Shapiro, Reference Jacobs and Shapiro2000). In this study, we find limited support for this claim by examining the effects of evidence and values-based messages on candidate support. These findings sit closely with several other recent studies that highlight the limits of targeted messaging strategies.
Like Bechtel et al. (Reference Bechtel, Hainmueller, Hangartner and Helbling2015), we examine messaging effects on a salient, contested issue and find little evidence that elites messages can produce sizable shifts in opinion. This serves to highlight an additional issue on which the potential for elite manipulation of public opinion appears limited.Footnote 9 In line with Broockman and Butler (Reference Broockman and Butler2017), we find that justifications offer a limited boost to politician support relative to a politician simply taking a position without further elaboration. Consistent with Hersh and Schaffner (Reference Hersh and Schaffner2013), we identify limited benefits to a targeted messaging strategy and show that mis-targeted messages introduce potential costs into the communication process. Here, this occurs when individuals respond negatively to values-laden messages that advocate for an issue position they oppose (see also Brewer, Reference Brewer2001).
These findings stand in contrast to other studies that emphasize the success of targeted messaging strategies (e.g., Grose et al., Reference Grose, Malhotra and Van Houweling2015; Jacobs and Shapiro, Reference Jacobs and Shapiro2000; McGraw et al., Reference McGraw, Timpone and Bruck1993) and highlight a limit to the effectiveness of these approaches on a particularly salient “easy” issue. While this shows that voters are resistant to elite manipulation on at least some issues, future research is needed to identify whether the resistance to targeting messaging strategies demonstrated here, or the more malleable public portrayed in several earlier studies, more closely represents public responses to these strategies in general.
APPENDIX A: TREATMENTS
Table A1 Justification Text
