Does issue importance attenuate partisan cue-taking?

Abstract

Are cues from party leaders so important that they can cause individuals to change their own issue positions to align with the party's position? Recent work on the importance of party cues suggests they do, especially given the literature on partisanship as a strong and persistent group identity. However, in this paper we test the limits of those partisan cues. Using a unique two-wave panel survey design we find that the effect of party cues is moderated by the prior level of importance individuals place on an issue. We find that when a person believes an issue area to be more important, party cues are less likely to move that citizen's position, particularly when the cue goes against partisan ideological norms. Our results show evidence that an individual's own issue positions—at least the important ones—can be resilient in the face of party cues.

While the correlation between partisanship and voters’ issue positions is well-established, there is significant debate surrounding the direction of the influence an individual's policy views have on the formation of a partisan attachment, and vice versa. Do the effects of socialization and group identity (Campbell et al., 1960; Green et al., 2004; Huddy et al., 2015; Achen and Bartels, 2016; Mason, 2018) overwhelm the importance of issues (Ansolabehere et al., 2008; Hillygus and Shields, 2008; Costa, forthcoming; Orr and Huber, 2020)? In other words, which is paramount, loyalty to party or commitments to issue positions?

Recently, Barber and Pope (2019) showed that people's partisan affect leads them to move—in both a liberal or conservative direction—when shown a policy cue from the president (and though it was more strongly demonstrated for Trump, they did find an effect among Democrats related to Obama). This party loyalist behavior came despite the ideological direction of the cue and occurred for both strong partisans and strong ideologues. Furthermore, this finding is one among many. Cohen (2003) had earlier showed the power of group Identity and people's “blindness” to that fact. Hill and Huber (2019) show that when people receive knowledge of the partisan composition of a Congressional roll call vote they change their own issue positions to better align with their party's elected officials. Lenz (2012) also shows that citizen attitudes are powerfully shaped by signals from party leaders.¹ However, these studies leave open the question of whether partisans’ issue positions on all issues are moveable by the party leader's current set of positions. Are there limitations to this powerful effect of partisanship?

There are reasons to believe cues can sometimes have little impact. Mullinix (2015) tested a combination of issue importance and polarization and found that the public is not blindly attentive to partisan cues on tax and education policy. Agadjanian (2020) shows that conflicting intra-party information also attenuated the impact of partisan cues. In this paper we take a broad cross-section of issues to see whether or not issue importance might attenuate partisan cue taking. In other words, is partisan cue-taking impacted by the degree to which a person cares about a particular issue? Building on the design of Barber and Pope (2019) we show that survey respondents adjust their issue positions to align with cues from President Trump. However, we find that respondents are less likely to do so on issues the respondent considers personally important when the cue is in the ideologically “wrong” direction (i.e. a liberal policy cue among Republicans). On the other hand, we find that when the ideological cue from President Trump aligns with the traditional party line (i.e. a conservative cue among Republicans) issue importance further amplifies the power of the ideological cue. In other words, when people indicate they care a great deal about an issue, they resist moving in the ideologically “wrong” direction, though they are quicker to move in the ideologically “correct” direction. These results show that while people are inclined to “follow the leader” of their party, issues — and issue importance — play an important role in the degree to which people do so.

Partisan Cues & Issue Importance

The idea that cues influence people's opinions extends back to Asch (1952) who showed that the source of the message changed how individuals interpreted the message and any accompanying information. More recently, Barber and Pope (2019) showed that a party leader's issue positions actually moved partisans in both a liberal or a conservative direction on various policies, depending on the ideological direction of the cue the partisans received. However, they did not directly test issue importance, which has received relatively less attention in the literature, despite being an important factor that may impact the effect of a cue on a person's policy views.

At the same time, prior research has shown that issue importance matters in a number of other contexts.² For example, public opinion research shows many members of the public care about at least a few issues (Converse, 1964); Petrocik, 1996) and that parties tend to “own” those issues in the public's mind (Egan, 2013). Krosnick (1988) finds a strong effect of issue importance for attitude stability over time, and Jacks and Devine (2000) show that people resist persuasion on issues that are important to them because people are able to more quickly bring to mind thoughts and feelings that help them defend their attitude against persuasion.³ Mummolo et al. (2019) note that people are only “conditionally” loyal to parties in the sense that important issues can drive a voter away from the normal partisan choice. Furthermore, research shows that during political campaigns, people choose to focus on partisan campaign material related to the issues they cared most about (Iyengar et al., 2008; Henderson, 2013), perhaps muting or amplifying the effects of new information. With these results dependent on heterogeneity in mind, we use a unique two-wave approach to first gather data on the importance respondents placed on a host of issues before introducing a partisan cue on that same issue at a later point in time.

Empirical Strategy

The data come from the 2018 Cooperative Congressional Election Study (CCES), where we conducted a two-wave panel survey in the weeks before the November general election and then immediately after the election. In the first wave of the study we measured issue importance across 20 different issues among 978 people.⁴

In the second wave of the study we asked respondents their position on the same issues that appeared on the first wave of the survey in a series of support (1), oppose (-1), don't know (0) questions.⁵ For each policy question, it was randomly determined whether the question would also include a statement about President Trump's position on the issue. The randomization took place at the respondent-issue level, and therefore each respondent was asked their views on the 20 policy issues where a random sample of those questions contained no presidential cue (6,763 respondent-issues), a cue indicating that President Trump had taken a liberal position on the policy (6,814 respondent-issues), or a cue indicating that President Trump had taken a conservative position on the policy (6,922 respondent-issues).⁶ The first wave took place several weeks before the second wave of the survey where we administered the treatment and recorded the outcome variable. This period of time between survey waves was crucial because it reduces the possibility that the questions of issue importance in wave 1 biased responses in wave 2 by priming particular issues in the minds of respondents.⁷

This experimental design in wave 2 is a near-replication of Barber and Pope's (2019) work where they presented respondents with similar cues from Donald Trump in both a liberal and conservative direction. The design has the particular virtue of, “Trump's non-ideological and ever-changing issue positions …allows us the unique opportunity to identify moments when issue content and party are in conflict. And this divergence allows us to identify which of these attachments appears to be more important in the minds of the typical citizen” (Barber and Pope, 2019, pg. 43). While we replicate the question design and use of Donald Trump as the cue giver, we use a number of additional issues beyond those that were included in the Barber and Pope (2019) study.⁸

An important virtue of this design is that by asking issue importance for each issue among each respondent, we have variation in the importance of issues by individual and not merely a measure of which issues are salient to the overall public. With this design, our unit of analysis is the individual-issue dyad, allowing for variation across and within individuals, which provides a great deal of statistical power while also allowing for estimating both between-subjects and within-subject effects. This design sets up a very simple test: as issue importance rises for each individual-issue dyad does the effect of an ideological cue change? If it does, then issue importance acts to attenuate/amplify the effect of elite cues. This is an important empirical improvement over previous designs which do not test issue importance for individual respondents and issues.

Results

The main findings of Barber and Pope (2019) replicate (see Table A2 of the supplemental materials), allowing us to focus on the significant variation in how much people respond to the cues based on individual levels of issue importance shown in Figure 1. A decomposition of that variance shows that approximately 70% of the variation in issue importance is due to within-subject differences in importance with the remaining 30% due to variation across respondents, establishing clear differences in how important people think issues are—even if they lean towards calling most issues important (48% of respondent-issue dyads were rated as ‘very important’, 30% ‘important’, 14% ‘neutral’, 5% ‘not that important’, and 3% ‘not at all important’).⁹ As a result, studies of issue importance and partisan cue taking that use aggregate measures of issue salience will not account for the significant variation across issues, within individuals. An individual-level measure of importance provides a much cleaner test of the theory and permits a within-subject design that accounts for respondent-specific features when conducting our analysis.

Figure 1. Aggregate Levels of Issue Importance. These average values are from questions asked of respondents in the first wave of the two-wave survey.

Figure 2 presents the results of a regression model that test whether partisan cues are impacted by higher levels of issue importance.¹⁰ In each model the dependent variable is measure of support for the liberal position on each policy question (1 = support, 0 = don't know, −1 = oppose).¹¹ The left “Liberal Trump” and right “Conservative Trump” panels indicate the ideological direction of the Trump issue cue treatment, with the comparison group being the control group that saw no cue. The issue importance variable is the five-point measure of issue importance with 5 being the highest level of importance. The ideological cue (Trump treatment) is then interacted with the measure of issue importance to test our hypothesis that increased issue importance will impact the strength of the Trump cue. The y-axis of Figure 2 shows the estimated impact of the Trump cue. Larger values indicate movement by respondents in a liberal direction while negative values indicate movement by respondents in a conservative direction (compared to respondents’ opinions when no Trump cue is presented).¹²

Figure 2. Average Treatment Effect Across Issue Importance. Note: The left panel shows the liberal Trump treatment effect and the right panel shows the conservative Trump treatment effect across different levels of issue importance for all respondents.

When President Trump signals that he has a liberal position on an issue (left panel), respondents follow the direction of cue and are more likely to provide a liberal response. However, people who care the most about an issue are the least likely to be moved as evidenced by the downward slope of the five points in the figure. The relationship is less clear when the cue from Trump is in a conservative direction (right panel), where the conservative cue moves all respondents in a conservative direction, but there is little impact across levels of issue importance.

While Figure 2 shows the treatment effects for all respondents, we have reason to believe that the impact of an ideological cue from President Donald Trump will be most meaningful among members of the Republican Party, as Barber and Pope (2019) show. Figure 3 shows the estimated average treatment effects for both the liberal and conservative Trump cues interacted with issue importance, but only among self-identified Republican respondents.¹³

Figure 3. Average Treatment Effect Across Issue Importance – Republican Respondents Only. Note: The left panel shows the liberal Trump treatment effect and the right panel shows the conservative Trump treatment effect across different levels of issue importance for Republican respondents.

The results in the left panel (the liberal trump Treatment) are similar to those discussed above in the pooled results, however, as hypothesized, the overall treatment effect is larger. Nevertheless, issue importance still dampens the power of the ideological cue. The estimated impact of the liberal cue for Republicans who care ‘not at all’ about the policy is 0.32 ([0.17, 0.46], 95% C.I.) while the impact of the liberal Trump cue for people who care ‘very much’ is only 0.22 ([0.13, 0.31], 95% C.I.). Thus, issue importance attenuates the power of the liberal Trump cue by approximately 30 percent.¹⁴

For the conservative treatment, however, we find that among Republicans issue importance has the opposite impact. Instead of dampening the strength of the conservative Trump cue, we find that those who indicate that they care ‘very much’ about the policy are the most likely to move in a conservative direction. The estimated impact of the conservative cue for Republicans who care ‘not at all’ about the policy is −0.06 ([−0.20, 0.08], 95% C.I.) while the impact of the conservative Trump cue for Republicans who care ‘very much’ is −0.18 ([−0.27, −0.10], 95% C.I.).¹⁵ Thus, the impact of issue importance in this case is to amplify the power of the conservative Trump cue by nearly 200%.

Figure 3 shows that the impact of caring about the issues works in different directions depending on the ideological direction of the cue presented by the party leader. When the party cue is in the counter-stereotypical ideological direction (i.e. a liberal cue from a Republican president), issue importance appears to work against the cue. However, when the cue is in the stereotypical ideological direction, issue importance appears to amplify the impact of the cue. These results, while novel, are consistent with other studies that show people's willingness to embrace information that aligns with their prior beliefs and resist information that runs counter to those priors (Zaller and Feldman, 1992). Furthermore, they conform with those who articulate the difference between issue importance versus opinion stability (Miller and Peterson, 2004), where here we see that importance increases opinion movement when the cue is pushing in the “correct” ideological direction.

Discussion

This research provides three important findings. First, we have replicated the original results of Barber and Pope (2019), confirming clear cue effects on policy views. Second, and more importantly, we find clear support for the idea that issue importance acts as a moderator to ideological cues. Finally, there is evidence that this result is asymmetric across the ideological direction of the cue. Issues that are subjectively more important to a respondent are less likely to be affected by a cue when the cue is counter to the traditional ideological direction of the party (i.e. liberal policies among Republicans). However, issue importance amplifies the cue when the cue aligns with the stereotypical ideological direction of the cue giver's party. This evidence, however, is somewhat preliminary because of the asymmetry of the party of the cue giver in our experiment. Observing the Democratic equivalent to Donald Trump would certainly strengthen this hypothesis and allow for a more complete test of the theory. However, the results shown here are still strong evidence that issue importance does matter.

There are two caveats to add to these conclusions. First, it is possible that there is heterogeneity in the clarity of the cue for people. Trump may be clear to one audience and less clear to a different audience—indeed this seems likely in real life. However, in an experiment is less of a concern. Most people probably received the treatment cue and simply accepted it at face value given their information set. Still, clarity and variability in the cue may be grounds for future research.

The second caveat is more difficult in that it involves a fairly deep identification problem. Is this effect about Trump or about partisanship? Frankly, without further work it is difficult to tell. And, even if that work were performed it might not settle the matter. Elected officials always carry partisan cues immediately with their reputation. There is no clean way to separated a cued official from a known party label—whether that is a president, or a member of Congress, or anyone who has a partisan label. He or she will always have that with them. Of course it is possible to simply cue the party label, but then it is not clear if the office is involved or not. We hope future work will use more cases to help distinguish which party officials have the strongest effects, but it is possible that a clean solution is not possible. What we remain confident about is that partisan presidents do have an effect, but that this effect is moderated by issue importance.

Among Republicans, the results for the interaction of issue importance with the Trump cue are large and were statistically significant across a wide variety of model specifications. In the case of a liberal cue, issue importance reduced the cue's impact by roughly one third the size of the main effect. In the case of a conservative Trump cue, issue importance increased the size of the cue's impact by more than double—strongly suggesting that issue importance influences party cue effects.

While the overall experiment shows that party leader cues have great power in shaping people's views, it is not the case that party leaders are entirely unconstrained in their ability to shape public perceptions. There are limits to what they can accomplish because the underlying beliefs of citizens play an important and independent moderating role. Citizens will only be pushed so far, in certain directions, and only for certain issues. This also suggests that issues and issue importance are in some ways less well-understood components of partisanship. Though, on average, partisanship may be an identity with relatively little issue content, there are likely circumstances under which people's partisanship is deeply connected to their issue positions. This provides clear support for the idea that issue importance should be considered in the study of ideology and partisanship.