Conformity to implicit social pressure: the role of political identity

Abstract

Although studies have shown that implicit social cues, such as images of watchful eyes, can elicit prosocial behavior, little research to date has examined individual differences in people’s susceptibility to such subtle social cues. For example, individuals with a conservative ideology typically value social conformity, obedience, and adherence to social norms more than liberals. To examine partisan heterogeneity, we analyze data from two large randomized field experiments on voting behavior. Results suggest that the impact of eyespots on voter mobilization is indeed likely driven by political identity, with a significant effect for Republicans but not Independents or Democrats. These findings are consistent with an emerging line of research revealing individual differences in how susceptible humans are to implicit social cues.

Keywords:

• Prosocial behavior
• eye images
• implicit social cues
• social norm
• field experiment

Implicit social cues, such as images of ‘watchful eyes’ have shown to promote prosocial behavior and cooperation in a variety of public good games (e.g., Burnham & Hare, 2007; Haley & Fessler, 2005) as well in many real-life social dilemmas, including mobilizing citizens to vote (Panagopoulos, 2014a, 2014b), increasing charitable giving (Bateson, Nettle, & Roberts, 2006; Powell, Roberts, & Nettle, 2012), and reducing littering and theft (Ernest-Jones, Nettle, & Bateson, 2011; Nettle, Nott, & Bateson, 2012). At the same time, the nature of the ‘watchful eye’ effect has become a source of active debate, as several recent studies have failed to replicate the effect (Carbon & Hesslinger, 2011; Fehr & Schneider, 2010; Matland & Murray, 2015; Northover, Pedersen, Cohen, & Andrews, in press) while other scholars argue that the presence of watchful eyes can enhance altruistic behavior but point toward boundary conditions, such as in-group vs. out-group membership (Mifune, Hashimoto, & Yamagishi, 2010), familiarity with the stimulus (Sparks & Barclay, 2013), limited effects in certain settings (Ekström, 2012), the role of contextual cues, such as lighting (Tane & Takezawa, 2011) and directed vs. averted gaze (Manesi, van Lange, & Pollet, 2016). Yet, even small effects can be impressive, however, especially when the results have practical significance (Panagopoulos, 2015) and when the required manipulation is minimal or when the behavior is typically difficult to influence (Prentice & Miller, 1992). Moreover, we argue that with some exceptions (e.g., Keller & Pfattheicher, 2011), examining variation in individual differences has been neglected in prior research and could help explain some of the observed inconsistencies in the efficacy of eyespot interventions. In general, watchful eyes are thought to be effective because people care about their public image and reputation (Bateson et al., 2006; Oda, Niwa, Honma, & Hiraishi, 2011), especially when there is a social cost associated with deviating from a norm (Cialdini & Goldstein, 2004). In fact, people are often willing to punish those who violate social norms (Fehr & Fischbacher, 2004) and the (implicit) threat of punishment has been a major force in maintaining cooperation in human societies (Fehr & Gächter, 2002; Richerson et al., 2014). Moreover, humans are particularly dependent on eye cues when it comes to social decision-making (Haxby, Hoffman, & Gobbini, 2002; Nummenmaa & Calder, 2009) and images that merely resemble human eyes can trigger involuntary neural responses to gaze, which often motivate subsequent prosocial conduct (Burnham & Hare, 2007). Although images of human eyes do not actually alter the level of anonymity, it is hypothesized that these brain structures evolved in settings where eye images came only from actual eyes and thus were associated with public behavior (Burnham, 2003; Burnham & Hare, 2007; Haley & Fessler, 2005; Rigdon, Ishii, Watabe, & Kitayama, 2009).

Nonetheless, it is reasonable to assume that not everyone is affected equally by (implicit) social pressure and some heterogeneity in response to eyespot treatments should be expected. For example, teenagers are generally more susceptible to social influence than adults (Sears, 1986). Similarly, people high in need for social approval are typically more susceptible to observer effects (van Rompay, Vonk, & Fransen, 2009). Yet, one particularly relevant and underexplored individual difference variable in the context of the watchful eye effect is the role of political ideology. Research has found that conservatives typically value social conformity more than liberals (Altemeyer, 1996; Hibbing, Smith, & Alford, 2013; Jost, Glaser, Kruglanski, & Sulloway, 2003). Conservatism has also systematically been associated with a personality trait known as conscientiousness (Carney, Jost, Gosling, & Potter, 2008), which, in turn, is associated with a desire to conform to social norms (Gerber, Huber, Doherty, & Dowling, 2011). Furthermore, whereas liberals tend to score higher on measures of rebelliousness and reactance, conservatives tend to score higher on indicators of conformity and obedience (Jost, Nosek, & Gosling, 2008).

Thus, because scholars have argued that social norms are often a stronger determinant of behavior for conservatives than for liberals (Fernandes & Mandel, 2014; Skitka & Tetlock, 1993), we set out test this hypothesis in the context of the watchful eye effect by analyzing data from two field experiments on voting behavior (Panagopoulos, 2014a, 2014b). Voting behavior is an ideal case study, as social influence mechanisms have proven to play an important role in voter mobilization (Davenport et al., 2010; Gerber, Green, & Larimer, 2008; Panagopoulos, 2010, 2013; Panagopoulos, Larimer, & Condon, 2014). In short, we hypothesize that while implicit social cues, such as watchful eyes, can promote prosocial behavior, the effect is likely conditional on how sensitive an individual is to subtle social cues in a particular decision context. In other words, across two field experiments, we expect a stronger effect of eyespot images on voter mobilization for Republican/conservative compared to Democratic/liberal voters.

Method

To evaluate our hypotheses, we analyze data from two different field experiments testing eyespot effects on voting in elections. In both studies, the units of observation are registered voters, and the outcome variable is individual voter turnout in the election, verified using official public records. The first experiment was conducted in Key West, Florida, during the 4 October 2011 mayoral election. The sample comprised the complete universe of registered voters residing in single-voter households in Key West (N = 13,452). The restriction to single-voter households in both experiments was done to minimize interference between units. Voters were randomly assigned to either a control group or to one of the three treatment conditions. Voters assigned to a treatment condition were sent one of the three versions of a postcard mailing within the week prior to the election. The text on the postcards was identical between the conditions (‘Do your civic duty and vote!’), but the image was varied randomly. Voters either received an image of a political (American flag), a nonpolitical (palm tree), or an implicit social pressure cue (eyes).

The second experiment, designed as a replication and extension, was conducted in Lexington, Kentucky, during the 8 November 2011 gubernatorial election. The complete experimental sample consisted of N = 71,593 registered voters residing in single-voter households in Lexington, KY. The design was identical to Panagopoulos (2014a), except that the gender of the eyespot image used in the key treatment was male rather than female. Voters were mailed a postcard within the week prior to the election and randomly assigned to a control group or to one of the same three treatment conditions. In addition to the control group, two other placebo conditions allowed for additional baselines to differentiate the influence of implicit social pressure cues from any general mobilization effect that may result from simple voter contact. Full details of both experiments are reported in Panagopoulos (2014a, 2014b).

Results

The main results of the two studies are presented in columns 1 and 6 of Table 1. The evidence reveals that only the treatments that featured eyespot images mobilized voters to the polls on Election Day at statistically significant levels. In the original (Key West) study, the estimates imply that the eyespot mailing raised turnout by 1.1% points on average relative to the control group that received no mailing, while the Lexington, KY study reveals turnout was 2.3% points higher on average in the treatment condition compared to the control. In both studies, the eyespot effects were statistically distinguishable from the control condition (no mailing) but also from both of the placebo (palm tree and flag) conditions.

Table 1. Partisan heterogeneity in eyespot effects in two studies.

Treatments	Study
	Key West, Florida					Lexington, Kentucky
	All	D	R	O	U	All	D	R	O	U
Implicit social pressure treatment: eyes	.011**	.006	.030***	−.003	.009	.023**	.014	.044**	.025	−.070
	(.005)	(.009)	(.012)	(.014)	(.008)	(.013)	(.018)	(.023)	(.039)	(.078)
Placebo treatment A: flag	−.002	.005	−.007	.010	.003	.015	.019	.029	−.038	−.098
	(.005)	(.009)	(.013)	(.015)	(.007)	(.013)	(.018)	(.024)	(.037)	(.092)
Placebo treatment B: palm tree	.001	.006	−.015	−.015	.017	−.009	−.007	−.011	−.023	.018
	(.005)	(.009)	(.013)	(.015)	(.008)	(.013)	(.018)	(.025)	(.037)	(.070)
N	13,542	5890	2848	690	4114	71,593	42,360	21,571	6082	1580
RMSE	.196	.221	.210	.120	.150	.422	.435	.416	.346	.366

Notes: Conditional Average Treatment Effects (CATE). Estimates (unstandardized coefficients) derived using OLS regression and represent average intent-to-treat effects (ITT). Dependent variable is voter turnout in the respective elections. Numbers in parentheses represent standard errors. *** signifies statistical significance at the p < .01 level, and ** at the p < .05 level, using one-tailed tests. Covariate-adjusted estimates only vary marginally and do not change overall significance of the results. Partisanship denoted as follows: Democrat (D); Republican (R); Unenrolled (U); and/or Other party (O).

Accordingly, Panagopoulos (2014a, 2014b) concluded that implicit social pressure cues (eyespot images) on postcard mailings were able to increase voter turnout in elections. Yet, the studies did not consider partisan differences in eyespot effects. As we discuss above, we expect the impact of watchful eyes to be moderated by political identity. We take advantage of the availability of information about individuals’ partisan affiliations in each experimental setting to examine whether there exists heterogeneity in the treatment effects attributable to eyespots. We emphasize that we rely on individual subjects’ actual partisan registrations, available in the official voter files, for the analyses that follow, not simply self-reported partisanship.

We proceed by partitioning the experimental samples in each experimental location by partisan affiliation. Subjects are assigned to one of the four partisan subgroups: Democrat (D), Republican (R), unenrolled (independent) (U), or ‘other party’ (O). The conditional average treatment effects for each subgroup are reported in corresponding, columns (2–5 and 7–10) in Table 1. Our analyses reveal evidence of heterogeneity in eyespot effects by partisan identity. Specifically, we find that the watchful eye effect is isolated to Republican registrants in both studies. No other partisan subgroup appears to have been successfully mobilized by the eyespot mailings. We also explored interaction terms between partisanship and treatment condition. Likelihood ratio tests comparing models with and without interactions were significant (study 1: LRchi2 = 73.70, p < .001; study 2: LRchi2: 469.09, p < .001). Evidently, the overall (ATE) effects reported in both studies (Panagopoulos, 2014a, 2014b) actually appear to be driven by the eyespots’ postcard mailings’ effects on Republicans, a finding that is consistent with our hypothesis and the theoretical arguments outlined above.

Discussion and conclusion

A growing body of research has shown that subtle (implicit) social cues, such as images of watchful eyes, can motivate cooperation and prosocial behavior in a range of societal contexts (e.g., Bateson et al., 2006; Ernest-Jones et al., 2011; Panagopoulos, 2014a, 2014b). At the same time, other research has found inconsistencies in the extent to which eyespots motivate cooperative behavior (e.g., Matland & Murray, 2015; Sparks & Barclay, 2013). We have argued that these observed differences may, in part, be due to a lack of consideration of potential heterogeneity in an individual’s propensity to respond to social pressure cues. Few studies have explored individual differences in the watchful eye effect and thus, our study makes an important contribution by being the first (to our knowledge) to examine the role of political identity.

Across two large randomized field experiments testing the effect of eye images on actual voting behavior, we find that the impact of eyespots on voter mobilization is indeed likely driven by partisanship. While a significant effect was observed for Republicans, no significant effect emerged for Independents, Democrats, or other political orientations. Notwithstanding evidence of positive social norm effects on voting behavior amongst Democrats as well (e.g., Gerber et al. 2008), our results are consistent with prior research in a number of important ways. First, our findings are broadly congruent with research in social and political psychology, which has repeatedly found significant associations between a conservative political ideology and a strong preference for social conformity, obedience, and adherence to prevailing social norms (Altemeyer, 1996; Carney et al., 2008; Hibbing et al., 2013; Jost et al., 2008; Skitka & Tetlock, 1993).

Having said this, we acknowledge and highlight that differences exist between political party identification and ideology. Yet, at the same time, our findings draw on verified voting records, which are more accurate than self-report-based survey measures. Moreover, prior research has found that (a) party affiliation is typically more strongly related to voters’ ideology than to their social identities (as defined by group memberships) and (b) the correlation and consistency between party identification and political ideology have strongly increased over the last decades (Abramowitz & Saunders, 2006; Jacobson & Carson, 2015). In fact, state-level data from the 2008 Congressional Cooperative Election Study (collected before the field experiments were conducted), reveal, using a 100-point measure of ideological self-placement, that actual major-party registration and self-reported ideology were strongly correlated in both Florida (.64, p < .01) and Kentucky (.55, p < .01), as in the US as a whole (.68, p < .01). We also find high consistency between mean ideology scores for registered Democrats and Republicans in both Kentucky and Florida (M_{Dem, KY} = 46.4, SD = 27.4 vs. M_Dem, _FL = 40.5, SD = 22.9) and M_{Rep, FL} = 76.2, SD = 19.9 vs. M_{Rep, KY} = 78.3, SD = 19.0).¹ In short, while some nuances between party identification and ideology clearly exist, both measures generally show good convergence.

Second, in a series of recent replication studies, Matland and Murray (2015) find that the empirical evidence for the watchful eye effect is rather weak. Interestingly, however, the only statistically significant eyespot effects reported by the authors (across five studies) were isolated to a politically conservative part of the country: Midland, Texas, where about half (47%) of the population is Republican. In contrast, in the current work, about one-fifth (20%) of the sample in the Key West study was Republican (43% Democrat), while 30% of the sample in Lexington was Republican (60% Democrat). Given that both samples were comprised mainly of Democratic voters, eyespot effects may have been stronger, on average, if the samples were more balanced in terms of partisanship. Thus, our findings suggest that a disproportionately small number of Republicans in null-result studies might make it harder to observe positive treatment effects, an artifact of partisan heterogeneity in eyespot effects. Together these findings add to a growing body of literature revealing that images of watchful eyes can indeed promote prosocial behavior, but that the effects are likely conditional on key individual differences and bounded by relevant contextual factors (e.g., Keller & Pfattheicher, 2011; Manesi et al., 2016; Mifune et al., 2010; Sparks & Barclay, 2013).

Lastly, we recognize that this study is not without limitations. For example, we acknowledge that other, unobserved factors could have influenced the association between political identity and conformity to watchful eyes. However, we deem this rather unlikely, given that recent research also indicates that it is not any civic duty context that triggers behavioral conformity, but the impression of social surveillance specifically (Gerber et al., 2008; Manesi et al., 2016). Nonetheless, we encourage future research to further explore the mechanisms by which watchful eyes differentially influence an individual’s susceptibility to implicit social cues.

Notwithstanding these limitations, we conclude by noting that our results are also broadly consistent with emerging findings from research in the context of other important social dilemmas. For example, studies find that exposing people to normative consensus cues has disproportionally positive effects on judgments about global warming for Republicans compared to Democrats (e.g., van der Linden, Leiserowitz, Feinberg, & Maibach, 2014, 2015). In short, this study not only sheds further light on the extent to which images of watchful eyes affect voter turnout in elections, but also speaks more generally to the finding that implicit social pressure cues may differentially influence human behavior and cooperation in important societal contexts.

Disclosure statement

No potential conflict of interest was reported by the authors.

Acknowledgments

The authors are grateful for the helpful comments of the editor and five anonymous reviewers.

Notes

1. 0 = Extremely Liberal; 100 = Extremely Conservative (CCES, 2008).