The dual nature of partisan prejudice: Morality and identity in a multiparty system

Rising hostility between members of opposing political factions has gained considerable attention in both academic and popular press. The adverse effects of this phenomenon are widely recognized, but its psychological antecedents remain the focus of ongoing debate in political psychology. Past research has honed in on two conflicting explanations: one highlights the extent to which people self-define as supporters of particular parties or candidates (the identity view), and another points toward the intensity with which they disagree on substantive matters of policy (the issues view). A nationally representative survey of 1051 eligible Spanish voters yielded support for both explanations. The perceived magnitude and nature of disagreement were associated with increased partisan prejudice, while controlling for partisan identification. Path analyses revealed that issue-based prejudice was more pronounced among ideologically extreme agents (β = 0.237, 95% CI [0.174, 0.300]) than toward extreme targets (β = 0.140, 95% CI [0.078, 0.201]), and replicated recent findings that identity-based prejudice is motivated primarily by non-instrumental factors (β = 0.286, 95% CI [0.230, 0.337]). Together, these results indicate that discrimination across party lines responds to two fundamentally distinct, though at times co-occurring, imperatives: to coalesce in ideologically homogeneous communities, and to protect one’s sense of partisan identity.


Introduction
Is polarization partly a matter of how we perceive others? Some studies seem to suggest something like that might be going on (Fiorina et al., 2008), even in the domain of morality and moral differences (Graham et al., 2012). According to this thesis, real moral differences would not be as important as perceived moral differences. In turn, misrepresented moral differences could be the basis of a whole series of dysfunctions in public debates and civic life. If moral and meta-ethical differences can promote intolerance, studying more systematically the misrepresentation of these differences becomes a priority. At the origin of these misrepresentations there may be different operating processes that should be better known. In what follows, we present some results of exploratory studies in relation to our study of the moral roots of partisan prejudice.
2 Do people misrepresent their ideological rivals?
In line with previous research (Graham et al., 2012), targets appeared to be misrepresented when compared to their self-reports. Target extremity vastly exceeded agent extremity in a paired t-test, t(1050) = 29.7, p < .001, dz = 0.92, and separate analyses by political party revealed that the discrepancy was larger for conservative than progressive targets: In particular, right-leaning participants appeared to perceive Unidos Podemos supporters as more progressive than the latter self-reported, t(751.7) = −2.63, p = .009, d = −0.19. In turn, left-leaning participants perceived both PP, t(342.2) = 14.97, p < .001, d = 1.11, and Ciudadanos, t(339.93) = 18.30, p < .001, d = 1.34, supporters as much more conservative than the latter self-reported-with differences exceeding one standard deviation.
We observed a corresponding discrepancy on our measure of metaethical beliefs: Respondents viewed targets as more objectivist than themselves, paired t(1050) = 17.55, p < .001, dz = 0.54. Right and center-right voters were viewed as much more objectivist than they reported, (d P P = 0.81, d Ciudadanos = 0.86), while the corresponding effect for left voters was moderate in size (d U nidosP odemos = 0.31).

Misrepresentation, moral objectivism and partisan prejudice
Participants' agent-objectivism correlates positively with their distortion scores, meaning that those participants which gave more intuitive objectivist responses to the sociomoral issues also tend to misrepresent more the political outgroup moral attitudes (r = .15, 95%CI[.09, .21], p < .001), which suggests a possible "false dissensus effect 2 " when characterizing the outgroup, r = .15, 95%CI[.09, .21], p < .001). After checking through linear regression models, this positive relationship held after controlling for several possible confounding factors (support for different political groups, agent-extremity, and degree of partisan identification).
We also observed by-issue differences in distortion. Distortion was greater on core issues (on which participants held objectivist beliefs; see Supplementary Table below). Why self-reports and rival assessments of moral values vary so markedly remains a puzzle for further work to unravel. In line with previous research, we found a left-right asymmetry: Conservatives were perceived as much more reactionary than they reported, while liberals were only perceived as somewhat more progressive than they reported being. However, we must also note the possibility that the observed discrepancies reflect systematic inaccuracies in self-report-for instance, if reactivity leads conservative participants to downplay their moral dissent and conviction. It could be that either 1) a 'misunderstanding account' where misperceptions about others' moral positions and moral motives fuel political hostility is the best explanation; or 2) a 'motivational account' where a motivation to advance one's political agenda also leads to strategically deforming and caricaturing others' moral positions is driving the observed effect; or 3) that precisely the fear of rejection and discrimination by others leads, at least some people, to publicly moderate their moral positions, partly hiding their private preferences. This latter 'social desirability' account would suggest that what we call distortion -the difference between attributed and self-reported moral judgments -might not be as high as registered by our way of measuring it, since at least some participants would be moderating their moral positions in response to social pressure. Additionally it could be that the 'misunderstanding account' is on the right track and that supporters of different parties do actually have different levels of accuracy and precision in their perception of political competitors' moral positions. Some of the sorting dynamics referred in previous research could supplement this account (Lang & Pearson-Merkowitz, 2015). 4 Do we observe left-right symmetry or asymmetry in partisan prejudice?
To evaluate whether progressives and conservatives differ in partisan prejudice, we report two analyses employing participants' moral values (i.e., conservative values < 0 < progressive values) and unfolded moral disagreement (preserving its sign: i.e., positive values = target more conservative than agent; negative values = target more progressive than agent).
In each model below, we enter both linear and quadratic terms. The quadratic term will describe any U-shaped differences-between centrists and ideologues on either side. The linear term will capture any left-right slope after accounting for any quadratic curvature. Thus, the quadratic term can be said to detect symmetric effects, while the linear term can be thought of as capturing asymmetric left-right differences.
First, we regress prejudice on unfolded-disagreement, unfolded-disagreement 2 , and both measures of objectivism. This model revealed a quadratic effect of disagreement, t = 3.69, p < .001, but no linear effect, t = 0.03, p = .98.