Evidence from online dating sites suggests that most users behave strategically: they remain silent about their political attitudes, more willing to divulge their weight than their politics

From 2017: The Moderating Effects of Marriage Across Party Lines. Shanto Iyengar, Tobias Konitzer. 2017. https://pdfs.semanticscholar.org/a55b/50f3de44529ee301c662aa42fb244e4ab992.pdf

“The Democrats, wherever you find ’em – in the media, think tanks, don’t care whereyou find ’em – they’re being consumed by it, folks. Theyre literally being eaten alive withan irrational, raw hatred.” – Rush Limbaugh, May 10, 2017.

In the aftermath of the 2016 election, the American electorate is hyperpolarized. Animustoward the out party is at an historic high. For the first time on record, the most frequentlyregistered feeling thermometer score for the opposing party (e.g. Democrats’ rating of theRepublican Party and vice-versa) in the 2016 American National Election Study was at theminimum, i.e. zero. Other indicators are similarly skewed toward the extremes. PresidentTrump’s approval drops precipitously from 80 percent among Republicans to under tenpercent for Democrats. Some six times as many Democrats than Republicans believe thatthe Trump campaign colluded with the Russians to sway the election (Washington Post Poll,April 26, 2017). Hostility directed at out groups is a fundamental barometer of group polarization. Classic studies on social distance (Bogardus, 1925), and the sense of social identity (Tajfel, 1970;Tajfel and Turner, 1979) have established that diverging sentiment for in- and out-groupmembers is inevitable. Group polarization defined in terms of differential affect for the inand out group occurs even when the basis for group affiliation is trivial and completelyunrelated to group interests.In the context of American politics, affective polarization deriving from political partyaffiliation is well documented, in stark contrast to ideological polarization, where the evidenceis mixed (compare Abramowitz 2010 with Fiorina, Abrams and Pope 2005). As for partisanaffect, data from the American National Election Surveys dating to the mid-1980s showsthat Democrats and Republicans not only increasingly dislike the opposing party, but alsoimpute negative qualities to its supporters (Iyengar, Sood and Lelkes, 2012).Out-group prejudice based on party identity exceeds the comparable bias directed atracial, religious, or cultural out groups (Muste, 2014; Iyengar and Westwood, 2015). Partisan2 affect has strengthened to the point where party identity is now a litmus test for interpersonalattraction. People prefer to associate with fellow partisans and are less trusting of partisan opponents (Iyengar and Westwood, 2015; Westwood et al., 2017). The most vivid evidence of increased social distance across the party divide concerns inter-party marriage. In the early 1960s, the percentage of partisans expressing concern over the prospect of their son or daughter marrying someone from the opposition party was in the single digits, but some forty-five years later it had risen to more than twenty-five percent (Iyengar, Sood and Lelkes,2012). Data from surveys of married couples, online dating sites, and national voter files confirmthat partisanship has become a key attribute underlying the selection of long-term partners (Huber and Malhotra, 2017; Iyengar, Konitzer and Tedin, 2017). Among recently marriedcouples in 1973, only 54 percent shared the same party affiliation. Forty years later, partisan agreement among this group had risen to 74 percent (Iyengar, Konitzer and Tedin, 2017).Survey results standing alone may not be the most meaningful measure of increasingpartisan animus. The expression of hostility based on partisanship is not subject to thesame social taboos as hostility based on other salient social divides (racial, religious, orethnic). Instead, hostility directed at the out party is deemed acceptable, even appropriate.Therefore, survey data could artificially elevate the significance of the partisan divide overother significant cleavages. But, importantly, there is considerable evidence of increasedpartisan animus outside the survey realm; this evidence is not subject to normative, consciousrestraints based on political correctness. Using a version of the Implicit Association Test,Iyengar and Westwood (2015) demonstrated that implicit bias directed at the out partyexceeded comparable bias based on race. They also showed that behavioral discriminationagainst partisan opponents in a variety of contexts exceeded discrimination based on othergroup cues, most notably, race. What explains the dramatic increase in affective polarization over the past few decades?3

The period in question (1985-2015) coincides with any number of major societal changes,including the vastly increased ethnic and cultural diversity of the population, the migration ofwhites from urban areas to the suburbs, the emergence of the South as a staunch Republicanregion, and the politicization of evangelical Christians. One possible explanation, sometimesreferred to as ”sorting,” is the increasing convergence of multiple salient social identities which reinforce each other. In other words, Democrats and Republicans differ not only int heir politics, but also in their ethnic, religious, gender, cultural, and regional identities.Sorting leads to overlapping group memberships and the increasing partisan homogeneity of primary and secondary groups is a further contributor to polarization. Family and kinship networks – key influences over the development of political attitudes – provide few opportunities for meaningful and long-term personal contact across party lines. As we notebelow, even at the level of secondary groups, defined on the basis of occupation, religion, orplace of residence, partisan homophily is extensive (for evidence on occupational similarity inpartisan affiliation, see Bonica, Chilton, and Sen 2016; the geographic sorting of the nation into Republican and Democratic enclaves is documented most recently in Chen and Cottrell2016 and Chen, Rodden et al. 2013.We investigate the role of interpersonal relations as a potential contributor to partisanpolarization. We focus on the family, the most important agent of political socialization.Comparing surveys of spouses conducted in the 1960s, the 1990s and the current era, wedemonstrate that over time, spousal disagreement – although clearly becoming less frequent– can act as a brake on polarization by fostering less hostile attitudes toward partisan opponents. The more heterogeneous the household, the less polarized the individual members.We replicate these survey findings a secondary analysis of a set of 2015 and 2016 field experiments that targeted registered voters in multiple states. Participants in these experiments completed surveys that included feeling thermometers and other measures of partisan affect. Because the surveys sampled multiple family members living at the same address, we can4 investigate the effects of household agreement on partisan affect. In fact, the data from these field studies converge with the surveys of spouses; mixed-party households are significantly less polarized.


Discussion

In our experience, it is unusual to observe differences of this scale in the behavioralsciences. In several instances, the order of magnitude of the difference in polarization between similar and dissimilar family pairs exceeded 200 percent! Even more unusual is thefact that our results survived multiple replications spanning different research designs, elec-toral contexts, and survey indicators of partisan affect. While we acknowledge the causalthreat posed by selection effects, i.e. potential spouses with polarized attitudes selectinginto homogeneous marriages, the matching analysis does little to bolster this alternativeexplanation.9In total, the story line is unmistakable; political disagreement within familyrelations discourages extreme evaluations of in and out groups, thus alleviating polarization.Of course, documenting the powerful effects of family diversity on partisan affect begsthe question of what exactly is the mechanism through which exposure to disagreementmoderates individuals’ evaluations of the parties. One possibility is that domestic tranquilityrequires the expression of opinions that respect the positions of significant others, makingindividuals more tolerant and accepting of disagreement. Alternatively, as suggested by theclassic “contact” hypothesis (Allport, 1954; Pettigrew, 1998, 1997), valued inter-personalrelations that cut across the party divide may serve to weaken negative stereotypes of theout group; for contrary evidence, however, see Enos (2014). Yet another possible mechanism,also suggested by previous research (Mutz, 2006), is that inter-personal contact heightensawareness of the values and arguments underlying the preferences of out party supporters,making the party appear less threatening. All these mediating mechanisms appear to becontingent on exposure to the opponent’s point of view; we would anticipate, accordingly,that the effects of family diversity on partisan evaluations will be enlarged when familymembers frequently converse about public affairs.While our results imply that cross-party family ties are a potential antidote to polariza-tion, it is important to keep in mind that this ”treatment” only impacts a relatively smallswath of American partisans. The most recent data on inter-marriage indicate that less9Although while matching might tackle the selection problem more appropriately then parametric re-gression analyses, we note that the identification assumption remains selection-on-observables.23 than twenty percent of partisans are exposed to disagreement. The likelihood that parti-sans’ children will diverge on political grounds is similarly remote; in a 2015 survey, 74.2percent of parent-offspring dyads agreed on their partisan affiliation (Iyengar, Konitzer andTedin, 2017). The critical question, therefore, concerns the ways in which society can lowerthe barriers to social exchange across the party divide.In theory, one solution to the problem of politically homogeneous networks is to weakenindividuals’ ability to signal their political affiliation. If all participants in the marriageor dating market were ”blind” to partisan affiliation, partner selection would be drivenprimarily by non-political attributes. Rational ”sellers” should deliberately conceal theirpolitical views when seeking out potential mates. In fact, the evidence from online datingsites suggests that most users of these sites behave strategically: they choose to remainsilent about their political attitudes. Research into the content of online daters’ personalprofiles shows that less than fifteen percent of online daters provided information abouttheir “political interests” and when they did reference politics in their personal profile, theyidentified themselves as “middle of the road” (Klofstad, McDermott and Hatemi, 2012). This same study shows, revealingly, that online daters are more willing to divulge their weightthan their political preferences.Online databases provide opportunities for people to sort into relationships on the basisof attributes extraneous to partisanship. Since some twenty percent of single individualsreport using online dating sites, technology might be a tool for dampening polarization.Yet, as Huber and Malhotra (2017) have recently demonstrated, individuals manage tounearth information about their prospective partner’s political views despite the lack oftransparency; so much so, that political ideology is the strongest predictor of successfulonline match making. The motivation to find a politically compatible mate is sufficient toovercome online daters’ lack of transparency about their politics.In closing, our results show that partisan attitudes are distinctly less polarized when close inter-personal ties are not based on the criterion of political similarity. For those seekingto reduce animus and conflict across party lines, it is important to design meeting placesor platforms on which people become less focused on questions of political identity as animportant basis for their inter-personal relations.