Modern democracy is unthinkable save in terms of the parties.
E.E. Schattschneider, The Semi-Sovereign People
As people reflect on whether they are Democrats or Republicans (or neither)…they ask themselves two questions: What kinds of social groups come to mind as I think about Democrats, Republicans, and Independents? Which assemblage of groups (if any) best describes me?
Green, Palmquist, and Schickler, Partisan Hearts and Minds
“Identity politics” has not traditionally been part of mainstream American political vernacular, but it emerged with a roar in 2016. Since Donald J. Trump announced his presidential candidacy, pundits on the left and right have debated the proper role of identity in mainstream party politics (e.g. Edsall 2017; WSJ Editorial Board 2017). As is often the case with abstruse elite debates, the public appears to have perked up only when the consequences of identity politics became distinctly tangible. Google searches for “identity politics” spiked in the weeks after Trump’s election and again in August 2017, immediately after the alt-right violence in Charlottesville. 1 These searches were not just coming from the coasts; citizens in the Rustbelt states that paved the way for Trump’s victory also wanted to know what this reactionary and apparently new form of politics was all about.
Contemporary pundits’ accounts tend to imply that only in the Age of Trump has identity politics taken root in mainstream American politics, as a divisive force in the Democratic Party (e.g. Lilla 2017) and a gauche populist movement within the Republican Party (e.g. Hopkins 2017). Ultimately, citizens looking to opinion leaders are liable to believe that radical interests that have “hijacked the parties” (Capehart 2017) and replaced a reasonable, inclusive, issue-based politics with conflict over immutable characteristics.
Group conflict indeed drives contemporary American politics. But contrary to punditry, the fusion of group conflict and partisan conflict is nothing new. As I will argue in this essay, the intense polarization of citizens across party lines stems in part from people’s beliefs about how the parties are composed by social groups. Democrats appear to dislike Republicans, and Republicans Democrats, because they see the other side as extraordinarily stereotypical and alien. For example, Democrats appear to believe (on average) that roughly 45% of Republicans are evangelical Christians, but just a third are. And Republicans appear to believe that over a third of Democrats are atheist or agnostic, but only around 10% are. More importantly, correcting these misperceptions reduces the dislike that citizens feel across party lines, demonstrating that these perceptions fuel partisan animus. In this sense, identity politics is not a new phenomenon in which a radical few have hijacked the parties. Instead, it’s what most of us do – and always have done – when evaluating political parties and their supporters.
Party Identities and Party Polarization
Party polarization is among the most notable developments in recent American political history. Among political elites, polarization is characterized by within-party homogeneity. Most Democratic politicians take liberal positions across issues, while most Republican politicians take consistently conservative positions (McCarty, Poole, and Rosenthal 2006). As a result, elite party polarization is usually seen as ideological: policy debates nearly always cleave along party lines and wedge issues have gone extinct.
Among Democratic and Republican supporters, however, party conflict is not particularly ideological (e.g. Fiorina, Abrams, and Pope 2005). Citizens’ policy preferences are far less “organized” than elites’ for a variety of reasons (e.g. Converse 1964; Ellis and Stimson 2012; Broockman 2016). For example, citizens don’t face the institutional pressures Members of Congress do to support their fellow partisans’ favorite policies (Clinton 2012) – and even if they tried to show similar solidarity, many rank-and-file partisans lack basic knowledge about which positions go with which parties (Freeder, Lenz, and Turney Forthcoming).
Instead, ordinary Democrats and Republicans are primarily polarized in terms of how they feel about each other. Partisans’ reported feelings toward their own party have been remarkably consistent throughout modern political history, hovering around 70 on the 101-point “feeling thermometer” used to assess survey respondents’ affect toward groups. Feelings toward the out-party, however, have plummeted between Watergate and today (Iyengar, Sood, and Lelkes 2012). In the mid-1970s, the typical party supporter rated his out-party a tepid 48. That figure that seems downright neighborly today. In 2016, out-party ratings hovered near 30, well below traditionally disliked groups such as “atheists” and “elected officials in Washington” (Pew 2014). Other survey measures confirm and illuminate affective polarization between Democrats and Republicans. Since the 1970s, citizens’ feelings toward out-party politicians – and trust in government, when controlled by the out-party – have plummeted, with potentially pernicious consequences for government (Hetherington and Rudolph 2015). Near-majorities of both parties say that the out-party is a “threat to the nation’s well-being” (Pew 2014). And a rising share of each party expresses dissatisfaction at the thought of a son or daughter marrying across party lines, a bellwether of partisan animus (Iyengar, Sood, and Lelkes 2012).
Most recently, a handful of studies have reported on the potential behavioral consequences of affective polarization. For example, partisans appear to discriminate along party lines: they show favoritism when evaluating scholarship candidates (Iyengar and Westwood 2014) and when trading goods and services online (McConnell et al. 2018). Perhaps most disturbing, partisans who most dislike the out-party are also most likely to believe unfounded conspiracy theories about that party, even controlling for in-party affinity and self-reported ideological dispositions (Roush 2017).
Political scientists increasingly care about identifying the causes of affective polarization. If we can isolate particular causes, the thinking goes, then perhaps we can develop appropriate “cures” for the pathologies noted above. But the search for causes of mass polarization has re-opened old debates about the nature of party identification itself (Johnston 2006). Polarization among citizens may manifest itself affectively, but the underlying roots of people’s feelings about the parties could still be cognitive. That is, inter-party animus may not stem merely from instinctive emotion, but instead from people’s explicit beliefs about the parties. The most stringent conceptions of a cognitive basis for partisanship contend that citizens’ policy preferences drive their partisanship (Markus 1979; Franklin 1991), and polarization stems from persistent disagreement over policies and values (Abramowitz 2010; Webster and Abramowitz 2017). Others in this camp contend that partisanship reflects retrospective considerations about the parties’ performance in government (Fiorina 1981; MacKuen, Erikson, and Stimson 1989). To these scholars, affective polarization could actually signal a healthy democracy. Another camp, however, concludes that partisanship is primarily affective in nature. These scholars argue that partisanship is more akin to a social identity than a considered, malleable decision (Campbell et al. 1960; Iyengar, Sood, and Lelkes 2012; Huddy, Mason, and Aaroe 2015) – and, thus, that we can explain polarization much the same way we can explain other types of social conflict: people derive self-worth from their sense of membership in social groups and deploy in-group favoritism and out-group chauvinism to boost their sense of group esteem (Brewer 1991; Greene 1999, 2004; Tajfel and Turner 2005). Under the strictest affective conception of partisanship, polarization is reactionary rather than considered, and perhaps better left to applied psychologists to “cure”.
A more nuanced approach asserts that partisanship is indeed a social identity, albeit quite distinct from those that come most easily to mind. Under this conception, citizens see parties as collectives of more fundamental groups in society and evaluate them according to how well the parties reflect their own identities (Green, Palmquist, and Schickler 2002). This conception implies that partisanship is both cognitive and affective in nature. It’s cognitive in that citizens have beliefs about the types of people who are Democrats and Republicans and reason according to these beliefs. 2 But these beliefs are not affectively neutral; people tend to have feelings about groups in society (Tajfel 1970). In sum, the group theory of parties contends that citizens form attachments to the Democratic or Republican Party – more akin to an identity than a standing decision about political preferences – based on their affect-laden social perceptions of Democrats and Republicans.
In their sweeping study of American democracy, Achen and Bartels (2016) summarize the odd status of the group theory of parties: “Most scholars of political behavior seem to agree that partisanship is both a form of social identity and, in significant part, a product of social identity. However, it is surprisingly difficult to provide clear empirical evidence for these propositions” (p. 234–235). Rarely in respectable academic circles does a dominant theory lack rigorous support. To be fair, circumstantial historical evidence exists: ethnic identities appear to have conditioned Bostonians’ partisanship, irrespective of social class, in the 1920s (Gamm 1989). Furthermore, white Americans’ responses to civil rights legislation is telling. Civil rights legislation rendered the Democratic Party’s policy image more clearly liberal on racial issues throughout the country, but in most parts of the US, white citizens’ partisanship remained stable. Only in the South, where Jim Crow laws had precluded black participation, did the parties’ social images change: newly enfranchised African-Americans joined the Democratic Party. Seemingly in response, Southern whites gradually left the Democratic Party for the GOP (Green, Palmquist, and Schickler 2002, figure 6.1). Importantly, for a significant stretch of the 20th century, white Southerners’ partisanship and their attitudes toward integration were only weakly related, indirectly suggesting that identity – rather than policy – fueled the Southern realignment (Achen and Bartels 2016, figure 9.4).
As the above illustrates, it is easier to marshal evidence against competing theories than direct support for the group theory of parties. Researchers generally rely on cross-sectional survey data, which provide weak leverage for evaluating theories about causes and effects. But even more fundamentally, alternative sources of causal leverage are elusive in this case. The gold standard for evaluating causality is the randomized controlled experiment, in which the researcher has control over the variable theorized to cause the effects of interest – also known as the X-variable, the independent variable, or, in the experimental context, the treatment. The researcher further has control over who receives the treatment and who is assigned to a comparison control group. Assignment is usually random to guarantee that participants’ “treatment status” is independent of the outcome of interest, also called the dependent (or “Y”) variable, so that any difference in outcomes can be attributed to the treatment (rather than a confounding variable). With that in mind, consider the difficulty in providing direct, individual-level evidence that “identity is the key moving force” behind Americans’ partisanship (Achen and Bartels 2016, p. 235). Experimentally manipulating attitudes toward social groups is quite hard – and manipulating people’s identities nearly impossible – because of the long-standing and durable nature of these orientations. How, if at all, can we get leverage on this research question?
Poetically, random assignment of another kind led to a path forward. In August 2013, I met Gaurav Sood when we were assigned poster expo space next to each other at the annual meeting of the American Political Science Association. That meeting developed into a series of e-mails, Skype conversations, and trans-Bay confabs regarding party stereotypes and polarization. One morning in Spring 2014 we deduced, consistent with theories of stereotyping, that if people reason about the parties according to their group composition, they are liable to overestimate the prevalence of party-stereotypical groups in “their” parties.
For example, people ought to overestimate the percentage of Republicans who are wealthy and the percentage of Democrats who are working class. Prototypes – mental images that represent categories – tend to reflect characteristics that distinguish the category, even if those characteristics aren’t overly common in the category (Rosch and Mervis 1975). For example, when people think of universities, they often conjure images of ivy-covered brick walls, even if the vast majority of campuses lack such a feature. We do so because, aside from Wrigley Field, few examples of non-academic buildings with ivy-covered walls come easily to mind. More broadly, this reflects humans’ tendency to focus on similarities, to the detriment of other relevant information, when making judgments – a phenomenon Tversky and Kahneman (1974) call the representativeness heuristic. If people think about the parties as amalgams of other social groups, we reasoned, they likely use the representativeness heuristic to paint their mental pictures of the parties, focusing largely on perceived overlap between categories like “the rich” and “Republicans” or “working class” and “Democrats.” Importantly, many groups that are party-stereotypical comprise just a sliver of the population – and, thus, they also comprise only a relatively larger sliver of “their” party. For example, just under 10% of Democratic Party supporters are atheist or agnostic (according to 2014 data). But a citizen trying to determine how secular the Democratic Party’s supporters are is liable to overlook these groups’ general rarity, focusing instead on atheists’ and agnostics’ notable tendency to be Democrats (rather than Republicans or independents). In doing so, they are liable to overestimate the prevalence of atheists and agnostics among Democrats.
We realized that if people indeed overestimate the prevalence of evangelical Christians in the GOP, or secular folks among Democratic supporters, then we could assess the effect of those misperceptions by clearing them up. That is, while we may not be able to manipulate people’s own identities, or even their feelings toward salient groups in society, we can manipulate their beliefs about how those groups compose the parties – the very thing said to drive political reasoning by the group theory of parties.
The Parties in Our Heads
To assess people’s perceptions of how the parties are composed, we conducted a series of surveys asking Americans to report their beliefs about the percentage of Democrats and Republicans who belong to party-stereotypical groups. We identified these groups from existing work on American partisanship, and they comport with groups later identified by Rothschild et al. (2018) as party-stereotypical. We asked respondents what percentage of Democrats they believed to be atheist or agnostic; black; lesbian, gay, or bisexual; and union members. 3 We also asked respondents what percentage of Republicans they believed to be at least 65 years old, evangelical Christians, Southerners, and $250,000+ annual earners. We estimated the actual quantities of interest – e.g. the percentage of Republicans who earn at least $250,000 per year – from the 2012 American National Election Study (and the 2014 Pew Religion and Public Life Project, for the religious party-group pairings). The primary study I’ll present here was conducted in 2014, when Gaurav and I surveyed 1000 American adults about their perceptions of how the parties are composed. Sampling was conducted by YouGov and resulted in a group of respondents quite similar in its own demographic makeup to the American adult population writ large. Our other studies probed the nature of these perceptions further and assessed their effects on partisan attitudes and polarization, providing new evidence on the group theory of parties. 4
38% of Republicans Belong to “The 1%”?
As it turned out, Americans’ perceptions of the parties were more skewed than we expected. Figure 1 plots the percentage of party supporters belonging to each of the four party-stereotypical groups against average perceptions from the 2014 YouGov survey. The figure plots these perceptions separately by respondent partisanship but, consistent with existing research (Green, Palmquist, and Schickler 2002), party images are relatively consistent across partisan groups – indeed, Democrats, Republicans, and independents alike vastly overestimate party-stereotypical groups’ numerical strength in “their” parties. For example, the top panel in Figure 1A depicts the percentage of Democratic Party supporters who identify as atheist or agnostic. In actuality, about 9% of Democrats are expressly secular. However, Democrats and independents overestimate this quantity by over 150% and Republicans by nearly 300%. Even more shocking, as Figure 1B’s top panel shows, just 2% of Republicans earn over $250,000 per year but Americans of all political persuasions estimated that between 30 and 45% of Republican supporters do. This is particularly surprising because both “$250,000 per year” and “the one percent” were colloquialisms for the economic elite around the time of our study; for 38% of Republicans (the average guess across all respondents) to belong to the economic one percent, a negative number of Democrats and independents would also need to belong!
People’s mental images of the out-party tend to be somewhat more stereotypical, but we should not confuse this with in-party accuracy. For example, Republicans appear to believe, on average, that 46% of Democrats are black – an overestimate of 22 percentage points, or nearly 100%. Democrats and independents do a bit better, but still report that roughly 40% of Democratic supporters are black. Similarly, Republicans and independents estimate that 35–40% of GOP supporters are eligible to collect Social Security – only 21% are – and Democrats are just slightly less accurate. Seven of the eight party-group pairings demonstrate this pattern, which is not particularly interesting; it comports with decades of work in social psychology on out-group stereotyping (e.g. Judd and Park 1988). What is interesting is the tremendous inaccuracy all respondents tended to show, regardless of partisanship. Relying on median perceptions – rather than means, which are more susceptible to the influence of outlying perceptions – fails to appreciably reduce the bias we observe, suggesting that a small number of large outliers cannot explain these results. More directly, 72% of the 8000 perceptions we observed were overestimates.
At this point, I suspect readers are figuratively shooting their hands up into the air, ready to offer a variety of compelling alternative explanations for the pattern I’ve described as genuine misperception. Upon observing the initial data, we did exactly the same. Much of our effort since 2014 has been devoted to testing these alternative explanations and better understanding the nature of responses to survey questions about numerical perceptions. On the latter, we do not believe that people actually walk around with the explicit belief “38% of Republicans belong to the economic elite.” Rather, evidence suggests that most people recognize wealthy people are significantly more likely to be Republican than Democratic, and through a series of mental gymnastics (e.g. Tversky and Kahneman 1974; Bordalo et al. 2016) reach the conclusion that some fuzzy but large percentage of Republicans is rich. These latent (or largely subconscious) beliefs are the essence of partisan identification, according to the group theory of parties. We argue that when people are asked for an explicit perception of the percentage of Republicans who earn at least $250,000 per year, their responses reflect these genuine-but-fuzzy beliefs about the share of Republicans who are in the economic elite, plus individual-level factors (e.g. ability to think in percentages), factors relevant to the group-party pairings (e.g. prevalence of stereotypes about Republicans being wealthy), and random measurement error (Ahler and Sood 2018a). Importantly, this conceptualization assumes that no systematically bizarre response tendencies exist en masse, which brings us back to the first concern: Do the numbers that respondents report actually reflect genuine – even if fuzzy – perceptions of groups’ compositional strength in the parties?
We address the most compelling alternative explanations in significant detail in Ahler and Sood (2018a,b), and greater information about what follows can be found there. But here is a quick-and-dirty summary. First, partisan trolling cannot explain these results. When we offered cash bonuses for correct responses, Democrats were no less likely to say that an inordinate number of Republicans are old, and Republicans were no less likely to say that gobs of Democrats are gay. Furthermore, people’s feelings toward the groups (as measured by 101-point “feeling thermometer” scales) fail to predict the share of that group they attribute to their out-party. Second, people’s difficulty with percentages (and math in general) can’t explain these results. We constrained respondents to 100 Republicans and required them to give us the number who earn “less than $50,000,” “between $50,000 and $99,999,” “between $100,000 and $249,000,” and “more than $250,000” (for example), with the requirement that these answers sum to 100. Reported perceptions of the percentage of partisans belonging to party-stereotypical groups only declined three percentage points; the average perception in the YouGov study was off by more than twenty points. Third, and perhaps most surprisingly, people’s ignorance of some of these characteristics’ rarity cannot explain these results. When we gave respondents the groups’ base rates – for example, that just 1% Americans earn at least $250,000 per year – the perceptions they reported were somewhat worse. Beyond these theoretical alternatives, we’ve assessed a number of possible issues related to survey design and instrumentation, all of which contribute substantively unimportant (if any) effects on people’s perceptions of party composition.
People appear to grossly misperceive the makeup of the Democratic and Republican mass parties, but is this actually consequential? That is, do these misperceptions affect citizens’ attitudes or behavior in a meaningful way, or are they merely fodder for a good laugh at academic cocktail parties? Ultimately, for our purposes here, do they fuel polarization the way the group theory of parties argues they should?
Correcting Misperceptions, Depolarizing Partisans
In 2015, we replicated our findings at the state level, discovering that Californians overestimate the likelihood that Democrats and Republicans in their own state hold party-stereotypical identities. We did so as part of the IGS-California Poll, a large, high quality, collaborative online survey conducted by the Institute of Governmental Studies at UC Berkeley with sampling through Survey Sampling International. Not only were California residents’ mental images inaccurate, they also correlated with political outcomes of interest collected as part of the broader survey. Californian Democrats who believe that a disproportionate number of state Republican supporters are wealthy or come from Orange County, for example, feel a significantly greater sense of polarization, as do Republicans who believe a disproportionate number of Democratic supporters are gay or Latino. Furthermore, these misperceptions correlate with one’s emotional investment in their own party’s successes and failures (Ahler and Sood 2018b). While these statistical associations support the group theory of parties, the old adage holds: “correlation is not causation.” For example, we might observe these correlations because of omitted variable bias – some unobserved factor, like black-and-white thinking, could cause people both to dislike the out-party and to overestimate out-party stereotypicality. Alternatively, reverse causation could be at work – dislike of the out-party could cause people to think more stereotypically.
To more conclusively examine whether these misperceptions cause polarization, we conducted a series of experimental studies on Amazon’s Mechanical Turk. As noted above, two key features distinguish experiments, also known as randomized controlled trials or A/B tests, from studies like our analysis of the IGS-California Poll. First, researchers achieve control over the causal variable of interest – in this case, perceptions of out-party composition – through the careful construction of treatments designed to manipulate those perceptions (or whatever the causal variable may be). Because the researcher – and not some unobserved process – manipulates the causal variable, differences in outcomes between conditions cannot be due to reverse causation. Second, researchers randomly assign experimental participants to treatments (or a comparison control group). Since assignment is random, it is statistically independent of all those pesky omitted variables discussed above in the context of the IGS-California Poll: we should expect our treatment and control groups to be quite similar prior to treatment. Any differences in the outcome variable – in this case, participants’ sense of polarization – may therefore be attributed to the treatment – in our case, people’s beliefs about how the out-party is composed.
Because Americans tend to be so inaccurate when responding to questions about how various groups compose the parties, designing a treatment was relatively straightforward. As before, we asked all survey respondents for the percentages of the out-party that they believed belonged to four party- stereotypical groups. In these studies, however, we presented a randomly-chosen group of respondents with the true percentages before all reported their feelings toward the out-party. 5 Since most people overestimate these percentages, our informational treatment should cause respondents to downwardly-revise their beliefs about the strength of stereotypical groups – groups to which most people tend not to belong – in their out-party. According to the group theory of parties, these corrections ought to reduce partisans’ sense of polarization from the other side.
We ran this experiment twice, using different measures of out-party dislike. In our first experiment, respondents indicated their feelings toward the out-party with a traditional 101-point feeling thermometer. In our second experiment, respondents answered four questions regarding how they would feel about various social interactions with out-party members – for example, a son or daughter marrying an out-party supporter. These questions are known as a social distance battery and their average is frequently used to measure affective polarization. We rescaled both measures so that scores close to 0 imply low affective polarization and scores close to 1 imply high levels of out-party animus. It’s worth noting that the bulk of scores on both measures fall in the 0.5-to-1 range – most people don’t especially like their out-party.
In both experiments, respondents whose misperceptions were corrected expressed significantly less animus toward the out-party. Out-party feeling thermometer ratings were nearly four points warmer in the “corrected” condition in the first experiment, and corrected respondents expressed a bit over one point less social distance in the second experiment. 6
Just as informative as these main results are the patterns within experimental conditions. Figure 2 summarizes these trends with locally-weighted scatterplot smoothing (LOWESS), which estimates average values of a dependent variable – in this case, out-party dislike – across overlapping “windows” of an independent variable – here, respondents’ beliefs about how stereotypical out-party members tend to be. The dashed vertical lines in each panel of Figure 2 represent the point at which respondents go from being “underestimators-on-average” to “overestimators-on-average” with regard to party-stereotypical groups’ strength in the out-party. Consistent with the aforementioned survey studies, large majorities of respondents in both experiments were “overestimators-on-average.” In both experiments, the minority of respondents who were relatively accurate (or were “underestimators-on-average”) expressed similar feelings toward the out-party, regardless of whether we corrected their perceptions. Among respondents to the right of Figure 2’s dashed line, however, as beliefs became more inaccurate, feelings toward the out-party diverged significantly according to experimental condition. Among uncorrected respondents, Democrats who saw large percentages of Republicans as rich and Republicans who believed large quantities of Democrats to be union members (for example) expressed significantly more animus toward their out-party than did respondents who were relatively accurate. That is, Figure 2’s LOWESS curves for uncorrected respondents are largely upward sloping: misperceptions about party composition predict out-party dislike. On the other hand, corrected respondents’ feelings toward their out-party show no systematic association with their previously-reported beliefs about that party’s composition. 7 Another way to think about this treatment effect is that the information appears to have most affected the most misinformed.
In sum, people appear to dislike their out-party – at least in part – because they see gobs of people who are stereotypical to that party and (on average) not like them when painting a mental picture of its supporters. When presented with information to the contrary, their attitudes toward that party depolarize a bit. To be fair, these effects are relatively small – Democrats in our “corrected” condition still rate Republicans quite low on the feeling thermometer, for example, and Republicans still aren’t particularly stoked at the prospect of gaining a Democratic son-in-law. While the estimated effects are not large, neither were our treatments especially powerful: they merely challenged the numbers that respondents reported to stand in for their implicit beliefs about party composition (e.g. that “a lot” of Republicans are rich). Future research on implications of the group theory of parties might use higher-impact experiments – perhaps with multimedia treatments, or even experiments in the lab or field involving interpersonal contact – to better understand how citizens’ mental images of the parties shape their attitudes and behaviors.
Principled Dislike, or Raw Affect?
How should we interpret these results? To many, they show a path toward calming a fraught political climate. According to this Pollyanna interpretation, if we could export our treatment to the real world and show Democrats and Republicans that the other side is not so different, we might mitigate polarization.
But others note the Cassandra interpretation: even if a real-world analog to our treatment made Democrats and Republicans hate each other less, we’d run a real risk of increasing social polarization on other dimensions. Telling Democrats, “Actually, only a third of Republicans are evangelical” – or telling Republicans, “Only a quarter of Democrats are black” – may lead folks to see the other side as more like them (on average) and thus reduce partisan anger, but in doing so it could also stoke religious discrimination, racism, and other flavors of group-centrism by further marginalizing already-marginalized groups. A real-world treatment could even reduce these groups’ representation in the parties and leave their members feeling politically homeless. 8 Before applying this research to real-world politics, we need to better understand how citizens’ perceptions of party composition drive their partisan attitudes.
To a great extent, the Pollyanna–Cassandra debate boils down to whether policy mediates the relationship between perceptions of party composition and affective polarization. When Democrats dislike Republicans because they’ve overestimated the share of evangelicals in the GOP, is it because they dislike evangelicals’ political causes (like school prayer and abortion restrictions)? Or is it simply that Democrats downright do not like Bible-thumpers?
If the pathway to partisan animus we identified is genuinely rooted in policy preferences, then the Pollyanna interpretation is likely correct. But compelling evidence to the contrary exists. For one, Iyengar, Sood, and Lelkes (2012) find meager support for policy’s influence on party affect, instead finding more for elite-mass transmission of polarization through campaigns. Additionally, Mason (2015, 2016) presents evidence that affective polarization and policy extremity are distinct, and further that policy commitments are not necessary for partisans to polarize as they have over the past two decades. Finally, in a paper published in 2014, I found that people’s perceptions of ideological polarization in the mass public can affect the extremity of their issue positions – but these perceptions of ideological extremity did not affect feelings toward self-professed liberals and conservatives, at least not in this experimental context (Ahler 2014). If policy commitments explained the relationship between social perceptions of the out-party and mass polarization, one would expect to see stronger evidence in the existing record.
But even if the Pollyanna interpretation is correct, we should not discount the role that identity politics plays in shaping partisan affect. Consistent with arguments that citizens reason about politics and policy through group-related heuristics (e.g. Lupia 1994), correcting the misperceptions we’ve noted leads partisans to see their out-party as less extreme on policy (Ahler and Sood 2018b). That is, people’s perceptions merely of who partisans are appear to color their beliefs about the typical Democrat’s and Republican’s issue positions. So even if a principled, policy-grounded basis for affective polarization exists, it’s quite likely to be overstated in citizens’ minds – since they tend to overestimate the prevalence of party- stereotypical partisans – and, thus, to give rise to undue partisan animus.
There are likely multiple roads to polarization. For some citizens, partisan affect may be a gut-level orientation (Iyengar and Westwood 2014). For others, it may reflect policy concerns. But for the bulk of citizens who are not true ideologues (e.g. Converse 1964; Fiorina, Abrams, and Pope 2005; Ellis and Stimson 2012; Ahler and Broockman Forthcoming), orientations toward groups in society – and perceptions of how those groups compose the parties – are liable to affect feelings toward Democrats and Republicans, one way or another.
The “New” Identity Politics?
Identity politics is nothing new. Beyond recognizing that, we must reject the notion that mass partisan conflict, for most of modern American history, has been loftier than base group affect. Mass partisanship is more complex than most identities. It exhibits the features of a social identity, as other essays in this issue note, but it also appears to be rooted in citizens’ orientations toward other, more fundamental groups in society. In this sense, Donald Trump’s campaign was not significant for bridging identity politics and party politics. Rather, it changed the political landscape by making explicit the implicit basis for citizens’ partisanship – our beliefs and feelings about groups in society.
Above, I presented evidence suggesting that citizens’ beliefs about how party-stereotypical groups compose the parties fuel polarization. In general, partisans tend to dislike groups that are representative of the out-party. They also – like all citizens – overestimate those party-representative groups’ shares in “their” party, and the more they do, the greater animus they feel toward the other side and its supporters. This is the essence of polarization among citizens (Iyengar, Sood, and Lelkes 2012). While elite political conflict is said to be increasing for ideological (e.g. McCarty, Poole, and Rosenthal 2006) or strategic (Lee 2009, 2016) reasons, citizens dig deeper into the trenches that have always torn us apart – long-standing cleavages based on race and ethnicity (Tesler and Sears 2010), faith and reason (Putnam and Campbell 2012), primal orientations and group affect (Hetherington and Weiler 2009).
We all like to claim that our partisanship is based on independently-reasoned policy preferences, but the evidence simply does not support our collective number of claims. Knowledge of public policy and the parties positions’ tends to be woeful (e.g. Achen and Bartels 2016). Citizens are more likely to adopt their party’s positions than to change their partisanship to match their previously-expressed preferences (Layman, Carsey, and Horowitz 2006; Lenz 2012). The incipient feelings we develop toward political symbols as children persist throughout our lives (Sears 1993). And, ultimately, partisan affect appears to be grounded more in identities than in policy preferences (e.g. Campbell et al. 1960; Mason 2015, 2016). We like to show the best, most reasoned part of ourselves to the world, but that is not necessarily the truest part, especially because we have an incredible capacity to justify gut-level conclusions ex-post (Haidt 2001). So while pundits have deemed 2016 the year that identity politics captured the parties, I argue that identity politics was always there. 2016 merely made its role in partisan politics baldly apparent.
Abramowitz, Alan I. 2010. The Disappearing Center: Engaged Citizens, Polarization, and American Democracy. New Haven, CT: Yale University Press. Google Scholar
Achen, Christopher H., and Larry M. Bartels. 2016. Democracy for Realists: Why Elections Do Not Produce Responsive Government. Princeton, NJ: Princeton University Press. Google Scholar
Ahler, Douglas J., and David E. Broockman. Forthcoming. “The Delegate Paradox: Why Polarized Politicians Can Represent Citizens Best.” The Journal of Politics. Google Scholar
Ahler, Douglas J., and Gaurav Sood. 2018a. “Measuring Perceptions of Shares of Groups.” In Misinformation and Mass Audiences, edited by Brian G. Southwell, Emily A. Thorson, and Laura Scheble. Austin, TX: University of Texas Press. Google Scholar
Ahler, Douglas J., and Gaurav Sood. 2018b. “The Parties in Our Heads: Misperceptions About Party Composition and Their Consequences.” The Journal of Politics 80 (3). Google Scholar
Berinsky, Adam J., Gergory A. Huber, and Gabriel S. Lenz. 2012. “Evaluating Online Labor Markets for Experimental Research: Amazon.com’s Mechanical Turk.” Political Analysis 20 (2): 351–368. CrossrefGoogle Scholar
Broockman, David E. 2016. “Approaches to Studying Representation.” Legislative Studies Quarterly, Forthcoming. Working paper. Google Scholar
Campbell, Angus, Philip E. Converse, Warren E. Miller, and Donald E. Stokes. 1960. The American Voter. Chicago: University of Chicago Press. Google Scholar
Capehart, Jonathan. 2017. “One Democrat knew Trump would win. Now she struggles to find a place in her own party.” The Washington Post 20 June. https://www.washingtonpost.com/blogs/post-partisan/wp/2017/06/20/one-democrat-knew-trump-would-win-now-debbie-dingell-struggles-to-find-a-place-i?utm_term=.c1022eaaff01.
Converse, Philip. 1964. “The Nature of Belief Systems in Mass Publics.” In Ideology and Discontent, edited by David Apter. New York: Free Press. Google Scholar
Edsall, Thomas B. 2017. “Donald Trump’s Identity Politics.” The New York Times 24 August. Google Scholar
Ellis, Christopher, and James A. Stimson. 2012. Ideology in America. New York: Cambridge University Press. Google Scholar
Fiorina, Morris P. 1981. Retrospective Voting in American National Elections. New Haven and London: Yale University Press. Google Scholar
Fiorina, Morris P., Samuel J. Abrams, and Jeremy Pope. 2005. Culture War? New York: Pearson Longman. Google Scholar
Freeder, Sean, Gabriel Lenz, and Shad Turney. Forthcoming. “The Importance of Knowing ‘What Goes With What’: Reinterpreting the Evidence on Policy Attitude Stability.” The Journal of Politics. Google Scholar
Gamm, Gerald H. 1989. The Making of New Deal Democrats: Voting Behavior and Realignment in Boston, 1920-1940. Chicago: University of Chicago Press. Google Scholar
Green, Donald, Bradley Palmquist, and Eric Schickler. 2002. Partisan Hearts & Minds: Political Parties and the Social Identities of Voters. New Haven, CT: Yale University Press. Google Scholar
Greene, Stephen H. 2004. “Social Identity Theory and Party Identification.” Social Science Quarterly 34 (2): 412–419. Google Scholar
Hetherington, Marc J., and Jonathan D. Weiler. 2009. Authoritarianism and Polarization in American Politics. New York: Cambridge University Press. Google Scholar
Hetherington, Marc J., and Thomas J. Rudolph. 2015. Why Washington Won’t Work: Polarization, Political Trust, and the Governing Crisis. Vol. 104. Chicago: University of Chicago Press. Google Scholar
Hopkins, Michael Starr. 2017. “Republicans and Their Identity Politics are Destroying America.” The Hill 6 Sep. http://thehill.com/blogs/pundits-blog/national-party-news/349428-republicans-and-their-identity-politics-are-destroying.
Huddy, Leonie, Lilliana Mason, and Lene Aaroe. 2015. “Expressive Partisanship: Campaign Involvement, Political Emotion, and Partisan Identity.” American Political Science Review 109 (1): 1–17. CrossrefGoogle Scholar
Iyengar, Shanto, and Sean Westwood. 2014. “Fear and Loathing Across Party Lines: New Evidence on Group Polarization.” American Journal of Political Science 59 (3): 690–707. Google Scholar
Judd, Charles M., and Bernadette Park. 1988. “Out-Group Homogeneity: Judgments of Variability at the Individual and Group Levels.” Journal of Personality and Social Psychology 54 (5): 778–788. CrossrefGoogle Scholar
Layman, Geoffrey C., Thomas M. Carsey, and Juliana Menasce Horowitz. 2006. “Party Polarization in American Politics: Characteristics, Causes, and Consequences.” Annual Review of Political Science 9: 83–110. CrossrefGoogle Scholar
Lee, Frances E. 2009. Beyond Ideology: Politics, Principles, and Partisanship in the U.S. Senate. Chicago: University of Chicago Press. Google Scholar
Lee, Frances E. 2016. Insecure Majorities: Congress and the Perpetual Campaign. Chicago: University of Chicago Press. Google Scholar
Lenz, Gabriel S. 2012. Follow the Leader? How Voters Respond to Politicians’ Policies and Performance. Chicago: University of Chicago Press. Google Scholar
Lilla, Mark. 2017. “How the Modern Addiction to Identity Politics has Fractured the Left.” New Statesman 18 Sep. https://www.newstatesman.com/politics/uk/2017/09/ how-modern-addiction-identity-politics-has-fractured-left.
Lupia, Arthur. 1994. “Shortcuts Versus Encyclopedias: Information and Voting Behavior in California Insurance Reform Elections.” American Political Science Review 88 (1): 63–76. CrossrefGoogle Scholar
Mason, Lilliana. 2015. ““I Disrespectfully Agree”: The Differential Effects of Partisan Sorting on Social and Issue Polarization.” American Journal of Political Science 59 (1): 128–145. CrossrefGoogle Scholar
McCarty, Nolan, Keith T. Poole, and Howard Rosenthal. 2006. Polarized America: The Dance of Ideology and Unequal Riches. Cambridge, MA: MIT Press. Google Scholar
McConnell, Christopher, Yotam Margalit, Neil Malhotra, and Matthew Levendusky. 2018. “The Economic Consequences of Partisanship in a Polarized Era.” American Journal of Political Science 62 (1): 5–18. CrossrefGoogle Scholar
Pew. 2014. Political Polarization in the American Public: How Increasing Ideological Uniformity and Partisan Antipathy Affect Politics, Compromise, and Everyday Life. Washington, DC: Pew Research Center. Google Scholar
Putnam, Robert D., and David E. Campbell. 2012. American Grace: How Religion Divides and Unites Us. New York: Simon & Schuster. Google Scholar
Rothschild, Jacob, Adam Howat, Richard Shafranek, and Ethan C. Busby. 2018. “Pigeonholing Partisans: Mass Stereotypes of Party Supporters and Partisan Polarization.” Working paper, Northwestern University. Google Scholar
Roush, Carolyn. 2017. “It’s Not Me, It’s You”: How Americans’ Animosity Toward Their Opponents Drives Modern Politics. PhD thesis Vanderbilt University. Google Scholar
Sears, David O. 1993. Symbolic Politics: A Socio-Psychological Theory. In Explorations in Political Psychology, edited by Shanto Iyengar and William J. McGuire, 113–149. Durham, NC: Duke University Press. Google Scholar
Stephens-Davidowitz, Seth. 2012. Everybody Lies: Big Data, New Data, and What the Internet Can Tell Us About Who We Really Are. New York: HarperCollins Publishers.Google Scholar
Tajfel, Henri, and John C. Turner. 2005. The Social Identity Theory of Intergroup Behavior. In Political Psychology. New York: Psychology Press. Google Scholar
Tesler, Michael, and David O. Sears. 2010. Obama’s Race: The 2008 Election and the Dream of a Post-Racial America. Chicago: University of Chicago Press. Google Scholar
WSJ Editorial Board. 2017. “The Poison of Identity Politics.” The Wall Street Journal 13 Aug. Google Scholar
In particular, they reached more than twice their previous all-time high relative volume in November 2016 and a similar level in August 2017. For the unfamiliar, stop reading this footnote now if you are not prepared to lose a few hours down a rabbit hole. Google Trends (trends.google.com) returns the relative interest in any topic you query, over the time period you specify, in the region you specify. Interest is measured as the frequency of Google searches for the keywords you enter. Note that you can only observe and compare relative popularity – you cannot observe the raw number of searches. See Seth Stephens-Davidowitz’s excellent treatise on big data, Everybody Lies, for more on this new approach for inferring human attitudes and behaviors.
An important finding is that party stereotypes tend to be quite consensual – Democrats, Republicans, and independents have similar beliefs about the types of people who belong to the two parties (Green, Palmquist, and Schickler 2002; Rothschild et al. 2018), and even how those groups compose the parties (Ahler and Sood 2018b).
One of these studies, discussed in detail below, was conducted with a sample that was representative of the Californian adult population. Our experimental studies were conducted on Amazon’s Mechanical Turk, a micro-task market for trading small services. Berinsky, Huber, and Lenz (2012) provide details about general features of the market. The labor pool is not representative of the broader American population, but we rely on these samples primarily for drawing causal inferences – e.g. the effects of misperceptions about out-party composition on feelings toward the out-party – rather than descriptive inferences about the typical American’s beliefs about party composition. (However, Mechanical Turk workers’ perceptions about party composition are quite similar to those we find in more representative samples.)
In this essay I present two comparison groups: a “corrected” group and an “uncorrected” group. Our design was actually somewhat more complex: among uncorrected respondents, half were explicitly asked for their perceptions of the stereotypical groups’ strength in their out-party, and these reports were not corrected. The other half were not asked about their perceptions about the out-party’s social composition until the end of the survey, well after the dependent measures had been collected. In all experiments we conducted, the mere act of asking about out-party composition appears to somewhat reduce partisan animus (or beliefs about out-party extremity), but these differences are not always significant.
This first difference is statistically significant at all conventional levels. The second is significant only with one-tailed tests, and its small size is worth noting. These differences are somewhat muted compared to those in Ahler and Sood (2018b) because, as noted in Footnote 5, the “uncorrected” condition here combines two experimental conditions, one of which potentially reduces partisan affect through its design.
As reported in Ahler and Sood (2018b), when summarizing these relationships with ordinary least squares regression (OLS), estimated slopes in the uncorrected condition are 0.00 and −0.00, respectively.
It’s worth noting that at the elite level, the parties polarized in part because of the Democrats’ embrace of Civil Rights and the death of the Dixiecrats with Jim Crow (e.g. McCarty, Poole, and Rosenthal 2006).