But Can She Make America Great Again? Threat, Stability, and Support for Female Candidates in the United States


Elizabeth N. Simas, University of Houston

Heading into Election Day 2016, major election forecasts indicated that Hillary Clinton was poised to become the first female president of the United States. Even the Los Angeles Times, a co-sponsor of perhaps the only national poll that showed a lead for Republican Donald Trump, predicted that Clinton would win with over 350 electoral votes. And yet, several key states carried by Barack Obama in 2012 (e.g. Florida, Ohio, and Pennsylvania) swung red, and Trump was named the president-elect. Both scholars and pundits have attempted to explain this outcome, with sexism and Clinton’s gender commonly arising as factors contributing to her defeat. While outright sexism may have harmed Clinton’s chances, I argue that her gender may also have worked against her in another, less overt way. More specifically, in a forthcoming article in Political Behavior, I argue that Trump’s calls to “Make America Great Again” may have triggered even the least sexist voters to desire male rather than female leadership.

My argument is rooted in system justification theory (SJT), which asserts that individuals have “social and psychological needs to imbue the status quo with legitimacy and see it as good, fair, natural, desirable, and even inevitable” (Jost, Banaji, and Nosek 2004, p.887; see also Jost and Banaji 1994). That is, people are motivated to defend the way things are in their social and political systems. This means that in the U.S., where male leadership is the norm in the political arena, [1]  high threat or persistent messages that America is in decline should decrease preferences for female politicians.

To test this theory, I embedded an experiment in a 1000-person module of the Cooperative Congressional Election Study. Respondents in this CCES module were randomly assigned to either the control group, which received no information, or one of two treatment groups that manipulated the level of system threat. The high threat treatment emphasized dissatisfaction with the current system and argued that the country has reached a low point in terms of social, economic, and political factors. The low threat treatment emphasized the opposite. After exposure to these treatments, individual were asked to rate Hillary Clinton on likability, sincerity, competence, and qualifications to be president. These four ratings were averaged to create one single score. As Figure 1 shows, those exposed to the high threat condition gave Clinton significantly lower ratings than those in the low threat condition. This difference is particularly evident among individuals who have a higher tendency toward system justification, further evidencing my general claim. 

Figure 1: Mean Trait Evaluations by Treatment Group and System Justification

simas*=different from the Low Threat Condition at p<.05; Plots represent group means and their 95% confidence intervals.

To test whether individuals had similar reactions in the voting booth, I also analyze their reported vote choice. In these analyses, I use perceptions of the performance of the national economy as a representation of perceived threat. The assumptions of the use of this measure are that those who respond that the nation’s economy has “gotten much better” should be akin to those exposed to a low system threat/economic stability treatment, while those who respond that the nation’s economy as “gotten much worse” should be similar to those exposed to a high system threat/economic instability treatment. My results are consistent with those from my experiment: as system justification increases, economic assessments have a much larger association with the probability of voting for Clinton. For an individual whose reported tendencies toward system justification high, a shift from the best to the worst economic assessment results in a significant .45 reduction in the probability of voting for Clinton. Additional analyses of both vote choice and evaluations of Clinton’s traits reveal that this occurs even among individuals who score low on modern or traditional measures of sexism. Thus, my results suggest that system justification motivations exerted unique effects and that sexism alone does not tell the whole story of how Clinton’s sex may have been a factor in the 2016 election.

But of course, Clinton is just one case. It may be the case that citizens would have failed to rally around any candidate with her issues and attributes, regardless of sex. In addition, Obama’s legacy and Clinton’s ties to the establishment may have made her the status quo and Trump the candidate change. In an attempt to address these limitations and bolster my argument, I draw on a second study conducted among students at the University of Houston in April 2017. The results from this study support my CCES findings in tow key ways. First, I find that those exposed to high vs. low threat give lower ratings to Clinton and higher ratings to Trump. Since these rating were given after Trump was serving as president, this helps address concerns that Clinton was viewed as the status quo in the previous analyses.

Second, an experiment using hypothetical mayor candidates finds that high system justifiers exposed to high vs. low threat gave significantly lower trait ratings to a Democratic candidate when she was portrayed as a woman rather than a man. Although we are somewhat limited in the conclusions we can draw from this single, hypothetical experiment, the results are still consistent with the idea that higher levels of system threat may harm a Democratic female candidate but help a Republican male candidate.

In total, my analyses consistently show that messages of threat vs. stability can at worst harm a female candidate, and at best simply remove advantages that she had over a male candidate. These findings not only shed additional light on the 2016 election, but also offer insights into the prospects for any female running for office. The resulting implication is somewhat of a catch-22 for female candidates. SJT implies that negative effects should diminish as more women get elected to office and female leadership becomes more of a norm, but yet, it may be difficult for women to do this if they continue to face bias when running during adverse economic or political conditions. Still, gender parity among elected officials in the U.S. is rising (albeit slowly) and thus, it is likely that a similar study conducted 25 years down the road would produce a different pattern results. Yet whether it does or not, the analyses presented here serve as an important reminder that outright sexism may not be the only factor working against female candidates.




[1] In the 116th Congress, men outnumber women 333 to 102 in the House and 75 to 25 in the Senate.

Changes in Candidate Evaluations over the Campaign Season: A Comparison of House, Senate, and Presidential Races

Patrick D. Tucker

Steven S. Smith

Campaigns devote considerable resources to persuading potential voters to support their own candidate over the opposition. Yet, less is well known about the dynamics of candidate support at the aggregate and individual level. Previous observational analysis finds that in presidential elections modest change in support occurs, while field experiments tend to demonstrate that campaigns are limited in their ability to persuade. These studies are limited in that they focus on either elections with large degrees of familiarity with candidates at the beginning of the campaign, or they focus on elections in which exposure to candidates is quite limited and familiarity is low. Thus, the questions remain, do citizens learn about candidates over the campaign season? Do they change their support for candidates from the beginning to the end of the campaign? And finally, if they do change their support for candidates, which types of voters and races are associated with change?

In this paper, we provide two contributions to the study of campaigns. First, we provide a direct comparison of change in candidate evaluations and support across House, Senate, and presidential races. As, these elections vary systematically in their level of competitiveness and salience, we expect citizens’ evaluations of candidates to vary across their type, as well. Second, with a unique panel that covers both the 2014 midterms and the 2016 general elections, we measure change in citizens’ evaluations of and support for candidates in these races.

Our data come from The American Panel Survey (TAPS). TAPS was an online nationally representative monthly panel administered by GfK-Knowledge Networks and the Weidenbaum Center on the Economy, Government, and Public Policy at Washington University in St. Louis from 2011 through 2018. At the end of each election’s primary campaign panelists were asked to evaluate the major party general election candidates and identify which candidate they would support in November. In the November wave, administered the day following the election, panelists were once again asked to provide their evaluations of the candidates and identify for which candidate they voted.

Over the course of the campaign, panelists significantly increased their certainty regarding the candidates’ policy positions and their placement on a traditional 7-point ideological scale. For example, in each wave panelists were asked to identify the policy positions of the candidates for 10 issues. In 2016, the average panelist provided 6.02 don’t know responses for a US Senate candidate at the end of the primary campaign. By November, however, this number had decreased to 5.04. We found improvement in willingness to identify positions at all levels of federal elections; while the average pre-general election average “don’t knows” for the presidential candidates was only 2.74 issues, by November the figure decreased to 2.07. These results were consistent with the percent of panelists who responded with “don’t know” when asked to place candidates on an ideological scale: for example, 41 percent of panelists were unwilling to place a Senate candidate on an ideological scale in the pre-general election period, while 27 percent of panelists did so immediately following the election.

While we were able to identify changes in evaluations and knowledge of candidates, we were less successful in finding transitions of support from one candidate to another. If panelists had announced their support for a candidate at the beginning of the general election campaign, they were remarkably unlikely to switch to the opposite party’s candidate. For example, in 2014 House races, 94.5 percent of those panelists who said they would vote for a given candidate at the end of the primary season, responded that they voted for the same candidate in the general election. This candidate-to-candidate transition was especially low in the 2016 presidential election. In our panel, only 1.6 percent of professed Clinton and Trump voters reported switching their vote to the other major party candidate in November. The greatest amount of movement over the campaign occurs among those who do not indicate support for a major party candidate at the beginning of the campaign. At the Senate and House level in both years, roughly one-third of all panelists who reported voting for a major party candidate in the general election indicated that they were undecided at the beginning of the general election campaign. For the presidential contest, this figure was much lower, 13.1 percent.

What predicts transition to a major party candidate? In a series of transition models estimated using a logit link function, we found mostly consistent support that presidential approval was a strong predictor of vote in House and Senate races. That is, initially undecided voters who approved of President Obama’s job performance were likely to vote for the Democratic candidate. Furthermore, we also found strong evidence that initially undecided voters and those voters who initially indicated that they supported the candidate of the opposite party were likely to vote for their own party’s candidate. That is, over the course of the campaign, voters tend to “come home” to their own party.

Our findings have important implications for the study of campaigns. We find that campaigns serve an important informational role for voters with respect to major party candidates, but we find little evidence that campaigns are associated with switching support from one candidate to another. Rather, our findings suggest that campaign season is associated with voters returning to their own parties, whether they were undecided, or if they were initially supporting the opposite party. Finally, we find that campaign effects are difficult to realize because voters tend to be most exposed to campaigns in which they have higher familiarity with candidates.




The Gendered Politics of Congressional Elections

Sarah Fulton, Texas A&M University

Kostanca Dhima, Texas A&M University


When Senator Elizabeth Warren was asked whether she believed that the men in the 2020 presidential race have a better chance of beating Trump solely because of their gender, she answered, “I believe they may think so, but they’d be wrong. […] What the data show now is that in competitive elections, women are out-performing men.”[1]

The question and Senator Warren’s answer reflects a paradox regarding the effects of candidate sex on voter evaluation: A variety of research shows that women are as likely as men to win,[2] and yet there is ample evidence that voters hold attitudes that disadvantage female candidates.[3] How can both of these things be true at the same time?

We argue that this paradox can be explained once we consider differences in candidate quality. Research suggests that barriers to entry for women in politics are greater than those for men – that women face a more difficult path to office and have to “work harder” to win elections,[4] resulting in more qualified female candidates on average.[5] Women also tend to hold themselves back as candidates until they view themselves as being able to win.[6]  So when we neglect differences in candidate qualification, it might appear like women do as well as men when they run but in reality, this only holds when women have a quality advantage. In other words, women do as well as men when they are more qualified but not when they are as qualified.

We demonstrate the effect of candidate quality on vote-choice and election outcomes for every two-party congressional race from 2006 to 2018. We find that female Democratic candidates are more qualified than their male counterparts, and that the qualification advantage explains why they are as likely (if not more likely) to attract the support of voters. When qualifications are held constant, female Democrats receive significantly fewer votes than their male co-partisans.

We also find that respondent gender and partisanship mediate the extent of the penalty, with male Republican and male independent voters being significantly more punitive. Female Democratic candidates can attract the support of male Republican and male independent voters when they have a qualifications advantage, but are penalized when they are merely “as qualified.”

Why would male Republican and male independents penalize female Democrats? Although we cannot say for certain what the nature of the penalty is, the penalty is not related to perceived ideological distance, to general perceptions that women are unfit for political office, and gender affinity. Since male Republican and male independents appear to be averse to something about female Democrats that fails to provoke a negative reaction from female voters (of any partisan stripe) and male Democrats, and fails to apply to female Republicans, one possible explanation is that “modern sexism” – resentment about men’s marginalization and loss of privilege – is at play. Female Democrats’ emphasis of gender egalitarian themes may cause a backlash among male voters, particularly Republicans and independents, but not female voters or Democrats, who are more likely to be sympathetic to those demands and/or to stand to benefit from them.

When we evaluate election outcomes, we again find that female Democratic candidates with a qualifications advantage are as likely as males to win elections; but are significantly less likely than males to win when qualifications are held constant. The proportion of male Republicans and male independents in a district determines the extent of the penalty, with women’s electoral prospects significantly declining as this proportion increases.

By showing that qualifications and male Republicans and male independents are consequential to women’s electoral fortunes both at the individual- and aggregate-level, we provide strong validation of the thesis that gender matters to voting behavior and election outcomes, over and above conventional explanations.

So yes, as Senator Warren responded, women can win, but they need to be highly qualified and strategic about the races in which they compete. Because male Republicans and male independent voters play a significant role in women’s electoral fortunes – over and above known factors in the literature – this group of voters should be taken into account by both operatives and scholars alike.



[1] https://www.cnn.com/videos/politics/2020/02/06/elizabeth-warren-new-hampshire-town-hall-male-candidates-odds-against-trump-sot-vpx.cnn

[2] Seltzer, Newman and Leighton, 1997; Darcy, Welch and Clark, 1994; Burrell, 1994; Darcy and Schramm, 1997; Lawless 2015

[3] Mendez and Osborne 2010; Dolan and Sanbonmatsu 2009; Schneider and Bos 2014; Fulton and Ondercin 2013; Sanbonmatsu 2002b

[4] Anzia and Berry 2011; Lazarus and Steigerwalt 2018; Pearson and McGhee 2013

[5] Fulton 2012; Fulton 2014

[6] Fulton et al. 2006

Electoral Participation in Iran’s Parliamentary Politics: Between Two Competing Explanations

Alireza Raisi, Emerson College

Why do citizens bother to vote in electoral authoritarianism when their votes do not change the main body of the ruling elite? The conventional wisdom of scholars commonly stresses the role of clientelistic exchanges in mobilizing citizens in electoral authoritarianism. In the broader context of transitional societies, scholars of comparative politics identify two general approaches to explain the determinants of voter mobilization. On one hand, some argue that the institutional setting, notably the electoral rule, determines the ways that politicians reach voters. On the other hand, the second approach highlights structural conditions such as the modernization stage as an underlying variable shaping the way that politicians mobilize voters.

Despite a body of empirical evidence from transitional democracies and authoritarian regimes, these studies highlight one determinant of voter mobilization at the expense of dismissing other elements and consequently overlook the complexity of electoral participation in authoritarian regimes. In particular, they downgrade the significant impact of micro-level power hierarchy and subnational state-building in shaping the pattern of mobilization. This paper studies the role of institutional and socioeconomic variables in shaping the pattern of participation in Iran’s parliamentary politics since the 2000s. Drawing from a statistical analysis of the parliamentary turnout, the study of several official reports and a historical examination of the pattern of mobilization in Iran, the paper argues that in provincial areas, subnational state machinery is the main driving force in Iran’s parliamentary politics. In contrast, parliamentary politics in urban areas has been mainly affected by national shifts in factional politics which ensue from controlled politics and programmatic policymaking. In local districts, Iran’s institutional setting, notably the electoral rule, accommodates the personal particularistic demands of local voters in MP-citizen linkage. In exercising the enormous discretionary power over the daily lives of provincial citizens, local state machinery impacts Iran’s parliamentary politics. This setting, which has been developed over an extended period of time, mediates the impact of structural variables such as socioeconomic indicators in Iran’s parliamentary turnout and creates institutional channels broader than the electoral system to determine mobilization in electoral authoritarianism. These findings of the paper align with theories that highlight the role of state-building in shaping the pattern of mobilization.

The paper has important contributions to the study of electoral authoritarianism. First, this study examines competing approaches in the study of participation in the particular context of Iran. The study concludes that the impacts of institutional and socioeconomic variables on the citizen-politician linkage and the pattern of participation are not mutually exclusive. Institutional arrangements may widen or narrow the scope of immaterial (such as personal) connections between citizens and politicians to accommodate particularistic demands. Second, despite scholarly efforts to study citizen-politician linkage in advanced industrialized nations and transitional democracies, this linkage in electoral authoritarianism has remained relatively unexamined. The findings of this paper shed light on the relationship between citizen and politician in Iran. Instead of clientelistic exchanges, the paper highlights the broader role of personal particularistic demands in shaping citizen-politician linkage in provincial districts.


Do Female Politicians Face Stronger Backlash for Corruption Allegations? Evidence from Survey-Experiments in Brazil and Mexico

Frederico Batista Pereira, University of North Carolina at Charlotte


Although very few women currently hold the positions of president and prime minister across countries, the twenty-first century has seen a rise in the proportion of women occupying the highest position of executive power across the globe. Along with this rise in female representation, we have also witnessed a trend of female chief executives struggling to maintain public support while facing corruption allegations and scandals. Two prominent cases are the recent impeachments of Dilma Rousseff in Brazil and Park Geon-hye in South Korea. The trend raises questions about whether female leaders receive differential treatment when they are perceived to be involved in corruption, especially since women are often seen by the public as more honest and trustworthy than men.

To investigate whether there is a stronger backlash for corruption against female leaders, and whether that feedback is related to beliefs about the purity and honesty of women relative to men, I conducted online survey-experiments in Brazil and Mexico, two large and diverse Latin American countries with high levels of public perceptions of corruption. The key difference between the two countries is the fact that, while Brazil has had a female president who was impeachment and perceived by many to be corrupt, Mexico has never elected a female president. Even though the simple two-case comparison between the two countries cannot offer a formal test of how context can moderate the gendered backlash effects, it can provide valuable insights about the phenomenon.

The survey-experiments provided participants with information about a hypothetical city councilor in the country. Participants received personal information (profession, marital status, and children) and a description of the experience and current job performance of the politician. The statement also informed participants that the politician received high evaluations and was awarded prizes for competency in office. The first experimental manipulation randomly assigned participants to read about a female or a male politician. The second manipulation randomly assigned participants to receive or not an additional statement including an allegation of corruption against the politician. The statement mentioned a hypothetical report from an organization investigating the politician’s past financial mismanagement of public resources and illicit enrichment.

The experiments also included three questions about different dimensions of the politician’s popularity, as well as questions about whether participants endorsed sexist views about women in politics, which included items on whether men are less honest than women, and if women have more purity than men. The main expectations that the studies seek to evaluate are whether the decrease in popularity due to the corruption accusation is larger for the female politician than for the male politician, and whether that drop is larger among participants who endorse stereotypes of women as more honest than men.

The results show that, while the decrease in popularity due to the corruption allegation is large in both Brazil and Mexico, the stronger backlash against the female politician is only observed in the latter. In Mexico, the decrease in popularity is six percentage points larger for the female politician than for the male politician. On the other hand, neither study finds that the backlash against the female politician accused of corruption is larger among the participants who endorse the stereotypes of women as pure and honest.

All in all, the findings are mixed about the differential backlash for corruption against female politicians. While the results from Mexico align with cross-national studies of executive approval in showing that corruption has larger negative effects on the popularity of female politicians, the null results in Brazil suggest that the backlash effect may depend on contextual factors. Among the many factors that may explain the differences between the two countries, three seem to be most salient. First, the impeachment of Rousseff could have provided a strong counter-example that reduced the relevance of the stereotypes about women’s purity and honesty. Second, recent anti-corruption efforts in Brazil may have decreased the overall levels of tolerance of corruption for both male and female politicians. Third, the availability of considerations about women at the local level may be stronger in Mexico, since the country has higher local-level female representation than Brazil. Nonetheless, the findings do not provide support for the claim the stereotypes of women’s purity and honesty explain the stronger backlash for corruption against female politicians in the contexts where it occurs. Future research is needed to understand the psychological mechanisms as well as the contextual factors driving the differential treatment against female politicians.






The Urban-Rural Gulf in American Political Behavior

James G. Gimpel

Nathan Lovin

Bryant Moy

Andrew Reeves

Expert and journalistic news coverage of recent U.S. elections regularly calls attention to the urban-rural divide in partisanship and voting. Commentators note that people living in the country commonly express different political preferences than those in the city. This is neither a new nor fleeting development as this divide has been present in politics for some time. Even so, one current of social scientific thinking downplays the role of space and place. Skeptics argue that if we had the proper inventory of individual characteristics, place characteristics such as urban and rural would evaporate, having no role to play in explaining political behavior. The urban-rural divide, they argue, is a shorthand description of what is more accurately characterized as a bundle of individual traits that describe voters in rural areas contrasted with a very different bundle of traits for those individuals in cities. This paper published in Political Behavior, challenges the idea that place and space are epiphenomenal, merely artifacts, or even bad substitutes for a proper accounting of individual level attributes.

Places commonly differ in terms of the composition of the people who reside there. No one is arguing about that fact. The question is whether composition is all there is, or whether place has an independent effect. Are two people, identical in their demographic profiles, likely to vary in their political behavior if one lives in a rural local and the other in an urban one? Investigating such a question has been difficult given the sample sizes of most surveys. Probability based sampling captures urban and suburban populations adequately, but rural areas are not well represented. With just a handful of respondents from the countryside, most polls lack the variation in geographic context that would enable convincing tests of the impact of place and geographic environment. We get around this limitation in by drawing upon pooled Gallup monthly surveys across a sixteen-year period to reach a sample size of over 100,000 respondents, including thousands of rural voters.

What motivates the urban-rural divide? We consider some mechanisms by which differences in urban and rural locales results in differences in political behavior. Cities act as engines of new ideas which often challenge more traditional orthodoxies. Sociological studies from the early twentieth century postulated that cities are favorable to greater psychological liberalism and are more progressive and accommodating of unconventional behaviors and belief. Others argued that big cities offered protection to those who have been stigmatized and bullied.

The basic idea is that all other things being equal, the residents of some locations will behave differently because of the characteristics of their locations. What citizens know and learn about politics is influenced by local settings and the social interactions within them, reinforced by repetition and routine. Importantly, a place or “neighborhood” effect registers an independent impact on any number of outcomes, controlling for individual characteristics. If such an effect does exist for political opinions and behavior, then, it should show up as statistically and substantively significant along with the usual individual characteristics long known to be related to those outcomes. But to offer a fair and reasonable test of place effects, one must represent geographies well by having a sufficient number of respondents situated in those places, something a metro area cluster sample may not do.

With this rationale in mind, we look at two attributes of geographic space closely tied to understandings of urban and rural difference: distance of respondents’ residences from the nearest urban centers; and the population density of the locations where they dwell. Distance limits interaction. Urban and rural populations differ, in large part, because the populations do not mingle regularly. Population concentration is thought to have a substantive impact on the psychological individualism of populations. The more populous a place is, the more people act in a reserved and indifferent manner toward one another, largely out of the need to limit the burden of getting to know large numbers of people. There is certainly liberation of expression in urban settings, explaining why they are havens of eccentricity and new ideas, but the cost appears to be loneliness and disorder. People don’t know their neighbors, and don’t much care. Conversely, relationships in rural areas have greater depth, and the neighbors care, but conformity pressures weigh more heavily, limiting freedom of expression.

In our data analysis it turns out that urban and rural differentiation in party identification cannot be reduced to an inventory of the commonly observed differences between the populations who inhabit these spaces.   Moreover, population density and distance exert an impact on party identification even after controlling for income, age, race and ethnicity, religious observance, and other items commonly employed in survey crosstabulations. The place effects are not so negligible, either. For example, two voters with otherwise similar backgrounds, one living in a city, the other living 165 miles outside the city, will differ in the probability of expressing Republican loyalty by about nine points. Density, similarly will alter the propensity to identify with the major parties, with those in the densest settlements fifteen points more likely to be Democrats than those living in the least dense settlements. Individual measures of income and race matter more than these place effects, but income and race are themselves shaped by the geography of settlement, suggesting that we are probably understating the impact of urban-rural location in the models we estimate, not overstating it.

In summary, people who occupy the same geographic spaces tend to be similar in their viewpoints, beliefs, and understandings. This is not explained away by virtue of the inherent properties of compositional indicators such as education, income, age, race and ethnicity, or religious practice. That there is selection into rural and urban locations does not deny the presence of a contextual effect, but is itself evidence that place does have an impact. This research points to the potent impact of socialization processes involved in the drawing of residents into particular neighborhoods and cities, and the ongoing conformity pressures that consolidate and reinforce local opinion majorities over enduring periods. Observers of contemporary U.S. politics should take note that local ecologies of mixed political opinion are not the norm.

In-group Love and Out-group Hate: White Racial Attitudes in Contemporary U.S. Elections


Ashley Jardina, Duke University

In the wake of the election of the nation’s first black president, Barack Obama, it became unequivocally apparent that despite Obama’s historic victory, the U.S. had not entered the post-racial era some more optimistic political commentators thought it had. Indeed, the record shows quite clearly that Obama was penalized electorally for his race, and over the course of his presidency, the American electorate became more, not less, polarized along partisan lines in part by way of white’s racial attitudes. What is more, the successful presidential campaign of Donald Trump, whose explicitly racist rhetoric was unrivaled by modern presidential candidates, officially put to an end any pretense that America had entered an age of racial peace and equality.

Many academics and observers of American politics have argued that Obama was punished, and Trump was rewarded, because of white American’s racial prejudice—the sense of outward hostility, resentment, or animus many white Americans possess toward people of color in the U.S., and particularly toward black Americans. In my recent contribution to Political Behavior, however, I propose that there are in fact two types of white racial attitudes that have mattered in recent years for American electoral politics: the out-group attitudes many whites have toward racial and ethnic minorities, and the sense of in-group favoritism or solidarity some whites have toward their own racial group. These attitudes are, to some extent, mutually exclusive. Many whites possess measurable degrees of out-group prejudice without also identifying with their racial group, and many white racial identifiers have low levels of out-group prejudice. In other words, there are two independent ways in which white racism can manifest—both via whites’ hostility toward people of color, and via whites’ desire to protect the power and privileges of their racial group. Both these attitudes, I argue, have influenced candidate support since at least the 2012 presidential election.

Most work in the social sciences has ignored the possibility that white Americans possess a racial identity or a sense of racial solidarity, arguing that because of their dominant status and their numerical majority, white Americans are able to take their race for granted. But in recent years, significant immigration to the U.S., demographic change, and the salience of racial and ethnic minorities in politics and the media, mean that some whites have come to see their group’s power and privileges threatened. As a result, white identity has become a salient political force. Obama, I suggest, is viewed as a source of that threat, and Trump, with his dog-whistle promises to “make America great again”, as an antidote. As a result, for whites, both out-group prejudice and in-group identity were attitudes brought to bear on their candidate evaluations in 2012 and 2016.

To test these claims, I draw on evidence from the 2012 and 2016 American National Election Study (ANES) Time Series Studies, and the 2016 and 2018 ANES Pilot studies—all nationally representative public opinion surveys. Using each survey, I examine the relationship between white racial resentment—a measure of anti-black affect coupled with the belief that black Americans do not subscribe to individualism—white racial identity, and both candidate evaluations and vote choice. I find that in 2012, even after taking into account other standard predictors of vote choice like partisanship, education, age, gender, and evaluations of the national economy, both white racial resentment and white racial identity were important predictors of vote choice, with more racially prejudiced and more racially identified whites significantly less likely to support Obama.

Predicted Vote Choice in 2012 by White Identity and Racial ResentmentSlide1

Using the 2016 ANES Pilot study, which was conducted in January of 2016 during the presidential primaries, I also find that both types of white racial attitudes were predictive of supporting Trump over any of the other Republican primary candidates. Trump was, it seems, especially unique in appealing to both sets of white Americans.

The Predicted Probability of Preferring Trump over other Republican Primary Candidates in 2016Slide2

Finally, I find that by the 2016 general election, racial resentment remained a strong predictor of vote choice, but white identity’s effect had been somewhat attenuated, perhaps by the growing number of media stories linking white identity politics to white supremacy. Nevertheless, we should not necessarily expect that this trend marks the end of white racial identity’s electoral power. In today’s increasingly polarizing environment, and as the U.S. continues to become more racially diverse, politicians will likely continue to appeal to both sets of attitudes, making the American public increasingly immune to any normative challenges to these appeals. We may therefore likely see both attitudes continue to matter in American politics for the foreseeable future.

Biases at the ballot box: do ethnic minority candidates get fewer votes?

Nicole S. Martin (University of Manchester) and Scott Blinder (University of Massachusetts, Amherst)

Ethnic minorities are an increasing proportion of voters and politicians in most established Western democracies, and the UK is no exception. Some research suggests that ethnic minority candidates can face a penalty at the ballot box. For example, one study showed that minority candidates in the 2010 UK general election – especially Muslim candidates – got about four percentage points fewer votes from white British voters than white British candidates did[1]. But much has changed in British politics since 2010, and there might be other reasons for this penalty (e.g. if Muslim candidates stood in worse seats than other candidates in 2010). We decided to investigate this using an experiment, where we randomly changed various things about hypothetical candidates – such as their ethnicity, policies, and gender – and measured how that affected whether people said they would vote for them or not.

We found that Pakistani candidates did receive fewer (hypothetical) votes purely on account of their ethnicity – about 6 points less than a white British candidate. Because this was an experiment, we can be sure that nothing else was different about these candidates except their ethnicity. Hearteningly, we did not find an overall vote penalty for black Caribbean candidates. This is because some voters react positively – and other negatively. Some voters are more inclined to support a black Caribbean candidate than a white British one, all else being equal. These voters (one quarter of our sample) like immigration, and have a strong commitment to not acting in a prejudiced way. On the other hand, other voters are disinclined to support a black Caribbean candidate. For instance, we found that a black Caribbean candidate would receive on average 3 points less than a white British candidate among people intending to vote Conservative in the 2017 general election.

Direct discrimination is only one reason to care about candidate’s ethnicity however; evidence from multiple countries suggests that who politicians are matters for what they do when they are in parliament. For example, ethnic minority MPs tend to ask more questions about immigration and minority rights than white British MPs[2]. Scholars call this ‘substantive representation’ of a group’s interests – rather than simply sharing the same demographic characteristic, which is called ‘descriptive representation’. Alongside straightforward prejudice, this might also explain why some white voters do not support minority candidates. Views differ starkly between white British and ethnic minority voters on (i) immigration and (ii) racism. Ethnic minorities are on average more likely to favour liberal immigration policy, and to think that the government should do more to address racism. We wanted to know if a candidate’s position on these issues affected their changes of being elected.

It turns out that voters care more about these policies than candidate ethnicity. A candidate who said that the UK should accept more refugees received 14 points less than a candidate with the most popular immigration policy. Perhaps surprisingly, this wasn’t the most restrictive immigration policy (“strongly limiting migration”); voters preferred candidates who wanted high-skilled immigration in shortage sectors. Voters also preferred candidates who wanted to enforce anti-social behavior laws to those who would prioritise racial equality laws. Since ethnic minorities in Britain are more supportive of refugees and concerned with anti-discrimination laws, this suggests that candidates promoting pro-minority policies might face an electoral backlash.

Ethnic minority candidates received more flack for expressing pro-minority and pro-migration positions; instead of the 4-point penalty for white British candidates supporting the enforcement of racial equality laws (instead of anti-social behaviour laws), Pakistani candidates were penalised 7 points, and black Caribbean candidates 8. This suggests that not only does the balance of opinion among voters rest against the preferred policy position of ethnic minority voters, but politicians from minority backgrounds who campaign to represent minority views on these issues will face higher hurdles in doing so than white campaigners.

Parties and campaigners have taken various approaches to increasing the diversity of MPs, including all-women shortlists, the Conservative party ‘A list’ of preferred candidates, and mentoring schemes. We tested whether voters cared about such efforts. Some candidates were described as having become a candidate after “getting involved in the political party”, whereas others as having been “included on a list of candidates from under-represented backgrounds”. The voters in our experiment preferred the candidate who had simply got involved to the one on the list by 4 points. Interestingly, this applied regardless of the candidate’s ethnicity.

Ethnic minority women can face a particularly high barrier in getting into politics, as they can face discrimination on the basis of both their gender and their ethnicity. One report estimated that Diane Abbott (a prominent Labour politician who is black) received 45% of the abusive tweets sent to women MPs in the six weeks before the 2017 UK general election.[3] Our results are slightly different however. We do not see any evidence overall of voters applying a double penalty for ethnic minority women, nor any penalty or advantage to being a female candidate in general.

After the December 2019 general election, 65 out of 650 MPs were from an ethnic minority background[4], which is in line with the 1 in 10 voters from an ethnic minority background[5]. Why then does it matter that some voters are disinclined to elect minority MPs? As we have seen, some voters are especially supportive of minority candidates, so doesn’t it even out in the end? We have two responses to this. Firstly, it took a long time for ethnic minority MPs to be elected in numbers matching the share of voters, and minorities remain under-represented in other parliaments across Europe and North America. Secondly, we believe that our experiment to candidates from different ethnic backgrounds can also tell us about how people from different ethnic backgrounds are accepted in society or not. Our results suggest that British voters are divided on whether ethnic minority MPs are a positive goal to be pursued, an outcome to be avoided, or an irrelevant consideration.

Slide1Figure 1: Average marginal effects of different characteristics on the likelihood of voting for a hypothetical candidate.

[1] Fisher, Stephen D., Anthony F. Heath, David Sanders, and Maria Sobolewska. “Candidate ethnicity and vote choice in Britain.” British Journal of Political Science 45, no. 4 (2015): 883-905.

[2] Saalfeld, Thomas, and Daniel Bischof. “Minority-ethnic MPs and the substantive representation of minority interests in the House of Commons, 2005–2011.” Parliamentary Affairs 66, no. 2 (2013): 305-328.

[3] https://www.amnesty.org.uk/online-violence-women-mps

[4] https://www.bbc.co.uk/news/election-2019-50808536

[5] https://www.runnymedetrust.org/uploads/2017%20Election%20Briefing.pdf

Negative and Positive Partisanship in the 2016 U.S. Presidential Elections


Alexa Bankert, University of Georgia

The 2016 presidential election was marked by widespread negativity: Instead of enthusiastically embracing their party’s presidential nominee, many voters used their vote to signal their opposition to the other party. In a poll conducted by the Pew Research Center in the months leading up to the 2016 elections, Americans overwhelmingly reported feeling angry with the other party while a much smaller share cited feeling enthusiastic about their own party. This asymmetry is not just an election artifact: In February 2018, another nationally representative poll revealed that 71% of Republicans and 63% of Democrats cite harm from the opposing party’s policies as a major reason to affiliate with their own party. These polls are indicative of the rise of negative partisanship among Americans.

The term ‘negative partisanship’ captures the notion that for some partisans, strong animosity towards the out-party is more central in shaping their political behavior than the endorsement of their own party. Thus, the partisanship is negative in the sense that it centers on the rejection of the out-party and its members, turning the exclusion from the political group – the “not being one of them” – into a meaningful political identity. Political scientists and political pundits alike have long treated negative partisanship as an inevitable byproduct of strong party attachments (i.e. positive partisanship) assuming that strong partisans are bound to disdain the other party.

In my Political Behavior article, I challenge this notion and develop as well as empirically validate a scale to measure negative partisanship. This scale entails items such as “I do not have much in common with supporters of this party” and “When people criticize this party, it makes me feel good”. I demonstrate that negative partisanship is, in fact, fairly distinct from positive partisanship. Moreover, both types of partisanship affect Americans’ political behavior in quite different ways. For example, opposition to bipartisan efforts is predicted by negative partisanship rather than positive partisanship, suggesting that strong partisans are not condemned to demonize the other party. At the same time, positive partisanship is the main driver of political engagement, which provides the reassuring insight that most Americans still become politically active to advance their in-party rather than to harm the other party.

Voting decisions, on the other hand, are more complicated, especially in a two-party system where a vote for one political party might feasibly be a vote against the other. My analyses show that among Democrats, the decision to vote for Clinton in the 2016 U.S. presidential elections was influenced by partisans’ attachment to the Democratic Party instead of their disdain for the Republican Party. In contrast, Republicans’ vote for Trump was influenced by both negative and positive partisanship. These results are aligned with survey data by the Pew Research Center revealing that 53% of prospective Trump voters said their vote was primarily a vote against Clinton while 44% said their vote was a vote for Trump. Democratic voters, on the other hand, did not report a similar split.

This asymmetry might be grounded in two different factors: First, among Democrats, Trump was not seen as a credible threat to a Democratic Party’s electoral chances. From this perspective, Republicans might have been more likely to develop a negative partisan identity since a victory of their own party was portrayed as unlikely. Second, party leaders can foster particular types of identity: The rhetoric of the Trump campaign emphasized differences between the Democrats and Republicans while the rhetoric of the Clinton campaign underscored unity (i.e. “stronger together”). Highlighting connections to similar others reduces the tendency to derogate the out-group.

Interestingly, these results flip for Democrats in the 2018 U.S. House elections: Positive partisanship is no longer predictive of Democrats’ vote decisions. Instead, negative partisanship – the rejection of the Republican Party – is the strongest predictor of voting for a Democratic Member of the House. Among Republicans, on the other hand, both positive and negative partisanship remain predictive of their House vote.

Future research might investigate the mechanisms that lead to the development of negative partisanship independently from or in conjunction with positive partisanship. While positive partisanship might facilitate the development of negative partisanship, it is not clear to what extent the reverse can be true. It seems feasible though that negative partisanship can, for example, serve as a superordinate identity that unites a diverse coalition of voters and thereby temporarily facilitates political mobilization. This could be the case for the Democrats in the 2020 general elections.

In conclusion, partisans are able to feel deeply connected to their party without feeling deep disdain for the other party. From a normative standpoint, this is good news for political parties in representative democracies and provides a strong impetus for political scientists to explore the factors that allow partisans to root for their team without vilifying their opponent.

Fake Claims of Fake News: Political Misinformation, Warnings, and the Tainted Truth Effect


Melanie Freeze and Mary Baumgartner, Peter Bruno, Jacob R. Gunderson, Josh Olin, Morgan Quinn Ross, and Justine Szafran

With the explosion of fact-checking organizations in the US over the last two decades and the strategic use of warnings by political elites on both sides of the aisle, misinformation warnings are becoming increasingly prevalent. One study found a quarter of Americans read a fact-checking article in the run-up to the 2016 presidential election. Warnings of fake news and tips on how to combat misinformation can be found on metro trains (see picture taken in Singapore by one the authors), media coverage, and presidential tweets. Political misinformation is deeply problematic for democracy and needs more attention and effort to counter its negative effects. However, our research reveals misinformation warnings themselves may also have a dark side as they can lead people to reject accurate information and may ultimately damage democratic processes.

Studies of eyewitness testimony in the field of social cognition and forensics have long been interested in exploring the influence of misinformation and warnings on memory, especially as they relate to eyewitness memory. The tainted truth effect occurs when misinformation warnings taint accurate information and lead people to discard correct information or, as the idiomatic expression states, “throw the baby out with the bathwater.”

Our research contributes to the relatively understudied topic of the tainted truth effect. We extend social cognition research on the tainted truth effect to a political setting, where post-event misinformation and information is presented in the form of political news. We first ask, after viewing a political event, how does later exposure to information or misinformation in a news article describing the political event alter individuals’ memory and recognition of the details from the original event? Second, when individuals are retrospectively exposed to a valid warning that the news article contained misinformation, are they able to discard the misinformation and remember the correct original event information? Third, do people discard accurate data when exposed to invalid warnings of misinformation? While all three research questions address important aspects of individual memory and information processing, the third question regarding the potential drawbacks of misinformation warnings, formally referred to as the tainted truth effect, is the focus of our research.

In our online survey experiment, 434 participants drawn from MTurk were asked to watch three one-minute floor speeches presented by relatively obscure members of Congress from both parties. The speeches covered a diverse range of issues from healthcare, foreign policy, and local infrastructure politics. Following a buffer period, participants were randomly assigned to one of six possible conditions (3 X 2 factorial design) that varied the content of the original event description presented in the news article (misleading, vague, or informative) and the presence of misinformation warning after the news article (warning, no warning). Finally, people were presented with a series of questions that assessed their memory of facts presented in the video clip and their perception of the information source.

Measures of memory, memory certainty, and perceptions of source credibility were calculated from these responses and the warning and description effects tested in a series of OLS regression models. Although the warning effects in the information and misinformation conditions are not statistically different from that of the control condition, warning effects do emerge within the post-event description treatment conditions and the differences produced by warning are significant across the two treatment conditions. Figure 1 presents the effect of warnings, both valid and invalid, on memory score for the experimental and fixed fact subsets. The tainted truth effect can be seen in the negative effect of warning on memory represented by the square/blue estimates. When warnings target or spill over to taint accurate information, individuals appear to reject that information even though it accurately described the original event.  In contrast, valid warnings had a positive effect on memory as they helped individuals discard false information. People who received a valid warning (green/triangle estimates) were significantly less likely to select misleading information than those who were not warned and more likely to correctly remember the original event facts.

Slide1Figure 1: Marginal Effect of Warning on Memory by Description Condition

Supporting the tainted truth effect, we also find evidence that retrospective, invalid misinformation warnings lead individuals to view the news as less credible (see Figure 2) and these invalid warnings are also associated with more memory uncertainty. Our findings generally align with the few studies that have previously examined this topic. However, our use of a diverse subject pool and political context provides insights into the influence of misinformation warnings on memory, memory uncertainty, and the perceived credibility of news that has been discounted by misinformation warnings. At face value, misinformation threatens democratic proceedings if it can influence and shape public opinion and social decisions. Yet our findings suggest we must also be careful to not ignore the negative consequences of invalid misinformation warnings, most notably their ability to reduce the credibility and acceptance of legitimate news.

Slide2Figure 2. Average Credibility of Post-event Description (Article) by Condition