Racial Isolation Drives Racial Voting: Evidence from the New South Africa

Daniel de Kadt

Melissa L. Sands

Why do voters support parties that are closely aligned with their racial identities? One set of under-explored hypotheses relate to the racial context in which voters reside. Voters’ lived experiences – who they encounter and interact with – may have profound consequences for their politics, preferences, and behaviors.

The case of South Africa, which transitioned to multiparty democracy in 1994 after almost 100 years of state-led segregation under colonial and apartheid rule, provides a unique laboratory for studying such context effects. Under the 1913 Natives Land Act, and subsequent legislation in the 1930s and 1950s, the vast majority of South Africa’s land was reserved for its minority white population. Black African, Coloured, and Indian South Africans were forced into disproportionately small, isolated spaces.

The collapse of apartheid, and the repeal of these segregationist laws, heralded massive shifts in the spatial distribution of different people in South Africa. Previously reserved white areas, especially those in urban and suburban spaces, became demographically mixed to differing degrees. Numerous factors shaped the migratory decisions made by black South Africans; for example, physical geography made certain spaces more likely to experience integration than others.

Given this context, we study whether white South Africans – the demographic minority but socio-economically dominant group – behave differently at the ballot box as a result of the racial contexts in which they live. To do so, we leverage data from a variety of sources. We combine data from South Africa’s final apartheid-era census in 1991 with geographic, census and electoral data from the post-apartheid period. We are thus able to examine how the demographic composition of South African neighborhoods, conditional on apartheid era demographics, is associated with electoral returns for different parties. We then combine a high-resolution cross-sectional survey dataset from South Africa’s Human Sciences Research Council with contemporary census data to test whether individual voting intentions and attitudes are associated with an individual’s demographic context.

Across both datasets we find that white South Africans who live in segregated areas – amongst almost exclusively other whites – are far more likely to support “white parties” like the Democratic Alliance or the Freedom Front Plus, than white voters who do not. Using natural geography as a statistical instrument, a difference-in-differences design, placebo tests using individual measures of racial resentment, and unique data on title deeds transfers for the immediate post-apartheid period, we show that the results are likely causal, and not simply the result of omitted variables bias or residential sorting.

Apartheid ranks among the most profound and disturbing acts of state-led social engineering in history, and our study provides new insights into its long term consequences. There is a great need for empirically rigorous work on the legacies of exclusionary economic and geographic structures. Our paper begins to unpack these consequences, providing insights and avenues for future research, and the newly constructed data sets will provide exciting opportunities for social scientists. Yet the findings have implications outside of South Africa; there are fundamental similarities between apartheid South Africa and many other countries. Racial isolation and segregation remain ubiquitous worldwide, and our study highlights the importance of social diversity in encouraging diverse politics.


How Labor Unions Increase Political Knowledge: Evidence from the United States

David Macdonald

Florida State University | @davidTmacdonald


Do labor unions inform their members? Despite their decline over the past several decades, unions remain an important political actor. Past research has found that unions have important political implications. For instance, union members, particularly those with less income and less education, are much more likely to participate in politics and more likely to turn out to vote. Stronger unions also reduce economic inequality, and lead to more equal representation of citizens by elected officials. Because unions are prominent political actors, it seems plausible that they can also increase their members’ political knowledge as well. However, we lack direct evidence on this relationship.

I argue that unions increase their members’ political knowledge, and that this occurs through two processes. First unions provide their members with direct sources of information through emails, newsletters, direct mail, campaign mobilization, blog posts, and social media. Second, unions reflect a workplace environment in which political discussion is more likely to occur. This means that union members, relative to their non-unionized counterparts are much more likely to be exposed to an information environment that facilitates learning about politics.

To test this, I use data from recent national election surveys, primarily the 2012 American National Election Study (ANES). I show that union members, particularly those without any college education, are significantly more politically knowledgeable than their non-union counterparts. I attribute this to the rich information environment that labor unions provide, and the efforts that they take to provide their members with political information. This helps to reduce the costs of seeking out political information, something that can be quite daunting for people with less formal education.


I used OLS regression to examine the relationship between union membership and political knowledge, across varying levels of educational attainment. Results from Figure 2 show that non-college educated union members are significantly more political knowledgeable than their non-union counterparts. Though the effect size is relatively modest, it is certainly not trivial, as shown in Table 4. Indeed, union membership reduces the “knowledge gap” between people without any college education and those with a college degree by approximately 34 percent. Results also show that this relationship is present in private and public sector unions, in the 1970s and 1980s as well as the 1990s and 2000s, and is stronger in states without right to work (RTW) legislation, where unions are larger, better organized, and more politically active.


This work has important implications. Organized labor has declined dramatically over the past several decades, due in part to economic globalization, but also by the policy decisions made by the federal and state governments. Of all the factors that are correlated with political knowledge, such as: age, education, gender, race, income, and interest in politics, union membership is the only one that can feasibly be influenced by politicians. Policies that weaken labor unions may end up depriving people, particularly those with less formal education, not only of a source of political mobilization, but also an important source of political information.

Partisan Dehumanization in American Politics

Erin C. Cassese, PhD

University of Delaware


Incivility in political discourse isn’t just a matter of insults and attacks – it is a multi-faceted phenomenon. One form that incivility takes is dehumanizing language and political rhetoric. Dehumanization involves the denial of human qualities and characteristics to individuals and groups, perceiving them instead as animalistic, mechanistic, or humanoid rather than fully human. Research in social psychology has linked dehumanization to negative intergroup attitudes and behaviors, including aggression and even violence. For example, research shows that dehumanization of undocumented immigrants and Muslims shapes punitive immigration attitudes and anti-terrorist policies, as well as support for the political leaders who endorse and campaign on these kinds of policies. While these relationships clearly have partisan implications, I wondered whether dehumanization is a distinctively partisan phenomenon; that is, do Americans also share a tendency to dehumanization their political opponents – specifically members of the opposition party?

Anecdotal evidence from political elites and pundits suggests that they might. For instance, Eric Trump said of Congressional Democrats who supported the Mueller probe “I’ve never seen hatred like this, and to me they’re not even people. It’s so, so sad, I mean morality is just gone, morals have flown out the window.” Following the vote to confirm Brett Kavanaugh to the U.S. Supreme Court, Todd Starnes of Fox News tweeted his thoughts about the Democratic outcry : “Those screaming animals in the Senate gallery should be tasered, handcuffed, and dragged out of the building.” Both of these comments reflect dehumanization in that they expressly deny the humanity of a particular group (in the case of Trump’s comment) or liken the group to non-human creatures (in the case of Starnes’s comment).

Is this kind of thinking primarily an elite phenomenon, or do rank-and-file partisans also dehumanize their political opponents? If they do, what are the consequences for partisan conflict? I address these questions in a forthcoming article at Political Behavior. To explore partisan dehumanization, I conducted two public opinion surveys just prior to the 2016 US presidential race. I found that partisans consistently rate their own party as more human than members of the opposing party – in both subtle and blatant ways. This tendency to dehumanize one’s political opponents was moderated by partisan identity strength, such that strongly-identified partisans more readily dehumanized their political opponents. This result is consistent with prior research showing that perceptual biases associated with partisanship are more pronounced among strongly identified partisans.

Partisan dehumanization was correlated with a preference for increased social distance from the opposing party and also perceptions that the opposing party is more morally distant from one’s own party. Both of these relationships point toward declining interpersonal tolerance and moral disengagement processes, whereby dehumanized people and groups are seen as lesser moral agents and thus less worthy of moral consideration. These results highlight the importance of attributions of humanity for social and political cognition and offer a new direction for research on negative partisanship and political polarization. They also point toward the broader consequences of incivility in political discourse.

Partisanship, political awareness, and retrospective evaluations, 1956-2016

Philip Edward Jones

University of Delaware

Voters often evaluate objective conditions through a partisan lens. Those who identify with the party of the president tend to see the economy as improving and the world becoming more peaceful than those who identify with the opposition. For example, at the end of the Obama presidency in 2016, around 46% of Democrats said the economy had improved over the past year compared to just 12% of Republicans.

We know less about how and why the size of these differences in partisans’ perceptions has changed over time, however. In a new piece in Political Behavior, I comb ANES survey data from 1956 to 2016 for measures of how respondents view changes in objective conditions over the past year. For each of the 103 questions included in these surveys, I calculate how much more likely in-partisans (those whose party occupies the White House) are than out-partisans to say that things have gotten better (these calculations adjust for respondent demographics using regression analysis).


Figure 1 shows these differences from 1956 to 2016, with a local average line superimposed. Differences between partisans have clearly not been constant over this time period. In the 1950s, in-partisans had on average a .19 greater probability than out-partisans of saying conditions had improved. These partisan differences declined by the 1970s (to an average of just .05) but then rebounded, increasing dramatically over the past few decades. In 2016, in-partisans had an average .21 greater probability of saying that things were getting better.

Contrary to the view of the 1950s and 60s as an era of muted partisanship, the results here suggest that the mass public was roughly as polarized in its evaluations as today. The period that stands out as unusual in these data is the 1970s, when partisan differences in perceptions were frequently minor or non-existent.

I also explore how the extent to which respondents pay attention to politics affects these results. While we might expect greater political awareness to reduce differences between partisans — since more engaged citizens should be more exposed to information about the actual state of the world — the opposite is true.


As shown in Figure 2, there are few differences between partisans with the lowest levels of political awareness across this time series. Among the least politically engaged, in- and out-partisans generally saw the world in the same way and were equally likely to say conditions were improving. Where we see the greatest disagreement among partisans is among the most politically aware. For these voters, in- and out-partisans consistently reach different conclusions about the state of the world — and it is this group of respondents that show the greatest changes over time.

Why are the most engaged the most likely to disagree about the state of the world? In further analysis, the article suggests that the more engaged a voter is, the more likely they are to internalize messaging from their party’s elites. As politicians have become increasingly polarized over the past few decades, voters who have been paying the most attention to politics have picked up on these cues and adopted increasingly polarized assessments of the world.

The results thus suggest that — absent more consensus among less polarized elites, which seems unlikely — stark partisan differences over the perceived state of the world, particularly amongst the most aware segment of the electorate, seem likely to continue.

Judicial Institutions and the Political Economy of Retirements

David A. Hughes (@davidhughes_phd)
Auburn University at Montgomery

What motivates a judge’s decision to retire from active service? Surely personal factors such as age, health, or financial stability help to guide individuals’ willingness to leave the bench. Indeed, research on the federal courts finds that judges are more likely to retire when they are more advanced in age or have vested in pension benefits.

If the role of the American jurist was only ministerial, then we might expect our analysis to end upon the examination of such personal factors as age and economic security. But judges are also policy-makers. Who replaces them upon their departure has critical implications for the types of policies courts are likely to render in the future.

Mindful that the judicial selection process is inherently political, federal court observers have long speculated that judges prefer to time their departures to coincide with administrations that share their own ideological perspectives.

For example, Sandra Day O’Connor preferred to retire in the early 2000s to spend more time with her husband. But on election night, when television networks initially called Florida for Al Gore, she is alleged to have remarked, “This is terrible,” because she thought she’d need to wait at least another four years to retire under a Republican administration.

The history of U.S. Supreme Court retirements is replete with such anecdotes. But when it comes to judicial retirements at the state level, scholars understand significantly less. For one thing, the historical record simply contains fewer such anecdotes compared to the federal courts. For another, judicial institutions at the state-level are far more heterogeneous than at the federal-level. This is principally due to the presence of competitive elections.

Unlike federal judges who have life tenure, elected state judges must, from time-to-time, stand before the electorate to win new terms of office. Some state supreme court justices run as often as every six years, and these contests can be rather competitive. Between 1990 and 2014, approximately 15 percent of all state supreme court incumbents lost their reelection efforts.

Elections complicate the retirement calculus because they introduce uncertainty into the judge’s decision-making process. This is especially true for those at risk of losing their reelection efforts. These individuals often lose because they have preferences inconsistent with the public’s. If judges retire to avoid a hostile electorate, then it is less likely they will do so under a co-partisan governor.

Indeed, recent scholarly work finds that justices on appointed state supreme courts time their departures with respect to their partisan alignment with appointing elites, while those on elected courts exhibit no such behavior. It appears, then, that the electoral connection encourages individuals unlikely to win their next election to go ahead and call it quits.

Compared to the politics of state court departures, scholars understand next to nothing about how pension benefits affect retirements. This is curious given the preponderant role such factors play at the federal level. Existing research should lead us to believe that any given state judge should be more likely to retire if she is eligible for her pension. But how might we expect electoral uncertainty to condition judges’ economic incentives to leave the bench?

To address these questions, I gathered data on 388 state supreme court justices’ retirement benefits across 26 years and 18 American states. I examined each justice’s eligibility for pension benefits in a given year of service, and if eligible, what percent of their salary they could earn if they were to retire that year.

Every state offers its judges some form of retirement plan. The federal judiciary uses the “Rule of 80,” which holds that judges may retire with full benefits once their age and years of creditable service total at least 80. While every state likewise defines eligibility and benefits using age and service, some make it easier for judges to become pension eligible than others, and many require judges to continue in service even after having already vested in order to increase the size of their benefits.

To study the effects of public pensions on judges’ retirement decisions, I examined the pension plans of ten elected and eight appointed state supreme courts. Using their age and years of creditable service, I then determined each judge’s pension eligibility and the size of their benefits.


Figure 1: Pension eligibility and size of benefits in 18 state supreme courts

As can be seen from the figure of justices’ retirement benefits, 87 percent of those who voluntarily retire are pension eligible (80 percent for elected and 95 percent for appointed courts), and the average justice earns approximately 59 percent of her active-status salary in retirement (53 percent for elected and 67 percent for appointed courts).

Using this new data, I weighed the likelihood of retirement in a given year against other factors such as partisan alignment with appointing elites, risk of electoral loss, etc. I studied the lengths of justices’ careers using a statistical method that examines the duration of events.

My analysis found that both elected and appointed state supreme court justices condition the timing of their retirements upon their pension eligibility. Nevertheless, the electoral connection hinders justices from earning as much in retirement as their peers on appointed courts. This is good evidence that state judges, like their federal counterparts, prefer to retire with pension benefits but that electoral risks or uncertainties inhibit their ability to increase these benefits.

I also found little evidence that state supreme court justices’ retirement decisions are politically motivated. This is somewhat surprising and inconsistent with prior work finding that appointed state supreme court justices retire when they are co-partisan with appointing elites, and elected justices retire when at risk of losing an election.

While my results are some of the first in the state courts literature to cast doubt on the theory of political retirements, they actually make this literature more consistent with studies conducted at the federal level. Analysis of large-N quantitative research on federal judicial departures demonstrates a clear split in the scholarly literature among those emphasizing the importance of political versus economic factors.

A Healthy Democracy? Evidence of Unequal Representation Across Health Status

Julianna Pacheco & Christopher Ojeda

Health affects nearly every aspect of our lives and recent research finds that politics is no different. American citizens who are in better health participate more often and hold different policy preferences than those who are in poor health. But do these health-related disparities in participation and opinions have political consequences? Are the voices of the those in good health privileged over those in poor health? And are some politicians more responsive to health-disparities among their constituents than are other politicians?

To answer these questions, we analyzed data from the 2012 Congressional Cooperative Election Study, which is a nationally representative survey of Americans. We compared the preferences of constituents to the votes of their member of Congress (MC) to determine how responsive politicians are to the public. We find that MCs are more likely to vote in line with the preferences of constituents in good health. This is especially true in districts where constituency preferences diverge across health status. We call this finding a “health bias” in representation.

But we didn’t stop there. We also wanted to know whether this unequal responsiveness cut across party lines and what factors might explain it. To determine if Republican or Democratic MCs were especially unequal in their response, we re-analyzed the data based on the party identification of the politician. We found extreme differences: Republican MCs were especially responsive to those in good health, while Democratic MCs were not any more responsive to one group or another. In short, the extent to which health is a basis for unequal responsiveness is entirely driven by Republicans.

We tested two explanations for this unequal responsiveness. The first explanation is based on health-disparities in participation. If those in good health participate more than those in poor health, then politicians might come to believe that their electoral fortunes depend on satisfying their constituents who are in good health. The second explanation is based on health-disparities in preferences. Shifts in health lead could lead to shifts in preferences that break a prior alignment between constituents and representatives. Both of these patterns could explain our findings, but in fact we find mixed evidence to support either explanation. More research is required to better understand why we observe this unequal responsiveness.

Over the past fifteen years, scholarship has highlighted the ways in which politicians privilege the voices of some citizens over others. This research has revealed that the rich, men, and White Americans are disproportionately represented in the roll call votes of representatives. To this line of research, we would add that health inequalities matter too. We hope that scholars continue to study the politics of health disparities and how they undermine the success of our democracy.



Compassionate Democrats and Tough Republicans: How Ideology Shapes Partisan Stereotypes

Scott Clifford

As partisan polarization has increased over the past decades, so has affective polarization in the mass public. People increasingly dislike their partisan opponents and hold negative stereotypes about their character and motivations. According to prominent theories of partisan identity, these negative stereotypes are merely rationalizations that serve to justify partisan discrimination and intolerance. However, an alternative explanation is that these stereotypes are largely accurate reflections of the increasingly clear ideological and value-based differences between the parties. I put these competing views to test in two studies and find support for the latter view – partisan stereotypes seem to reflect the parties’ values.

In the first study, I used a national sample to examine the character traits that partisans desire and the traits that partisans perceive in each other. The results show that partisans place different weights on several aspects of moral character. Democrats placed relatively more value on being compassionate and fair-minded, while Republicans placed more emphasis on being loyal, obedient, and wholesome. However, these trait preferences were much more strongly related to a person’s ideology than to their partisan identity. Stereotypes of partisans were largely consistent with these differences as well. People saw Democrats as relatively more compassionate and fair-minded, while Republicans were seen as relatively more loyal, obedient, and wholesome.

In a second study, I used an experiment to examine whether ideology is driving these partisan stereotypes. I asked survey respondents to judge the character of a series of fictional individuals. The partisanship and ideological stances of these fictional individuals were randomly assigned, allowing us to tease apart the source of partisan stereotypes. The results show that people draw reliable trait inferences from partisan cues, which is consistent with the first study. Democrats were perceived as relatively more compassionate and fair-minded, while Republicans were seen as relatively more patriotic, tough, and wholesome. However, ideological information about the fictional individuals had much larger effects than information about their partisan identities, and the effects of partisan cues were diminished when ideological information was provided. These results suggest that partisan stereotypes are not mere rationalizations, but largely accurate reflections of the ideological and value-based differences between the parties.