Interdisciplinary Workshops on Politics and Policy

Using over one million survey responses and machine learning to learn what happened in the 2016 and 2018 election, why nobody saw Trump coming, and how data science is being used in the 2020 election

Like all computerized systems, ballot-marking devices (BMDs) can be hacked, misprogrammed, and misconfigured. BMD printout might not reflect what the BMD screen or audio conveyed to the voter. If voters complain that BMDs misbehaved, officials have no way to tell whether BMDs malfunctioned, the voters erred, or the voters are attempting to cast doubt on the election. Several approaches to testing BMDs have been proposed. In pre-election logic and accuracy (L&A) tests, trusted agents input known test patterns into the BMD and check whether the printout matches. In parallel or live testing, trusted agents use the BMDs on election day, emulating voters. In passive testing, trusted agents monitor the rate at which voters “spoil” ballots and request another opportunity to mark a ballot: an anomalously high rate might result from BMD malfunctions. In practice, none of these methods can protect against outcome-altering problems. L&A testing is ineffective against malware in part because BMDs “know” the time and date of the test and the election. Neither L&A nor parallel testing can probe even a small fraction of the combinations of voter preferences, device settings, ballot language, duration of voter interaction, input and output interfaces, and other variables that could comprise enough votes to change outcomes. Under mild assumptions, to develop a model of voter interactions with BMDs accurate enough to ensure that parallel tests could reliably detect changes to 5% of the votes (which could change margins by 10% or more) would require monitoring the behavior of more than a million voters in each jurisdiction in minute detail—but the median turnout by jurisdiction in the U.S. is under 3000 voters, and 2/3 of U.S. jurisdictions have fewer than 43,000 active voters. Moreover, all voter privacy would be lost. Given an accurate model of voter behavior, the number of tests required is still larger than the turnout in a typical U.S. jurisdiction. Even if less testing sufficed, it would require extra BMDs, new infrastructure for creating test interactions and reporting test results, additional polling-place staff, and more training. Under optimistic assumptions, passive testing that has a 99% chance of detecting a 1% change to the margin with a 1% false alarm rate is impossible in jurisdictions with fewer than about 1 million voters, even if the “normal” spoiled ballot rate were known exactly and did not vary from election to election and place to place. Passive testing would also require training and infrastructure to monitor the spoiled ballot rate in real time. And if parallel or passive testing discovers a problem, the only remedy is a new election: there is no way to reconstruct the correct election result from an untrustworthy paper trail. Minimizing the number of votes cast using BMDs is prudent election administration.

This project investigates the role of place-based identification in influencing Americans’ racial attitudes and policy preferences. Specifically, I argue that Southern identity (i.e., identification with the American South) is an influential but omitted factor in the study of political behavior across racial groups. In this project, I create a novel survey measurement of Southern identity and assess its impact on public opinion. Contrary to the extant literature, this work argues that Southern identity has political consequences for the opinion formation of Black Americans as well as for White Americans. I expect that Southern identity will be associated with group-centric racial beliefs reflecting the perceived communalistic nature of Southern culture. Analyses from three original surveys suggest that Southern identity influences both Black and Whites to adopt distinct racial beliefs different from their non-southern racial group members. These results hint at a challenge to the claim that Southern identity among Black Americans is not as politically relevant as it is for White Americans. This work also speaks to the need for more nuanced approaches to understanding American’s racial beliefs across race and place.

Do refugees reshape long-term political behavior in receiving areas? To investigate this question, I examine the expulsion of ethnic Germans from Eastern Europe into West Germany at the end of WWII. Expellees were strangers to the cultural practices in their new surroundings. Tensions with natives forced expellees to rely on each other and helped foster a strong group identity. I argue that this shared identity, coupled with political circumstances specific to Germany, engendered support for the radical right among expellees. Using district-level data from 32 elections spanning 100 years, I find that communities which received greater shares of expellees remain more supportive of the radical right in the short, medium, and long term. This legacy of forced migration responds to changes in the political context within Germany, and is driven primarily by districts that received greater shares of expellees who were not citizens of the Third Reich during the interwar period. 

Despite a large comparative literature on party machines in distributive politics, few studies have systematically examined how party leaders select local brokers to staff their party organizations. We provide a theoretical framework for studying these selection decisions. We argue that patrons must balance two key concerns: a broker’s efficacy among clients and their loyalty to party and patron. We test the relative importance of these concerns through a conjoint experiment conducted with 343 local political patrons across two Indian cities. Briefly, we find patrons strongly prefer loyal brokers, and not simply brokers who are popular with clients. We suggest this preference reflects the high threat of broker exit under conditions of inter-party competition and intra-party factionalism. Further, we find that patrons value a broker’s everyday problem-solving efficacy more highly than their election-time mobilizing efficacy. We validate our experimental findings against actual broker promotion patterns in our study cities, drawing on data from a unique survey of 629 brokers operating within migrant slums.

In this paper we reexamine the role of racial group consciousness (RGC) in explaining why Black Americans choose to engage in costly, to the individual, political action. Attempting to add clarity to decades of inconsistent and at times contradictory findings, we argue that the effect of RGC at inspiring political action among racial minorities is conditional on 1) the relevance of the political activity to achieving a group-based ends, and 2) individual capacity to assume the cost associated with engaging in the activity. Given these conditions, we designed a series of behavioral experiments that vary the group relevance of political action while holding capacity to engage constant. We find that while nearly all measures of RGC exhibit a consistently strong relationship with engagement in low-cost political behavior (stated support or intent to support), only RGC beliefs that directly capture perceptions of discrimination reliably explain Black Americans’ willingness to engage in costly, to the individual, group based political behavior. This work has important implications for how we understand the role that race plays in black political decision making.

How do militants and governments strategically adapt their military tactics in asymmetric conflict contexts? This paper presents patterns from the new Palestinian-Israeli  Insurgency & Militarism (PA’ILIM) dataset, which includes information on approximately 180,000 Israeli military (and settler) actions and over 5,000 acts of Palestinian political violence from 2009 to 2018. This event-level dataset uses local Israeli and Palestinian NGO reports to capture fine-grained information on patterns of violence in the Israeli-Palestinian conflict, including more low-level negative interactions between Israelis and Palestinians than is typically covered in datasets reliant on news stories. As such, PA’ILIM provides a fuller picture of the broad repertoire of violence used by both sides as well as the daily realities of violence for the communities involved in this long-standing conflict. To illustrate the utility of this dataset, I use the data to examine changes in patterns of violence over time, demonstrating how the unique strategic dilemmas facing each actor manifest in distinct tactical choices. This dataset will be useful to scholars interested in the Israeli-Palestinian conflict and, more generally, in the conduct of asymmetric warfare, strategic adaptation in conflict, state violence or repression, and repertoires of militant violence.  

How do perceptions of belonging or lack of belonging to American society influence political interest and political engagement? To date, there have been few inquiries that systematically investigate notions of perceived belonging to U.S. society and the political ramifications of these predispositions. This project addresses this puzzle and investigates how a sense of perceived social inclusion or exclusion influences political engagement among Latinos, the largest, one of the fastest growing and most pivotal groups in American politics. By bridging literatures in political science, sociology and psychology, this book project offers a novel framework centering on the idea that notions of belonging are fundamentally tied to political attitudes and political behavior for members of marginalized groups. I argue that members of marginalized groups develop different understandings of inclusion in the U.S. according to their every-day experiences, and that these perceptions have the potential of conditioning their political attitudes and political behaviors. To advance this argument, this multi-method project, leverages surveys, experiments and in-depth interviews.  I develop a new set of items to measure perceptions of inclusion and exclusion from U.S. society. I examine these items in over six state and national surveys including the 2016 and 2020 Collaborative Multi-racial Post-election Surveys (CMPS) as well as the 2018 Cooperative Congressional Election Study (CCES). Relying on multiple survey experiments, I assess the extent to which elite messaging can shift perceptions of belonging, and how shifts in these predispositions have behavioral consequences. Belonging and claims of membership have been at the core of the political struggle of racial and ethnic groups in America, but as a psychological construct the notion of perceived belonging has received little attention in political science. Through a novel framework rooted in interdisciplinary perspectives and by empirically testing a new measure of perceived belonging to U.S. society, this project makes a significant contribution to our understanding of the factors that shape political behavior among members of marginalized and stigmatized groups.

I develop a theory of presidential behavior based on the idea that what they actually accomplish doesn’t matter. Most scholarship on the American presidency argues that presidents care about what the government they lead accomplishes in terms of policy. I argue the power of the contemporary presidency is mostly found in the ability to put on a compelling show of governance for some critical audience, and that the substantive impacts of policy are not important. Presidential initiatives often masquerade as policymaking intended to alter the status quo—but yet, leave that status quo untouched. This is because the president’s accountability relationship with the public does not require them to be effectual, and their motivations to leave a historical legacy are typically overstated. Presidential actions are mostly useful ideological signaling and credit claiming devices that help them retain office. As a result, the constitutional and legal barriers to action enforced by the separation of powers are less important for explaining presidents’ behavior than previously thought. 

Why do political parties drop out of coalition negotiations? Before coalition governments are established, political parties engage in extensive coalition talks in which the coalition’s policy agenda and the distribution of ministerial offices are negotiated. Although they may have spent weeks or even months in negotiations with potential parties, parties do abandon coalition talks, thereby significantly delaying government formation. Despite the central importance of timely government formation for the stability and performance of political systems, we know very little about why parties abandon coalition talks, giving up attractive ministerial posts and risking public blaming. We argue that parties learn about the policy horizon of the other parties during the negotiation process as they are conducted away from the scrutiny of the public and media. When there is no overlap between a party’s own policy horizon and that of its negotiation partners, the likelihood of a party dropping out of coalition talks increases. To test our theoretical argument comparatively and over time, we have compiled a comprehensive dataset covering 380 political parties in 27 countries following 160 elections from 1991 until 2017. Our results have important implications for our understanding of coalition governments, democratic representation and the stability of democratic systems.

Can endorsements persuade voters to transcend politicized ethnic divisions to support candidates from other groups? We argue that endorsements can serve as a form of vicarious contact, allowing voters to observe cooperative interactions between coethnic and non-coethnic politicians. This vicarious contact encourages voters to positively update their beliefs about how they will be treated by a non-coethnic in office. In assessing this claim, we provide evidence from Kenya, where simulated radio news segments were used to experimentally manipulate the ethnic relationship among voters, endorsers, and candidates. We find that voters who hear endorsements from their own coethnics, rather than non-coethnics, are significantly more likely to vote for a non-coethnic candidate. We further find evidence suggesting this result is mediated by coethnic trust: the trust premium enjoyed by coethnic endorsers is transferred to non-coethnic candidates. Voters believe endorsed candidates are more likely to provide non-discriminatory representation.

In recent years, thousands have protested against police violence in the United States; yet, police continue to enjoy legitimacy among many Americans. We develop a theoretical framework to understand how people form beliefs about who deserves violence from the state and test it on a large sample of White and Black Americans (N=11,166). The results reveal the multiple ways race affects deservingness evaluations. Respondents’ own racial identities and attitudes matter more than the race of the person targeted by police. Both White and Black respondents who expresshigher levels of racial resentment are more likely to blame detainees and less likely to blame officers. Despite thesesimilarities, Black respondents are 58% more likely than White respondents to signal that police abuse is undeserved by supporting a financial settlement for the victim. This study provides further evidence that perceptions of who deserves state violence are racially dependent.

This talk will explore the factors driving public attention and mobilization in the wake of officer-involved killings. Analyses of survey data and data on victims of officer-involved killings reveals that less than half of victims of officer-involved killings receive public attention (measured as trending on Google) and only about 8 percent were protested. Only a small subset of victims can drive political mobilization: exposure to killings of black, low culpability, high visibility victims increased political interest and voter turnout. These victims were more likely to be protested as well.

Scholars warn that affective polarization undermines democratic norms and accountability. If citizens increasingly detest the other party’s supporters, are they more likely to endorse norm violations, overlook copartisan politicians’ shortcomings, oppose compromise, adopt their party’s views, or misperceive economic and public health conditions? A large, influential literature speculates as such. However, such speculation remains difficult to test. We argue the contrary: affective polarization’s consequences should be generally confined to interpersonal domains, with more circumscribed political implications. We support this argument with unique experiments which exogenously manipulate citizens’ levels of affective polarization and trace downstream consequences, such as their reaction to information about their actual representatives. In our experiments (N = 9,836) we produce the equivalent of three decades change in affective polarization, but find no evidence that these changes in affective polarization influence a broad range of political behaviors—only interpersonal attitudes. Our results raise doubt about the view that reducing affective polarization would meaningfully bolster democratic norms or accountability.

During periods of potential political opening, citizens experience repression and dissent events. Are people who are exposed to repression demobilized, or does repression increase some individuals’ grievances or resolve? When individuals see others’ expressing dissent, are they more or less likely to also participate? This paper draws on a unique panel dataset from the months spanning Zimbabwe’s 2018 election, during which citizens experienced unexpected acts of state repression and opposition protest. Using a two-way fixed effects model, we assess how citizens’ dissent intentions change after exposure to repression and dissent events both locally and through social media, and measure three types of individual-level mechanisms that are hypothesized to drive decision-making about dissent. We find that exposure to both repression and dissent is mobilizing for opposition and non-partisan citizens. We find evidence for different types of mechanisms for repression and dissent events: our evidence on dissent is in line with basic information updating, while for repression relational and emotional mechanisms may mobilize citizens despite increased perceived costs of dissent. For ruling party supporters, we find that dissent events lead to counter-mobilization, and that exposure to repression has little effect. Local repression and dissent events are consistently more strongly related to dissent intentions than social media exposure. Our findings contribute to existing debates about the effectiveness of repression, while providing new insights into how individuals make decisions during periods of contentious mobilization.

Why do ethnic minorities tend to be geographically concentrated? This paper argues that ethnic segregation can be a strategic response to targeted exploitation. Focusing on the Mexican case —where only 6% of localities contain 80% of the indigenous population, we develop a research design akin to spatial regression discontinuity to trace the long-run impact of exposure to colonial officials on contemporary ethnic segregation. Our identification strategy takes advantage of long-gone historical district boundaries from the 16th to the late 18th centuries to compare localities that varied only in their distance to their colonial capitals. Despite being otherwise similar, we show that localities farther from colonial officials in the past hold a disproportionate share of indigenous population today relative to localities that were closer. Consistent with the idea that minorities sought refuge from the reach of colonial administrators, our results are stronger in areas where the potential for extraction in colonial times is higher, where indigenous institutions are more prevalent, where indigenous resistance in the form of rebellions and judicial lawsuits was more intense, and where the presence of religious missionaries potentially shielded indigenous settlements. Our findings provide evidence that ethnic segregation arises and persist over time as a combination of top-down coercion and bottom-down agency.

Presidential patronage occupies a central role in accounts of administrative capacity, state development, and party politics. Yet existing scholarship lacks systematic evidence about how patronage considerations affected the composition of the federal workforce in the United States. We address this omission with new individual-level data on the personnel who occupied more than 145,000 positions in the Department of Interior between 1849 and 1905. Using panel data on the allocation of bureaucratic personnel, we find that a state’s electoral support for the presidential administration significantly increased the hiring rates of state residents. This finding is robust across a range of model specifications and measurement strategies and when accounting for a variety of congressional characteristics. In additional analyses, we show that the connection between electoral politics and allocation of patronage was significantly attenuated following the adoption of federal civil service reforms. Our findings document the role of presidential parties in staffing the federal bureaucracy in the nineteenth century, provide evidence of potential mechanisms through which patronage operated, and suggest how agency characteristics and institutional reforms contributed to the weakening of presidential patronage.