To interpret elections, social scientists and media pundits often ask: how much did particular groups, or voting blocs, contribute to a candidate’s vote total? Analysts often answer this question by regressing vote choice on voters’ attributes, interpreting changes in coefficient magnitude across elections as shifts in support. We show, however, that this analysis fails to take into account the prevalence of the group in the electorate and the rate that group turns out to vote — yielding quantities that are not directly useful for understanding election outcomes. To avoid this base-rate fallacy, we introduce a set of tools for estimating where candidates received votes and how voting bloc patterns differ from prior elections. We apply these tools to study US elections, demonstrating that there is little evidence that Black and Hispanic voters shifted to Republicans in the 2020 election and that Donald Trump’s support was concentrated among voters with moderate attitudes towards racial outgroups.

• New York Times
• The Economist
• FiveThirtyEight

• The Science of Politics

This paper subsumes a previous working paper circulated as "Who Put Trump in the White House? Explaining the Contribution of Voting Blocs to Trump’s Victory." Original draft available here.

Can exposure to successful celebrities from a stigmatized group reduce prejudice toward that group writ large? We study the sudden and phenomenal rise to fame of Liverpool F.C. soccer star Mohamed Salah, a visibly Muslim player. We estimate the causal effect of Salah joining Liverpool F.C. on Islamophobia using hate crime reports throughout England, 15 million tweets from British soccer fans, and a survey experiment of Liverpool fans. We find that Merseyside (home to Liverpool F.C.) experienced a 16% drop in hate crimes, compared to a synthetic control. There is no similar effect for other types of crime. We also find that Liverpool fans halved their rates of posting anti-Muslim tweets relative to fans of other top-flight clubs. The survey experiment suggests that the salience of Salah's Muslim identity is important for reducing prejudice against Muslims more broadly. Our findings indicate that positive exposure to outgroup role models can decrease prejudice.

• The Economist
• The Times
• Freakonomics Podcast
• HuffPost
• Vice
• The Independent
• Yahoo!
• Bleacher Report
• Fox News
• Daily Mail
• Sky News
• Liverpool Echo
• Stanford Magazine
• Al Jazeera (Arabic)
• Spiegel Online (German)

• Middle East Political Science Podcast

How much does self-interest drive Americans’ policy attitudes? Survey research typically finds that self-interest’s role is minimal. Such conclusions are typically reached by examining attitudes toward federal policies that present diffuse costs and low stakes. We consider a starker test-case of self-interest: controversies surrounding development of dense and affordable housing in Americans’ communities. Liberal homeowners, especially, must cope with dissonance between their egalitarian ideology and a desire to protect their home values and quality of life. They often embrace liberal housing goals and redistributive housing policies, but join conservatives in opposing dense housing in their own communities. Two survey experiments show that liberal homeowners are cross-pressured, and barely more likely than conservative homeowners to support dense housing development. Messages appealing to homeowners’ self-interest reduce support further, while countervailing appeals about housing’s benefits to low- and middle-income families barely offset the negative effect. We discuss implications for the politics of equality of opportunity in state and local politics.

• New York Times
• The Atlantic
• New York Magazine
• Foreign Affairs

In the literatures on public opinion and legislative behavior, there are debates over (1) how constrained preferences are and (2) whether they are captured by a single left-right spectrum or require multiple dimensions. But insufficient formalization has led scholars to equate a lack of constraint with multidimensional preferences. In this paper, we refine the concepts of constraint and dimensionality in a formal framework and describe how they translate into separate observable implications for political preferences. We use this discussion to motivate a cross-validation estimator that measures constraint and dimensionality in the context of canonical ideal point models. Using data from the public and politicians, we find that American political preferences are one-dimensional, but there is more constraint among politicians than among the mass public. Furthermore, we show that differences between politicians and the public are not explained by differences in agendas or the incentives faced by the actors.

How pervasive is partisan sorting and polarization over public policies in the American public? We examine whether the barriers of partisan sorting and polarization seen in national politics extend to important local policies that shape economic development. To describe the extent of partisan sorting and polarization over local development policies, we employ conjoint survey experiments in representative surveys of eight U.S. metropolitan areas and a hierarchical modeling strategy for studying heterogeneity across respondents. We find that strong partisans are sorted by party in some of their policy opinions, but rarely polarized. The same voters who disagree about national issues have similar preferences about local development issues suggesting a greater scope for bipartisan problem solving at the local level.

• New York Times
• CityLab

Why have economically declining regions turned toward right-wing parties? To explain this puzzling phenomenon, we develop a theory linking internal migration, localized social institutions (e.g., family and friend networks), and voters’ preferences for social insurance. We start with the observation that social ties provide insurance against adverse life events, such as job loss, and highlight two implications. First, those with strong social networks prefer lower spending on social insurance, because they have access to informal insurance that acts as a substitute for public programs. Second, social ties discourage people from moving, even when better economic opportunities are available in other regions. Combining these mechanisms, we argue that the effect of economic shocks on a region’s politics depends on the strength of social ties. Regions with dense social ties have muted migratory responses to negative shocks relative to regions with weaker ties. Further, those who remain in declining regions are more conservative than those who migrate — resulting in an electorate with lower demand for social insurance. Macro-level analysis of American election results, import shocks, and migration data provide empirical support for the theory’s predictions. An original survey corroborates the micro-level mechanisms. The results have important implications for understanding right-wing populist support in economically declining regions in the U.S. and other post-industrial countries.

Over the past 40 years of American politics, college-educated white voters have defected from the Republican Party, while the white working class has become a reliable source of Republican support. I study the issue basis of this realignment. To do so, I generate over-time estimates of public opinion on four broad issue domains from 1984 to 2020 and develop a theoretical framework to understand how issue attitudes translate into electoral coalitions. Using this framework, I find that both economic and cultural issues have contributed to the observed realignment. College-educated white voters have become increasingly liberal on economic issues since the mid-2000s; college-educated voters now express more liberal views than working class voters on every issue domain. Over the same time period, cultural issues have become more important for the voting decisions of the working class. The increasing weight placed on non-economic issues means that the conservative cultural attitudes of white working class voters translate to Republican support at a higher rate than in the past. Together, these findings suggest a nuanced role for economic and cultural issues in structuring political coalitions. Educational realignment has deep roots across issue domains, suggesting that the new coalitions are likely to be stable into the foreseeable future.

• Washington Post
• City Journal

• The Science of Politics

Politics in the United States has become increasingly nationalized and polarized, raising the question: Are politicians responsive to the local concerns of their constituents? I investigate this question by examining how local economic conditions affect the expressed priorities of congressional candidates, drawing on the content of televised campaign advertisements between 2000 and 2016. Exploiting over-time changes in local unemployment rates, I find that the issue content of campaign ads varies substantially with local conditions. Specifically, candidates in high-unemployment areas devote more attention to jobs and employment and less to the safety net. Democrats also discuss business less in high-unemployment areas. The magnitude of the effects varies by party in a way consistent with theories of strategic emphasis. These findings suggest that, rather than politics being uniform throughout the country, candidates are responsive to the conditions in their districts — but this responsiveness depends in part on the national-level political environment. Economic geography therefore acts as a constraint on the nationalization of politics.

The internet has made it easier than ever for citizens to voice their opinion to their elected representatives. However, officials may infer that constituents who write to them via lower-effort online mediums care less about the issues than those who communicate in person. To test how policymakers evaluate messages from constituents, we fielded a national survey of local U.S. policymakers to examine responsiveness to different types of messages. Our findings indicate that online communication presents a double-edged sword: while it lowers the effort needed for constituents to communicate, officials discount information conveyed through the internet relative to traditional ones. We examine this trade-off using an embedded conjoint experiment. Our results indicate that a social media message would have to be sent by more than 47 constituents for it to exceed the value of a single face-to-face meeting. These findings illustrate that, all else equal, in-person meetings remain the most effective form of grassroots communication. Still, online communication can be an effective choice to the extent that it facilitates a sufficient increase in overall levels of constituent engagement.

Throughout the United States, an increasing number of states are adopting laws that make it easier for voters to vote by mail (VBM). While research has addressed the potential effects of these changes on turnout, little attention has been paid to other aspects of voter behavior. In this paper, I argue that mail voting can decrease ballot roll-off — the tendency of voters to selectively abstain from voting in some races on the ballot. Roll-off is very common, with roll-off rates often exceeding 10%. Exploiting county-level variation in implementation of mandatory vote-by-mail laws in Washington, I show that VBM is associated with an aggregate decrease in roll-off in a variety of down-ballot races. The results persist even after several mail elections. These findings are consistent with absentee voting causing voters to take more time to educate themselves about their choices.

A core question in the study of democratic politics is what factors influence the decision to run for office. A full accounting of the process of candidate emergence requires understanding the individual-level factors that influence potential candidates. Yet, existing studies typically focus on a single factor in isolation or study aggregate outcomes, rather than individual-level decisions. To overcome these limitations, we embed a conjoint experiment into a survey of local officials — a population from which candidates for from which candidates for higher office often emerge. We vary election scenarios and measure interest in running for the given office. Politicians are more sensitive to variation in the fundraising burden than any other factor considered — including legislative salary. Politicians are also deterred by the presence of an incumbent and by negative advertising. We find little evidence that they are directly responsive to their opponent’s ideology.

The United States is facing a housing affordability crisis that disproportionately burdens low-income households. Local government regulations and institutions often make it difficult to build more housing. Research has documented how residents use participatory institutions to stymie new housing development, but the role of local politicians is less clear. I investigate whether there is an electoral incentive for local officials to limit affordable housing development. To answer this question, I merge a geocoded dataset of new affordable housing placed in service with precinct-level city council election results in Chicago and New York City. Exploiting within-district variation in the distance to new housing, I find no evidence that politicians are punished for new housing in either city. Supplemental analyses using survey data in a wider geographic area corroborate this result. I provide evidence on the role of the information environment in explaining this null result: most online news coverage of new affordable housing developments is positive or neutral in tone. These results further spotlight the role of non-electoral participatory institutions as the primary political mechanism driving housing policy.

Researchers commonly use multilevel regression and poststratification (MRP) to es- timate opinion in small geographies and to adjust unrepresentative samples. Yet, MRP estimates are still subject to error, whether due to modeling choices or non-ignorable survey error caused by declining response rates and differential nonresponse. We show how auxiliary quantities whose marginal distributions are known at the geography of interest — such as election outcomes — can be leveraged to account for this error. We propose jointly modeling the geographic-level correlation between the auxiliary calibra- tion variables and responses to survey items of interest, whose truth is unknown. We then use the estimated correlations to adjust the estimates of target variables. This procedure relies on the assumption that the correlations observed in the sample are the same as in the population. To validate the approach, we use a pre-election poll to examine vote intention in three statewide races in Michigan in 2022. We calibrate estimates to one election and examine errors for the remaining races. We find that calibration decreases the average error by two-thirds — suggesting that the method has the potential to greatly increase the accuracy and value of MRP estimates.

In the study of elections, there is substantial interest in the ability of campaigns and the media to influence the criteria that voters use to evaluate politicians. Despite this widespread interest in “activation” or “priming,” there is a lack of consensus on the exact estimand being studied. Further, the literature has not formally analyzed the conditions under which the activation estimand can be identified. In this paper, we draw on the literature to formalize three distinct conceptions of activation. We use these formalizations to analyze commonly used observational and experimental research designs. A key result is that two ignorability assumptions are necessary to assess whether campaigns or the media caused voters to change the weight they place on an issue in their voting decisions. A weaker formulation of activation, based on prediction, relaxes one of the assumptions, at the cost of reduced theoretical interpretability. Our framework organizes the literature on activation and priming, and may help spur development of new research designs that improve estimation of activation effects.