Publications

Confusing ballots muddle the connection between voter intentions and votes, diminishing the ability of elections to facilitate representation in political institutions. This motivates our examination of the 2018 midterm election in Florida, where the ballot used in Broward County yielded an abnormally high number of undervotes in Florida’s United States Senate race. We offer cross-sectional and temporal analyses that eliminate explanations for Broward’s Senate undervote that do not turn on ballot design. Respectively, these analyses compare Broward County and its precincts to other counties and their precincts and compare elections in 2016 with those in 2018. Our purview also extends beyond Florida to states that had Senate and gubernatorial elections in 2018. We generate counterfactual estimates of Senate vote totals had Broward County used a conventional ballot in 2018, and our counterfactual results lie in statistical purgatory. They show neither that the Broward County ballot was pivotal to the Senate election outcome nor rule out this possibility.

Unequal participation and unequal internet access are both structural inequalities across the American states, factors that keep certain demographic groups in a continued marginalized status. According to the Current Population Survey (2015), nearly 20% of American households do not have access to the internet. Not having internet access hampers the ability to vote, as individuals without such are less motivated and less able to acquire information on how and for whom to vote. The demographic groups who are least likely to have internet access – African Americans and Hispanics, low income, low education, and rural – are also less likely to vote. A key question is whether government policies that make internet access more widespread also increase the likelihood that individuals from these groups will become voters. To answer this question, this paper assesses whether online voter registration, a law whose effect is strongly structured by broadband or high-speed internet access, has a more positive impact on voter registration and turnout in areas with on average more internet access. To evaluate this hypothesis, this study utilizes 2014 and 2016 Current Population Survey data on individuals from across the American states. This is combined with state-level broadband availability data aggregated from the census block level from the Federal Communications Commission Varying internet access by state and using online voter registration policies as a treatment variable, this study determines whether state internet availability structures the impact of this law on political participation. Ultimately, the results of this study reveal the degree to which the impact of online voter registration policies is contingent on the availability in the American states. 

Party loyalty in U.S. Congressional elections has reached heights unprecedented in the post-war era. Theories of partisanship as informational cues would predict that ticket splitting from national partisanship should be even more rare in low-information elections. Yet, here I show that ticket splitting in state and local offices is often higher than in Congress. I use cast vote records from voting machines that overcome ecological inference challenges, and develop a clustering algorithm to summarize such ballot data. For example, about a third of South Carolina Trump voters form a bloc whose probability of ticket splitting is 5 percent for Congress, but 32 percent for county council and 50 percent for sheriff. I show that a model with candidate quality differentials can explain these patterns: Even in a nationalized era, some voters cross party lines to vote for the more experienced and higher quality candidate in state and local elections.

The administration of elections depends crucially upon the quality and integrity of voter registration databases. In addition, political scientists are increasingly using these databases in their research. However, these databases are dynamic, and may be subject to external manipulation and unintentional errors. In this paper, using data from Orange County, California, we develop two methods for evaluating the quality of voter registration data as it changes over time: (1) generating audit data by repeated record linkage across periodic snapshots of a given database, and monitoring it for sudden anomalous changes; and (2) identifying duplicates via an efficient, automated duplicate detection, and tracking new duplicates and deduplication efforts over time. We show that the generated data can serve not only to evaluate voter file quality and election integrity, but also as a novel source of data on election administration practices.

In recent years, a number of states have passed some version of automated voter registration (AVR). Implementation varies, but the core idea is to more aggressively promote voter registration as an option at the Department of Motor Vehicles. Evaluation of the impact of AVR has been limited thus far, mostly because AVR implementation has itself been limited. But post-reform data are now available for a number of states, and others that are considering adopting the reform are wondering what the effects of the laws have been in these states. What would registration and turnout have looked like in AVR states had the reform not been implemented? In this paper we take advantage of the first election cycle with significant post-AVR data across a range of states to explore the registration effects of AVR. We employ both difference-in-differences and synthetic control approaches to identify causality. Registration effects so far appear to have been solid overall and larger for Latinos. Evidence for Asian Americans and young people is more ambiguous, as is evidence for effects in individual states. We conclude with thoughts on future directions to help develop better estimates in these areas.

In voting, maps are traditionally used to indicate voter turnout and/or election results with respect to political parties. This paper explores the use of geospatial choropleth maps to analyze voter arrival patterns through a case study of using logs from electronic poll books (EPB) across the State of Rhode Island and Providence Plantations (RI). The EPB transaction logs record various metrics, such as the precinct number, location, and timestamp for every voter that checks in on Election Day. Geographically referenced jurisdiction datasets were plotted in ArcGIS and combined with the EPB transaction log data for the entire state’s 2018 Midterm elections to create the choropleth maps. The choropleth maps were shaded based on the percentage of total check-ins observed during selected time windows throughout the day.  

Analyses were undertaken to assess the visual representation of arrival densities for both the state and its major metropolitan area. Arrival observations statewide were highlighted and expanded in conjunction with known jurisdiction profiles. At the town/city level arrival patterns were identified based urban/suburban and rural areas. A precinct level analysis was performed in the metropolitan area and revealed differing arrival patterns within the City of Providence. General observations are provided based on visual inspection.  Identifying specific precinct groups with similar overarching trends of community voting behavior will require computationally based clustering methods. Future considerations of how data visualization of arrival patterns via EPBs are discussed. 

In 2005, the Texas Legislature allowed counties to move from precinct level voting to county-wide “vote centers” – locations in a county where all voters will vote, regardless of their address.  Vote centers are theoretically less expensive to administer and conveneit for many voters, but less is known about the impacts on specific communities.  Using Texas’ registered voters’ list from the Secretary of State’s Voting Division the project will use Geographic Information Systems (GIS) to calculate estimated driving distances and times from each registered voter’s residence to the voter’s vote center location. The restuls show that the displacement of traditional precinct level voting and the increase in distance between polling locations takes a greater toll on voter turnout for voters in rural counties and Latinos. 

 Street-level bureaucrats operate under fairly limited oversight, and are also the actors that de facto set the terms for policy implementation. This leaves room for discrepancies between the theoretical intent of a law and how it is actually applied by its most direct arbiters. In American elections, poll workers serve in this street-level bureaucrat position, bridging the gap between legal voting requirements required by state governments and citizen experiences at the polls. Using an examination of voter eligibility requirements and in-depth interviews conducted with poll workers in the city of Chicago—a jurisdiction where poll workers are asked to verify voter identity by signature—this paper examines what constitutes poll worker decisions about voter eligibility. Respondents offered detailed—and varied—information about what they look for when comparing signatures. Reflecting ambiguity in state election code, they also had different ideas about how to navigate mismatching signatures—ranging from relatively lenient responses to strict enforcement. Notably, they suggested that working in the same precinct and with the same people from election to election made the processing of voters easier. 

Post-election audits are thought to bolster voters’ confidence in elections, but it is unclear which aspects of audits drive public trust in election results and why. In a set of survey experiments fielded by YouGov to a sample of 2,000 Americans, we used both factorial and conjoint designs to understand which attributes of election audits are most important for increasing voter confidence in legitimate election results. Overall, we find that what an audit finds is much less important than how the audit is conducted, so long as the audit does not uncover exceptionally large errors. Structural features of the audit, like who conducts it and how its results are announced, turn out to be more consequential to voter evaluations of election results than the actual number of discrepancies found. Although voters are quick to pick up partisan cues about audits, this has not produced a broader polarization around election audits, and voters rationally do not punish an election in which the winner was called correctly for a few mis-counted votes. Our findings suggest that election administrators can bolster voter confidence through the design of election audits, without serious fear that turning up small numbers of errors will harm voter confidence.