Publications

How do we ensure the accuracy and integrity of a statewide voter registration database, which often depends on aggregating decentralized, sub-state data with different list maintenance practices? We present Bayesian multivariate multilevel model to account for common patterns in local data while detecting anomalous patterns, using Florida as our example. We use monthly snapshots of state’s voter database to estimate countywide change rates for multiple response variables (e.g., changes in voter’s partisan affiliation), and then jointly model their changes. We show that there is much heterogeneity in how counties manage voter lists, resulting in very different patterns in additions, deletions, or changes of records. Our method identifies several Florida counties with anomalous rates of changes in the 2016 election.

Voter list maintenance has received increasing attention in the popular press and from advocacy groups in the past few years. Little scholarly work, however, has detailed who is removed - or "purged" - despite no change in their legal eligibility to vote. This paper leverages data from North Carolina to investigate the characteristics of purged individuals between 2010 and 2016. Although we find that minority voters were less likely than white voters to be purged, they were significantly more likely to cast a provisional ballot after being purged despite no change in their eligibility to vote. Purged minorities who cast a provisional ballot were also less likely to have their provisional ballots counted than white voters. This paper presents the first evidence that imprecise voter list maintenance disproportionately disenfranchises voters of color.

Maine recently became the first state to implement instant-runoff voting, or ranked choice voting (RCV) in U.S. Senate and House elections. Before a court ruling, the state also successfully conducted RCV primary elections for governor in the summer of 2018. With seven candidates vying for the nomination, and as the first time voters in the state would try RCV, this race offered an opportunity to examine voting behavior to see how voters acted in a new electoral system.

This paper explores three areas of RCV voting through the analysis of Maine’s 2018 gubernatorial primary. First, this paper offers a descriptive analysis of the ballots cast in the primary, examining various voting propensities and patterns. Second, the paper attempts to address questions about the confusing nature of RCV by looking broadly at ballot errors. Third, the paper addresses the criticism that RCV does not always guarantee majority winners by examining ballot exhaustion in close detail. As the extant literature on RCV has looked at many of these areas separately, it is worthwhile to examine Maine in a broad context rather than through reexaminations of any one single avenue. Using this method, we can see how the inclusion of new data from Maine might bolster or weaken current arguments about RCV. In sum, voting behavior in Maine seems to be in line with many other studies. Rates of single-shot voting seem to be lower than reported in other locations, suggesting more participation in ranking, but the rate of completely ranked ballots was also lower, likely due to the large number of rankings allowed. Error rates seemed to be in line with other studies, but we find that the actual effect of any ballot errors was minimal. Finally, the number of completely blank ballots was at the low end of one study, but it is difficult to find comparable data. These blank ballots should be the subject of additional study.  

An increasing number of states have adopted laws that require voters to show photo identification to vote. We show that the deterrent effect of strict ID laws on turnout persists even after the laws are repealed. To assess the persistent effect of ID laws on turnout we leverage administrative data from North Carolina and a photo ID law that was in effect for a primary election, but not the subsequent general election. Using exact matching and a difference-in-differences design, we show that the photo ID law caused a 1 percentage point turnout decrease for voters without a North Carolina ID law in the primary election. After the law was suspended this effect persisted: those without an ID were 2.6 percentage points less likely to turnout in the general election. The general election effect is robust to a variety of alternative explanations and we show is consistent with aggregate analyses that find a null effect of voter ID laws. Our results suggest that photo ID laws’ deterrent effect persists because voters lack information about the changing requirements for voting, creating confusion that keeps them from voting.

Felon disenfranchisement laws restrict the voting rights of more than 6 million US Citizens. Beyond the effects on voter turnout and electoral outcomes, how do these laws affect individual-level attitudes and behaviors? This study implements two field experiments embedded within panel surveys conducted before and after statewide elections in Ohio and Virginia. The survey population is composed of US citizens with felony convictions who were once disenfranchised, but are now either eligible to vote, or to have their voting rights restored. Experimental treatments provide varying assistance with the restoration of voting rights and voter registration. Treated subjects report stronger trust in government and the criminal justice system, and an increased willingness to cooperate with law enforcement. The results suggest that reversing disenfranchisement causes newly enfranchised citizens to increase their pro-democratic attitudes and behaviors - all of which are predictors of reduced crime and recidivism. 

The most commonly accepted model of public attitudes toward election rules assumes that citizens follow the cues of their preferred party’s elites and support rules that would benefit that party in elections. This paper proposes an alternative model in which most citizens prefer fair electoral institutions at the expense of partisan interest when that choice is made explicit, while a minority of committed partisans are driven by partisanship. To test this theory I use two survey experiments and the specific case of redistricting to determine how the presence of party labels and evidence of the opposing party behaving unfairly affect citizens’ choice between a “partisan gerrymander” district map and a “nonpartisan fair” map. The first experiment finds that while introducing party labels makes partisans more likely on average to choose a gerrymandered map, a clear majority of partisans choose a nonpartisan map across all experimental conditions. Only the those citizens who strongly identify as members of a political party or score highly on a measure of negative partisanship are likely to choose partisanship over fairness. The second experiment finds that presenting Democrats with evidence of egregious Republican gerrymandering causes them to be more likely to support similar pro-Democratic gerrymandering, but the reverse was not true for Republicans. 

The topic of district size was so important in the early American republic that even before it passed what would become the first amendment, the 1st Congress approved an amendment to limit the size of constituencies in the House of Representatives. In contrast to the amendment providing freedom of speech, assembly and religion, this earlier amendment dictating a ratio of representation failed to receive the requisite support among the states. Absent this constitutional provision, the number of representatives and thus district population is governed by statute, the most recent of which caps the size of the House of Representatives at 435. With over 300 million people living in the United States, each representative is responsible for more than 700,000 constituents.

The framers were rightly concerned about how one person could provide adequate representation to so many people; however, the level of apportionment used by the United States has other ill-effects. In particular, we show that low levels of apportionment (a high ratio of constituents to representatives) exacerbates bias in electoral outcomes. We show that increasing the population while holding constant the number of representatives can lead to more bias in electoral outcomes. We test this claim using Monte Carlo simulations of hypothetical districts under varying levels of apportionment. Using a computer algorithm we produce tens of thousands of alternative maps of congressional districts under existing levels of apportionment and under levels of apportionment in which there are more members of the House. Using this large set of hypothetical districts, we find that as the number of representatives increases, the expected level of bias decreases.

In this paper, I examine the effects of vote-by-mail on voter behavior and voter knowledge. I argue that vote-by-mail electoral systems result in a more informed electorate, because voters have additional time with their ballot and access to resources to conduct research about races on the ballot that they know nothing about. I present the results of two empirical studies that support this prediction. First, I find that all-mail elections in Utah cause a 6.368 percentage point decrease in straight ticket voting. This is consistent with the logic that voters spend more time with their ballots when voting by mail relative to when they are voting at a polling place. Second, I estimate the effects of vote-by-mail on voter knowledge using an original repeated cross-sectional survey that was fielded during the 2018 general election in California. The research design exploits the implementation of the California Voters Choice Act (VCA), which resulted in five counties in the state switching to an election system in which all voters in the counties are sent a mail-in ballot. I find that the VCA causes an increase in voter knowledge and an increase in time that voters spend gathering information about the election. However, the reform does not affect the prevalence of political discussion or levels of knowledge about the party identification and ideology of candidates.

Objective. In the historic 2018 Georgia gubernatorial election, which for the first time featured a black Democratic woman against a white Republican man, we assess opinions toward voting-related concerns and their impact on voters’ preferences. Methods. We surveyed political behavior with a representative sample of likely Georgia voters to gauge their perceptions of voting-related improprieties and their possible effect on vote choice. Results. Our analysis makes it clear that not only did voters divide over election-related concerns, which dominated the narrative of the gubernatorial contest, but opinions on this matter strongly influenced the choice for governor. Conclusions. Reflecting a national development, but amplified in Georgia because demographic changes and growing minority participation are increasing competition, election administration has become highly politicized and it was the focal point of the 2018 gubernatorial campaign and the principal short-term factor shaping mass voting behavior.

Since the 2016 American election, issues of cyber-security have been at the forefront of public debate on electoral integrity. Issues of disinformation, fake news, voter privacy, and database hacking have been of top interest to policymakers, scholars and the public alike. One key theme that underlines this debate is how the use of new technologies impact the perceptions of voters and their trust in the electoral process. Even if elections are made secure from hacking and manipulation, just the perception of this threat as prevalent will impact voters’ confidence in elections and democracy more generally.  

This paper responds to the question: How does the use of electoral technology at the voting booth impact voters’ trust in the electoral process? This project focuses in on the use of voting machines at polling places, one of the most basic and well-known forms of technology used in elections. While election administrators use technology throughout the electoral cycle, for everything from voter registration to announcing the results, the voter most intimately experience the use of technology in elections through their experience at the ballot box, with the use of direct recording electronic voting mechanisms, optical scanning machines or other forms of e-voting technology.  

This paper seeks to provide additional evidence on the impact of these technologies on voter trust, by considering the use of technology in in-person voting in the United States. The American experience is particularly useful to study since the technology used at the polls varies widely between American counties, due to a de-centralized system of election administration. Taking advantage of the differences in electoral technology used between and within state, this paper considers the impact of voting technology on citizen trust in American elections in the 2014 and 2016 American elections. It considers the predictors of public survey responses on trust in elections. It tests whether there is a relationship between the technology used and voter confidence in their vote and concerns about hacking. The results speak to current debates about the use of technology in elections and their influence on public trust.