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.
2023 Papers
The number of vote-by-mail (VBM) ballots in the 2020 presidential election reached record highs. Reasons for the sudden increase in voters using VBM ballots range from accessibility to social distancing restrictions and other COVID-19 pandemic health and safety concerns. In conjunction with other circumstances encircling the presidential election, the reporting of VBM results was delayed, which led to mistrust and contributed to mis- , dis-, and malinformation surrounding the election results. Ultimately, it was difficult for election officials to predict how long it would take to verify and tabulate VBM ballots, especially due to the rapid adoption and expansion of VBM. To understand the VBM verification and tabulation system, this research utilizes time studies to define rates, distributions, and processing times for VBM processes in order to support election officials in preparing for VBM results reporting in future elections. The data explored in this study were collected from several counties in the greater Salt Lake City region during the 2022 midterm election. These data consisted of time studies on manual and machine-supported VBM process steps, including, but not limited to, ballot arrival, signature verification, ballot extraction, tabulation, and adjudication. Through statistical methods, processing times, processing rates, and representative probability distributions for each VBM process step are defined. These data can assist in predicting the necessary workforce and forecasting the time to report
election results with existing equipment. The results aid in supporting election officials and administrators in making data-enabled decisions for future election planning and scheduling.
Does experience with voting by mail reduce mail-in voter fraud beliefs? Can personal experience counter partisan forces when the two conflict? I answer these questions by examining mail-in voter fraud beliefs in a single-state study following the 2020 presidential election, in which mail-in voter fraud became a more intense political and partisan issue. I focus on the ways in which partisan affiliation and personal experience may shape voters’ attitudes about mail-in voter fraud. Building on Zaller’s (1992) RAS model and Cramer and Toff’s (2017) framework of personal experience, I develop a theory to explain how firsthand experience can mitigate partisanship. Using OLS regression and instrumental variable analyses to test this theory, I find suggestive, causal evidence that voting by mail decreases beliefs about the prevalence of voter fraud. My results also confirm that affiliation with the Republican Party increases voter fraud beliefs. Including an interaction effect in the analysis provides no support for my hypothesis that the effect of voting by mail is conditioned on partisanship. I conclude from these findings that voting by mail can indeed mitigate the effect of partisanship on beliefs about fraud, despite the strength of partisan ties and polarization of the current political era. I consider the potential policy implications of these findings and argue that efforts to combat misinformation about mail-in voter fraud should consider interventions that increase voters’ experience with voting by mail.
In 2020, Alaska voters approved a ballot measure that implemented Ranked Choice Voting (RCV) in statewide and state legislative contests, as well as Congressional races. The 2022 election was the first cycle in which voters and election officials experienced voting using RCV rules. This paper examines the information landscape surrounding the RCV elections, including how organizations, candidates, political parties, and election officials communicate with voters in 2022 about the new rules. We use data gathered from tweets and newspaper coverage surrounding the election, as well as through interviews with Alaskan stakeholders. The paper also examines election results of primary and general elections from 2014 through 2022 to examine whether RCV created increased voter confusion. Finally, we note that Alaska provides unique opportunities to analyze RCV understanding among language minority communities (Tagalog, Spanish, and Native Alaskan). Therefore, we will also analyze if voters’ understanding of RCV rules varies by language communities, in addition to other demographic factors among voters.
This paper presents a new measure of gerrymandering, called the Reverse Gerrymandering Index (RGIx), and uses it to analyze the evolution of packed districts in the US House of Representatives from 1872 to 2022. Unlike the efficiency gap and partisan symmetry measures, which are influenced by vote variation across the entire state, the mainstay of RGIx is that it estimates the gerrymandering level of each district by comparing its partisan vote distribution with the partisan vote distribution of the districts adjacent to it. From there, RGIx can be used to identify highly gerrymandered districts, measure regional or historical variation in gerrymandering, or as a variable in statistical models, to name just a few applications. It can also calculate gerrymandering scores for an unlimited number of parties per district, making it applicable to comparative electoral systems research. In this paper, I use RGIx to show that there were few packed congressional districts for nearly a century after the American Civil War, that gerrymandering began rising rapidly in federal elections in the 1960s, and that it began to shift dramatically in a pro-Republican direction starting in the 1980s.
In the past two decades, voter registration policies have changed to provide more options for citizens to register to vote through state elections websites, automatically through interactions with state agencies, on the day of elections or early voting, and before they turn 18 years of age. Despite these expansions in the methods available for citizens to register to vote, voter registration has remained significant barrier to participation for young voters and voters from historically underrepresented backgrounds. In this study, we seek to evaluate the extent to which expansions in voter registration policies in recent years have impacted voter registration and turnout in the United States, especially among young voters and racial or ethnic minority voters.
Using a nationally-representative voter file sample and an analysis of state-level policies, we conduct an empirical, multi-level analysis to estimate the potential impacts of expanded voter registration policies (specifically Automatic Voter Registration (AVR), Online Voter Registration (OVR), Same-day Voter Registration (SDR), and pre-registration of youth before they turn 18 (pre-reg)) on voter registration and turnout while controlling for both individual and contextual state- and county-level factors that are known to impact voting behavior.
We find that although there is a correlation between SDR and turnout, and between OVR and registration rates, these relationships are not significant after controlling for confounding factors at the state and county level. However, when we look at subgroups of interest, there are significant relationships between policy and outcomes. For youth, same day registration is positively associated with voting, but negatively associated with registration, which may indicate that SDR is supporting those voters most susceptible to being dropped from the rolls. OVR is associated with higher registration rates among youth and negatively associated with turnout. This suggests that while OVR makes registration more accessible to young people, it is not mobilizing them to get to the polls. For people of color, we see a similar negative relationship between SDR and registration, which again may mean that the most vulnerable parts of the population are not staying on the rolls between elections. While these findings are somewhat complex to parse, they do offer some initial evidence for the role that expansive registration policies can play in reaching traditionally underrepresented voters and merit further study.
Cities offer a unique context for the study of redistricting because the national partisan divide is often less relevant and because most U.S. cities feature large minority populations. The latter characteristic is important because minorities regularly lobby for majority-minority districts in their cities. Despite their perception as an important tool for minority empowerment, it is unclear what conditions facilitate the creation of majority-minority districts. In this paper, we have taken geospatial data of over 100 city council district maps, merged them with census demographic information, and used an MCMC-based redistricting simulator to draw a representative sample of the underlying distribution of plausible maps within each city. We demonstrate that when majority-minority districts are viable, cities tend to implement more of them than are drawn in the average race-neutral simulation. This is true of both Black-majority and Latine-majority districts. We also find that citizenship and segregation rates are fundamental determinants of the number of Latine-majority and Black-majority districts than can be drawn, as well as the number implemented.
Do local election officials descriptively and substantively represent their constituents? Election officials are uniquely situated to influence participation rates and alleviate persistent racial and ethnic disparities in voter participation. Yet recent surveys of election officials have found them to be overwhelmingly white. Using a newly collected panel of local election officials across hundreds of counties and over two decades, a series of race
imputation methods, and large scale administrative and vendor datasets on turnout and race, along with a differences-in-differences design, I test whether minority election officials increase turnout and registration rates of their non-white constituents. Additionally, I examine whether minorities administer elections differently. I find that descriptive representation of Black voters is increasing among election officials, and that minority and white election officials administer elections in similar ways. These findings have implications for the importance of representation among local election officials and may provide insight into reducing the racial turnout gap.
Politicians and pundits have made trust in the administration of American elections an issue of political disagreement. Combining politicization with inflexible partisan polarization could undermine an essential condition of democracy: the peaceful transfer of power following elections. Can messaging about trust in elections break through partisan polarization? Partnering with election officials from Los Angeles County, Colorado, Georgia, and Texas, we used messaging experiments with nearly 8,500 Americans following the 2022 US midterm elections to measure the impact on trust in elections. We find that state and local election officials are particularly effective at increasing trust in their own state elections. Our pooled estimate suggests that one 30-second official advertisement increases trust in local elections by about one-fifth of the pre-treatment difference between Democrats and Republicans. Videos explaining protections on election integrity in Arizona and Virginia increase trust that our national sample reports in elections administered outside their own state. Our results suggest election officials can break through partisan politics and play an important role in rebuilding trust in the democratic process.
What does it take to make American voters feel more confident in the electoral process? Recent work has explored questions along these lines, assessing voter trust as a function of information diets, endorsements of the electoral process by co-partisan elites, past experiences, modes of voting, and election outcomes. We investigate whether public opinion about the accuracy and security of elections in America are anchored by how much is spent on them. Applying this “price-quality” heuristic to the context of elections, we specifically test whether increased funding for elections increases voter confidence. Using a preregistered survey experiment fielded by YouGov on a sample of 2,000 American voters, we provide novel insights into what voters know about
the sources of election funding, how they evaluate the competing fiscal demands of local governments, how they prioritize various tasks of election administration, and their support for proposals to increase elections funding. To our knowledge, this study represents the first instance in which such questions have been asked in an experimental context. The overall pattern of results suggest that voters are generally misinformed about how elections are funded; voters are divided on how election administrators can improve elections; and while voters generally view current levels of spending on election
as excessive and are not motivated to broadly increase funding, spending on elections nevertheless factors into evaluations of election quality. Taken together, these findings shed light on what voters think about election administration and the capacity for money to shape attitudes about the electoral process.
We examine changes to 175,311 precincts between the 2016 and 2020 general presidential elections. Our data are the result of a unique effort to collect precinct boundaries that has never been accomplished before: on a national scale across multiple elections. We observe that precincts that underwent major changes – beyond minor changes that are generally reflective of city and town annexations affecting small populations – contain a greater share of Hispanic and Black residents and are more likely to be found in denser population areas than those that do not change. We find precincts that underwent major changes on average experienced slightly lower turnout rate increases in 2020 than those that did not change.
Voting Buddy (votingbuddy.com) was launched during the 2022 election cycle to help voters with their decision-making. The tool asks users five “Myers Briggs” style questions related to political ideology. The five questions include an overview question, and more detailed questions regarding sentiments towards non-Americans (domestically and abroad), social welfare and fiscal policy, social policy, and industrialization and corporate policy. The tool then provides textual and graphical assessments of each user and matches each user with like-minded politicians/candidates. The tool can match users with politicians/candidates nationwide or within Congressional voting districts based upon user provided zip codes. Prior to the 2022 election, Voting Buddy was preloaded with assessments of all U.S. Senators and Representatives and all of their opponents (including third party and non-affiliated candidates). Because Voting Buddy’s core team includes political scientists, educators, and engineers, Voting Buddy’s algorithms, assessments, and comparisons were all found to be impartial by Voting Buddy users (voters and news media). The proposed paper will discuss Voting Buddy’s role to date in helping to foster an informed and engaged voting public. Examples include Voting Buddy’s use at voter registration events to engage voters, engaging first-time voters (high school/college students) with Voting Buddy, etc. In addition, the paper will discuss ways that Voting Buddy can be used in the future to inform and engage the voting public. This discussion will also project the election-related implications (related to turnout, voters more confidently participating in the election process, etc.) if Voting Buddy is successful.
COVID-19 caused worldwide disruption to virtually every aspect of human life, including elections. This study assesses the impact of potential COVID exposure, convenience voting policies, and partisanship on voter behavior in the 2020 US general election. Using a new data set comprising county and state data, we demonstrate that countywide COVID-death rates depressed turnout from 2016 levels. COVID mortalities, partisanship, and the availability of different balloting options contributed to changes in the use of mail and early-in person voting. Early spikes in COVID deaths had the largest impact, suggesting once voters chose whether or how to vote, they kept to their decisions, despite the availability of new information about declining infection rates, new vaccines, and improved treatments.