Leading up to the 2020 general election, state election boards grew concerned that the coronavirus disease 2019 (COVID-19) pandemic might drive voters away from the polls or that crowded polling stations would spread the virus and lead to a wave of new cases, hospitalizations, and deaths. In an effort to safely conduct the election, many states changed their voting laws by implementing automatic voter registration, removing excuse requirements for absentee ballots, and expanding early voting windows. These changes, meant to encourage turnout and protect public health, were expensive to implement, politically contentious, or both. But did the changes have the desired effects?
This report examines the impact of voting laws on voter turnout and choice of voting method (referred to from here on as voting method) in the 2020 election and the effects of in-person voting on the spread of COVID-19. However, it does not directly estimate the effect of state election processes on the transmission of COVID-19 during and following the 2020 presidential election. Below, we describe the relationships between voting laws and in-person voting on Election Day, and those between in-person voting and COVID-19 transmission.
Although voting laws and their effects on voters' behaviors have long been an area of study for social scientists, the COVID-19 pandemic introduced additional risks to in-person voting and might have made flexible voting options, such as absentee ballots, more appealing to would-be voters. Throughout the report, we use an index to measure states' aggregate election flexibility that was created in a previously published RAND report, An Assessment of State Voting Processes: Preparing for Elections During a Pandemic (Kavanagh et al., 2020). The RAND flexibility index documents states' election processes and summarizes their flexibility along three dimensions: voting registration, early voting, and remote voting. States' election flexibility is rated as one of four scores: very low, low, medium, or high, which we use as the main measure in our analyses. Because the prior report rated states' election flexibility as of July 2020 and many states changed their voting laws between July and November 2020, we update states' election flexibility ratings for the November 2020 election.
In addition to studying the relationship between states' election flexibility scores and voter turnout using descriptive statistics, in the primary analysis, we use a panel of county-level voter turnout and a two-way fixed effects model to estimate the difference in the effectiveness of flexible voting policies in 2020 relative to their importance in past elections.
We then test whether voting laws changed the methods that voters used to cast their ballots. If voting laws kept polls from becoming overcrowded on Election Day, they might benefit public health by mitigating the spread of COVID-19. We use data from Aristotle, a campaign finance and reporting firm, that includes information on voting method in 2020. Although the data are large and contain rich information about voters' characteristics (e.g., income, race/ethnicity), this information is only available for the 2020 presidential election—a feature that prohibits the use of panel data methods (e.g., two-way fixed effects) to identify the effects of voting laws. Instead, we construct an empirical model that controls for several important covariates (i.e., factors that influence voters' choice of voting method but are not of direct interest). Only under restrictive assumptions does the model produce estimates of the causal effect of election processes on voting method. Instead, it is best to interpret the findings as robust conditional correlations.
To identify visits to polling stations, we use anonymized cell phone data and records of polling station locations from SafeGraph and the Center for Public Integrity, respectively. The data are a cross section, which again limits the statistical tools available to us. We incorporate several important covariates into our empirical model to estimate how states' election flexibility affects the number of long visits to polling stations on Election Day. Finally, in our test of in-person voting's effect on COVID-19 spread, we employ both difference-in-differences and event study regressions to measure the static and dynamic causal effects of in-person voting during the pandemic.
It is our hope that the empirical evidence presented in this report will help inform future debate and policy on voting laws by documenting the effects of various election policies during a public health crisis. This report does not assess the effect of flexible voting laws on the incidence of voter fraud or public perception of election integrity.1
Trends in Voting Processes
States relaxed their voting laws leading up to the election. Fifteen states implemented changes to their voting processes, which led to increases in their aggregate election flexibility score between the concluding research date from Kavanagh et al., 2020, and Election Day on November 3, 2020. Figure 1 maps states' election flexibility scores on Election Day.
The flexibility of states' election processes on Election Day increased voter turnout. Residents in states with higher election flexibility scores were 0.9 percentage points more likely to cast a ballot in 2020 than were residents in states with marginally lower scores. The ease of acquiring absentee ballots likely explains the higher turnout: Voters were, on average, 20 percentage points more likely to mail in their ballot than voters in less-flexible states.
Voting Behaviors by Demographic
The responsiveness to voting laws varied by demographic. Democrats were more likely to vote and far more likely to vote by mail than Republicans when offered more-flexible election processes. Among all ethnic groups considered, Black voters were least responsive to changes in election processes. Although estimates suggest that Asian voters were 27 percent more likely to use an absentee ballot when their state's election flexibility score increases, Black voters were only about 15 percent more likely to do so. We suspect that distrust in the security/efficacy of absentee ballots might explain, to some degree, both Republicans' and Black voters' hesitancy to vote by mail, but we do not explicitly test this hypothesis.
In-Person Voting and COVID-19 Spread
States with less-flexible election processes had a larger share of their voters cast ballots in person on Election Day. This difference had consequences: U.S. counties with higher per capita in-person Election Day voting had higher rates of new COVID-19 cases in the first four weeks after the election. If America had completely eliminated in-person voting, our model predicts there would have been roughly 500,000 fewer new COVID-19 cases in the nine weeks following the election.
Taken together, we conclude that voting laws can mitigate the spread of COVID-19 and increase voter turnout by making remote voting more accessible. The results provide evidence of the importance of flexible voting laws—particularly for elections held during a pandemic.
However, this report is not a full cost-benefit analysis of election processes. We do not closely examine the financial costs that many laws and processes entail, or whether absentee ballots lead to greater incidence of voter fraud—though extant research shows voter fraud to be an exceptionally rare phenomenon, whether the ballot is cast in person or mailed in (Bump, 2016).
This caveat aside, the report does identify and, to the extent possible, quantifies three effects of flexible voting laws: (1) large increases in voter turnout, (2) reductions in in-person voting on Election Day, and (3) slower spread of COVID-19 in the weeks after the election. State legislatures should weigh these benefits as they assess the many proposals to change voting laws.