key: cord-0062927-4wgvqt5e
authors: Kousser, Thad; Hill, Seth; Lockhart, Mackenzie; Merolla, Jennifer L.; Romero, Mindy
title: How do Americans want elections to be run during the COVID-19 crisis?
date: 2021-04-28
journal: nan
DOI: 10.1177/20531680211012228
sha: f34ebb83470713560862cb3626ed04de951963f0
doc_id: 62927
cord_uid: 4wgvqt5e

To inform the vital conversation among the nation’s political leaders, elections administrators, and scholars about how to hold a safe, accessible, and fair election in November 2020, this article reports how a sample of 5612 eligible American voters, surveyed 8–10 April, wanted to see the election run during the COVID-19 crisis. We embedded a randomized experiment presenting respondents with truthful summaries of the projections of two teams of scientists about the pandemic. Our descriptive findings show that in November 2020, four in 10 eligible voters would have preferred to cast their ballot by mail rather than in person and that a majority of respondents favored policies expanding mail voting. Our experimental findings show that respondents who read the scientific projections were more likely to prefer voting by mail, were more likely to trust that a mail ballot would be counted accurately, and were more likely to favor holding the election entirely by mail.

With the COVID-19 pandemic dramatically disrupting everyday life in the United States, state and local election officials sought to adapt the way that they administer elections to ensure the safety of voters and their workers, while still securing access to the ballot and the integrity of elections. Sixteen states delayed their presidential primary elections because of the public health crisis. US Senators Amy Klobuchar and Ron Wyden introduced legislation to expand voting by mail and adopt other new procedures in the November presidential contest. 1 Leading election law scholars advanced proposals to expand voting by mail (Hasen, 2020) and political science experts on election administration provided guidance (Persily and Stewart, 2020) . By the end of August, 20 states had made a change to their vote by mail policies, allowing at least 83% of American voters to cast a ballot by mail. 2 Courts across the nation, including the US Supreme Court, were asked to rule on cases addressing voting changes in response to COVID-19, 3 and President Trump himself repeatedly weighed in on the topic, asserting that "Mail ballots are very dangerous for this country because of cheaters." 4 Important to informing this vital public debate is the voice of America's voters themselves. In this article, we present the results of a survey that we fielded on 8-10 April, 2020 asking a sample of 5612 eligible American voters for their views on how to run November's elections. We asked a series of questions designed to determine how voters would like to cast their own ballots in the November 2020 election, their confidence in how accurately their own ballots and the ballots of others will be counted if cast through different modes of voting, and their preferences about potential changes in federal policy and funding levels. We report the results of this original survey below and make the data available for replication (data available at https://doi.org/10.7910/DVN/3GFZ9L).

Because the extent of the COVID-19 crisis and how it would unfold over the coming months was uncertain to eligible voters in April 2020, just as it was for the policymakers and election officials who must plan for a general election, we embedded a randomized experiment in this survey presenting alternative future scenarios. We presented two treatment groups of respondents with truthful summaries of the projections of two widely-cited teams of scientists, with one team projecting a peak of the public health crisis in the spring of 2020 and the other projecting that its impact would peak in the fall if social distancing measures were relaxed at that time. A control group received no projections.

This design allows us to address two research questions. The first is descriptive: How do eligible American voters want to see the November 2020 election run-with predominantly in-person voting options, through the mail, or with multiple voting options allowed? Our second question leverages the randomized experiment to make causal inferences about whether eligible voters are influenced by the projections of experts, adapting their personal preferences and policy positions based on predictions about when the COVID-19 crisis will peak. Our hypotheses, which we filed in a pre-analysis plan at the EGAP Registry, predict that exposure to the scientific projections would affect how voters view the election: we expected that both treatments would make respondents more likely to prefer to vote by mail themselves and to support policies that allow this option. We expected to observe the strongest effects for the treatment projecting a fall peak in the crisis.

One aim of this research design is to provide policymakers with conditional guidance, presenting data on how voters want to cast their own ballots and to see election rules change if different scenarios come to fruition. As health experts learn more about the likely timing of the peak impact of COVID-19, election officials may adjust their planning. Our findings under the two treatments can present guidance about how the public views voting under either a spring or a fall peak in the impact of the crisis.

The findings also speak to the political science literatures about voting by mail, voting under personal risk, and whether voters are relatively myopic or forward thinking. A well-developed literature on mail ballot voting has used observational methods (Berinsky et al., 2001; Gerber et al., 2013; Banducci, 2000, 2001; Oliver 1996; Traugott 2001 ; see also Gronke et al. 2008 for a review); natural experiments (Kousser and Mullin, 2007) ; and field experimental methods (Arceneaux et al., 2012) to study which types of voters prefer to cast mail ballots and how shifts to this mode of voting impact political participation. New research explores its effects on the partisan composition of the electorate, showing that mandatory vote-by-mail elections lead to a similar electorate as elections with polling place and mail options (Barber and Holbein, 2020; Thompson et al., 2020) ; but that a partisan gap over voting by mail has appeared and broadened over the course of the pandemic (Lockhart et al., 2020) . Our study contributes to the literature studying voting by mail in the past by assessing how voters viewed this option in 2020, when a major public health crisis posed new risks for voting in person.

Because our experimental design focuses respondents' attention on this crisis, it allows us to observe whether their preferred mode of voting is influenced by health risks. While the risks are of a different nature and scale, voting during a pandemic can be conceptually connected to studies of voting and violence in India (Wilkinson, 2004) and Afghanistan (Weidmann and Callen, 2013) and to the broader literature on elections held under violent conditions (Ellman and Wantchekon, 2000) . And because our two treatments vary the timing of the health risk, any differential effects brought by the "spring peak" and "fall peak" treatments can speak to the literature on how forward thinking voters are (Healy and Malhotra, 2009 ). Among the eligible voters whom we surveyed, we asked are voting intentions about an election taking place more than six months in the future responsive to different predicted scenarios, or is the American electorate at this time looking toward the November 2020 elections solely through the lens of present conditions? The estimated effects of our experimental treatments will shed light on these questions.

We begin by providing the details of our survey and our experimental treatment, and then outline our expectations about how responses may change under our experimental treatments. We then present the overall views of our respondents along with the responses for our control and treatment groups to three sets of questions: How do eligible voters prefer to cast their own ballots-by mailing in their ballot, or by casting it at a polling place on Election Day or at an early voting center?-and whether they would be comfortable waiting in line at a polling place or willing to work as a poll worker, both with and without social distancing measures in place? Which mode of voting gives eligible voters the greatest confidence that ballots will be counted accurately, both for their own votes and the votes of others? What federal policy and funding changes do voters support, including policies that would allow any registered voter to request a mail ballot, require election officials to send a mail ballot to every registrant, or proposals to shift elections entirely to mail ballot voting? Our experimental findings show that respondents who read the scientific projections were more likely to prefer voting by mail (by 6.3 to 7.0 percentage points, depending on the treatment); were more likely to trust that a mail ballot would be counted accurately (by 2.9 to 3.7 percentage points); and were more likely to favor holding the election entirely by mail (by 3.3 to 4.0 percentage points).

We fielded a survey from April 8-10, 2020 asking a nationally diverse sample of 5612 eligible American voters for their views on the upcoming November election, along with other questions on their political and policy views asked after our battery of elections questions. We list the full text of our elections questions and answers in our online appendix, and we post all of our data and replication files at https://doi.org/10.7910/DVN/3GFZ9L. We recruited our sample through the online platform Lucid Fulcrum Exchange, with all respondents completing their surveys online through a Qualtrics instrument designed by the authors. Coppock and McClellan (2019: 1) find that "demographic and experimental findings on Lucid track well with US national benchmarks." Our survey instrument and methodology was reviewed and certified as exempt from IRB review by the UC San Diego Human Research Protections Program before fielding.

Our sample frame is the citizen voting age population of the United States, based on the characteristics of that frame reported in the 2018 American Community Survey. So that our sample of respondents would reflect this larger population, we sampled to meet targets of respondents matching the distributions of key demographic characteristics of voting age citizens: gender, age, education levels, race, ethnicity, and region. We created survey weights based on those targets, using gender on its own, the joint distribution of age by education, 5 and the joint distribution of race by ethnicity. 6 All of our reported results are based on these weights.

In order to ensure that our sample contained a sufficiently large number of respondents in minority racial and ethnicity groups in order to conduct future analyses of the potential for disparate impact of voting regulations, we drew a large sample of 5612 eligible voters nationwide. Even without using weights, our pool of respondents is quite diverse: around 70.7% are White, 11.9% are Black, 10.6% are Hispanic, 5.6% are Asian, and 1.2% reported being of another race. Each of these figures is within two percentage points of the estimates for the nation's citizen age voting population reported in United States Census Bureau (2020). The respondents were also a politically diverse group of eligible voters: asked to recall the 2016 presidential election, 30.1% reported that they did not cast a ballot, and of those who did, 47.8% reported voting for Hillary Clinton, 47.5% reported voting for Donald Trump, and 4.7% supported another candidate. Looking ahead to the November 2020 election, 81.6% of respondents reported that they either "Definitely" (68.4%) or "Probably" (13.2%) intended to vote.

This article reports, first, overall preferences for our battery of questions about the 2020 elections, showing how a sample of America's eligible electorate viewed them in the midst of the COVID-19 crisis. Then we separately report mean responses for three equally-sized groups exposed to different information about the outbreak. We provided each treatment group with truthful summaries of the projections of one of two widely-cited teams of scientists, one at the University of Washington and the other at Imperial College London. One of the teams projects a peak of the public health crisis in the spring of 2020 (the "spring peak" treatment), while the other projects that its impact will peak in the fall if social distancing measures are relaxed at that time (the "fall peak" treatment). The control group received neither forecast. Before answering questions about their preferences on voting, respondents in our survey were randomized into one of the three conditions below: 7 1. The spring peak treatment: "While no one can be certain how the COVID-19 outbreak will progress in the United States, one well-respected team of scientists at a leading university has projected that if social distancing measures are widely adopted, the effects of the virus will reach their peak in April, then gradually decline throughout the spring and into the summer." 2. The fall peak treatment: "While no one can be certain how the COVID-19 outbreak will progress in the United States, one well-respected team of scientists at a leading university has projected that if social distancing measures are widely adopted now but are lifted during the early fall, a new surge in cases will come and the effects of the virus will reach their peak in November or December." 3. A control group that was not presented with any predictions.

Because respondents were randomized into the three groups, we can attribute any observed differences in their subsequent answers to the information provided in the treatments. As in a typical survey experiment, we rely on randomization to isolate the causal effect of our treatment (Sniderman, 1996) . In our appendix, we report the results of a balance check that confirms that our three groups were distributed similarly across conditions on a host of demographic characteristics The appendix also includes a manipulation check showing that most respondents generally acquired the information about the projected timing of the peak. Unlike most survey experiments, because we are interested in both the treatment effect and the overall distribution of responses, we report all results using the survey weights.

Based on our pre-analysis plan registered at the EGAP Registry (the plan is available at https://osf.io/ewdyn), we hypothesize that exposure to either of the treatments providing information about the COVID-19 crisis will affect respondents, making them more likely to prefer to cast a mail ballot themselves and to support policies that would allow this option for others. Our postulated mechanism here is that any information that brings public health concerns to the top of respondents' minds will cue them to think about stay-at-home and social distancing directives (Zaller, 1992) . When threats related to issues such as public health are salient, individuals will be more inclined to support policies that provide protection (Albertson and Gadarian, 2015; Kam and Estes, 2016) . This should make individuals more likely to prefer a safer option of voting at home rather than in person, unless social distancing can be achieved in a polling place.

Second, we expect to see the strongest experimental effects for the projected "fall peak" treatment, which will cue voters to pay particularly close attention to the potential safety concerns of in-person voting options in November. Under conditions of heightened anxiety, as is the case with COVID-19, 8 individuals should be more likely to pay attention to information, process that information carefully, and use that information in the formation of policy preferences and in their behavior (e.g. Albertson and Gadarian, 2015; Huddy et al., 2007; Valentino et al., 2008) . They should therefore be more inclined to prefer voting by mail when they are paying attention to a peak of the fall instead of a peak in the spring. These differential treatment effects may not be strong; indeed, Healy and Malhotra's (2009) work on voter myopia might suggest that they would be null, with voters viewing elections through a current lens. Still, our firm prior, registered in our pre-analysis plan, is that scientific projections which highlight the potential that COVID-19 could still be a major public health threat in November when the general election takes place, will prompt respondents to be more supportive of mail ballot options at that time.

Finally, we expect that all of these treatment effects will carry through from personal preferences to policy preferences, as eligible voters first contemplate how they would like to cast a ballot and then view policy proposals in light of their personal choice. While findings in the literature are mixed on whether people connect their self-interest with policy preferences (for a discussion see Lau and Heldman, 2009) , scholarship suggests they are more likely to do so when the stakes are clear (Chong et al., 2001) , when the issue is highly salient in the environment (Lau and Heldman, 2009) , and when policies are transparent, presented in a clear way and are easy to understand (Chong et al., 2001) . The stakes are certainly high in the case of COVID-19, and policies around supporting vote by mail are fairly clear and easy to understand. To test these hypotheses, we will focus in the data analysis below on the proportion of respondents favoring voting by mail, comparing the two treatment and control groups. Hypothesis 1. Exposure to either treatment providing projections about the COVID-19 crisis will make respondents more likely to favor mail ballot voting and the policies that promote it. Hypothesis 2. Exposure to the treatment projecting that the COVID-19 crisis could peak in the fall will make respondents especially likely to favor mail ballot voting and the policies that promote it.

We anticipate that administering a survey that prompts respondents to think about COVID-19 in the midst of this public health crisis is generally unlikely to yield strong treatment effects. In April 2020, the pandemic is likely to have been at the top of the minds of all respondents, including those in the control group. All respondents are therefore "pre-treated" (see Druckman and Leeper, 2012) . The treatment effects that we report below, then, may well be conservative estimates of the more general impact of raising potential public health threats when asking questions about elections.

Personal preferences for casting ballots Figure 1 presents our first set of results. As with all of our figures, it lists the full survey text of each question in the top left corner of each graph, then reports the mean percentage of respondents in the control and two treatment groups who selected a given answer in gray columns and the overall percentage in the white column. We report our point estimates of these percentages at the base of each column with error bars representing 95% confidence intervals around these estimates (these are roughly analogous to the margin of error of the survey). All confidence intervals and p-values are adjusted for multiple comparisons with a Bonferroni correction, with standard errors based on the 15 analyses that we conduct.

We begin by exploring how individuals preferred to cast their own ballot in November. We find that 39.6% of respondents overall chose mailing in a ballot sent to them a month before Election Day, rather than voting at a traditional polling place or at a professionally-staffed early voting center. This expressed preference differs from the past actions of voters in our sample. Before providing any of the groups with an experimental treatment, we asked respondents whether they knew their polling place location or whether they vote by mail. Only 12.8% respondents indicated that they typically "vote by mail or not at a polling place."

Turning to the survey experiment, the proportion of respondents who most preferred to cast a ballot by mail was six to seven percentage points higher for those who read either of the scientific projections about the COVID-19 outbreak, consistent with our Hypothesis 1. While 35.2% of respondents in the control condition selected voting by mail as their most preferred way to cast a ballot, 41.5% of those provided the "spring peak" prompt and 42.2% of those reading the "fall peak" projection chose this option, with both treatments having effects that were statistically significant at above the 95% confidence level compared to the control group. Each of these effects is not only statistically significant but quite large in scale; the "spring peak" projection brought a 6.3 percentage point increase in expressed preferences for voting by mail, while the fall peak shifted the mode through which voters prefer to vote by a full seven percentage points. These observed treatment effects are strong, especially given that all respondents in April of 2020 were functionally "pre-treated" (see Druckman and Leeper, 2012) by the nation's intense focus at that time on COVID-19. The fact that these estimated treatment effects are similar in scale for the two projections provides support for Hypothesis 1 but not for Hypothesis 2, which expected that the "fall peak" treatment would exert the strongest effects here. We did not observe any treatment effects when we invited respondents, at the conclusion of the survey, to visit a federal website providing information about how to apply for a mail ballot in their state. Overall, 27.4% of respondents selected this option. It may be that people were not interested in implementing steps toward voting by mail so far out from November. While the treatment influenced how people would like to cast their ballot, it did not affect whether they intended to vote.

We also observed no treatment effect when we asked respondents whether they were likely to cast a ballot in November if voting in person was their only option (and, subsequently, if voting by mail was their only option). For both questions, we removed from our analysis those who had, earlier in the survey, indicated that they "definitely" or "probably" would not vote in November, in order to focus on the plans of potential voters. As the final two graphs in Figure 1 show, overall 3.3% responded that they "probably" or "definitely" would not vote if an in-person option was their way to cast a ballot, while 5.2% reported that they would not vote if a mail ballot was their only option. 9 In summary, respondents who read projections about the COVID-19 outbreak shifted significantly toward preferring to vote by mail, but this information did not affect their likelihood of visiting a website about this option or alter their intent to participate in November. This stated commitment to voting, even under a public health crisis, is consistent with existing work that demonstrates a strong social norm that citizens should vote in elections. We can see this in public opinion surveys, where social-desirability bias leads individuals to inflate their reports of voting even if they did not cast a ballot (e.g. Ansolabehere and Hersh, 2012; Enamorado and Imai, 2019; Jackman and Spahn, 2019; Silver et al., 1986) . We also see this in get out the vote (GOTV) studies which show that exerting social pressure messages in mailers is much more effective in increasing turnout than mailers that do not exert social pressure around the social norm of voting (e.g., Abrajano and Panagopoulous 2011; Gerber, 2016; Gerber et al., 2008; Mann, 2010 ; for a review see Green and Gerber, 2019) .

We also asked voters about their comfort levels with waiting in line at a polling place or working as a poll worker, either with or without social distancing measures implemented at the polling place. We asked both versions of these questions to the full sample, in sequence. As Figure 2 shows, when we asked about comfort levels at polling places that did not implement social distancing, we observe a substantively strong and statistically significant effect for the "fall peak" treatment, in the expected direction. Only 51.4% of respondents receiving that treatment said that they would be comfortable waiting in line at such a polling place, compared with 58.8% of respondents receiving the "spring peak" treatment and 59.8% of respondents in the control group. Similarly, respondents who read the fall peak projections were much less likely to be willing to work as a poll worker (30.7%) than respondents in the spring peak (36.1%) or control (36.6%) conditions.

Both of these fall peak treatment effects were significantly well above the 95% confidence level. These findings are consistent with our Hypothesis 2 but not with Hypothesis 1: when respondents anticipated their comfort in November with waiting in line or working at polling places without strict social distancing measures in place, only the "fall peak" predictions influenced them.

When we asked these questions again, but specified that polling places would follow a number of specific social distancing measures recommended by the National Association for the Advancement of Colored People (see https://naacp. org/coronavirus/voter-access-and-participation-duringcoronavirus-pandemic/), respondents in all groups expressed higher levels of comfort and no differences across treatment groups emerged. Overall, 74.8% of respondents would be comfortable waiting in line at a polling place that implemented extensive social distancing measures and 48.3% would be willing to work at such a poll. Finally, 78.6% of respondents would be comfortable dropping off their ballot at professionally staffed drivethrough locations, with no treatment differences observed.

The two questions reported in Figure 3 ask respondents how confident they are that their own ballot will be counted correctly if it is cast either by mail, at a polling place, or at a professionally staffed early vote center. Overall, respondents are less likely to respond that voting by mail gives them more confidence than the other two ways to cast a ballot: 29.6% select voting by mail for their own ballot, and 27.2% are most confident in the integrity of results if other voters cast ballots by mail. It is important to note that many mail ballots are rejected: an estimated 430,000 mail ballots were not tabulated in the United States in the 2018 General Election, and in recent Florida elections, "younger voters, voters not registered with a major political party, and voters in need of assistance when voting are disproportionately likely to have their VBM ballots not count" (Baringer et al., 2020: 1) .

Turning to the survey experiment, both of our treatments increase the likelihood that respondents identify mail ballots as the mode that they have the most confidence in, for each of these questions. These treatment effects, which In the upcoming November 3 election, would you be comfortable dropping off your ballot at a drive−through location professionally staffed by your county elections office, if it was reconfigured in order to adhere to social distancing protocol? range in scale from three to five percentage points, are consistent with our Hypothesis 1. Our interpretation of this finding draws on what we observed in Figure 1 ; respondents who read either treatment about the COVID-19 projections were significantly more likely to select voting by mail as their preferred mode of voting. What we observe here is that they are also more likely to identify that mode as the one that they trust the most for the integrity of counting their ballot and the ballots of others. This likely reflects a psychological process where individuals reason backwards from policy preferences to justifications (Kunda, 1990; Sniderman et al., 1991) .

Finally, in Figure 4 we report how respondents viewed some of the major policy proposals that had been advanced at the national and state level to increase opportunities to cast mail ballots in the November 2020 election. Our expectation was that those in the treated conditions would be more supportive of policies to increase the ease of voting by mail. The first question asked whether they would support national legislation directing all states to send a mail ballot to any voter who requests one, which was a central aspect of legislation introduced on 18 March 2020 by US Senators Amy Klobuchar and Ron Wyden, and it was also among the proposals advanced by the Brennan Center (2020). Overall, 74.7% of respondents who took a position supported this approach (in our policy questions, we calculate proportions based on "Yes" and "No" responses, removing "Unsure" responses). The next question summarized a proposal that moves one step further by sending a vote by mail ballot to every voter, even if they do not request one in advance. Overall, 63.9% of respondents support this option. On neither of these questions did we observe any significant treatment effects.

When we asked respondents about the most controversial proposal-national legislation directing all states to shift entirely to voting by mail-we did observe strong treatment effects consistent with Hypothesis 1. Respondents exposed to either of the COVID-19 projections were significantly more likely than those in the control condition to support this shift toward mandatory voting by mail. Support for this proposal registered at 54.5% for respondents who received no prompt about the outbreak, 58.5% for those in the "spring peak" treatment, and 57.8% for respondents in the "fall peak" treatment.

Our last question asked about support for increased federal funding to support vote by mail systems in state and local governments. We informed respondents that "The Brennan Center has estimated that the cost of supporting all of these measures would be US$2 billion. The recent COVID-19 Stimulus Bill passed by Congress provides US$400 million in grants to states." We asked if they favored "Additional funding to increase total federal support for the 2020 elections to US$2 billion," keeping funding at its current level of US$400 million, or reducing it. There were no observed treatment effects, and overall, 35.3% of respondents favored increasing funding.

In our online Appendix, Figure A1 presents additional analysis of further questions about how to implement voting by mail and online voter registration. For those four questions, we observed no treatment effects.

This article reports on how a sample that reflects the demographic characteristics of eligible American voters viewed the November 2020 elections. We present descriptive findings showing that four in ten eligible voters preferred to cast their ballot by mail rather than in person in November and that a majority of respondents favored policies that would expand vote by mail options. Our experimental findings show that respondents who read the COVID-19 projections were more likely to prefer voting by mail, to express discomfort with waiting in line or working at a polling place that did not practice social distancing, more likely to trust that a mail ballot would be counted accurately, and more likely to support a proposal to hold the 2020 election by mail.

Our goal is not to interpret these findings or advocate specific policy implications; rather, we seek to provide data to inform the vital national conversation about how to plan for elections under COVID-19. It is important to note that whether the expressed preferences for a voting method result in actual partisan differences in voting behavior were revealed in November 2020. Our experimental approach can allow elections officials to see how voters' preferences and perspectives may change depending on two potential scenarios for the future path of infections. The confidence that our sample of eligible voters expresses about their ballots being counted correctly, along with their views on policies now being proposed, can inform policymakers at the national, state, and local levels as well as election reform advocates. The strong and significant treatment effects that we observe in our experiment can also answer broader political science questions about how everyday Americans address physical risk when it comes to their exercise of democracy. To address the potentially disparate impacts of COVID-19 risks on different groups in our nation, there is a need for further study of divergent trends in how demographic and partisan groups answer these questions and respond to the experiment. We are now conducting these analyses, and invite others to do so with our data, to provide a factual basis for a critical national discussion.

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article. direct links to both of the studies containing these projections.

is Who Americans Trust about Coronavirus Information," Washington Post, March 20, 2020. Available at https://www. washingtonpost.com/politics/2020/03/20/were-all-anxiousabout-pandemic-who-do-americans-want-hear/ (accessed 29 April 2020). 9. To determine which voters might be deterred from casting a ballot if their options were restricted, we explored whether voters who expressed strong overall intentions to vote in November would still plan to vote if they only had one option about how to cast their ballot. Among those who expressed, earlier in the survey, that they "definitely" planned to vote in November, only 1.5% responded that they definitely would not vote if voting by mail was their only option. Among those who expressed, earlier in the survey, that they "may or may not" vote in November, 5.9% responded that they definitely would not vote if voting by mail was their only option.

This publication was made possible (in part) by a grant from the Carnegie Corporation of New York. The statements made and views expressed are solely the responsibility of the author.

Democrats Introduce Bill to Promote Mail-In Voting Amid Coronavirus Crisis

At Least 83% of American Voters Can Cast Ballots by Mail in the Fall

Why the Supreme Court Made Wisconsin Vote During the Coronavirus Crisis

Trump is Pushing a False Argument on Vote-by-Mail Fraud. Here are the Facts

Our bins for the ages of respondents are 18-24, 25-44, 45-64, and 65 and older, and for education they are high school or younger, some college, bachelors, or graduate degree

Our bins for race are White, Black, Asian, and Other, and our bins for ethnicity are Hispanic or not Hispanic

In debriefing materials, we provided respondents who received one of the treatments with a direct link to the study that was summarized for their group, and provided all respondents with References Abrajano M and Panagopoulos C (2011) Does language matter? The impact of Spanish versus English-language GOTV efforts on Latino turnout

Anxious Politics: Democratic Citizenship in a Threatening World

Validation: what big data reveal about survey misreporting and the real electorate

Get out the vote by mail: a randomized experiment testing the effects of mobilization in traditional and vote-by-mail precincts

The participatory and partisan impacts of mandatory vote-by-mail

Voting by mail and ballot rejection

Who votes by mail? A dynamic model of the individual-level consequences of voting-by-mail systems

How to protect the 2020 vote from the coronavirus

When self-interest matters

Validating the demographic, political, psychological, and experimental results obtained from a new source of online survey respondents

Learning more from political communication experiments: pretreatment and its effects

Electoral competition under the threat of political unrest

Validating self-reported turnout by linking public opinion surveys with administrative records

Why people vote: estimating the social returns to voting

Social pressure and voter turnout: evidence from a large-scale field experiment

Get out the Vote

Convenience voting

Trump is wrong about the dangers of absentee ballots

Myopic voters and natural disaster policy

On the distinct political effects of anxiety and anger

Why does the American National Election Study overestimate voter turnout

Disgust sensitivity and public demand for protection

Going postal: how all-mail elections influence turnout

Absentee voting, mobilization, and participation

Does voting by mail increase participation? Using matching to analyze a natural experiment

The case for motivated reasoning

Self-interest, public attitudes, and support for public policy

America's electorate is increasingly polarized along partisan lines about voting by mail during the COVID-19 crisis

Is there a backlash to social pressure? A largescale field experiment on voter mobilization

The effects of eligibility restrictions and party activity on absentee voting and overall turnout

Ten recommendations to assure a healthy and trustworthy 2020 election

Who overreports voting?

Innovations in experimental design in attitude surveys

Reasoning and Choice: Explorations in Political Psychology

Universal vote-by-mail has no impact on partisan turnout or vote share

Citizen Voting Age Population by Race and Ethnicity

Is a worried citizen a good citizen? Emotions, political information seeking, and learning via the internet

Violence and election fraud: evidence from Afghanistan

Votes and Violence

The Nature and Origins of Mass Opinion

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The authors are grateful to the University of California Office of the President's Multicampus Research Program for providing funding for the New Electorate Project, Grant ID MRP-17-454899.

Thad Kousser https://orcid.org/0000-0002-5489-5837Supplemental Material

The online appendix is available at: http://rap.sagepub.com/ content/by/supplemental-data Notes