Elderly voters sit outside in camp chairs while waiting in line to cast ballots

Voters wait in line to cast ballots on the first day of early voting in New City, a New York City suburb, New York, October 24, 2020

Photo by Mike Segar/Reuters

Our recent report, Attitudes on Voting in 2020 (Kavanagh, Gibson, and Hodgson, 2020a), described Americans' intentions to vote in the November 2020 election as determined by a survey conducted in May this year. We examined the likelihood that they would vote and how they planned to vote—whether in person or by mail. To assess how attitudes have changed since then, we asked the same set of questions in August to a subsample of 1,841 respondents from the original survey and compared the responses.[1]

Changes in Intention to Vote

Changes Across Respondents

We asked respondents to report their intention to vote in percentage terms from 0 to 100.[2] Figure 1 shows the results of our survey in May 2020 and August 2020. We consider four categories: nearly certain voters (intention to vote greater than 90 percent), uncertain voters (intention to vote between 51 and 90 percent), uncertain nonvoters (intention to vote between 10 and 50 percent); and nearly certain nonvoters (intention to vote less than 10 percent).

Figure 1. Changes in Intention to Vote, May–August 2020

May August
Nearly certain nonvoters (intention to vote less than 10%) 12% 16%
Uncertain nonvoters (intention to vote between 10% and 50%) 11% 5%
Uncertain voters (intention to vote between 51% and 90%) 11% 11%
Nearly certain voters (intention to vote greater than 90%) 66% 67%

The greatest change in voter intent at the aggregate level between May and August is a decline in the number of uncertain nonvoters (from 11 percent to 5 percent of the total) and an increase in the number of nearly certain nonvoters (from 12 percent to 16 percent). We also observe a small increase in the percentage of voters who reported in August that they were almost certain to vote. Given the relatively small size of these changes, the relative stability at the aggregate level in intention to vote between the two periods is worth noting. This does not rule out the possibility that the continuing coronavirus disease 2019 (COVID-19) pandemic (or any other situational factor) is depressing intention to vote; but, at the very least, any decline resulting from the pandemic was small enough between May and August to be offset by other factors, such as enthusiasm for a candidate or issue.

Changes at the Individual Level

Stability at the aggregate level, however, could be the result of some people increasing their intention to vote and others decreasing their intention to vote. To explore this possibility, we examined changes in attitude at the individual level. These results are shown in Figure 2. The key findings were as follows:

  • There was little change among those who were nearly certain voters and nearly certain nonvoters in May. Almost all respondents who reported in May that they were nearly certain to vote (intention above 90 percent) reported the same in August. Of the remainder, the majority expressed an intention above 50 percent (uncertain voters). Almost three-quarters of those who said in May that they were nearly certain not to vote (intention of less than 10%) reported the same in August.
  • Change in intention was more common among those who were uncertain voters and uncertain nonvoters in May. Thirty-seven percent of those who were uncertain voters in May remained the same in August. Of the remainder, 30 percent became more certain that they would vote and 30 percent reported that they were nearly certain nonvoters. For those who were uncertain nonvoters in May, slightly more than 30 percent reported the same in August, but about 20 percent said that they were nearly certain to vote and about 30 percent said that they were nearly certain to not vote. These notable changes underscore the fact that voters updated their intentions as 2020 has progressed.

Despite some stability at the aggregate level, then, there is a good deal of change in intention to vote at the individual level. Most of this change occurs among those who were uncertain voters or uncertain nonvoters in May. Between May and August, some portion of both groups become more certain in their intent, shifting either to the nearly certain voter or the nearly certain nonvoter category (a total of 55 percent of those who were uncertain voters and nonvoters in May). Almost one-half of those who expressed some degree of uncertainty about their intention to vote in May remained uncertain in August. Once again, although it is possible that the pandemic is depressing turnout among some voters, we do not see a clear or consistent decline that would suggest widespread effects of the pandemic.

Figure 2. Changes in Intention to Vote at the Individual Level, May–August 2020

May August
Greater than 90% 66% 67%
51–90% 11% 12%
10–50% 11% 5%
Less than 10% 12% 16%

Who Has Changed Their Intention to Vote Since May?

The third question that we asked about intention to vote is whether certain individuals are more or less likely to report higher (or lower) intentions to vote in August than in May. Table 1 identifies the characteristics associated with both types of change.

Shifts in voting intention depended most significantly on political affiliation and perceptions of safety. We observed that Republicans were more likely than Democrats (as well as independents and other affiliations) to report an increase in voter intention between May and August; Democrats were correspondingly more likely to report a decrease in intention to vote. Second, we found that those voters who live in the Midwest appeared most likely to show an increase in intention to vote (controlling for political affiliation), which is notable from the perspective of election outcomes but is not clearly linked to the pandemic context.[3]

The relationship between intention to vote and perceptions of safety is more complex. First, those who reported in either May or August that they felt confident that the elections would be conducted safely reported an increased intention to vote in August compared with their intention in May. Interestingly, it did not matter in which of the two surveys (August or May) they expressed feelings of safety. Even voters who expressed perceptions of safety in May but not in August reported an increased intention to vote. There are several possible explanations for this. One is that factors outside the perception of safety (such as enthusiasm for a specific candidate or issue) are driving attitudes toward voting. Another possibility is that voters have increased intention to vote despite safety concerns because of limited availability of mail-in options in their states or information about state safety protocols.

Those who said that they did not expect to be able to vote safely in person in both May and August were most likely to report a decline in their intention to vote. Most often, these shifts were into the nearly certain nonvoter category (intention below 10 percent). Finally, those who reported high perceptions of safety in both May and August had the highest intentions to vote, although we did see some downward trend among these voters out of the nearly certain voter category (intention above 90 percent). Still, these respondents remain more likely to vote than those with less expectation of safety, and, overall, the total number of voters shifting from above 90 percent to a lower intention is small.

Table 1. Characteristics Associated with Changes in Intention to Vote, May–August 2020

Increased Intention

  • Republicans (compared with Democrats, independents, and other affiliations)a
  • Midwestern residence (compared with Northeast residence)
  • Those who reported feeling safe in May or August survey

Decreased Intention

  • Democrats (compared with Republicans)a
  • American Indian or Alaska Native respondents (compared with White respondents)
  • Those who did not feel safe in both May and August surveys

a We use self-identified party affiliation rather than official registration.

Changes in Intended Voting Method

Changes Across Respondents

In both May and August, we asked respondents how they planned to vote, whether by mail or in person.[4] Figure 3 compares intentions to vote by mail and in person as reported by our respondents in May and in August. For the purpose of this analysis, we combined all mail-in voting options (e.g., absentee, mail-in) into one category, and we consider only respondents who reported an intention to vote of at least 51 percent (i.e., the nearly certain and uncertain voters referred to in the last section). We limited the sample in this way to focus on those respondents who are more likely to actually vote in November (we call these respondents likely voters). However, the trends in the reduced sample (with a total of 1,695 people) are very similar to those in the full sample.

In August 2020, we found that 47 percent of likely voters planned to vote by mail (up from 42 percent in May 2020) and 52 percent of likely voters said that they would vote in person (down from 56 percent in May 2020). Although we see some changes in intended voting methods between May and August, the change is more dramatic when we compare 2020 intentions with those recorded from past elections. When asked about how they voted in the most recent past election, 69 percent of these same voters reported that they had voted in person.

One key observation is that intended methods of voting vary considerably depending on political affiliation. Republicans were more likely than Democrats to say that they would vote in person, and this difference appears more significant for the November 2020 election than in the past. For Democrats who said that they were likely to vote, less than one-half who said that they voted in person in their last reported election expect to do so in November 2020.

Figure 3. Changes in Intended Voting Method Among Likely Voters, May–August 2020

May August
Other 2% 1%
Vote by mail 42% 47%
In-person vote 56% 52%

Changes at the Individual Level

We can also look at changes at the individual level by comparing whether likely voters changed their intended voting method between the May and August iterations of the survey. The results are shown in Figure 4, and the key findings are as follows:

  • First, most likely voters did not change their minds about how they would vote in November. We observe that most respondents who reported in May that they were likely to vote and would vote in person reported the same in August (81 percent). The same is true for the majority of those who reported in May that they would vote by mail (83 percent).[5]
  • There is some switching between in-person and mail-in voting intentions across the two periods. About 19 percent of those likely voters who planned to vote in person in May reported in August that they would instead vote by mail. Seventeen percent of those likely voters who said in May that they would vote by mail reported in August that they would vote in person. These two groups are roughly equal in size.
  • Finally, for the small group of likely voters who replied "other" in May to the question about voting method, we find that 72 percent reported in August that they would vote by mail, and 25 percent reported that they would vote in person. This is a large increase, though the total number of people in this group remains small. Notably, almost all likely voters who reported "other" in May shifted to different methods in August, and very few respondents shifted into the "other" category from May to August. The large number of likely voters who shifted out of the "other" category might suggest that voters who answered "other" in May were expressing uncertainty about how they would vote and that they had made a decision by August.

On net, then, we observe a shift toward more mail-in voting among likely voters, and this increase comes both from those who planned to vote in person and those who reported "other" as their voting method. Still, the number of planned in-person voters remains higher than the number of planned mail-in voters, and some respondents shifted into the in-person category since May despite any possible health risks.

Figure 4. Changes in Intended Voting Method for Likely Voters at the Individual Level, May–August 2020

May August
In-Person Vote 56% 52%
By Mail 42% 47%
Other 2% 1%

Who Changed Their Intended Voting Method Since May?

Finally, we consider the types of people who switched their intended methods of voting between May and August. Table 2 identifies characteristics of those most likely to have shifted their intended voting method between May and August.[6]

Across all factors considered, political affiliation and perceptions of safety were the strongest predictors of the likelihood of shifting to mail-in voting. Race or ethnicity is also important. Here, we observe several key trends:

  • Political affiliation: Democrats were more likely to switch between May and August from an intention to vote in person to voting by mail; Republicans were more likely to shift in the opposite direction, from mail-in to in-person voting.
  • Race and ethnicity: Hispanic voters were more likely than non-Hispanic voters to shift from in-person to mail-in voting; Black or African American voters were more likely than White voters to report a shift toward in-person voting.
  • Perceptions of safety: We observed that likely voters who said that they did not expect to be physically safe at the polls in November 2020 but did expect that their vote would be counted were more likely than other voters to switch from in-person to mail-in voting.
  • Age: Older voters were more likely to shift to mail-in voting than younger voters. This might reflect their greater health risks should they contract COVID-19.

Table 2. Characteristics Associated with Changes in Intended Voting Method

Switch to Mail-In

  • Older voters (compared with younger voters)
  • Hispanic voters (compared with non-Hispanic voters)
  • voters who have safety concerns but expect their vote to be counted
  • Democrats (compared with Republicans)

Switch to In-Person

  • Black or African American voters (compared with White voters)
  • Republicans (compared with Democrats)

Concluding Thoughts

Changes in intention to vote and intended voting method have been modest since May 2020 but are notable nonetheless. When considering intention to vote, we found a drop in intention for a group of respondents who had been uncertain voters in May and in August expressed that they were nearly certain not to vote. When considering intended voting method among those who reported they were likely to vote, we observed a continued shift toward mail-in voting. Individual characteristics most commonly associated with changes in intention to vote and changes in intended voting method appear to be perceptions of safety, political affiliation, and race or ethnicity.

Perceptions of safety are clearly related to intention to vote. Those with low perceptions of safety across our May and August surveys are among the least likely to vote and showed a decline in intention between the two surveys. Those who reported that they expected voting in person to be physically safe in one of the two periods, in contrast, increased their intention to vote. However, those with safety concerns were also more likely to shift to mail-in voting, perhaps as a mitigation of those concerns. Political affiliation is also associated with a likelihood of shifting to mail-in voting, with self-identified Democrats being more likely to shift. Finally, we find that race or ethnicity matters in selection of a voting method. Hispanic respondents were increasingly likely to shift toward mail-in voting; Black or African American respondents were more likely to shift toward in-person voting. Given the higher concerns about accurate vote counts among these voters, this trend is important (Bowler et al., 2015; Kavanagh, Gibson, and Hodgson, 2020a).

Past work has shown that voting intentions continue to shift right up to Election Day for some voters, especially those who are uncertain about their intentions (Achen and Blais, 2015; Glaser, 1958). The ongoing pandemic and the constant changes in state election procedures for November 2020 could increase the amount of churn this year. We will examine this in the next report in this series, which will consider data that we collected in October.

References

  • Achen, Christopher H., and André Blais, "Intention to Vote, Reported Vote and Validated Vote," in Johan A. Elkink and David M. Farrell, eds., The Act of Voting: Identities, Institutions and Locale, London: Routledge, 2015, pp. 195–209.
  • Bowler, Shaun, Thomas Brunell, Todd Donovan, and Paul Gronke, "Election Administration and Perceptions of Fair Elections," Electoral Studies, Vol. 38, 2015, pp. 1–9.
  • Centers for Disease Control and Prevention, "CDC COVID Data Tracker," webpage, undated. As of October 4, 2020: https://covid.cdc.gov/covid-data-tracker/#cases_casesper100k
  • Glaser, William A., "Intention and Voting Turnout," American Political Science Review, Vol. 52, No. 4, 1958, pp. 1030–1040.
  • Kavanagh, Jennifer, C. Ben Gibson, and Quentin E. Hodgson, Attitudes on Voting in 2020: Preparing for Elections During a Pandemic, Santa Monica, Calif.: RAND Corporation, RR-A112-9, 2020a. As of October 20, 2020: https://www.rand.org/pubs/research_reports/RRA112-9.html
  • Kavanagh, Jennifer, C. Ben Gibson, and Quentin E. Hodgson, Voter Attitudes Toward the 2020 Election: August 2020 Update, Santa Monica, Calif.: RAND Corporation, RR-A112-16, 2020b. As of October 30, 2020: https://www.rand.org/pubs/research_reports/RRA112-16.html

Notes

  • [1] Because voting intentions will continue to shift as the election gets closer, we expect that there have been changes since these data were collected. The next report in this series will consider data that we collected in October. We will also conduct a post-election survey.
  • [2] The question read, "What is the percent chance that you will vote in the 2020 Presidential election? The percent chance can be thought of as the number of chances out of 100. You can use any number between 0 and 100. For example, numbers like 2 and 5 percent may be 'almost no chance', 20 percent or so may mean 'not much chance', a 45 or 55 percent chance may be a 'pretty even chance,' 80 percent or so may mean a 'very good chance,' and 95 or 98 percent chance may be 'almost certain.'" For descriptive statistics, see the technical appendix to this report.
  • [3] The Midwest region encompasses Illinois, Indiana, Iowa, Kansas, Michigan, Minnesota, Missouri, Nebraska, North Dakota, Ohio, South Dakota, and Wisconsin. Rates of COVID-19 in these states remained moderate or low through the survey period, though they have surged since the beginning of September (Centers for Disease Control and Prevention, undated).
  • [4] The exact question asked was, "Assuming you do vote, how do you expect to cast your ballot?" Possible responses were: In person at a polling place; Will request and return an absentee ballot because I cannot vote in person on Election Day; Vote by mail (returned ballot by postal service); Vote by mail (returned ballot to polling place or to local registrar's office); and Other.
  • [5] We find that this change is statistically significant at the 0.05 level using a chi-square test.
  • [6] This analysis uses only respondents with a likelihood of voting above 50 percent.

In this report, we consider the voting intentions of 1,841 survey respondents for the November 2020 U.S. election. Specifically, we explore how the likelihood of voting and the intended methods of voting among respondents changed between May and August 2020. We consider how intentions changed and which populations are more or less likely to have changed their minds about whether and how they will vote in November. We also explore how perceptions of safety and integrity are related to intention to vote. The data used in this report come from a RAND Corporation American Life Panel survey conducted in August 2020.

We appreciate the support of Michael D. Rich, Terrence Kelly, Henry H. Willis, Jordan R. Fischbach, Matthew D. Baird, Erica Robles, and Steph Bingley. We are also grateful for feedback from our reviewers, Brian Michael Jenkins and Michael S. Pollard. Their contributions significantly improved the report. We thank Arwen Bicknell for her expert editorial assistance on this report. We also appreciate the efforts of Beth Bernstein and Jeremy Rawitch in the publication process.

This report is part of RAND's Countering Truth Decay initiative, which is focused on restoring the role of facts, data, and analysis in U.S. political and civil discourse and the policymaking process. The original 2018 report, Truth Decay: An Initial Exploration of the Diminishing Role of Facts and Analysis in American Public Life, by Jennifer Kavanagh and Michael D. Rich, laid out a research agenda for studying and developing solutions to the Truth Decay challenge. Truth Decay worsens when individuals lose trust in institutions that could serve as sources of factual information. Legitimate and safe elections can be a first step toward building and maintaining a government that people trust. This report considers public perceptions of the safety and integrity of elections in pandemic conditions, and their views of local preparedness for the elections.

Funding for RAND's Countering Truth Decay initiative is provided by gifts from RAND supporters and income from operations. RAND would like to recognize the Joel and Joanne Mogy Truth Decay Fellowship, established by the Mogys in 2020 to support research on Truth Decay, civics, and democracy. The authors drew from the Mogys' generous gift to fund this project.

This report is part of the RAND Corporation research report series. RAND reports present research findings and objective analysis that address the challenges facing the public and private sectors. All RAND reports undergo rigorous peer review to ensure high standards for research quality and objectivity.

The RAND Corporation is a nonprofit institution that helps improve policy and decisionmaking through research and analysis. RAND's publications do not necessarily reflect the opinions of its research clients and sponsors.