Ballot Privacy, Voter Coercion and the 2020 US General Election
Download data and study materials from OSF
Kings College London
University of Rochester
University of Rochester
Sample size: 3365
Field period: 06/16/2021-10/19/2021
Nearly half (46%) of respondents to a Pew Survey reported voting by mail in the 2020 general election, often citing concerns about exposure to COVID-19 as a reason for their decision. While voting remotely can certainly be a way to mitigate the risk of contracting disease, it may expose voters to other potentially harmful circumstances. Specifically, in the absence of truly private voting spaces and election officials who can enforce voters’ rights, voters who opt to vote by mail may find themselves vulnerable to threats, intimidation, and other coercion designed to affect both their decisions to turn out and their vote choice. Studies outside of the United States have found that intimidation at the hands of family members, friends, employers and others is a meaningful risk for voters – with women, dependent family members, and voters with lower education most vulnerable to these forms of coercion. In this study, we assess the extent to which voters in the 2020 U.S. Presidential election faced such intimidation using a novel list experiment. We also compare the extent to which people who did not vote, people who voted by mail, and people who voted in person perceived themselves to be at risk. Finally, we provide important descriptive information about the types of environments in which people voted regardless of whether they chose to vote in-person or remotely, and whether they filled out their ballots while in discussion with other people or whether voting was truly a solitary exercise for voters in 2020.
Hypotheses based on demographic and resource factors: Voters with demographic characteristics likely to make them more dependent on others in their family are more likely to be vulnerable to intimidation within the family.
H1a: Women are more likely than men to experience familial electoral intimidation
H1b: Young adults and retirement-age adults are more likely to experience familial electoral intimidation, such that age will be related to this variable in a curvilinear way
H1c: Those with lower education levels are more likely to experience familial electoral intimidation
H1d: Those without independent income are more likely to experience familial electoral intimidation
Hypotheses based on cultural factors: Relatively traditional cultures (e.g., certain religious communities) tend to be more hierarchical, and hierarchical power relations within the family are conducive to coercive control and intimidation.
H2a: More religious people are more likely to experience familial electoral intimidation.
H2b: First generation immigrants from regions were clientelism is common are more likely to experience electoral intimidation.
Hypothesis based on institutional factors: Votes completed outside official polling places are more vulnerable to intimidation, as there are no formal safeguards to protect ballot secrecy when voting takes place in the private sphere. States that currently have laws that suppress turnout, or have historically practiced turnout suppression are more likely to harbor individual and groups that will seek to undertake intimidation at public polling places.
H3: People who vote by mail are more likely to experience familial electoral intimidation than those who vote in person.
H4: People who live in states with voter ID laws are more likely to experience or perceive electoral intimidation at public polling places.
H5: People who live in states That have ever been on the Department of Justice’s pre-clearance list are more likely to experience electoral intimidation at public polling places.
Hypotheses based on political factors: In households that are homogeneous in partisan terms, there is no need for family members to seek to alter the votes of others, whereas in heterogeneous households, this is a relevant possibility.
H6: Those living with people who have partisan preferences that differ from their own are more likely to experience familial electoral intimidation.
The intervention was a list experiment that appeared in different formats to nonvoters, people who voted in person, and people who voted by mail. For nonvoters, in addition to providing a list of possible reasons why people might not vote, the treatment group saw the option "Someone threatened or intimidated me to prevent me from voting (this might include, for example: threatening to physically hurt you, threatening to fire you from your job or cut back your earnings, or threatening to impose other serious hardship)." People who voted (either in person or by mail), in addition to a list of reasons they might have voted for their chosen 2020 presidential candidate, the option "Someone threatened or intimidated me to make me vote for a specific candidate (this might include, for example: threatening to physically hurt you, threatening to fire you from your job or cut back your earnings, or threatening to impose other serious hardship)." We also included direct questions about coercion in the study.
The key outcome variable in our list experiment was the number of statements from the available choices that respondents said reflected their experiences. These are integers from 0 to 6 or 7 depending on the mode of voting. We also looked at binary responses ("yes/no") to the direct versions of the questions "If Did someone threaten or intimidate you to prevent you from voting in the 2020 presidential election/Did someone threaten or intimidate you to make you vote for a specific candidate? "
Summary of Results
Analysis is still ongoing, but the results of the list experiment are null across the board. Test statistics for differences in means for a t-test comparing the average numbers of true statements selected among treated and control units among in-person voters is actually t = -0.56, p = 0.57; for mail voters t = -0.61 with p = 0.54 and for nonvoters it is t = -0.49 with p = 0.624. None of these are statistically distinguishable from zero, and if anything the directions of the point estimates suggest that people presented with an opportunity to indicate someone threatened them to influence or suppress their vote were actually less likely to say that happened. That's a good thing for democracy! Our analysis into whether this holds for specific subgroups of respondents is ongoing. We did find in the descriptive data, however, that people who voted by mail were much more likely to report having a real-time discussion of their ballot with someone else.