With concerns rising over a potentially contested presidential election next week, a new study by Stanford researchers suggests that worries over one type of election-day fraud are overblown: ballots cast in the name of dead people.
In a working paper
posted this week, a team of Stanford political scientists analyze roughly 4.5 million voter records from the state of Washington for evidence of ballot fraud involving deceased individuals. They find 14 cases where a ballot may have been stolen and submitted on behalf of someone who had died, and even these cases may not have been fraud-related.
“We’re talking about 0.0003 percent of all voters over an 8-year period,” says Andrew Hall
, a senior fellow at the Stanford Institute for Economic Policy Research (SIEPR) who conducted the study.
According to Hall, who is a professor of political science in the school of Humanities and Sciences, the research marks one of the few studies on deceased voter fraud. He says that the findings, although specific to Washington, cast doubt on claims that votes submitted on behalf of dead people are a widespread problem. The insight also comes as the COVID-19 pandemic has led states to ramp up vote-by-mail programs in order to give people the opportunity to vote from the safety of their homes if they wish.
President Trump, for one, has called into question the legitimacy of mail-in ballots — going so far as to say that the potential for fraud could be “catastrophic.” And in August, Donald Trump Jr., the president’s son, promoted false claims that 8 percent of all votes in this year’s Michigan primary were cast using dead people’s mail-in-ballots.
Hall says that the findings, though specific to one state, have potentially broader implications.
“As states expand vote-by-mail during the pandemic, many people have raised that concern that mail-in ballots are sent to dead people, stolen, and counted in elections,” he says. “We find that this type of fraud is likely to be extremely rare in states that take basic precautions.”
Hall conducted the study along with seven current and former Stanford political scientists: PhD students Jennifer Wu, Chenoa Yorgason, and Hanna Folsz; PhD candidates Cassandra Handan-Nader and Tobias Nowacki; and, Andrew Myers, a SIEPR predoctoral research fellow. Daniel Thompson, a Stanford PhD who is now an assistant professor at UCLA, was vital to the research as well.
A detailed portrait of voters
Voter fraud, of course, has long been a source of controversy in American elections. Bribery, duplicate ballots, false registrations and other methods have all been mentioned over time as potential threats to the legitimacy of democratically-elected government. Research, however, has uncovered very little evidence that these issues are common in modern American elections, in part because states have taken effective steps to thwart them.
While election chicanery can be hard to prove, instances of impersonating a dead person are easier to catch, Hall says.
“Who votes and who dies are both matters of public record,” he says. “This makes this particular claim directly testable.” In fact, like many states, the state of Washington itself checks public death records and removes deceased voters from the rolls, taking special care to prevent dead people’s ballots from being counted. While the state of Washington already has the data to conclude that this form of fraud is rare, Hall points out that “it is important for independent academic researchers to evaluate the claims of state authorities about the security of our elections.”
Academic research on the issue has been scant, but at least one in-depth study, of the 2006 general election in Georgia, found “essentially zero” cases of it.
The state of Washington turns out to be an ideal laboratory for investigating deceased voter fraud. Since 2011, all of its counties have used universal vote-by-mail, , a program in which every registered voter is mailed a ballot. This is distinct from absentee voting, in which voters can opt into voting by mail.
Washington stands out, too, for the quality of its data on voters, which include names and dates of birth, and voter histories for a long period of time. The state also makes public its official death records. The result, the study authors write, is a “nearly-perfect” snapshot of state voters over time.
In the end, Hall and his team gathered information on every distinct voter from 2011 to 2018. The researchers then linked voter records to death records on the basis of full name, gender, and county of residence, collecting dates of birth from online obituaries to address possible false positives in which multiple people share the same name within the same county.
They found that, of the 4.5 million voters studied, suspicious ballots were cast for 14 who had died long before election day and whose information in voter rolls and death records were an exact match. When researchers relaxed requirements that middle names match, they found there were only 43 additional cases that were questionable, though these cases are more likely to be false positives.
“We can’t even say for sure that these 14 instances are fraud,” Halls says. The problem could be clerical errors or two individuals who have the exact same name and birth date. “What we can say for sure is that, at least in Washington, this form of fraud is extraordinarily rare.”
Hall cautions that the findings are specific to Washington and don’t say anything about other types of potential voter fraud. “And we’re not saying anything about whether vote-by-mail is ‘good’ or ‘bad’ — just that this one particular claim about it seems false,” he says.
For criminal minds, daunting challenges
Hall also says that the results can’t be immediately generalized to other states. One reason why is that Washington goes to great lengths to prevent people from mailing in fake ballots, including through the use of unique barcodes for validating eligibility and rigorous measures to confirm that a recently deceased resident is removed from its voter list, and has developed these measures over many years of practice. Nevertheless, basic precautions that all states take likely make this form of fraud unlikely in any circumstance.
“Given how hard this fraud is to carry out at scale, and how large the penalties are, it’s not surprising that very, very few people attempt it,” Hall says. “Although our results can’t be immediately generalized, this type of fraud is likely to be vanishingly rare for states that take similar precautions.”