Our new paper uses administrative data from WA to evaluate the popular claim that dead people’s ballots are fraudulently cast at high rates in vote-by-mail elections. We find that this type of fraud is extremely rare (thread)
https://stanforddpl.org/publication/wu_et_al_2020_dead_voting/wu_et_al_2020_dead_voting.pdf
As states expand vote-by-mail during COVID-19, many people have raised the concern that mail-in ballots are sent to dead people, stolen, and counted in elections. For example, Donald Trump Jr. has raised this concern about Michigan: https://twitter.com/DonaldJTrumpJr/status/1294734395736236034?s=20
The president, along with the Heritage Foundation, has also made this claim. Or, take a look at this tongue-in-cheek tweet from Mike Huckabee along the same lines. https://twitter.com/GovMikeHuckabee/status/1320104112420212739?s=20
Evaluating claims like these is important because the legitimacy of our democratically elected government relies in part on people believing in our election outcomes.
The claim that dead people’s ballots are counted in elections is directly testable. Who votes and who dies are both matters of public record. We collect this data for the state of WA, one of the most prominent universal vote-by-mail states, and a state with great public data.
We match death records to records of whose ballots were counted, using individuals’ full names, county of residence, gender, and age. We then manually collect dates of birth from online obituaries to separate out duplicates from potential cases of dead people voting.
From 2011 through 2018, we estimate that there are 14 *potential* cases of dead people’s ballots being counted as votes (0.0003% of voters). Even these cases may be clerical errors or cases of people sharing the same name and birth date. This form of fraud is really, really rare.
We also carry out an automated technique to check our work, weaken standards for matching names, and make sure we’re not missing other obvious cases of fraud. We find no plausible evidence that this fraud is anything other than extremely rare.
It’s important to understand *why* this fraud is rare. It’s really hard to carry out at scale. It requires knowing when and where to find ballots, forging signatures, and avoiding the state’s built-in monitoring of public death records. And it’s a serious felony if you’re caught.
Given how hard it is to carry out at scale, and how large the penalties are, it’s not surprising that very, very few people attempt this kind of voter fraud.
Our paper is not meant to endorse or reject vote-by-mail as a way to run our elections. There are many other considerations in evaluating it. Our purpose is to evaluate the specific claim that it is vulnerable to allowing dead people to vote. At least in the state of WA, it’s not
While our results are focused on WA, they are likely to extend to any state that takes similar precautions. In any such state, this form of fraud is likely to be vanishingly rare.
This paper is the result of a huge team effort with many awesome co-authors. @imjenwu @TobiasNowacki @danmthomp @jesselyoder as well as Chenoa Yorgason, Hanna Folsz, Sandy Handan-Nader, and Andy Myers.
I’m incredibly proud of the work this group has been doing to use large-scale data and cutting-edge social science methods to answer important questions about our elections. The tradeoff between high-quality data and size of question is overrated.
This is work in progress and we're looking forward to comments.
Perhaps of interest to @KimWyman12 @secstatewa
Here is a nice, accessible write-up of the study: https://twitter.com/SIEPR/status/1321824535411261440?s=20
You can follow @andrewbhall.
Tip: mention @twtextapp on a Twitter thread with the keyword “unroll” to get a link to it.

Latest Threads Unrolled: