[Image via jazelc]

MIT Professor Charles Stewart has updated a 2010 paper he wrote about the problem of “lost” vote-by-mail votes – and it has some interesting observations about the growing use of the mail ballots, especially in 2020 during the coronavirus pandemic. He shared his thoughts on Election Updates:

A “lost vote” occurs when a voter does all that is asked of her, and yet her vote is uncounted in the final tally. Estimating the magnitude of lost votes in American presidential elections has followed the work of the Caltech/MIT Voting Technology Project (VTP), which initially estimated the magnitude of lost votes in the 2000 presidential election—due to failures of voter registration, polling-place management, and voting technologies—to be between 4 and 6 million out of 107 million cast that year.

Because of data and conceptual limitations, lost vote estimates have tended to focus on in-person voting, ignoring lost votes due to mail ballots. In a paper I recently finished, I revisited an article I wrote in 2010 that attempted to fill the hole in our understanding of lost votes, by considering mail votes in the 2008 election. That paper estimated that as many as 22% of mail ballots were “lost”—as defined by the VTP—in that election. Despite the fact that I opined in the article that this was clearly an over-estimate, this 22% statistic has been repeated without the caveats that appear in the article…

Charles has revisited that earlier paper because VBM has changed, both in policy and in practice, and because improved data collection has provide a richer base of evidence for his analysis:

Over the past decade, it has been suggested that I should reconsider that earlier article, for two reasons. First, mail balloting has become much more complicated, with states adopting a variety of approaches to mail balloting. Each approach, from excuse-required absentee balloting to vote-by-mail, creates unique risks for and protections against lost votes. I owe it to the evolving policy to align my thinking with the new reality. Second, the data have become better than they were in 2010. A reconsideration should reflect that new data.

The new paper draft is posted here in advance of publication this fall, but Charles shares the major takeaways:

  1. “Lost vote” is a term of art that draws our attention to the gap between a voter’s intention to vote—in this case by mail—and the completion of the intention. In no way does it refer to ballots that have been physically lost, in the literal sense that there are millions of ballots sitting in a trash heap somewhere.

  2. The number of lost mail votes in 2016 was more like 4% of mail ballots, not 22%. The principal source was rejected ballots, which has garnered plenty of attention in the 2020 primaries, for instance, because they arrived late. The next-largest cause was a heightened residual vote rate, that is, over- and undervotes. The smallest contributor, which is also the most difficult to estimate, is problems with the postal service and the non-delivery of requested ballots.

  3. The states that have the most expansive vote-by-mail laws have the lowest lost-vote rates. This is because no requests for absentee ballots are lost in the mail in these states and because vote-by-mail states reject a much smaller fraction of returned mail ballots than states that require voters to explicitly request them.

  4. Conversely, states that require voters to request absentee ballots have higher lost-vote rates, mostly because these states are more likely to reject them when received.

  5. The fact that 22% of the ballots that were mailed to voters in the vote-by-mail states in 2016 were not returned for counting is due almost entirely to voter abstention, nothing more.

  6. The biggest empirical puzzle remains why 7% of voters in excuse-required states and 14% of voters in no-excuse states who requested a mail ballot never returned one. If 99.5% of the mail gets delivered within the window of postal service standards, this can’t be because ballots are getting lost by the USPS. But, these percentages seem too high to be explained simply by ballot requesters getting cold feet.

  7. The states that will expand the use of mail ballots the most in 2020 will be among those with the greatest ballot-rejection rates in 2016. New York’s 2016 rejection rate was over 10%, which is entirely consistent with reports currently coming out of the state from the primary.

  8. One sign of hope is that the heightened scrutiny of mail ballot rejections, including some court-case settlements, may keep rejection rates in check in November. Georgia is a good example. In 2016, its mail-ballot rejection rate was 6.9%. In the recent primary, it was closer to 1%.

  9. Voting by mail is risky.  So is voting in person, especially in the age of COVID-19.  The risks are of a different nature.  It is the responsibility of election officials to try and minimize voting risks as much as they can.  It is the responsibility of voters to weigh the risks of voting, and to vote using the mode they feel the most comfortable with.

As focus on vote-by-mail intensifies in 2020, so will concerns about the possibility of “lost votes.” Thanks as always to Charles for this analysis, which provides an important analytical backdrop to the debates and preparations surrounding the increase in VBM in 2020. Stay tuned …