Which is more secure? To vote in person, or by mail?

But voting by mail has some other practical risks that few have noted. It’s more complex than voting in person, administratively speaking. As a result, those who cast their ballots by mail might be less likely to have their votes counted than those who vote in person.

In an article forthcoming in the Harvard Data Science Review, I have worked to quantify how much riskier it is for someone to vote by mail than in person. Depending on the state in which a citizen is voting, the increased risk of having your vote lost — meaning, not counted in the election — ranges from 3.5 percent to 4.9 percent.

“Lost votes” is a term coined by the Caltech/MIT Voting Technology Project (VTP) in its 2001 report “Voting: What Is/What Could Be.” Here’s what it means. Suppose a voter wakes up on Election Day, fully intending to vote, and does everything required to do so. If the intention is thwarted, that is a lost vote. For instance, she might arrive at the polls at 6 p.m. and finds the line is so long that she leaves and can’t return to cast a ballot by 7 p.m. when the polls close.

If she intends to vote by mail, here’s what might go wrong. In states where a voter must apply for a mail ballot, the ballot application could get lost in the mail; the local election office could lose the application or deny it; the ballot might not make it back to the voter, for instance, getting lost in the mail; and the marked ballot might not make it from the voter back to the local election office. Even if the ballot arrives, it could be rejected because it arrived late or lacked a signature — the two most common reasons for rejection. Finally, the ballot could have an error that she could have caught had she voted in person....

Estimating these risks is difficult. It’s nearly impossible to find hard data on whether the ballot request makes it to the local election office. I approximated this by relying on USPS performance reports, which indicate that approximately 0.4 percent of first-class mail and 0.7 percent of marketing mail fail to reach its destination within three days of the delivery standards set by the Postal Service. I assumed that 99.9 percent of the time, when local election officials do receive a valid application, they respond by mailing out a ballot. If we apply the 0.4 percent failed delivery figure both to the voter’s application for and return of a mail ballot, and then add 0.1 percent chance of it getting lost in the office and the 0.7 percent failed delivery figure for the election office’s response, this suggests that 1.6 percent of mail-ballot applications fail to deliver a ballot to the voter.