Alex Halderman, the highly respected computer science professor asking the Clinton campaign to seek recounts, posted an explanation early this morning. His explanation is persuasive as a general condemnation of electronic voting in any election, especially a close one. He shows how frighteningly easy it is to hack voting machines. But he does not offer specific statistical claims about the 2016 presidential race, other than noting the Russian email hacking. He is right that we should audit and verify as a matter of public policy. Why not? But there is still no specific anecdotal or statistical evidence suggesting an actual hack or vote-rigging in this election. He does not mention the “7%” statistical shift in electronic voting towards Trump, and that’s the factual claim that seems to be generating media interest. Nate Cohn at the NY Times confirms what I’d been posting yesterday: the general pro-GOP correlation with electronic voting is probably attributable to race and class by precinct, rather than hacking. White precincts are slightly more likely to have electronic voting. See link in comments, H/t Richard Kim.
Here is my main problem with focusing on electronic voting as the way the 2016 election was stolen: the key battleground states used different voting systems. Yes, Wisconsin and Pennsylvania use electronic voting, as do many other battleground states. But Michigan (and Minnesota, NH, and Iowa) use only optical scan paper ballots, not electronic/computerized voting. Michigan’s margin and demographics were very similar to Wisconsin’s and Pennsylvania’s, and the three states’ histories of presidential voting and 2016 polling were roughly similar. In fact, Michigan had been more blue than Pennsylvania and Wisconsin, so if anything, Michigan reflects an even bigger shift to the GOP in the Midwest, regardless of voting technology. Other states with paper ballots show a consistent pattern: the state-by-state polling was inaccurately pro-Clinton, underestimating the white turnout. Minnesota (a paper ballot state) had been a deep blue state for two decades, but was razor thin this year (about 1% for Clinton, a dramatic shift to the GOP). New Hampshire (paper) was closer on election day than the polling averages. This election seems not to be a story of a failure of voting technology, but a failure of polling technology.
One more point: for a recount to make a difference, Clinton would have to overcome substantial deficits in all three states (MI, WI, and PA — 70,000 in PA alone!). And to be fair, Trump would be entitled to his own recount of New Hampshire and Minnesota. The evidence of hacking would have to be overwhelming to make any difference, and in the end, the House of Representatives would resolve such a contested election… For Trump. Let’s have an audit to verify the security of our elections. Let’s fight for voting rights and fight to get rid of insecure electronic voting. But please don’t circulate conspiracy theories of how this election was stolen, until we see detailed, hard statistical evidence suggesting otherwise. We lost the election. Let’s not lose focus on this corrupt administration and its far-right-wing transition.