President-elect Donald Trump at a rally at the Wisconsin State Fair Exposition Center in West Allis on Dec. 13. (Evan Vucci/Associated Press)

Here are things we know about elections in the United States generally and the election in 2016 in particular.

  1. We know that in-person voter fraud is an exceedingly rare occurrence.
  2. We know that laws meant to combat those rare instances of illegal voting often have the effect (intentionally or not) of reducing turnout more heavily from groups that tend to vote Democratic.
  3. We know that Donald Trump is president because, while he lost the popular vote, he won three states by about 78,000 total votes: Michigan, Pennsylvania and Wisconsin.
  4. We know that a margin that narrow — 0.56 percent of the votes cast in those states — could possibly have been turned around by any number of things.

At Mother Jones magazine, Ari Berman makes the case that one thing in particular might have helped turn around No. 4 on our list. Specifically, No. 2.

“Clinton’s stunning loss in Wisconsin was blamed on her failure to campaign in the state, and the depressed turnout was attributed to a lack of enthusiasm for either candidate,” he writes. “The impact of Wisconsin’s voter ID law received almost no attention. When it did, it was often dismissive.”

Berman points to a number of anecdotal examples of people prevented from voting thanks to the state’s voter ID law and a few studies. One that he cites was covered by The Washington Post, and found that perhaps a quarter of black Wisconsinites in the state’s two largest counties may have been turned away from voting or discouraged from voting by lack of ID. Black voters are often disproportionately affected by such laws — and black voters generally vote heavily for Democratic candidates.

Berman’s colleague Kevin Drum offered a simpler demonstration of the possible effects of the new law. He picked up data from the Brookings Institution and made a chart showing the drop in black voter turnout in these same three states and nationally.

That chart, however, has some significant flaws.

It’s based on data from the Census Bureau, which, as Michael McDonald of the United States Elections Project pointed out on Twitter, has significant flaws.

“The Census Bureau counts all persons with missing data as ‘did not vote’ in their survey,” he wrote. “This is a terrible practice. No other pollster treats their missing data this way.” That matters in Wisconsin in particular, he pointed out, because the percentage of black respondents in that state who had missing data in 2016 was 36 percent — four times higher than in 2012. “When the Census Bureau counts missing data as ‘did not vote,’ Wis. African-American appear to alarmingly declined [in turnout] when, in fact, fewer answered the survey,” he wrote.

To what extent did black voter turnout drop in Wisconsin?

“When I correct for non-response and over-report bias, I estimate Wisconsin African-American turnout declined 7.5 points,” McDonald said, “while nationally the decline was the same 7.5 points.”

We can look at this another way. The political data company L2 Political pulled data for The Post on individual black voters in the three key states. Wisconsin has fewer black voters than Michigan or Pennsylvania. But on the metric at issue, that decline from 2012, the numbers in Wisconsin aren’t really remarkable. While a much higher percentage of black voters registered in Wisconsin didn’t vote in either 2012 or 2016, according to L2’s data, the drop-off from 2012 to 2016 was about the same in each state. In Michigan, 10.8 percent of black voters registered in the state voted in 2012 but not 2016. In Pennsylvania, it was 10.6 percent. In Wisconsin, 10.7 percent.

(We asked L2 why the number of voters who’d skipped both elections was so high in Wisconsin, but it’s not clear what the reason is. Courts limited the voter ID law in the state before the 2012 election, so it’s likely not that. For our immediate purposes — comparing 2012 to 2016 — it doesn’t really matter.)

We know that neighborhoods with high densities of black residents shifted Republican in 2016 relative to 2012. As Brookings notes in its analysis, this is a function both of increased turnout from white voters and decreased turnout from black voters. It’s something that happened pretty universally across the country, and not just in Wisconsin or other states with strict voter ID laws.

Why this drop? Why the decline that McDonald notes of 7.5 points in turnout?

In 2008 and 2012, the black voter turnout rate surged past that of white voters, according to Census Bureau data adjusted by McDonald. In 2016, it receded back to where it was in 2004, slightly below turnout for white voters.

One probable reason for this was the change at the top of the ticket. Trump likes to brag that he got more support from black voters than any recent Republican candidate — but he’s the only Republican candidate in more than a decade who wasn’t running against the first black president.

Berman’s likely not wrong that voter ID laws nationally and in Wisconsin likely tamped down support that would have gone to Hillary Clinton. What we don’t really know, though, is the extent to which those laws may have affected the outcome of the race. Even that study showing that a quarter of black voters in two counties might have been discouraged from casting a ballot wouldn’t have been enough to flip the state to Clinton.

Voter fraud is not a widespread problem, particularly at the national level, and there’s a valid debate over whether voter ID laws are mostly a vehicle to suppress Democratic votes. But it’s very hard to say, given the data at hand, that Clinton would have won Wisconsin without such a law, much less the presidency.