On election night the data point I spent fruitless minutes looking for was some sort of estimate of total voter turnout. There are obvious reasons the number is difficult to identify at that point (or even now), including mail ballots that may or may not arrive or be quickly tabulated. But the more we learn about this midterm, the more likely it becomes that we set records for low turnout. Here’s how Robert Siegel succinctly summed it up on NPR yesterday:

It’s well known that the turnout in midterm elections is a lot lower than in presidential years. In the past two presidential elections, around 130 million Americans cast votes. In the midterms four years ago, 91 million voters took part. That number represented 42 percent of the American population of voting age. This year – well, the numbers aren’t official yet, and political scientist Michael McDonald of the University of Florida who tracks such things told me today that the total is likely to rise a little as late arriving mail-in ballots and other missing votes are counted.

But as of today, according to numbers from the Associated Press, a bit over 83 million people voted. As a share of the voting-eligible population, that is 36.6 percent. It was a lot higher than that in Colorado and New Hampshire where there were very hard-fought Senate races.

But if after all the votes are counted, if the national turnout rate doesn’t reach 38.1 percent, it would be the lowest turnout since the midterm elections of 1942. And as Michael McDonald points out, that was in the middle of the Second World War.

Think about that for a minute. In an election that everyone obsessively said was “all about turnout,” in which there were a goodly number of competitive statewide races, and where Democrats staked all their hopes on a big shiny expensive state-of-the-art GOTV project, and Republicans reportedly undertook their best turnout effort yet, it’s still likely a record was set for low turnout. What would have happened if nobody had tried to turn out the vote? How much worse could it have gotten?

It’s certain, of course, that turnout wasn’t down uniformly, even in a given state, but that turnout was down more for Democratic-prone voting groups. That’s what “midterm falloff” is all about–the problem the DSCC’s Bannock Street Project was confidently aimed at mitigating, and the problem Republicans were deliberately trying to exacerbate with various efforts to make voting more difficult, especially for young and minority voters. What I’m now wondering more and more is whether midterm falloff doesn’t just have a differential effect between but also within demographic categories, meaning that young and minority folk more likely to vote Republican are also more likely to vote (this would explain the improved GOP performance in categories of voters they transparently are not appealing to).

In the end, it’s not easy to figure out what exactly gets marginal voters over the line–or doesn’t. Maybe, as I said in my TPMCafe column late Tuesday night, in an allusion to HRC, “it take[s] a presidential candidate to build this particular village.” But it’s a puzzle Democrats need to solve if they ever want to win a midterm election, unless they have a plan for significantly boosting their appeal to old white folks.

Ed Kilgore

Ed Kilgore is a political columnist for New York and managing editor at the Democratic Strategist website. He was a contributing writer at the Washington Monthly from January 2012 until November 2015, and was the principal contributor to the Political Animal blog.