Many ultralow-turnout special elections this year appear to have favored Democrats heavily. Take the first Democratic flip, in New York’s 9th Assembly district. Registered Democrats outnumbered registered Republicans on Election Day, 40 percent to 39 percent, according to data from L2, a nonpartisan vendor of voter files. But in the 2016 general election, Republicans had a 16-point registration advantage, 44 percent to 28 percent. There were even more extreme shifts this year.

This kind of huge difference (17 points) in the partisan composition of the electorate (as measured by registration) is not unprecedented in a low-turnout special election. Only around 10,000 people voted in the race for New York’s 9th; just 4,000 did in Oklahoma’s 37th. It is much more difficult or even impossible to pull off that kind of advantage in a high-turnout election.

That’s probably why the results in the higher-turnout Virginia and New Jersey general elections — in both the governors’ and state legislative contests — looked quite different. The results were far more correlated with recent presidential contests. There were no wild, outlying results, even though there were more individual legislative contests in Virginia and New Jersey than in all of this year’s special elections. This tight relationship between presidential vote choice and state legislative results is reminiscent of typical congressional elections.