Voters wait in line to cast ballots during early voting Oct. 20 in Sandy Springs, Ga. (John Bazemore/Associated Press)

Cook Political Report’s Dave Wasserman made an interesting observation on Twitter on Tuesday evening.

He was responding directly to the results of a special election in Georgia’s 6th Congressional District. As expected, based on who’d already returned ballots, Democrat Jon Ossoff jumped out to an early lead when the state published the results from early voting. But as votes from Election Day itself were counted, Ossoff’s lead narrowed quickly. He went from having more than 70 percent of the vote to less than 50 percent as the night wore on — not enough to avoid a runoff race in June.

Wasserman was also referring, though, to any number of other recent, similar examples. Perhaps the most infamous recent example was in Florida in November. The campaign of Democrat Hillary Clinton came into Election Day confident that it would carry the Sunshine State, given the dominance her campaign had shown in early voting. But, again, the Election Day tally worked against her, as detailed in a new book assessing how her campaign stumbled. A Florida win meant the presidency. An Election Day surge for Republican Donald Trump ended that dream.

Those two examples aside, it’s worth considering Wasserman’s assessment more broadly. Are Georgia and Florida simply outliers? Is it the case more broadly that early voting tends to favor the Democrat in a race more heavily?

Looking at last year, the answer is clearly “yes.”

Not every state publishes election results broken out by the type of vote cast. At least six states do, though, and in five of those six, Clinton fared better in early voting than she did in votes cast on Election Day.

The sole exception was Louisiana, but even there it was close. Clinton won 38.8 percent of the vote in early balloting, doing slightly better — 40.2 percent — on Nov. 8. Everywhere else, her margins were better beforehand. In Maryland, she did 15 points better in early voting. In Iowa, 17. In North Carolina, almost 10. In Georgia? Just over a percentage point.

It can be hard to evaluate early voting compared to Election Day voting without specific numbers from the state, but we can estimate it. If we look at the number of early ballots returned by members of each party — data collected by Michael McDonald’s U.S. Election Project — and compare it to the actual results in the state, a similar pattern emerges. Looking only at ballots cast by members of the Democratic or Republican parties and comparing those to the margins of the vote for Clinton and Trump, in nearly every case the percent of ballots returned by Democrats in early voting was larger than the actual Clinton vote on Election Day — suggesting heavier support for Trump in votes cast Nov. 8.

This is imprecise, of course. In West Virginia, for example, a lot of Democrats cast early ballots for Trump, which skews the results. And notice that this analysis would suggest that Clinton fared worse in early voting, while the actual state data shows that she didn’t. Still, the pattern matches Wasserman’s assumption. More Democrats voting early than the final results would indicate.

There are several implications to this. One is that we’ll see a lot more election nights like the one in November and like Tuesday, with Democrats jumping out to an early lead, and Republicans eventually eating away at it. Humans are weird creatures, and we tend to view election night returns like a football game, in which one candidate or the other is waging a comeback. Of course, that’s not the case; the actual football analogy would be if referees simply released the score of the game in random one-point increments after it had ended.

The more important implication is how it frames efforts to curtail early voting. If early voting is used more heavily by Democrats, it stands to reason that cutting early voting would reduce the number of Democrats who vote.

Which itself explains another manifestation of partisan polarization: opposition to early voting efforts by Republicans.

Election results for each state: Arizona, Georgia, Iowa, Louisiana, Maryland and North Carolina.