President Trump’s campaign travel this week included two unexpected stops. On Wednesday, he flew to Iowa for a rally in Des Moines. On Friday, he has two rallies scheduled, including one in Georgia.

These are states he won by nine points and five points respectively four years ago. If his campaign thinks they need to be shored up with less than three weeks to go until the election, well, that’s not a great sign.

But they do need to be shored up. In each of FiveThirtyEight’s averages of polls from those states, former vice president Joe Biden has a narrow, one-point lead. That’s the sort of margin that might allow Trump to wrest the states back into his win column, but if he’s having to fight over states he won by nine points four years ago, it suggests a bleak result in the various states that he won by the skin of his teeth.

In fact, as we reported last week, the state-level polling averages currently suggest that Trump will lose the election by a wide margin. Even if those polls are off by a decent-size margin, Trump still loses. If the polling average in each state holds as of writing and is as far from the mark as it was in 2016, Biden will flip five states (though not Georgia and Iowa) and win by 100 electoral votes.

Explore how poll shifts affect current polling averages below.

Poll shift?

Or uniform error of 0 points.

Your browser cannot display this graph.

Again, a lot of the states are still close and can easily swing back to Trump. But a general shift in the president’s favor probably necessitates an overall national shift — something that to this point simply hasn’t been a feature of the 2020 race.

Of the last 13 presidential contests (from 1968 to 2016), the candidate leading with 20 days to go won the popular vote on all but two occasions. One of those was 2000, when the final popular vote total swung to Democrat Al Gore — but not in enough places to earn him a victory in the electoral college. The other time was in 2012, when polls underestimated Barack Obama’s national support.

That analogy no doubt holds some appeal to Trump: an incumbent president who earned more support than expected in his first race and who trailed in his reelection bid. But with 20 days to go in 2012, Obama trailed by only 1.6 points, according to FiveThirtyEight’s average, not the 10.5 points by which Trump currently trails. There have been big shifts in the final weeks of a campaign, but the biggest gap between the polling average 20 days out and the final result was that 5.5-point swing in 2012. Give Trump 5.5 points nationally and he’s still trailing by only a little less than Jimmy Carter was at the same point in 1980.

The Carter analogy no doubt does not hold appeal for Trump.

There’s another complicating factor with which Trump has to contend. According to data compiled by the University of Florida’s Michael McDonald, more than 14.4 million ballots have already been cast this year. That’s more than 10 percent of the ballots which were cast in 2016. At a similar point in 2016, 16 days from the election, the number of ballots cast early totaled less than 5 percent of the 2012 total. (In 2008 and 2012, the number of ballots tallied by McDonald made up about a quarter of the prior election’s total.)

Why does that matter? Because those votes are done, over, set in stone. If something massive happens now, some of those voters might be able to invalidate the votes they already cast, but it’s not a simple proposition and not something that can be done in every state. So Trump’s campaign can be confident that there are already 14 million-odd votes which have been submitted at a point where he trails nationally by about 10 points. Most of those votes are likely from Democrats, given the dynamics of voting by mail this year, but those are nonetheless votes that have been banked.

Put this together and the picture for Trump is bleak. He trails by a wide margin nationally at a point when trailing by a wide margin generally reflects the actual result. He trails in key swing states and is close in states that he won easily four years ago. And millions of votes have already been submitted, making them largely immune to any last-minute shift in the contest.

As always, we reiterate that Trump can certainly win the race. But a presidential contest in which the Republican incumbent is trying to make sure he can win the agricultural state of Iowa is not a contest in which things are going the way he would wish.