The political pundit class is abuzz with claims that President Donald Trump’s reelection effort is getting crushed by a massive early voting turnout by supporters of Vice President Biden. But is that really the case?

In Florida, one of 19 states featuring both voter registration by party affiliation and reporting of early voting numbers, Democrats jumped out to a 490,000-voter lead on Oct. 21 based on the party affiliation of those voting early. But just two days later, Republicans surged to the polls, cutting that lead down to 387,000, a 21% reduction.

The fact is that early voting numbers are a usually a poor guide to predicting an election—and now this is especially the case in the wake of COVID-19 and how voters view the virus. This is because polling has shown a wide partisan fear gap regarding the virus, with Republicans generally looking to return to some semblance of normalcy while Democrats are far more likely to worry about public gatherings.

As a result, Democrats are turning towards mail-in ballots and voting early when, they believe, they can take advantage of less crowded polling places. Republicans, on the other hand, hearing that up to 1-in-20 mail-in ballots can fail to convert to a vote due to postal and processing failures, are therefore trooping to the polls in person—with many intended to cast a vote in Election Day.

The fact that early voting is a poor predictor of elections hasn’t stopped commentators from trying. But aside from the effect of COVID-19 on voters’ minds, there are two other big considerations when looking to early votes as a clue.

First, since only a small share of states report both the level of early voting and partisan voter affiliation, it can be hard to ascertain any meaningful trend. For instance, as of Oct. 24, some 13.1 million Democrats had already voted in 19 states, including the battleground states of Florida, North Carolina, Nevada, and Pennsylvania, while in those same states only 7.4 million Republicans had cast an early ballot in-person or by mail. Almost 6 million voters had no party affiliation. That would appear to mean lights out for Trump. But consider, California alone accounts for 27.3% of these early Democrat votes—Biden could win California by 10 million votes and the result there would have no effect on Pennsylvania and its 20 Electoral College votes.

Second, and even more importantly, just because a registered Democrat in Pennsylvania or Florida voted, doesn’t mean she voted for Biden. Registered Democrats outnumbered Republicans in Florida, North Carolina, and Pennsylvania in 2016, but Trump won those states anyway. This was partly due to voters crossing party lines. So many Democrat voters did that in 1980 and 1984 that they were dubbed “Reagan Democrats.” The same thing happened in 2016 for Trump and is likely to happen again in 2020.

In Texas, voters completed their 12th day of early voting on Saturday, Oct. 24, marking the last day 2020 can be directly compared to 2016 (since Gov. Greg Abbott added an extra week of early voting to allow for less crowded polling places).

Among 15 large urban counties reporting early and by-mail results that can be compared to 2016, we see overall turnout of 43.75% compared to 46.08%, a decline of 2.33%, or 5% proportionately. Raw turnout is up a bit to 4,863,482 from 4,498,431, an 8.1% increase. Here it’s important to note three things: Texas added about 14% new registered voters since 2016 as the state grew 7.3%; there are still six days of early voting remaining; and, typically, the first two days and the last two days of early voting are the heaviest. Therefore, it’s almost a certainty that a larger share of Texans will vote early by mail or in-person than in 2016.

While voting in person is running a little behind where it was in 2016 after 12 days, voting by mail is setting records, with 5.15% of voters in the urban counties voting by mail as of Oct. 24 compared to 3.19% in 2016. Overall, 4.93% of registered Texas voters have used the mail to cast a ballot, totaling 835,096 votes. In the 15 urban counties alone, this amounts to 261,000 more votes by mail than in 2016.

Texas’ 2020 early voting numbers have been called a “surge” by some, many of whom are pushing the scenario of a Biden upset in Texas. This narrative is attractive because it raises the specter of a decisive Biden win and a larger realignment of American electoral politics in the Democrats’ favor. But a few factors point against this being the case, even as a new Dallas Morning News/UT Tyler poll shows President Trump down 3 points to Vice President Biden.

Most importantly, if Texas were as close as this poll suggests, Biden would have campaigned in the state (he hasn’t set foot in Texas and neither has his running mate) and not canceled most of his much-ballyhooed $5.9 million TV ad buy there.

In addition, the poll shows a strong 8-point lead for Sen. John Cornyn, suggesting some of Trump’s apparent weakness in the poll is due to the “Shy Trump” voter phenomenon wherein some Trump voters are reluctant to admit their support for the President.

Further, the last day of polling was Oct. 20, two days before Biden admitted during the last presidential debate that he wanted to end oil—not a winning policy position in Texas (or in Pennsylvania for that matter). Lastly, the poll showed ominous signs for Biden with undecideds, about 5% of those polled and considered likely voters, breaking more than 2-to-1 for Trump with independent voters leaning for Trump 3-to-1.

Most U.S. elections are won or lost at the margins. Undecided voters, especially independents, appear to be tilting towards Trump. This suggests his campaign could close strong, heavily outperforming the polls in 2020—just as he did in 2016.