NBC News has created a useful early voting resource that tracks in-person and by-mail balloting at the state level nationwide. What it tells us doesn’t match what we’re hearing from the pollsters—who continue to insist that Texas is turning blue, or at least a little purplish.

As of Oct. 29, NBC News has analyzed Texas’ early and by mail voting through Oct. 26, finding that 54% of those voting early are Republican vs. 36% Democrat and 10% other or unknown. This gives the Republicans up to an 18% lead in early voting in Texas, or 8% if all the unknown and first-time voters are assigned to the Democrats.

The day before, NBC calculated the margin as 53% to 37%.

While Texas doesn’t have party registration, NBC News used commercial databases to infer partisan turnout. Texans’ primary voting history is commercially available, providing a strong indicator of party affiliation.

Texas counties reported that 8.4 million Texans have voted as of Oct. 28., meaning that about 600,000 Texans have voted since NBC’s analysis with data from two days before.

Because polls suggest that Republicans are more likely than Democrats to vote in-person and on Election Day (due to a wide partisan gap in the level of fear over COVID-19), it can be projected that the estimated share of the Republican vote still has room to climb in Texas.

Before going deeper into an analysis of what this might mean for Texas in 2020, a few words of caution. Early voting typically favors one major party or another, with Democrats or Republicans generating an edge with mail or in-person early voting on a state-by-state basis. But the pandemic has likely upended that. As a result, early voting numbers need to be even more carefully considered.

For example, Pennsylvania has seen about a 10-fold increase in mail-in ballot applications in 2020 compared to 2016, with a little more than 3 million ballots mailed out. Of these, almost 2 million have been returned, with NBC reporting a 46% margin in favor of the Democrats, 68% to 22%, with 10% of the by-mail ballots having either no party or a third party affiliation. Almost 6.2 million Pennsylvanians voted in 2016. With a higher turnout expected, almost 40% of the Keystone State’s 2020 vote may be by mail.

Unlike in Texas, where only 11% of the early vote is by mail, 100% of the early vote in Pennsylvania is by mail.

Of course, partisan affiliation of the voter doesn’t necessarily indicate who the voter is casting their ballot for—after all, the term “Reagan Democrat” came into the political lexicon for a reason. And, in 2020, there are indications that President Trump’s reelection effort will be boosted by a large number of Democrat votes in states like Pennsylvania.

Digging into the Texas numbers, it’s entirely likely that the identified Republican share of voters will climb from 54% to 56% by Election Day. Being conservative and assigning all the new and unidentified voters to the Democratic or minor party column would generate an estimated turnout of about 56% Republican to 41% Democrat with 3% other.

Just because a voter has a party affiliation doesn’t mean that voter will reliably follow through. In 2020 there has been plenty of ink—or electrons—spilled on the topic, with the “suburban woman” vote for Biden on one side and increased support for Trump among Black and Hispanic voters on the other.

Still, it’s increasingly hard to square polling in Texas with the actual early turnout numbers. This is especially the case when looking at polls that show high party loyalty to the major party candidates. The Real Clear Politics average of polls in Texas show Trump with a 2.6% margin. Of the five polls conducted entirely in October, this includes results ranging from the Dallas Morning News’ Biden at +3, to Rasmussen’s at Trump +7.

If Texas voters adhere to their party affiliations, the early voting there could indicate that President Trump is on track to meet or exceed his 9 point win over Hillary Clinton in 2016, meaning the polls are understating support for Trump by an average of about 7%. If that does end up being the case, it could mean two things: political polling has gotten worse since 2016; and Texas isn’t turning blue anytime soon.