New York Times polls are betting on a political realignment

Should you buy it?

The past quarter-century of American politics has been dominated by two major trends. One is simply increasing political polarization: red states have perpetually become redder, and blue states bluer. In 1996, only 7 states were decided by 20 or more percentage points. But in 2020, 19 states were, leading to many wasted votes.

The other is increasing polarization along educational lines. As recently as 2006¹, there was basically no difference in voting Democratic or Republican based on whether a voter had graduated from college. But in 2020, Joe Biden won 54 percent of the vote among white voters who’d graduated from college but just 37 percent who hadn’t, according to estimates from the data firm Catalist.²

These educational splits have typically been lesser among nonwhite voters. But the increasing educational divide is coming into tension with the most longstanding feature of American politics: racial polarization. Since the Civil Rights Era, Black voters have been the building blocks of the Democratic coalition, voting for them in overwhelming numbers. Hispanic and Asian American voters have been more swingy but have usually at least leaned toward the Democratic Party.

However, Black and Hispanic voters are more working class — less likely to have completed college degrees — than white ones. So in principle, a continued increase in educational polarization would lead to erosion in Democratic support among these groups, but gains with white ones. From an Electoral College standpoint, this would actually be a good trade for Democrats since white voters are overrepresented in their impact on the Electoral College relative to their share of the overall voting population.

And in some polls, that’s exactly what we’re seeing. Donald Trump’s polling against Joe Biden had consistently shown him making huge gains with Black and Hispanic voters, especially younger voters and those without college degrees. Kamala Harris’s numbers have improved among these groups, but the same trend persists, including in the most high-quality surveys. For instance, a New York Times/Siena College oversample of Black voters from their recent national survey showed Trump winning 15 percent of the Black vote, up from 9 percent in 2020. That’s not great — Black voters are still very, very Democratic — but it’s potentially enough to make the difference in states like Georgia. 

Meanwhile, a recent national NBC News/Telemundo/CNBC poll exclusively of Hispanic voters showed Harris winning them by only 14 points. Keep in mind that there’s never any hard evidence of how any racial group votes — we have to rely on exit polls and ecological inference, which themselves can be subject to error — but that’s a considerably narrower margin than in recent elections. In 2020, Biden won Latinos by 28 points, according to the AP Votecast exit poll, and 26 points per Catalist’s data. However, Harris is slightly improving on Biden’s performance in many if not most polls of white voters.

If these trends are real, then you’d also expect to see shifts in the electoral landscape, with Harris making gains in the whiter states, including the Blue Wall battlegrounds of Michigan, Wisconsin and Pennsylvania — but Trump improving his performance in the more racially diverse Sun Belt. And you do get exactly that in some surveys. A pair of NYT/Siena polls on Saturday found Harris leading Trump by 4 points in Pennsylvania but Trump up by 6 in Arizona.

But while these numbers are similar to previous NYT/Siena of these states, they differ from the polling averages there, which show much closer races in both cases — and they also differ from certain other high-quality surveys. Just this week, for instance, Wall Street Journal polls showed Trump 1 point ahead in Pennsylvania but Harris leading by 2 in Arizona. So let’s take deeper look and see if we can untangle the mystery.