How to Interpret Polling Showing Biden’s Loss of Nonwhite Support

Is President Biden really struggling as badly among nonwhite voters — especially Black voters — as the polls say?

I’ve seen plenty of skepticism. Among nonwhite voters, a Democratic presidential candidate hasn’t fared as badly as those polls suggest in a presidential election result since the enactment of the Civil Rights Act in 1964. In the case of Black voters, the disparity between the usual support for Democrats — around 90 percent or more — and the recent polling showing it in the 70s or even the 60s just seems too much to accept. Some skeptics believe they’ve seen results like this before, only for Republican strength to vanish on Election Day.

But if we compare the polls with those from previous election cycles, Mr. Biden’s early weakness looks serious. His support among Black, Hispanic and other nonwhite voters is well beneath previous lows for Democrats in pre-election polls over the last several decades — including the polls from the last presidential election. Yet at the same time, his weakness is put in better perspective when judged against prior polls, rather than the final election results.

Here’s how you should interpret what the polling really means for Mr. Biden’s eventual support among nonwhite and especially Black voters.

Election results are the wrong benchmark

A major source of skepticism of Mr. Biden’s weakness among nonwhite voters is the sheer magnitude of the drop-off, based on the difference between the early poll results among registered voters and the estimated final results in post-election studies, like the exit polls.

It’s an understandable comparison, but it’s a bad one. Millions of people are undecided in polling today, while all voters have made up their minds in these post-election studies. The registered voter polling also includes millions of people who won’t ultimately vote; the post-election studies typically include only actual voters.

These two factors — undecided voters and low-turnout voters — help explain many seemingly weird differences between pre-election polls and the post-election studies.

For illustration, consider the following from our New York Times/Siena College polling:

• Mr. Biden leads, 72 percent to 11 percent, among Black registered voters over the last year.

• Mr. Biden’s lead among Black voters jumps to 79-11 if undecided voters are assigned based on how they say they voted in 2020.

• He leads by 76-10 among Black voters with a record of participating in the 2020 general election.

• His lead among 2020 voters jumps to 84-10 if we allocate undecided voters based on their self-reported 2020 vote preference.

  For comparison, this same group of Black voters who turned out in 2020 reported backing Mr. Biden over Donald J. Trump, 89-7, in the last election.

The upshot: The gap between post-election studies and registered voter polls narrows considerably after accounting for the inherent differences between the two measures — undecided voters and turnout.

This lesson isn’t limited to Black voters. To take a different example, Mr. Biden leads by just 46-34 among young registered voters in our polling over the last year, but he leads by 57-35 among young validated 2020 voters if we assign undecided voters based on their 2020 vote preference. His lead among Hispanic voters grows from 47-35 to 56-36 with the same approach. Among Asian American, Native American, multiracial and other nonwhite voters who aren’t Black and Hispanic, it goes up to 50-39, from 40-39.

Of course, we can’t assume that Black, Hispanic, young or any voters will turn out as they did in 2020. We can’t assume that undecided voters will return to their 2020 preferences, either. The point is that the differences between pre-election registered voter polls and the final post-election studies explain many of the differences between survey results by subgroup and your expectations.

If you must compare the crosstabs from registered voter polls with the final election studies, here’s a tip: Focus on major party vote share. In the case of Black voters, Mr. Biden has a 71-12 lead, so that means he has 86 percent of the major party vote in our Times/Siena polling, 71/(71+12) = 86. That roughly five- or six-point shift in major party vote share is a lot likelier to reflect reality than comparing his 59-point margin among decided voters (71-12 = 59) with his 80-point margin from 2020.

Why major party vote share? The logic is simple. Imagine that today 17 percent of eventual Biden voters are undecided and 17 percent of eventual Trump voters are undecided. What would that mean for a poll of voters who will eventually vote 86 to 14? They would be 71 to 12 in the polls today.

Mr. Biden’s polling weakness is unusual

There’s another aspect of the skeptics case that I’m less sympathetic toward: the idea that we always see this kind of weakness among nonwhite voters, and it just never materializes.

If you look back at polling from prior cycles, it becomes clear that Mr. Biden today really is quite a bit weaker than previous Democrats in registered voter polling from prior cycles.

If there’s any consolation for Mr. Biden, it’s that the drop-off is a bit smaller in our Times/Siena polling: In fall 2020, our polls gave Mr. Biden an 81-6 lead among Black registered voters, compared with the aforementioned 71-12 in a compilation of the last four Times/Siena polls.

The story is similar among Hispanic voters, who did not show similar levels of support for prior Republican candidates.

Now it’s possible that these 2020 figures were overly rosy for Mr. Biden, given that the polling more generally overestimated his support that year. Perhaps you can knock Mr. Biden’s major party vote share down two points in 2020. Either way, it seems clear he’s running well behind where he stood in the run-up to the 2020 election, while his opponent is running at least five points ahead.

This is a smaller shift than the 20-plus point change implied by the comparison between the polls and the final election studies, but it’s still quite significant. It’s also quite comparable to other demographic shifts in recent years, like Mr. Trump’s gains among white working-class voters in 2016 or his gains among Hispanic voters in 2020. At this time in both cycles, no one imagined that Mr. Trump would make 40-point gains in Obama counties in rural Iowa, and then big improvements near the Rio Grande four years later. In the end, he gained about seven points of major party vote share among these groups nationwide — about the same shift we see in the polling today.

Turnout is another option

If you’re still skeptical that Mr. Trump can make gains among nonwhite voters, it’s worth remembering that there’s another possibility: Many disillusioned or disaffected nonwhite voters might just stay home.

That possibility seems especially plausible today, with so much of Mr. Biden’s weakness concentrated among younger voters and those without a robust track record of voting. That’s exactly what happened in the last midterm election, when the Black share of the electorate fell to multi-decade lows amid weak polling for Democrats.

Looking back over the last few decades, there’s a clear relationship between the racial turnout gap — the difference between white and Black turnout — and the proportion of Black registered voters who back Democrats in pre-election polls since 1980. Or put differently: When Black voters don’t support Democrats, they tend not to vote.

It’s possible that the Black voters who back Mr. Trump in the polls today will ultimately show up for him next November. But for now, when I see Mr. Biden’s share among Black voters slip into the 60s and 70s in the polls, I mostly see yet another decline in the Black share of the electorate, at least “if the election were held today.”

If there’s any good news for Mr. Biden here, it’s that the election is still 14 months away.