Democrats' polling denialism won't fix their minority voter problem
Even the alternative explanations are really bad for Biden
Donald Trump leads in the average of the most recent polls by 2.2 points. Despite the talking up of a better series of polls for Biden, Trump is ahead by 4 points in the latest Wall Street Journal poll. Nikki Haley leads Biden in the same poll by 17 points.
The response from the Democratic smart set has been to engage in various forms of polling denialism. The recent polls are an example of Mad Poll Disease. Another theory is that respondents are engaged in “expressive responding,” a clever-sounding explanation for why poll results don’t actually mean what they say. Proponents of this theory argue that members of base Democratic groups are using the polls to vent their frustrations with Biden, but would never actually abandon him.
Then there are the savvy Democratic pros whose aura of sophistication increases every time they sling the term “bedwetter” — the Obama-era term of derision for the nervous nellies who doubt the nominee’s strategy — in emails to their buddies at Playbook. They are supremely confident in their own ability to engineer a campaign that will reverse current polling trends. And they should be! They are political operatives. It’s their job to say this, and then to try and make it happen.
The case made by the pros is a pretty simple one you always hear from incumbents: We haven’t had an actual campaign yet, and wait till the Biden campaign fires its billion-dollar messaging Death Star. The campaign is not yet about Trump because he’s not really in the news that much (seriously?). The election will become a choice, not a referendum on the incumbent.
If this sounds familiar, it’s because this is what the Trump campaign was saying four years ago. And it’s what every incumbent’s team ever says, clinging desperately to the fading incumbency advantage. The last time we heard this, in 2020, the Biden camp had a pretty adroit response: the Death Star got blown up.
Biden’s defenders shouldn’t have to try this hard. The fact that polls a year out are never perfectly predictive of the final results does not need to be pointed out—it’s perfectly obvious to anyone with knowledge of recent history.
Campaigns matter, for sure. But which side the campaign will favor is essentially unknowable a year out. Republicans will have their own billions. Trump could get convicted, and probably have bigger tail risk as a result, but voter perceptions of Biden’s mental acuity are already so precarious that even just one or two slip-ups could prove a deal-breaker.
The polls are no more and no less than the midpoint of a very fat-tailed probability distribution of what’s likely to happen next year. Polls are not perfectly predictive. They miss. But there’s no way to guess ahead of time which way the miss will go. Chances are 50-50 that things for Biden will improve. They could also get worse. Trying to outguess the polls is usually a fools’ errand.
What to make of Republicans’ surprising strength among young and minority voters?
The Discourse has predictably honed in on Trump’s polling gains among younger and minority voters—drawing closer among millennial and Gen Z voters, topping 20 percent of Black voters, and drawing within single digits among Hispanics. These gains mostly extend to other Republicans in head-to-head matchups against Biden, so they are not exclusively a Trump phenomenon.
These nonwhite shifts really stand out in Adam Carlson’s aggregation of polling crosstabs published over at Split Ticket:
As the author of a book that makes the argument for nonwhite electoral realignment along the same populist lines that redrew the political map in 2016, these numbers, if they hold, will suggest that the realignment has taken hold much more quickly than even I’d predicted.
The numbers—a 31 point shift right among African Americans and 20 points among Latinos—are indeed stark. And for their critics, it’s this starkness that undermines them. Black voters simply don’t shift this much from election to election, they argue. In no presidential election since 1980 has a Republican won more than 12 percent of the Black vote. The shifts being portrayed in the polls right now are unthinkable, so they can’t be right. And if you mentally undo (or is it, unskew?) them, you’re basically looking at a rerun of 2020, with a 2-3 point Biden victory in the popular vote and a very close race in the Electoral College.
I don’t disagree that the polls have a too-good-to-be-true quality about them. But even if Trump doesn’t win the rainbow populist landslide indicated by current polls, the alternative explanations Democrats are giving for these numbers still suggest that Biden is in deep trouble.
And what are these explanations? That core Democratic groups are frustrated by the economy, that they’re unenthused by Biden both as a status quo politician and because of his age, that Biden supporters aren’t motivated to answer pollsters’ phone calls, that Democrats may have a turnout problem rather than a persuasion problem.
Any of these are bad omens for an incumbent running for re-election. Taken together, they’re hardly any better for Biden than a prima facie reading of the polls themselves.
They all amount to an admission that Democrats are struggling mightily to mobilize core constituencies. If turnout declines in 2024, a fair certainty, Black, Latino, and youth turnout will take a disproportionate hit. And if turnout declines even further among Biden 2020 voters in these constituencies—which all the numbers right now are pointing to—then you start to get something resembling the big persuasion shifts the polls are now showing. You don’t get extra credit for losing votes on turnout instead of persuasion.
Many of the polling deniers are operating under the assumption that Biden’s victory in 2020 was robust and he can afford to lose a couple of points in support, just like Obama did in 2012. But Biden’s victory in 2020 was nothing like Obama’s in 2008. Biden’s victory in the Electoral College was the most precarious since George W. Bush’s in 2000, where a 0.6 point shift in the national popular vote would have been enough to re-elect Trump even as he would have lost the popular vote by 6 million votes. With a Republican Electoral College advantage borne of working class realignment, Biden simply doesn’t have much room for error.
Biden’s weakness among Black and Latino voters is without precedent for a modern Democrat
Democrats like Carlson wave off Biden’s polling struggles with Black voters by saying that the polls always underestimate Black voter support for Democrats. That’s a misdirection. Yes, polls do tend to understate the percentage of the Black vote Democrats will get, as they’ll likely take the lion’s share of the undecideds. The same dynamic applies in the other direction: polls generally underestimate the share of the white working class, rural, and red state vote Republicans get. Undecideds generally fall out based on the overall leanings of the group.
But the current polls are showing something wholly different. It’s not a matter of voters parking themselves in the undecided column, but Trump and other Republicans nearly doubling Trump’s 2020 share of the African American vote. That’s evidence of a real persuasion shift, at least in the current polls. Only 10 percent of African Americans in the last New York Times/Siena poll reported voting for Trump in 2020, a number less than half of what he gets now.
So, it’s not really true that polls have significantly understated Democrats in the same way before. High quality pre-election polls in 2020 showed Trump getting just 7-9 percent of the Black vote.
Nor is the claimed exception in 2022 all that it seems. Then, polls were said to have underestimated Democratic margins by 27 points, with Republicans predicted to get 16 percent of the Black vote. But this Bleeding Heartland analysis gets a 27 point error by cherrypicking the worst possible number for Republicans from the post-election polls, the Pew validated voter study, which showed Republicans getting just 5 percent, below Obama 2012 levels. Both the traditional exit poll or AP VoteCast, with much higher sample sizes, which put Republicans at 13 and 14 percent respectively, which is more consistent with the steady erosion we’ve seen for Democrats in post-Obama elections.
Perhaps we will see a reversion back to traditional Democratic support levels as the choice gets nearer. That’s always a possibility. You can’t expect perfection from the polls. But we shouldn’t misstate what the polls today are showing. Biden’s Black voter results are really without precedent for a Democrat in modern polling. His Latino polling numbers would be the worst Democratic numbers seen since John Kerry only won the Latino vote by 9 points in 2004, a number widely acknowledged to be too optimistic for Republicans. There’s no sugarcoating that this is bad for Biden.
If you’re looking for corroboration in hard data, one can look to voter registration trends in both Florida and North Carolina, two states where voters register by party and where voters self-report their race on the voter file. They show a large uptick in the share of new Black voters registering as Republicans on a two-party basis after the 2020 election. I’ve extended the data on voter registration to 1960, right before the Civil Rights-era realignment, to show just how unprecedented these figures are. A similar analysis conducted of Hispanic voters prior to the 2020 election was predictive of a shift towards Trump. But what about good Democratic results in special elections? In 2020, we didn’t really see any signs of a Hispanic shift towards Trump until the presidential election itself, powered by infrequent voters with little track record of support for either party.
Polls often understate—rather than overstate—demographic shifts
Just as polls nationally and within demographic groups are just as likely to overstate as understate Democrats or Republicans, so it is with forecasting demographic shifts. Polls prior to the 2020 election showed only a modest shift of 8-10 points among Hispanic voters, and the actual shift turned out to be in the range of 12-18 points, with 20+ point precinct-level shifts visible throughout the country.
Likewise, pre-election polls in 2016 projected a white working class swing towards Trump, but it was vastly undersold. Yes, Trump led in Iowa and Ohio, but few people were connecting the dots to suggest this meant trouble for Hillary Clinton in demographically similar states like Wisconsin and Michigan.
That’s not to say that the already dramatic shifts manifest in today’s 2024 polls will come be even more jaw-dropping than those today. Polls also forecast a big swing back to Biden among seniors and white working class voters in 2020 that didn’t really materialize. At this point, the best way to interpret the polls are as directional signals. The chances are greater that Biden will do somewhat worse overall, and with Black and Latino voters specifically, than he did in 2020. Unless something dramatic changes, it’s hard to see how he does better with these key groups than he did in 2020. And even if only a more muted shift of the current polls manifests on Election Day, that’s still a point in favor of the multiracial populist realignment.
Nominated for best line in survey analysis in 2023: "You don’t get extra credit for losing votes on turnout instead of persuasion." 😄