The Election Forecast That Team Biden Likes Is Just a Ticking Time Bomb
FiveThirtyEight’s dead-heat prediction will self-destruct in roughly two months.
Something strange is going on in Poll World.
Simply put, FiveThirtyEight’s model gives Joe Biden a much better chance of beating Donald Trump this fall than the other leading models out there do—and it has had Biden’s odds improving since his rambling, unfocused performance in the June 27 presidential debate.
As of Wednesday, for example:
-
FiveThirtyEight’s model gives Biden a 54-in-100 chance of winning the election. (Which is to say, of winning enough states that he would win the Electoral College, not of winning the popular vote.)
-
By comparison, polling expert Nate Silver—who founded FiveThirtyEight but left last year—has his own model that gives Biden a … let’s say significantly lower chance of winning. (To see the exact number, you’ll have to subscribe to Silver’s Substack.)
-
The Economist’s model gives Biden a 1-in-4 chance of winning the election.
-
The New York Times’ polling averages show Trump leading by 3 points or more in six swing states that Biden won in 2020 (Arizona, Georgia, Michigan, Nevada, Pennsylvania, and Wisconsin)—margins that, if accurate, would correspond with a huge Trump win in the Electoral College.
The gap between FiveThirtyEight and its competitors has outsize importance given how many Democrats are arguing that Biden should end his reelection campaign because polls show he is headed toward a major loss.
Biden has responded by claiming repeatedly that the race is in fact a dead heat, and FiveThirtyEight, which is probably the best-known election forecaster, seems to validate that position. On Tuesday, former Biden chief of staff Ron Klain tweeted an image of the site’s electoral-vote forecast as evidence that the donors and elected Democrats who say the president lacks a “path to victory” are incorrect. Extremely pro-Biden pollster Simon Rosenberg has said something similar.
On the other hand, no one else agrees. And other poll experts have started to ask FiveThirtyEight’s current director, G. Elliott Morris, direct questions about his methodology. (Slate reached out to FiveThirtyEight and its parent company, ABC, which declined to comment.) So what’s going on? I asked David Broockman, a professor of political science at UC–Berkeley who teaches a quantitative methods class, for his thoughts.
“I don’t have the code for their model so I don’t know for sure, but my understanding is that they basically are putting all the weight on the fundamentals right now, so almost no matter how bad the polls get they would show the same answer,” Broockman wrote in reply. In forecasting, fundamentals refers to factors independent of polls—unemployment rates, inflation data, incumbency—that historically correlate with election outcomes. Biden is an incumbent overseeing a growing economy, which should mean he’d be sailing smoothly to reelection.
But he’s not. And other pollsters appear to be weighting fundamentals less heavily, which means they put more importance on polls, in which Biden is (unambiguously) losing.
Perhaps more concerning for Biden, Broockman says, is that there’s good reason to believe that fundamentals won’t help predict outcomes in this race the same way they have historically. “The rationale would be that historically the early polls aren’t super predictive. But that’s because people don’t know the candidates yet,” he said. “Hard to imagine people not knowing how they feel about these two!”
If Biden’s fundamentals situation is favorable but he’s still losing, that seems as if it would be evidence that a generic replacement candidate could be expected to do better than him, not worse—especially given that he’s notably limited in his effectiveness as a communicator. (Put another way, previous candidates who enjoyed the same fundamental advantages that Biden does were probably all more capable of making a case for themselves on television.) Beyond that, statistically minded observers have raised additional questions about quirks in FiveThirtyEight’s state-specific predictions that suggest that the site is overstating Biden’s chances for reasons that go beyond basic weighting choices.
Either way, FiveThirtyEight’s model will get less favorable to Biden as Election Day approaches and it gradually weighs fundamentals less and less heavily. (The other models will do this as well.)
Even the people who work at FiveThirtyEight are saying as much. Morris wrote on Twitter/X that an version of the site’s model that places less emphasis on fundamentals currently gives Trump, not Biden, a 57 percent chance to win. (The same tweet adds that “none of this will matter much by Labor Day,” as polling gains predictive value relative to other information.) He has also written that Klain’s citation of his model to justify Biden’s continued presence in the race is, quote, “weak sauce.” On that, at least, all the polling people seem to agree.
Discover more from CaveNews Times
Subscribe to get the latest posts sent to your email.