Things are looking a bit bleak for Biden’s re-election at the moment. As of this writing, the forecast most bullish on Biden, FiveThirtyEight, only gives him a 51% chance of remaining in the white house. Other forecasters, like the Economist, Jack Khersting, Nate Silver, and even me,1 are more bearish, hovering around a 30% chance of re-election. In the aftermath of his poor showing at the first presidential debate with Trump, there has been an outcry online wondering if the Democrats would be better served by Biden stepping down. Notably, the New York Times and Nate Silver have suggested that a new nominee take his place in order to improve the chances of a Democrat occupying the white house next year.
1 I plan on releasing publicly next week!
The arguments for removing Biden as the nominee in the wake of the debate, however, fail to support the suggestion with any electoral analysis beyond “Biden’s age is a problem for voters.” I’ve been working on a presidential forecast for over a little over a year now, so I have a unique opportunity to fill this analysis gap.
Per my model, which, in short,2 mixes polls with economic data, approval ratings, and historical forecasts, Biden has about a 28% chance of winning. As an incumbent running for re-election, this is not great. Replacing him with a (hopefully) stronger alternative sounds like an appealing proposition to Democrats in theory. In reality, however, Biden’s hypothetical replacement doesn’t fare any better and only has a 20% chance of defeating Trump.
2 Upon launch, I’ll share the full methodology in detail.
This is largely due to the fact that Biden is the incumbent president and incumbency is still worth something, electorally.3 Specifically, incumbency it’s worth about 2.5% in terms of the two-party voteshare. In many states, this doesn’t change the expected outcome. Biden losing by 40% in Oklahoma is not meaningfully different than another Democrat losing by 42%. In tightly contested states, however, a 2.5% change in expected voteshare has a massive impact on the probability of winning.
3 Trump’s loss as an incumbent was a rare occurrence in the modern election era!
There are sound reasons to believe that Biden should retire — he is very old and has a very high-stress, high impact job. But the belief that Biden should retire to improve the chance of a Democratic win in November isn’t supported by this analysis. Biden’s age is a huge concern for voters and an electoral liability contributing to his grim chances, but the outlook for an as-of-yet-named hypothetical replacement candidate is even worse.
Of course, this analysis hinges on this model being a reasonable one. I think it is, but am open to critiques or alternatives — you can view the source for this page to explore for yourself. The important thing is that a model provides a framework for evaluating the world in a principled manner, even if the result is counter-intuitive. In Nate Silver’s own words:
But what a model ideally provides you with is a structured way of thinking through a challenging problem. Because our intuitions, even if they’re right, often lack any statistical precision.
–Nate Silver, in The presidential election isn’t a tossup
This article was re-published on July 2nd, after having found a bug in the model code used when the article was originally written on June 28th. The bugged version underestimated the correlation across states, resulting in a tighter distribution of electoral college votes.
While the overall conclusion is the same — a generic replacement is less likely to win than Biden — the severity of the change is different. As originally written, Biden had a 24% chance of winning and a generic replacement had a 7% chance of winning. In the corrected version, Biden has a 28% chance of winning, while a generic replacement has a 20% chance of winning.
Citation
@online{rieke2024,
author = {Rieke, Mark},
title = {No, {Biden} Shouldn’t Be Replaced as the {Democratic}
Nominee},
date = {2024-06-28},
url = {https://www.thedatadiary.net/posts/2024-06-28-debate/},
langid = {en}
}