Biden's position in the polls is* improving against Trump — here's my theory for why
* Or probably, at least. But it may be more about Trump than about Biden.
In case you missed it, yesterday I announced plans for the 2024 presidential election model, which will launch in June for paying subscribers. I won’t give you too hard a sell until we get closer to the release date. But if you’re interested in the forecast — plus the second half of this article, which is one of the posts I run for paid subscribers each month — you can sign up using the link below.
To a first approximation, not all that much has changed in the presidential race.
If you’d asked me six weeks ago — “hey Nate, what’s going to happen in the election?” — I’d tell you that Donald Trump would probably win an election held today, but that there’s still a long way to go and November is more or less a toss-up.
And if you asked me that today, I’d tell you the same thing.
Now, as you might imagine, this question isn’t hypothetical. People ask me this sort of thing all the time. And usually they aren’t satisfied with the “toss-up” answer and probe for further specificity — “c’mon man, who’s really going to win?”. Six months ago, if you’d given me a free bet on one of the candidates, I’d have taken Joe Biden. Six weeks ago, I’d have taken Trump. Today, I’d take … I’m not sure.
That’s not me being wishy-washy. It’s the honest and inevitable response when the probability of a forthcoming event is about 50/50. Psychologically, Biden being, say, a 53/47 favorite instead of a 53/47 underdog feels a lot different — but in practice, it’s the same damned thing. You do not want to be one of those people on the DC party circuit who rounds up a 53 percent chance to 100 percent. Twitter can be even worse about this. Some people are obsessed with who is ahead in national polling averages today — but today doesn’t matter. November 5 does. And national polling doesn’t really matter either — the Electoral College does.
Nonetheless, if you’re the sort of reader who cares about the difference between a 53/47 edge in one direction or the other — and if you’re a Silver Bulletin regular, you probably are — it looks to me like Biden’s numbers against Trump have improved by a hair, probably to a degree that isn’t just statistical noise.
Here’s a compilation of every national poll I could find from 538 or RCP that meets the following criteria: the pollster surveyed the race at some point after the State of the Union address on March 7, and also polled the race at some point between January 1 and March 7, so we have a before-and-after comparison. If a pollster surveyed the race both with and without third-party candidates like RFK Jr. included, I just averaged the two versions of the poll together.1
Between 23 polling organizations that surveyed the race in both periods — who said polling is dead? — Biden now trails Trump by an average of 0.8 percentage points, as compared to trailing by 2.2 points in the older version of the same polls.
This is a relatively visible change because Biden is now more often showing leads in national polls. He had just 4 outright leads (i.e. not counting ties) in the old batch of polls, as compared with 9 leads in the new batch. That’s still not great for Biden, because Trump retains an Electoral College advantage and has remained ahead in the large majority of swing state polls — something people obsessed with hyping up every Biden lead in a national poll conveniently ignore. Still, Biden has plenty of time to right the ship — is this at least a meaningful step in the right direction?
Well, assume that the 23 polls have an average sample size of 1,000 respondents each — the combined sample size is 23,000 people. The margin of error for the Trump-Biden gap2 in a sample of 23,000 respondents is about 1.3 percentage points. So you could represent the data like this:
“Old” polls: Margin between Trump +0.9 and Trump +3.5
“New” polls”: Margin between Biden +0.5 and Trump +2.1
As you can see, there’s some overlap between these — enough that Biden lost ground in about a third of the polls, even though his standing improved overall. So it’s possible this is just statistical noise. But it probably isn’t, particularly with this sort of apples-to-apples comparison where we’re looking at pairs of polls from the same firms. And there’s no particular reason to be anchored to the null hypothesis here — it’s not like polling in January and February is some particularly meaningful indicator.
So I’m going to proceed through the rest of the article assuming that the shift is real. What caused it?
I’ve seen quite a few theories proposed:
It’s improving consumer perceptions about the economy
It was the State of the Union
It wasn’t the State of the Union itself, but the change in the tone of media coverage after the State of the Union
It’s Trump’s trials getting underway and Trump otherwise being more visible
It’s Biden campaigning more vigorously after having been a little soporific before
It’s the primaries being over and/or voters being more resigned to the Trump-Biden matchup