SBSQ #8: Trump might be a felon before the election. Does that mean the polls are worthless?
Plus: when should you hedge your bets? And what's it like being an extremely minor nerd celebrity?
Welcome to the April edition of Silver Bulletin Subscriber Questions, the place for paid subscribers to ask me questions about pretty much anything. Typically, I’ve been going with one question that gets a rather long response, and then 4-5 other “lightning round” questions. But I’m going to mix it up this month and with three medium-length responses instead. Feel free to use the comments to chat about this month’s responses or to ask questions for May.
One more thing: I’m hiring for an Assistant Elections Analyst, someone to help with running the presidential election forecast once it launches in June, plus other data-related tasks. The response has been very encouraging so far, so I’m not planning on blasting this out on social media or posting it on job boards. But if you do know somebody who you think would be the perfect fit, don’t hesitate to forward it to their attention. The deadline to apply is Thursday (May 2).
In this month’s edition:
Do election models adequately account for Trump’s conviction risk?
When should you literally hedge your bets?
What’s it like being an extremely minor nerd celebrity?
Do election models adequately account for Trump’s conviction risk?
David Abbott asks:
As a long-term connoisseur of Nate’s election model, I’m not sure it has much utility this year. It simply can’t capture the risks Trump faces in his criminal trials. The model has no idea that Trump is facing four felony prosecutions. It doesn’t know when they will occur or what optics they will create. Furthermore, we have a n of 0 when it comes to the effect of a felony trial on a presidential candidate.
The model is robust to ordinary news cycle stuff. The use of fat-tailed t distributions is pretty robust. However, when a model doesn’t account for what is probably the biggest variable in the system, what does it really tell us?
We’re verging into “more a comment than a question” territory here, David — but I’ll let it go since it’s a smart comment/question. Indeed, traders at Polymarket think there’s an 81 percent chance that Trump will be convicted on at least one felony charge before the election.
It’s possible I’ll take a longer/deeper run at this question again at some point in the newsletter. But my current impression is this:
I fairly strongly disagree with the n=0 framing — that is, with the idea that is some “black swan” event that might render the model useless;
Nevertheless, it certainly rises to the level of my mental list of “factors that could be material and aren’t well accounted for by the model”;
However, it can be hard to come up with a list of such mental adjustments without engaging in confirmation bias.
Look, whenever someone says an election development is unprecedented, I’m suspicious, because people always say that about every election, as this XKCD cartoon reminds us:
Granted, most of these XKCD examples are far more trivial than the issue you’re bringing up, David — the possibility of Trump being convicted of a felony before November, most likely as a result of the New York hush money case. But it’s hard to define an objective threshold for “special circumstances so unusual that you just have to throw the model out”. And the general heuristic advice is that people are too liberal about this — too eager to engage in special pleading in declaring some reason why the ordinary rules don’t apply.
Also, it can be easy to forget about special circumstances that work in the opposite direction. For instance, there’s never been an 81-year-old running for re-election before: Biden’s age is a huge deal to voters (and in my opinion reasonably so). So that’s also n=0.
Getting less abstract, it’s certainly not crazy to think a Trump conviction could have a material effect. If you ask voters in polls, in fact, a decent number of them say it could sway their votes. Look, I want to keep an open mind here. It is an unprecedented circumstance, that’s fair to say. Even if elections are mostly about vibes/optics, the vibes of an actual felony conviction and the assorted imagery and headlines could be pretty bad for Trump. And even if it only changed the mind of 1 in 100 voters, that would materially affect the results.
But people are generally bad about answering hypothetical questions. If a pollster called me and asked what my plans were for tomorrow morning, I’d say I was going to wake up early and go for a 3-mile run, try that new breakfast sandwich place I’ve been meaning to hit up, and then knock out a killer newsletter post. And every now and then I have a morning like that. But for the most part we’re creatures of habit. And habitually I’m a late-riser. If I can get one productive thing done before 10 a.m., it’s a win. Your empirical self is usually a more reliable guide to your future behavior than your aspirational self.
And the public seems to be pretty habituated to tolerating Trump’s various schemes and scams, and his even actual alleged crimes. I mean, if people are willing to vote for Trump despite January 6, are they really going to care about Stormy Daniels? If people cared that much about these criminal cases, would Trump be ahead in the Electoral College in the first place?
I’d just say this. As an experienced gambler, I often find myself making a mental list of “tiebreakers”. (If I’m deciding whether to call down a potential bluff in poker, for instance, then my intuition about my opponent’s physical demeanor will sometimes make the difference if the decision is otherwise close.) And that’s where the Trump legal risk slots in for me: it’s a tiebreaker favoring Biden. It’s one of a number of reasons I think the election in November is a tossup even though Trump would be favored in an election today. But it’s not a reason to chuck out the model or to treat this as an n=0 circumstance.
When should you literally hedge your bets?
Speaking of gambling, here’s our next question from friend-of-the-newsletter Matt Glassman: