Election season is high season for the pundits and the prediction game. An U.S. presidential election is to politics what the world cup or the world series is to sports, that is, the biggest, most expensive, and hardest fought competition. As a result, we benefit from a multitude of models.
I have to admit to you that personally, I have always found it more interesting to think about what should be done in order to win an election, rather than to predict the outcome of it. My job is to make presidents and if I have been working on a race for a year or longer, I can wait another couple of days until I know the result.
I became aware of Nate Silver probably some ten years ago during an election cycle when he accurately predicted almost every senatorial race. A few years later, his model however also predicted that Hillary Clinton had about a 90% chance of winning and we all know how it turned out. And even four years ago, I remember listening to a podcast where he argued that the idea of a hidden Trump voter is ridiculous. Joe Biden ended up winning, yes, but let’s be honest that it was way closer than what most pundits predicted.
Trafalgar Group, a Republican pollster, was one of the few not to underestimate Trump’s appeal in 2020, but then was far off during the midterms of 2022 underestimating Democrats.
More generally speaking, I think that some pollsters in the U.S. use ridiculously small sample sizes. I think it’s ill-advised to look at averages of polling, and I would argue that if you weigh your raw data, you’re sort of admitting that you can’t do random sampling, which is of course the heart and essence of polling.
While one pollster may get lucky, I don’t think that there is any amount of polling that can accurately predict the outcome of this election. As I have long argued, if we’re not in a situation where pollsters are now overcompensating for having underestimated Trump before, we will have to live with the fact that this is a close race.
As for the models focusing on fundamentals, I’m quoting Prof. Lichtman’s keys to the White House extensively in my book, Beat the Incumbent. I however also think that he is sort of tweaking his model for it to fit the circumstances and I would argue that some of the keys actually depend on campaigning. And specifically regarding the 2024 election, I’m afraid that inflation is not represented strongly enough among the keys.
The point I’m trying to make is that for me, it’s not a matter of choosing a model or approach over another one. Different approaches lead to different results and it’s up to us to make the best out of it. Each one has its merits and its weaknesses. My take is that you’re probably not as smart as you may think you are when you get it all right. And you’re not as dumb as everybody (on social media) says you are when you get it all wrong.
Reference
- Lichtman, Allan (1996). The keys to the White House : a surefire guide to predicting the next president. Lanham: Madison Books.
- Perron, Louis (2024). Beat the incumbent. Proven Strategies and Tactics to Win Elections. New York: Radius Book Group.
Picture: ChatGPT