An Unfair, Post-Hoc Critique of Nate Silver
…I recently estimated Trump’s chance of becoming the GOP nominee at 2 percent.
Trump isn’t the nominee yet, but I think we can still say mistakes were likely made by Nate Silver with his assessment. If there are people who predicted Trump’s success, they were not known to me. But Silver’s specificity makes him an unfairly easy target for finger-pointing, which I will now do. I doubt we are the only parallel universe out of 50 that has Trump winning the nomination.
Silver arrived at 2% by arguing that there were six independent barriers to Trump’s winning the nomination. Even if Trump had a 50% chance of clearing any one of the hurdles, the probability of passing all the hurdles was only 0.5^6, or 1.5625%.
The problem was, these six hurdles were not really independent. Let’s review them.
- Trump’s media exposure advantage
could be broken once there’s some news about another candidate. Every Republican on stage will have the opportunity to make news in the debate tonight, for instance. It’s possible we’ll still be talking about the Trump surge in a few weeks, but it’s also possible that we’ll be contemplating the Ben Carson or Ted Cruz or Chris Christie surge instead.
- Low-information voters (who tend to be Trump supporters) will either become informed and no longer support Trump or not vote since they don’t care about politics enough to do so.
- Trump will disappoint in Iowa and New Hampshire, taking away favorable media coverage from him.
- Candidates like Trump without establishment support
would soon be bypassed as the rest of the field consolidated down to one or two other establishment-backed alternatives.
- Trump will not be able to accumulate delegates because of
Poor organization in caucus states, poor understanding of delegate rules, [and] no support from super delegates
- Delegate and convention rules “[work] against candidates like Trump.”
Though Silver considers these to be six independent events, they are not independent. #1 is about media coverage, but #3 is also about media coverage.
#2 is about Trump’s support being from those outside of the set of people typically interested in politics, but this has significant overlap with #5 and #6. #4, #5, and #6 in particular seem to blur together. Trump not having establishment support is the same thing as saying he has poor organization, which is the same thing as saying he won’t be able to work the convention rules to his favor. To cross any one of these barriers is to cross all of them, but Silver instead multiplies the probabilities together to end up with a small number.
There’s another problem. Silver applies his skepticism in a one-sided manner. Most obvious is his assumption in #4 that he will be overtaken once the field is reduced “down to one or two other establishment-backed alternatives.” We only got down to two just now; Silver didn’t express skepticism that candidates would be reluctant to exit the race.
How ought we judge Silver for this? Virtually everyone at the time assigned Trump a 0% probability of winning. 2% is actually closer than his peers! But somehow, it is easier to point and laugh at a 2% estimate than a 0% one, especially when the 2% comes with a justification and the 0% is just words. Silver, even when wrong, managed to be interesting. Let’s give him that.
Image by Randy Stewart