The Palin Question
By The possibility that McCain’s choice of Palin really did significantly affect the 2008 elections is one I find genuinely fascinating.
The question is obviously a good example of the fact that it’s essentially impossible to prove social science hypotheses involving causation, with the small number of trials making things even more difficult. What’s interesting about Johnston and Thorson’s data is that normally, when discussing the effect of VP choices, there isn’t even a correlation that could make a causal effect plausible. So, ultimately, it comes down to whose story you find convincing. Focusing solely on the facts of the 2008 race, I find the Johnston/Thorson explanation more satisfying. But we also have to place this in the context of previous data showing both that VP picks matter very little and that campaigns in general are overrated in their effect on elections, which strengthens the argument of skeptics.
The other thing to say is that even if Palin did have a substantial negative effect, it’s not even clear that McCain’s gamble was irrational — after all, he was almost certainly going to lose anyway, and losing by a bigger margin doesn’t really matter. It’s possible that the VP choices of trailing candidates might be less risk-averse in the future, leading to VP choices mattering more. On the other hand, aside from marginal home state effects there still isn’t a good plausible example of a VP having a significant positive effect, so the Palin debacle may make future selections even more risk-averse.