If you’re not reading Nate Silver’s 538 Blog over at the NY Times, you probably should be – it’s a comprehensive analysis of all the Presidential polls out there, and I’ve extolled it’s virtues before.
As of right now, Silver is predicting a 77% chance that Obama will win the election in November. That number will change between now and November, and I’d be surprised if it weren’t somewhere lower than that come Election Day. Either way though, the Conventional Wisdom on Election Day will almost certainly be that Obama is the favorite to win.
Let’s say Romney wins though – he pulls an epic upset and gets exactly the right combination of Electoral math to carry the day. There will undoubtedly be many on both sides that will decry the shoddy work done by Silver and the ridiculousness of his 77% predictions. The Right will declare it a stunning blow against the Lame-stream Media.
Only here’s the thing – a Romney victory wouldn’t be proof that Silver is wrong. A 77% chance is basically 3 out of 4 – meaning that 1 out of 4 times it will be wrong. If I tell you that it’s unlikely you can flip a coin and get two heads in a row, does it follow that I’m wrong, just because you in fact get two heads in a row? If the weather guy says there’s a 75% chance of rain, and it doesn’t rain, does that mean he was necessarily wrong?
So yeah, take political predictions, including Silver’s with a grain of salt. The difference between Silver and most pundits, however, is that his methods are transparent and reliance on statistics is relatively robust. I predicted a long time before it was obvious that Obama would win, so I’m not surprised that the percentage is up to 77% at this point – but if Silver and I were wrong, it will just mean that I’m a jackass, not that statistics are a bad lens with which to view politics.
In short, a Hedgehog looks at a prediction that ended up being wrong and assumes the underlying method is flawed. A Fox remembers that even a 99% chance means there’s a 1% chance of being off.