Preach It, Brother Nate!

trizzlor says:

While Silver has a point, I think he overstates the idea that policy can be understood and characterized by just following the data. A great example of that is Wonkblog, which prides itself on following the data, but still makes common ideological decisions in their interpretations. They will often describe Republican economic proposals that cut social programs as putting a burden on the poor, which completely disregards the fundamental conservative principle that the safety net and regulation are a burden on the poor in and of themselves. Silver is mostly doing prediction, so he’s more immune to this, but he still chose to use a model that aggregates polls instead of a model based on the fundamentals (for example, which would have predicted an Obama loss). That’s an ideological prior on how voting will work, one that’s been effective in predicting two elections (so P=0.25), but a prior nonetheless. I think it’s very rare to find scientists that don’t have *some* ideological guess at how the underlying/hidden model works (and those scientists are usually not very good); what’s important is being open about those priors. But by pretending that he just follows the data, Silver is actually obscuring his own priors.

Of course, I’ll always take Silver and Wonkblog over Tom Friedman looking at one cross-tab on a minuscule poll (or worse, talking to a cab driver on the way to the airport) and writing up another rehash of The World is Flat, as he does nearly every week. But that’s not a particularly high benchmark.Report