The powers of data prediction

I was a skeptic, but  I have begun to hang on Nate Silver’s every prognostication, election addict that I am. More importantly, some conflict forecasting work we’ve been doing in Liberia has (so far) unexpectedly successful. More on that in the next weeks.

In the meantime, others have been forecasting more weighty matters:

Hunch then looks for statistical correlations between the information that all of its users provide, revealing fascinating links between people’s seemingly unrelated preferences. For instance, Hunch has revealed that people who enjoy dancing are more apt to want to buy a Mac, that people who like The Count on Sesame Street tend to support legalizing marijuana, that pug owners are often fans of The Shawshank Redemption, and that users who prefer aisle seats on planes “spend more money on other people than themselves.”

From “How Visa predicts divorce” in the Daily Beast.

Three of the Hunch forecasts above fit my profile.

5 thoughts on “The powers of data prediction

  1. I followed Nate in 2008 — and was engrossed. I am much less so. I’m finding him insipid these days. Finding ways to predict chances of conflict is useful. Predicting who’s going to win an election is irritating. Nate was at his best explaining the limitations of various pollsters and describing trends and developing stats for looking at what was happening. Now he’s into justifying his statistical methods and explaining away mistakes with a lengthy discussion of whether there’s a new paradigm. That answer: there may be a new paradigm if in fact Newt wins and shows the old paradigm not longer holds. Oh that’s helpful. Now Nate’s probability to win Florida has dropped from something like 95% post NH, to 45% post South Carolina? Maybe the world is a little more complicated — and getting more so — than his models. Chris, I look to you to take on this kind of blather, even in it’s thoughtful veneer.

  2. Of course I meant Mitt Romney’s changes of winning Florida, not Nate’s, which we can reliably predict as zero.