Tuesday, April 13, 2010

Learning About Inflation from Physicists

Statistically significant correlations don't mean certainty. If you run 100 regressions, on average one of them will be significant at the 1% level. We all pay lip service to this notion and most of us really believe it only when an "it's just a single paper" result comes out that disagrees with our priors.

Some physicists developed a test for the amount of correlation one would expect to find even if two series are actually unrelated. When they apply it to inflation forecasts
Their results show that inflation is predictable only one month in advance. Look ahead two months and the mathematics shows no predictability at all. “Adding more data just doesn’t lead to more predictability as some economists would hope,” says Bouchaud.
More of this article here. More on problems using the statistics. Hat tip: Sumner who tends to agree. The Sumner link also has a wonderful discussion of the housing market bubble, how the bubble spreads from one city to another, and the role government regulation and particular zoning boards play in causing, exacerbating, and perpetuating housing bubbles.

No comments:

Post a Comment