Monday, December 31, 2012


From Krugman, on Romer/Bernstein. This is about more than the inflation bet. This is about a lot of people that try to confuse the public about the scientific status of the social sciences:

"’s really important to distinguish between fundamental predictions of a model and predictions that an economist happens to make that don’t really come from the model. The prediction that huge increases in the monetary base will cause large increases in the price level, and that big government deficits will cause big increases in interest rates, are more or less inescapable if your model of the economy is one in which recessions are supply-side problems, not the result of inadequate demand. Conversely, the prediction that neither of these things will happen if the economy is in a liquidity trap is a fundamental prediction of Keynesian models. On the other hand, the unfortunate Romer-Bernstein prediction of a fairly rapid bounceback from recession reflected judgements about future private spending that had nothing much to do with Keynesian fundamentals, and therefore sheds no light on whether those fundamentals are correct."


  1. Let me say first that I have not been following all this. However, I should just (for now at least) want to caution people about using the words "forecasting" and "prediction" interchangeably. In economics, some people produce large forecasting models designed to literally predict the future course of macroeconomic aggregates. Then there are macro models that seek to explain past events (usually macro-aggregates). Economists will say that these models predict (actually retrodict)events. All models are much better at retrodiction than at literal forecasting.

    Mario Rizzo

    1. This is an important point, Mario - thanks for that.

      Forecasting macro models are limited but probably necessary. Our best bet is to do with these what we do with the weather forecast: use it to decide whether we should bring an umbrella or not, but don't expect miracles.

      I think there is something here about qualitative vs. quantitative predictions too, though.


All anonymous comments will be deleted. Consistent pseudonyms are fine.