Tuesday, December 1, 2009

The loss aversion of economic forecasters

Macroeconomic forecasting is tricky business, not just because it is difficult to predict the future. As a forecasters, you are risk averse and do not want to be caught dead wrong when you make a forecast that deviates from the consensus. Suppose you model (or your gut) tells you that GDP goes the other way from all other forecasters. If it turns out you were right, you win big, but the others have the excuse that they were the consensus. If you are wrong, you lose much more, as you deviated from the consensus. The best strategy is then to go with the consensus, whatever you actually found.

Is there really such an asymmetry in the forecaster's loss function? Jörg Döpke, Ulrich Fritsche and Boriss Siliverstovs explore this looking at 17 forecasters in Germany. They do not quite study the game I describe above, rather whether there is asymmetry in the under- and over-estimation of inflation and the growth rate. I found my question more interesting, but anyway, they find little evidence of asymmetry. They also find that forecasters work rationally, i.e., their forecasts are unbiased and they use all the information available. Overall encouraging results that should give very little space for playing the game I mentioned first.


Anonymous said...

They might still be playing the game, but forecasting is good enough on average that it can't be detected - i.e. the consensus is mostly right.

Graeme said...

There has certainly been a very similar problem with sell side equities analysts' forecasts (in addition a bias towards optimism with regard to recommendations)