Sunday, September 10, 2006

Useless Economic Forecasters

Nouriel Rouibini answers the argument against his view that the U.S. is going to have a recession in 2007, namely that the owerwhelming majority of other economic forecasters says otherwise: he points out that in March 2001, when the previous recession had already started, no less than 95% of all economic forecasters said there would be no recession!

So, agree or disagree with Rouibini in this specific case, but the fact that consensus estimate is for no recession is certainly not a valid argument for disagreeing with him.

He also links to this IMF paper that shows that almost all private sector forecasters are useless in predicting recessions. Of the 60 recession episodes in the surveyed countries only two (Russia and Ukraine in 1996)were predicted in April the previous year, and only 3 of 60 were predicted by October the previous year. In both April and October the previous year, 60 out of 60 forecasts were too optimistic (i.e. even the 2-3 analysts who predicted a recession thought it would be milder than it turned out to be).

As for predictions of U.S. recessions more specifically, none of the post-war recessions were predicted by Wall Street consensus forecasters even by the time the recession had already started, showing that the consensus forecast at the onset of the 2001 recession that there would be no recession were no one-time error.

Over the years that I have followed economic forecasts I've noticed the pattern that consensus forecasts are virtually always for next years growth to be what is believed to be the long-term structural growth rate, meaning that during strong economic booms, they nearly always forecast too low growth, whereas during economic slowdowns and recession they predict far too high growth.

These errors are particularly interesting given how these economists generally adhere to the Friedmamite "positive economics" view that the descriptive realism of theories is irrelevant, and that the only thing that matters is the ability to predict things. Well, it seems that the use of descriptively unrealistic assumptions have done all those highly paid Wall Street forecaster no good in terms of predictive ability. They are still not any better than economists with realistic assumptions or even dart-throwing monkeys with regard to predictions.

As the paper notes: there are two main hypothesis as to why forecasters are so useless when it comes to predicting recessions: 1. They have insufficient information 2. Most are employed by banks and brokerage firms, and as recessions are bad for business, bearish forecasts are also bad for business and so forecasters are deliberately predicting too strong economic outcomes to fool customers.

To test this one would have too look on to what extent the inability of Wall Street forecasters to predict recessions is related to their forecast errors reflect a reluctance to predict deviations from the trend (which would support hypothesis number 1) and to what extent their forecast errors are biased in a bullish way. The paper provides no data to enable us to make any certain conclusions with that respect, and I have found no other study of this subject elsewhere. My hunch is that it is probably in fact a combination of the two. In my experience there is a bullish bias in economic forecasts, but as the fact that forecasters also fail to forecast temporary growth accelerations illustrate, it also reflects a inability to forecast deviations from the trend.


Post a Comment

<< Home