Monday, April 14, 2014

Forecasting Recessions: Economists' Record Of Failure "Unblemished"

From:  http://www.voxeu.org/article/predicting-
economic-turning-points
"All my economist's say, 'On the one hand or the other.'  What I need is a one-handed economist!" -- Harry Truman.

Sorry, Harry, apparently even that won't help.

A new study out the Centre for Economic Policy Research in the UK indicates that, for the last thirty years at least, "the record of failure to predict recessions is virtually unblemished."

Ouch!

Take a look at the chart to the right.  It is a little hard to interpret but it starts about 2 years prior to the onset of an average of all the recessions over the last 20 years.  The top blue line represents the "normal" evolution of forecasts regarding GDP growth, that is, in a non-recessionary environment.  On average this is about 3% per year across all of the countries studied.

The forecasts from economists - the red bars - start out pretty close to this norm but begin to drop below the norm at the 8-10 month point.  While, on average, the forecasts continue to decline over the year preceding a recession, they still miss the mark (albeit slightly) even at the end of the year.  In other words, they get less wrong by the end of the year but they are still all - as in all - wrong.  The authors indicate that this paper replicates the results found by a 1990's paper that looked at the same effect over an earlier time period.  The effect is even worse when looking at recessions that develop after banking crises.
Note:  The bottom blue line which shows the actual average GDP growth is positive because, as the authors point out:  "on average, growth is not negative during recessions in advanced economies because the dating of recession episodes is based on the quarterly data and annual growth tends to remains positive during many recessions."  'Nuff said.
The authors also add that there are three schools of thought about why these forecasts are so uniformly incorrect:  Economists don't have enough information, don't have the incentive or aren't good enough Bayesians (i.e. hold on to their priors too long) to make accurate forecasts.  The jury is still out with regard to the actual reason but the effect seems like the kind of thing an intel analyst would want to account for when using macroeconomic forecasts in other than business analyses.

(Tip of the Hat to Allen T. for the link!)

No comments: