Bad Forecasts

Michael Mandel

Floyd Norris has a piece in the NYT today pointing how consensus economic forecasts are often wrong (I have a similar piece here). Norris writes

But lately Wall Street's consensus forecast has seemed more likely to miss than to hit, a fact that investors may want to take into account before they sell their homes and await the bargains that will inevitably follow the great housing bust.

Norris is actually being far too kind. It's not just lately--forecasting in general is just a disaster.

Here's a couple of reasons

a) Anything which is market-traded--including interest rates and exchange rates--is almost certain to be unforecastable. The simple reason: If there's a consensus forecast that everyone believes, that forecast gets built into the market. For example, suppose that Sally and Ted are both sure that the dollar is going to weaken next year. Then they sell dollars up to the point where one of them is not so sure, and the chances of a rise or a fall are evenly balanced.

b) GDP growth, which is not market traded, is unforecastable in practice because economists don't have a good way of predicting the shifts in underlying productivity growth.

Take a look at this chart.


Each year BW does a very fine survey of economic forecasters, asking them for their predictions over the next year (for example, the survey done in late 2005 asked for their forecast for GDP growth from the fourth quarter of 2005 to the fourth quarter of 2006)(see here for the latest forecast survey). Ordinarily there is a wide spread between the high and low forecasts, as you might expect.

However, out of the past 14 years of surveys, there were six years (1991, 1995, 1996, 1997, 1998, and 2000) when actual growth fell outside the range between high and low. There was another year (1999) where measured growth fell at the end of the range.

That is to say, 40% of the time the actual growth fell outside the range of forecasts. What happened was that forecasters didn't anticipate the sharp acceleration of structural productivity. In fact, it took them several years to believe that it was real. If and when structural productivity decelerates, there will be big forecasting errors as well.

The net result: Because productivity growth cannot be accurately forecast, neither can GDP growth.

Before it's here, it's on the Bloomberg Terminal.