Wednesday, July 05, 2006

How accurate do you think we have to be? 

Daniel Gross looks at the forecasting record of the Wall Street Journal and sees the glass as half-empty:
In January, the economists collectively said they thought 3-month Treasury bills would yield 4.63 percent on June 30; they were at 5.00 percent. In January, the economists preicted 10-year Treasury notes would be at 4.9 percent; they were at 5.15 percent; in January, they predicted that 1st quarter GDP would come in at 3.6 percent; it was 5.6 percent; in January, they projected that May's CPI would show a 3.1 percent gain from May 2005; it showed a 4.2 percent gain from May 2005. Again, this is no rap on the economists in the survey--a very smart bunch who have all sorts of interesting and insighftul things to say about the economy. Rather, it's a rap on the notion that we should give a great deal of weight to their short-term forecasts.
When I train students in forecasting, one thing they learn is that when they forecast they will be wrong -- so the best thing to have is a framework for understanding why and when to change the forecast. The GDP miss was wide but likely due to transitory effects of a lowered GDP number in the last quarter. I like Barry's idea of averaging two quarters, though the most recent release of GDP figures makes that average a much more satisfying 3.65%. That's better than expected, so if you were off on the strength of the US economy you'd also be a little low on inflation and interest rates. So all of this paragraph Gross has found comes, to my mind, from a single source: broad-based growth in aggregate demand in the last quarter.

I think the revisions to GDP -- which is what the forecasters are being graded on, not the initial estimates -- qualify as true surprises. I'm not sure how accurate Mr. Gross thinks forecasters should be, but I wouldn't be dismissive of the most recent effort. With the forecasters predicting 3% growth in third quarter GDP, a 3.1% inflation rate, and both short and long run rates around 5.25%, very few predictions are for a recession. If one does hit in early 2007, then you can start calling forecasters really wrong.