In Defense Of Overly Optimistic Forecasts

Via Brad Plumer, “A recent paper by Harvard economist Jeffrey Frankel examined the growth predictions made by government agencies across 33 countries and found that the projections were consistently over-optimistic.”

There are some obvious problems with bad forecasting, but I do think it’s worth sticking up a bit for over-optimistic forecasts. The essence of the problem is that government forecasters never predict recessions, and when recessions do happen, they don’t predict long grinding jobless recoveries. They assume, in other words, that macroeconomic management will do its job even though in practice it doesn’t. But I’m not sure what else they should do. When Google Maps estimates driving time from D.C. to Pittsburgh, it doesn’t account for the fact that you might drive recklessly fast or else that you might get pulled over by the cops for driving recklessly. Attempting to build user error into the model would be very confusing, and in the government case I think would constitute a kind of “soft bigotry of low expectations.” It already makes me angry, in fact, when I see government officials point to charts showing that the 1990, 2001, and 2008 recessions were all followed by “jobless recoveries” and then kind of shrug. These are charts about policy failure, and policymakers shouldn’t be planning to fail they should be trying to hit forecasts.

Instead of trying harder to improve our long-run budget forecasts, what we ought to be doing is discounting long-run budget forecasts much more in our evaluation of what belongs on the policy agenda. People in Washington treat it as obvious that a CBO projection of Social Security outlays in the year 2065 is a very serious subject to talk about, but that the legality of banning genetically engineered super-sprinters from Olympic track & field events in the year 2065 is frivolous. But I’m much, much, much more certain that we’ll have genetically engineered sprinters in the future than I am about the future productivity growth rate.