02 March 2011

Budget Predictions and Bias Correction

The WSJ reports on a study released by the Pew Center on the States on the accuracy of state budget revenue forecasts.  In short -- the forecasts are not so good:
In a report Tuesday, economists at the Pew Center on the States and the Nelson A. Rockefeller Institute of Government conclude that states increasingly overestimate their tax revenue during tough economic times. The report's authors calculate that during the depth of the latest recession in 2009, income forecasts by all 50 states overshot reality by a total $49 billion.
There is a pretty obvious fix here -- the forecasters should continue doing what they have been doing, and decision makers need just apply a bias correction based on the historical performance of the forecasts.  In this case the bias correction could be a function of relevant economic variables that are correlated with past forecast performance.  Such a bias correction would improve the skill of the revenue forecasts.

Forecasts that are predictably wrong can be just as useful as those that are predictably right.  The full report from Pew can be found here.


  1. From the figure it's not clear that they're biased over the long run. The issue is that they miss the cyclical turning points, which is exactly what you expect of trend-extrapolating forecasts.

    There may not be any systematic correction that yields better forecasts, and if there were, no one would believe it anyway. What they really need is a budget policy that's robust to forecast error.

  2. There's also the problem for states that rely on highly progressive income taxes (soak the rich) that receipts of those taxes fall very much faster than sales and property tax revenue when the economy is in decline.