HICKS: What we can learn from forecasting blunders

For those of us with the audacity to publish economic forecasts, it has been a bad couple of years. It may be best to begin with a look in the mirror.

In December 2008, I forecast the U.S. and Indiana economies shrinking throughout the coming year, with job losses and declines in all sectors except health care. I also said the economy would hit bottom in the summer of 2009, but that recovery—especially in the labor market—would be slow. In December 2009, I repeated this exercise, predicting slow growth in everything but the labor force beginning last winter. So how’d I do?

I am pleased to say my timing of the turnaround and the relative decline in each sector appears to have been nearly spot on. Sadly, I did a wretched job of estimating the depth of either jobs or income losses here in Indiana. My estimates on income declines were off by half, and job losses the same. The timing is important, but the depth of the decline is far more critical. It is small comfort that I am not alone. The Presidents Council of Economic Advisors, as fine a group of economists as you’ll find, did worse. They forecasted current unemployment rates to be roughly 3.5 percentage points lower than they are now. So what next?

A New York Times columnist recently wrote about the issue of economic forecasts, arguing that economists would soon do away with their models. That was precisely wrong. Instead, we are going back to the models, keeping the parts that work and remedying the errors.

My forecast contains about 125 equations that rely heavily on the experience of recent history. This recent history predicted a more modest downturn than we had. I’ve already rewritten the models to incorporate a longer memory. I hope to get it better next time, and I am sure I am not alone. But that begs the question, just what should economists be expected to know and how should we explain it?

I think there is one easy principle: honesty. We ought to be honest about what we do and do not know, and how prone to error an estimate might be. This last part is the most difficult to communicate. We could do like weather forecasters and give probabilities, but that doesn’t make public budgeting any easier. Perhaps we shouldn’t say the unemployment rate will be 9.5 percent, when we really mean it will be between 9.1 and 9.9 percent.

We can forecast things like the annual labor force, population and household income for broad regions with great certainty. The smaller the region and the longer the time horizon, the less certain are forecasts. Partly this is because data is less accurate and available for small places, but also small regions suffer more random changes, which are devilishly hard to model. Ironically, the things we wish to know with the most certainty are often the things we can forecast the least well. Just like the weather.•

__________

Hicks is director of the Center for Business and Economic Research at Ball State University. His column appears weekly. He can be reached at cber@bsu.edu.