Why can’t economists predict for *beep*?

Why can’t economists predict for *beep*?

“When will economists just admit that they can’t predict for *beep*?”

When Özyeğin University marketing professor Steven H. Seggie asked this rhetorical question in a tweet after the U.S. jobs report on Sept. 5, I correctly predicted Turkey’s July industrial production figures in my Sept. 8 column to prove him wrong. I have to admit I cheated. Anyone familiar with Turkish data would have expected a negative annual growth: There were simply three less working days in July 2014 compared to the same month of the previous year.

In fact, I would agree with Steven that economists can’t predict for *beep*, for the simple reason that forecasting is very hard: The economy is a complex beast, and economists are not in agreement over the economic or statistical models that produce the variables such as growth, inflation or asset prices that they try to forecast. Moreover, most of these variables, as well as other data used in their models, will often not have been measured correctly.

But I could help you choose a good forecaster. For starters, forecasting is as much science as it is art. For example, British economist David Hendry works on methods for modeling unanticipated changes to some parameters of the forecaster’s model - what economists call structural breaks. A good forecaster should be familiar with these new approaches and applying them.

There are also incentive problems. The average forecaster will follow the herd: She doesn’t want to look like an idiot if everyone else gets it right, and if the consensus is wrong, she can argue that no one got it right. During my stint as a market economist, I felt this pressure whenever one of my forecasts was way off. I once took the risk and made a risky inflation call, ending up handing our bond trader a neat profit. A good forecaster should not need to herd.

Again judging from my own experience, you should prefer forecasters who don’t have to provide a forecast for every data out there and who are not under a strict deadline to do so. There were times when I just threw numbers in for the Central Bank’s monthly expectations survey; I just didn’t have the time to predict all the two dozen variables in the survey.

The Financial Times’ Tim Harford recently summarized the results of psychology research that identified the common traits of “superforecasters,” i.e. people who predict the future better than others.

Learning from past mistakes, working in teams and adopting a skeptical approach, as well as knowing a bit of probability, seem to go a long way.

But the best forecaster is the one who can put her money where her mouth is. Market economists can lose their jobs if they are consistently off the mark, but as I explained above, that gives incentives to herd, not to provide great forecasts. If banks required their economists to take positions on their forecasts, or based bonuses on the accuracy of their predictions, we would definitely see better forecasts.

It just occurred to me that I satisfy most of these conditions. Maybe, I should start a forecasting consultancy. I could call it TIC Forecasting: Talk is Cheap, our forecasts are not!