!VAR HAGENDOORN

Politics and Econometrics Don't Really Match

One of my professors used to say that politics and econometrics don't really match, which is in part due to terminology. Politicians want numbers plain and simple. They want to be able to say that next year's economic growth will be 2.7 percent or that the jobless number will fall by 121,000 if their policies are adopted. Econometricians do produce such numbers, but always with a confidence (or uncertainty) interval. Now a "confidence interval" let alone an "uncertainty interval" does not sound very encouraging, especially when, in the case of the economic growth example, it ranges from say 0.7 to 3.7.

I had to think of this when I read the latest US unemployment figures, which were published by the US Labor Department last week. What caught my eye is that the job growth number for September was revised from an originally reported 51,000 to 148,000. That's nearly three times as much!

So remember this when you read jobless data, predictions of economic growth etc. Strictly speaking the prediction itself is only of value when you know the confidence interval. And the US unemployment figures aren't even predictions...