Friday, December 2, 2011

New rule: if you are going to call yourself an economist, you need to know the meaning of a confidence interval

I was listening to NPR on the drive in to work this morning, and a heard a man who was labeled "an economist," say that the job growth numbers were disappointing, because measured job growth in November was 120,000, whereas the consensus forecast for job growth was 130,000.

The Bureau of Labor Statistics puts the 90 percent confidence interval of the monthly job growth estimate for the estalishment survey at 100,000.  This means the standard error of the estimate is about 56,000, so the difference between the BLS estimate and the consensus forecast number was less than .2 standard errors, which is essentially zero.  I suppose Mr. Economist would be happy had the number come in at 140,000, which would have been above expectations.

One needn't have a Ph.D. in economics to understand confidence intervals--one undergraduate course in statatistics will do the trick.  Yet week after week, I hear people who call themselves economists yammering on about small movements in numbers that are as likely noise as anything else.