"The average unemployed American has been out of work for almost 40 weeks."
Economist Paul Krugman, in a New York Times column
We turned to the Bureau of Labor Statistics, the federal agency that is the official source of employment statistics.
BLS publishes statistics for "average weeks unemployed" for unemployed workers. For the most recent month available, May 2011, the figure was 39.7 weeks — just as Krugman said.
We should note BLS publishes a second statistic that measures a similar concept — "median weeks unemployed." The terms "median" and "average" refer to different ways of determining what is a typical case.
Statisticians arrive at a "median" by listing all of the numbers by size and then locating the one at the exact midpoint of the list. Half of all examples on the list will be bigger than the median, and half will be smaller. By contrast, an average is determined by adding up all of the numbers on that list and dividing by the total number of items on the list.
As it turns out, the "median weeks unemployed" statistic produces a number that's quite different. The median unemployed worker was out of work for 22 weeks in May 2011 — a little more than half of the average length of unemployment. This result likely happened because there were a modest number of people with very, very long unemployment spells, as well as a much larger number who had relatively short periods of unemployment.
Does this difference matter in analyzing the accuracy of Krugman's statement? Not much.
For starters, Krugman's underlying point — that long-term unemployment is a serious problem right now — remains true regardless of which measure you use. The "average weeks unemployment" figure hit a record high in May, while the "median weeks unemployed" figure hit its fourth-highest level ever.
Also, the difference between average and median weeks unemployed has been more or less steady over the years — pretty close to the same two-to-one ratio we see today. So today's divergence between the two statistics is nothing unusual by historical standards.
Krugman could have used a statistic that's shows a much shorter duration for unemployment. But we think he's on safe ground using the 39.7 weeks figure. We rate his statement True.
Edited for print. For the full version — and to read other rulings — go to PolitiFact.com.