Rauchway knows this but wants to measure unemployment using an alternative series which shows a lower unemployment rate in 1938 (12.5%). Nothing wrong with that but there’s no reason to call people who use the official series liars.
The problem is, Tabarrok quotes the old, bicentennial edition of HSUS. And in that edition, you find Lebergott’s unemployment data, which was assembled before Michael Darby, Robert Margo, and David Weir’s work. As Tabarrok should know, and as readers of this site do know, in the current edition of HSUS, when you look for an unemployment time series, you find Weir’s.
Rauchway thinks that counting people on work-relief as unemployed is a right-wing plot. If so, it is a right-wing plot that exists to this day because people who are on workfare, the modern version of work relief, are also counted as unemployed.
Here is what Weir says about that:
For 1931 to 1943, I accept Lebergott’s employment estimates as accurate, except for a major conceptual conflict regarding the classification of federal emergency relief workers. Darby challenged the standard classification followed by the census, the CPS, and Lebergott that counted such workers unemployed. Lebergott has argued eloquently that counting them as unemployed is a more accurate depiction of the failure of the private economy to generate unemployment. Margo has found that the labor supply behavior of relief workers shared some characteristics of both employed and unemployed workers, and suggests that at least some should probably be classified as employed. In the absence of a clear basis for distinguishing employed from unemployed relief workers, I agree with Darby that counting all relief workers as employed is more consistent with modern theoretical interpretations of unemployment, so I include them as government workers.
It’s true, HSUS should put Weir’s series, which is in its current edition, on the web instead of the old series.
But I believe all this means that I am right, and Tabarrok is wrong.