Employment, Unemployment, and Hours: A Comment on Farmer's "Unemployment and Hours Are Very Different Creatures"

Roger Farmer recently argued that, as the title of his blog post says, “Unemployment and Hours Are Very Different Creatures.”  I’m not so sure I agree. 

Using data for the 1965-2016 period, he charts the unemployment rate and the labor force participation rate in Chart 1.  The LFPR rises fairly steadily from 1965 until late 1999 and then begins to decline.  The UR, by contrast, rises in recessions and falls in recoveries, with no particular long-term trend.  He suggests—and this seems clear—that there is no relationship between the LFPR and the UR.

In Chart 2, he shows average weekly hours in the private sector (AveWH) an the unemployment rate.  For most of the 1965-2016 period AveWH are declining; again, there is no apparent relationship with changes in the UR.

The chart below indexes the employment-population ratio and (1 minus the unemployment rate—a 7% unemployment rate is therefore 93% here) using 2007 as the base period [I_EPR and I_(1-UR)] in order to  chart them along with an index, not of Average Weekly hours, but of Aggregate Weekly Hours (AggWH), which, on the BLS web site is presented with a base period of 2007 = 100.  (Data on AggWH are available beginning in March 2006).  All three of these are indexed to 2007, and, because I transformed the unemployment rate, we would expect the indexes to rise or fall together—if they are related at all.  I also included an index of Average Weekly Hours (AveWE) worked.

 

Well, they do appear to be related, and fairly closely—except for the index of AveWE.  Why do I find something different from Farmer?  His hours measure is Average Weekly Hours.  If, during downturns, employers respond by laying workers off and maintaining the average work week, we would see Aggregate Weekly hours to fall in recessions and rise in recoveries, even as Average Weekly Hours change little or not at all.  And, based on the 2006-2016 period, laying off workers and maintaining the work week is exactly what happened.  And we would expect the EPR to fall in recessions and rise in recovery, and (1 — UR) to fall in recessions (UR rises) and rise in recoveries (UR falls).

I only have one business cycle here (Farmer has 6 in his data series), so there’s no guarantee that what I have found will be replicated in subsequent cycles (or would appear in prior cycles if I had aggregate hours data).  But I’d be willing to bet on it.