Unemployment and Hours are Very Different Creatures

In a post in 2014 that was fairly scathing about the RBC model, I made the claim that long-run shifts in the unemployment rate are caused by 'animal spirits'. My confidence in making that claim is supported by empirical work which establishes that the unemployment rate and the real value of stock-market wealth are both non-stationary but co-integrated.

In words, neither the unemployment rate, nor the stock market (measured relative to money wages) shows any tendency to return to a fixed number. But a weighted sum of unemployment and the stock market (called a co-integrating vector) DOES return to a fixed number. In a previous post I put it this way: the unemployment rate and the stock market are like two drunks walking down the street tied together by a rope. They can never get too far apart from each other.

Brad Delong thinks that it would be better to assert that long-run shifts in the employment-to-population ratio are caused by 'animal spirits'.  I don't think he is right. 

If by 'employment' we means hours in paid employment, the employment-to-population ratio changes for three reasons. More or less people enter the labor force. More or less people in the labor force find jobs. And those people with jobs vary the number of hours they work. 

 Chart 1: Labor Force Participation and Unemployment.   (c) Source: Prosperity for All, 2016, Oxford University Press page 54.

Chart 1: Labor Force Participation and Unemployment.   (c) Source: Prosperity for All, 2016, Oxford University Press page 54.

Chart 1 shows the unemployment rate in blue plotted on the left axis and the labor force participation rate (LFPR) in red plotted on the right axis. My take on this graph is that most of the movements in the participation rate are secular. They do not vary much during recessions (plotted as the grey shaded areas), relative to their movements over long periods of time. I suspect that most of the low-frequency movements in participation can be explained by demographic changes and sociological factors that caused women to enter the labor force en masse in the 1960s.  Research from the St. Louis Fed supports this claim.

 Chart 2: Hours and Unemployment.  (c) Source: Prosperity for All, 2016, Oxford University Press page 53.

Chart 2: Hours and Unemployment.  (c) Source: Prosperity for All, 2016, Oxford University Press page 53.

The other possible reason for changes in the employment-to-population ratio is that it is caused by changes in hours. Chart 2, which plots average hours per week in the private sector in red, measured on the right axis, and unemployment in blue on the left axis, suggests that hours also do not have much to do with recessions. The main feature of average hours per week is that they have followed a steady secular decline from roughly 39 hours per week in 1964 to 34 hours per week in 2016. As American workers became wealthier, they chose to work less. 

In two recent empirical papers, here and here, I studied the connection between the real value of the stock market and the unemployment rate. Charts 1 and 2 explain why I chose the unemployment rate, rather than the employment-to-population ratio, as my object of study.  

You can read much more about this research in my book Prosperity for All which is available NOW on kindle and can be purchased as a real old-fashioned print book from OUP here (coming to amazon very soon).