Reassessing the Beveridge Curve “Shift” Four Years Later
Early on in the current recovery, economists and policymakers were worried about a potential shift in the Beveridge curve—an empirical relationship between job openings and unemployment that is viewed as a measure of the efficiency with which the labor market is matching unemployed workers to the available openings. The data at the time suggested to some that a shift was occurring that would indicate that efficiency was falling: Job vacancies were rising, but the unemployment rate was not declining, fueling a debate about a structural problem in the labor markets.
Exactly four years ago, we touched upon this issue here, and argued that it was too early to call what had happened a shift. Most of the arguments for a shift were based on data from the Job Openings and Labor Turnover Survey (JOLTS), which had only started measuring economy-wide job openings in December 2000. Examining an alternative data source that went back to the 1950s, we concluded that the Beveridge curve behavior we were seeing in this recovery was typical of recoveries in general, and that the curve would likely follow its historical business-cycle pattern going forward, erasing any evidence of a shift.
Well, four years later, we have 16 more quarterly data points to inform us. The figure below plots the evolution of the Beveridge curve over the recovery so far, along with three prior business cycles. It is safe to say that what seemed like a shift in the Beveridge curve ended up being another manifestation of the “normal” dynamics of unemployment and vacancies in the United States. The Great Recession that hit the global economy in 2008-09 caused a significant rise in unemployment everywhere, including the US. Massive job losses in a short period of time along with drastically low labor demand left us with an unemployment rate of 10 percent at the peak in October 2009. Since then, however, unemployment has declined to 6.2 percent.
Over the same period, firms gradually adjusted to the recovery and raised their demand for labor as well. We have two alternative sources of job openings for the U.S. over this period, one from the JOLTS and one from the Conference Board Help-Wanted Online Index (HWOL), and both show this is the case. Since July 2010, job openings increased by 59 percent according to JOLTS and 42 percent according to the HWOL. More importantly, improvements in the job openings data were more pronounced—especially for the HWOL data—early in the recovery relative to the improvement in the unemployment rate. This difference in timing is behind the counter-clockwise loop in the Beveridge curve.
Observers have followed the Beveridge curve during the recession and the recovery to glean some insight into potential structural changes in the labor market. Whether or not a shift implies an actual structural change—specifically, a decline in the matching efficiency of the labor market—is still debatable. However, one thing is clear: there is no shift to begin with. We believe that this debate and the ensuing evidence showed us that inferences about complicated and ill-defined concepts such as structural change in the labor market cannot be made by just looking for a break in the simple (and reduced-form) empirical relationships between macroeconomic aggregates in the midst of a deep and long recession.