In the November 8 issue of the San Francisco Fed’s Economic Letter, Rob Valletta and Katherine Kuang look at the regional and sectoral behavior of employment and find little evidence for the mismatch employment thesis currently making the rounds. If there were a serious mismatch problem, you’d expect to find major disparities in employment across geography and industry: healthy regions or sectors would show shortages of workers, and sickly ones would show surpluses. But in fact you don’t see that: regional and sectoral variation is little different now from past cycles. Valletta and Kuang conclude that the major reason for persistently high unemployment is cyclical and not structural.
The San Francisco Fed study looks at Beveridge curves. Beveridge curves plot unemployment against job vacancies. In general, the more vacancies there are, the lower the rate of unemployment. Minneapolis Fed President Narayana Kochlerlakota’s argument, outlined in a speech he made over the summer, depended heavily on the idea that the Beveridge curve suggests that the unemployment rate should be about 1.5–2.0 points below where it is, which is what leads him to the mismatch conclusion. His curve, though, was based entirely on only the most recent decade’s data.
But there’s nothing at all stable about this relationship over time. The Valletta–Kuang paper includes a graph showing how much the Beveridge curve has wiggled around over the decades. For example, Valetta and Kuang point out that the curve shifted about 4 points to the right between 1960 and the early 1980s before shifting back about 2.5 points by 1989, and that the variation in the NAIRU during that period was considerably smaller than the movement in the curve would suggest. We thought this worth a closer look.
Unfortunately, the JOLTS data only begins in 2000. Most earlier work on the unemployment–vacancy relationship used the Conference Board’s old Help Wanted (HW) index as a proxy for openings. In an earlier paper, Valletta used the period of overlap between the HW series and the JOLTS data to estimate a consistent vacancy rate series going back to 1960. He graciously shared that with us.
In the graph below you will see a set of Beveridge curves by decade—they show that a major cause of the noisiness in the long-term relation is that the vacancy–unemployment connection shifts over time. The r2’s for the decade regressions are a lot better than for the whole series, with most in the high 0.80s/low 0.90s. But there are also some noisy decades: the r2 for the 1970s is just 0.39. The 1980s aren’t so great either, with an r2 of 0.77.
The labor market is usually thought to be functioning better the further down and to the left (i.e., closer to the origin) the Beveridge curve is on the graph. The further out the curve is, the more friction, like regional or sectoral mismatches. The graph for the 2000s (which is based on data through June 2009, when the relationship prevailing earlier in the decade started breaking down) shows a much better-functioning labor market than in earlier decades, especially the 1980s.
Until, that is, we get to recent history, as shown by the dot marking the data for November 2010. It’s about two unemployment points to the right of where the curve suggests it “should” be. But it’s actually lower than the 1980s curve would predict, and not that much higher than the 1970s curve. And those were decades of major structural change in the U.S. economy—periods of major financial and real sector shocks, far more severe than those of the following two decades.
So the recent breakdown in the Beveridge curve looks to us more like a reflection of a major financial and psychological shock (one that has left employers extremely shy about hiring) than some fresh mismatch in the U.S. labor market. The fresh mismatch theory seems especially odd in light of the fact that things seemed to be functioning so well (meaning the Beveridge curve was close to the origin) for so much of the decade until the economy fell apart. The novelty is the having fallen apart, not some recent change in the labor force.