March 22, 2011
The unemployment puzzle--or at least one of them
Brad DeLong (hat tip Mark Thoma) looks at the frustratingly slow progress in U.S. labor markets and offers an intriguing hypothesis about the changing nature of American recessions and recovery cycles. First, some relevant framing facts:
Between 1950 and 1990—back in the old days of Federal Reserve inflation-fighting recessions—the unemployment rate in the United States would on average close 32.4% of the gap between its initial value and its natural rate over the course of a year. If the United States unemployment rate had started to follow such a path after its fall 2009 peak, we would now have an unemployment rate of 8.3% instead of 8.9%.
"But there is the unfortunate fact that none of the United States net reduction in the unemployment rate over the past year comes from increases in the employment-to-population ratio, and all of it comes from decreases in labor force participation…
"…[W]e note that between 1950 and 1990—back in the old days of Federal Reserve inflation-fighting recessions—the employment to population rate in the United States would on average rise an extra 0.227 in a year for each year that the unemployment rate was above its natural rate. If the United States employment-to-population ratio had started to follow such a path after its fall 2009 peak, we would now have an employment-to-population ratio of 59.7% instead of 58.4%. It would indeed be morning in America, instead of the current economic state."
The reference to "the old days of Federal Reserve inflation-fighting" is the key to DeLong’s story of how recent recessions, and their aftermath, differ from most of our postwar experience:
"The obvious hypothesis for why this current recovery—and the last two recoveries—in the United States have proceeded at a sub-par pace is that the speed of recovery is linked to the causes of the downturn. A pre-1990 recession was triggered by a Federal Reserve decision that it was time to switch policy from business-as-usual to inflation-fighting. The Federal Reserve would then cause a liquidity squeeze and so distort the constellation of asset prices to make much construction, sizable amounts of other investment, and some consumption goods unaffordable to demand and hence unprofitable to produce. The resulting excess supply of goods, services, and labor would lead inflation to fall.
"After the Federal Reserve had achieved its inflation-fighting goal, however, it would end the liquidity squeeze. Asset prices and incomes would return to normal. And all the lines of business that had been profitable before the downturn would be profitable once again. From an entrepreneurial standpoint, therefore, the problems of recovery were straightforward: simply pick up where you left off and do what you had used to do.
"After the most recent downturn, however—and to a lesser extent after its two predecessors—things have been different. The downturn was not caused by a liquidity squeeze. The Federal Reserve cannot wave is [its] wand and return asset prices to their pre-downturn configuration. The entrepreneurial problems of recovery are much more complex: not to recall what it used to be profitable to produce but rather to figure out what new things it will be profitable to produce in the future."
I am sympathetic to this view as the dynamics of employment recoveries do seem much different in the post-1990 era than in the pre-1990 era. To provide yet another example, on average it took 10 months to recover all the jobs lost during the recessions of the period between 1950 and 1989. In contrast, it took 23 months to recover the jobs lost following the 1990–91 recession and 38 months following the 2001 recession. Right now we are 20 months from the official end of our most recent contraction and still almost 5.5 percent below the pre-recession employment peak. (A stark graphical reminder from Calculated Risk is featured at Economics Roundtable.)
What's more, the mechanism DeLong appeals to rings true. That is, prior to the 1980s, downturns in the economy were dominated by intentional tightening by the Federal Reserve to contain inflation. In the post-Volcker era, however, structural shocks appear to be the dominant story.
But the deeper one digs into the labor market facts, the deeper the questions become. Let me offer up an old macroblog favorite, the Beveridge curve. We have noted in the past that the Beveridge curve—which measures the relationship between the unemployment rate and job openings created by businesses (or vacancies)—seems to have shifted outward over the course of this recovery. In words, relative to the jobs available, unemployment is higher than we would have predicted based on the vacancy/unemployment relationship that prevailed in the 10 years prior to last year.
My view has been that this shifting of the Beveridge curve relationship represents structural issues in the U.S. labor market. A counterargument has been that such shifting is typical of the early phases of a recovery. And if the 1990-91 and 2001 recessions are the best analogs to the current cycle, we might well expect the relationship between job availability and unemployment to return to the prerecession norm as the efficiency of the job-matching process recovers along with other aspects of the economy. In fact, we might even expect the job-matching pace to improve over time as illustrated by the following chart that plots the time path of job openings and the unemployment rate (with the 1990–2001 and 2001–07 periods highlighted in green and red respectively):
The fact that all the points highlighted in green lie to the left of the points highlighted in red means that, for a given number of job openings, unemployment rates were lower during most of the past decade than they were in the 1990s. And there is certainly no evidence that outward shifts in the Beveridge curve are permanent.
Here, though, is where the search for unified business cycle theories gets a little tough. The return of the Beveridge curve to its prerecession norm is a distinctly post-1980 phenomenon. Excepting the 1960–61 episode, the recessions in the 1950–80 era are consistently associated with seemingly permanent rightward shifts in the Beveridge curve:
On the other hand, the aftermath of the inflation-fighting 1981–82 recession represented a clear shift, as the vacancy/unemployment relationship improved dramatically (albeit slowly):
How can we explain, then, that the supposedly demand-driven recessions of the pre-1990 era were associated with seemingly structural changes in the Beveridge curve relationship (in the "wrong" direction prior to 1980 episodes, in the "right" direction after the inflation-wringing contraction of 1981–82).
And what about this time around? Here's where we stand:
Is the recent shift in the Beveridge curve here to stay, or transitory as appears to have been the case in the last two recessions?
Can I get back to you on that?
By Dave Altig
Senior vice president and research director at the Atlanta Fed
An aside: A warm welcome to our new blogging brethren at the New York Fed's Liberty Street Economics.
Update: If you prefer your Beveridge curve history in movie form, Roger Farmer's your man.
TrackBack URL for this entry:
Listed below are links to blogs that reference The unemployment puzzle--or at least one of them: