The Atlanta Fed's macroblog provides commentary and analysis on economic topics including monetary policy, macroeconomic developments, inflation, labor economics, and financial issues.

Authors for macroblog are Dave Altig, John Robertson, and other Atlanta Fed economists and researchers.

« What makes forecasting tough | Main | Optimism…pessimism…and a bit of perspective »

August 18, 2010

Just how curious is that Beveridge curve?

A few weeks back I made note of the following:

"Since the second quarter of last year, the unemployment rate has far exceeded the level that would be predicted by the average correlation between unemployment and job vacancies over the past decade."

The focal point of that comment was the so-called Beveridge curve described by the Cleveland Fed's Murat Tasci and John Lindner as follows:

"The Beveridge curve is an empirical relationship between job openings (vacancies) and unemployment. It serves as a simple representation of how efficient labor markets are in terms of matching unemployed workers to available job openings in the aggregate economy."

Since my last post, the U.S. Bureau of Labor Statistics (BLS) published the June edition of its Job Openings and Labor Turnover Survey (JOLTS). Just as not much changed in June relative to May, either with respect to job openings or the unemployment rate, not much changed with the Beveridge curve:


(A monthly version of this picture can be found in the JOLTS graphs and highlights published on the BLS Web site.)

One of the observations made in my previous post was that the apparent shifting of the Beveridge curve—in other words, the observation that given recent experience the number of unemployed individuals seems high relative to the number of available jobs—might be explained by extended unemployment benefits, but only if you are willing to accept estimates of the policy's impact that are on the high end. I referenced a few Federal Reserve papers—here and here—but they only included estimates on the lower end. Several people have asked (in the comments section of my earlier post and in private e-mails) where the higher-end estimates come from. One of these is from an article titled "The Economic Effects of Unemployment Insurance" by Shigeru Fujita, which is forthcoming (but not yet published) in the Philadelphia Fed's Business Review. (Shigeru estimates that extended unemployment benefits raise the unemployment rate by 1.5 percentage points, enough to explain the lion's share of the Beveridge curve shift.)

Tasci and Lindner, in the article mentioned earlier, offer up a few other observations. First, in the last several months labor market statistics have in general been distorted by the entry and exit of significant numbers of temporary Census workers. Second, it does appear to be the case that the current rise in the unemployment relative to job openings is just a standard characteristic of the early phases of a recovery. On this point they provide this chart …


… along with this explanation:

"One important observation is that a longer-term look at the Beveridge curve shows that the dynamics we have seen recently are not an exception, but are common during the recovery phase of business cycles. As the economy starts improving, it takes time to deplete unemployment, even though job openings are relatively quick to adjust.

"Hence, cyclical changes may not necessarily present themselves as… a neat movement along the curve. During and after recessions in the postwar period, the Beveridge curve has generally followed a pattern of shifting to the right during a recovery. One potential reason for this could be that even though some unemployed workers start filling the available job openings, workers who had left the labor force might get encouraged by the recovery and start looking for a job, thereby keeping the unemployment high. While the Census may have skewed the data for this recovery, the path of the curve going forward looks poised to follow in the footsteps of previous recessionary periods."

Those sentiments have been echoed more informally by Robert Waldmann, by Andy Harless, and at Free Exchange. And they may prove to be exactly right. But as Tasci and Lindner conclude:

"Firm conclusions will only be able to be drawn as more data are generated."

By Dave Altig, senior vice president and research director at the Atlanta Fed

August 18, 2010 in Business Cycles , Labor Markets | Permalink


TrackBack URL for this entry:

Listed below are links to blogs that reference Just how curious is that Beveridge curve? :


oh yes! "the economy starts improving, it takes time to deplete unemployment" however, no one know how much time it will take to overcome that.

Posted by: Crude John | August 18, 2010 at 11:11 PM

Does this have anything to do with structural unemployment?
Or the argument here is this seeming anomaly is entirely related to extended benefits ?
The latter seems a bit optimistic to me

Posted by: C Jones | August 19, 2010 at 09:20 AM

With all due respect your observation is pure BS. The idea that unemployment benefits lead to higher unemployment statistics is searching for an anomaly where none exist.

I've been unemployed for 2 years now, I owned my own business so I receive no unemployment benefits and am not counted in the unemployment surveys. Maybe I'm counted in the U6 numbers but not sure how they would capture my data or categorize me.

In plain English, the unemployment statistics, like most current government statistics, are woefully inadequate to capture the real employment picture in the USA.

Looking for a statistical anomaly in the unemployment data is like checking a sandbag next to the breach in the levee to see if it's adding to the problem.

Posted by: OrganicGeorge | August 19, 2010 at 10:59 AM

Mr. Altig, with all due respect, I am pretty sure all these observations by experts and highly intelectual people are a bit too narrow.

"It serves as a simple representation of how efficient labor markets are..."

Right there is the fallacy. There is no such a thing as "efficient labor markets". I know at least one case of job demand for a wind turbine producer who hasn't been able to cover his needs regarding his workforce because in the area there are no individuals with the right specialized skills. Similarly, I know of another individual who welds special steel pipes for x-ray machines and can't find a job in Miami, even when there are probably no more than 3 others that can do such job in all South Florida. Plenty of work for him in New York, though.

So the point is, jobs are localized in the vast majority of cases and that makes for an inefficient market. It is not like buyers and sellers are all in the same market and have all the information available. The mismatch is compounded now because a large number of jobs lost are of the lower skills type (say retail service) all the while the economy is moving towards a revival of manufacturing with strength in exports. Now, how long will it take to add new specialized skills to the unemployed before we may start to see a match between those have no jobs and those who are offering one?

In the long run, everything becomes efficient by force of nature. But we need to eat today and everyday.

Posted by: Boy Plunger | August 24, 2010 at 01:37 AM

Firms have also pared back on employees over the last 2 years and reorganized for efficiency. Those that are working find themselves doing the work of more than 1 person. Those with hiring authority are in no different a position. Sure, they may have job openings. But, they will only hire those that have done the exact same job recently, so as to add business value quickly. Managers today do not have adequate time to mentor a new employee. Skills that two years ago were close enough are now not enough. This may explain part of the Beveridge curve variation.

Posted by: CD | August 24, 2010 at 01:43 PM

It's hard to accept Economists refuse to tie inflation measurements to population samplings & will not explain how we are to continually grow GDP from a number reached only by years of leveraged gambling.

Posted by: bailey | August 25, 2010 at 10:58 AM

You say:

"Second, it does appear to be the case that the current rise in the unemployment relative to job openings is just a standard characteristic of the early phases of a recovery."

This is true and is indeed fairly obvious when one thinks about it. But that doesn't mean that the BC hasn't shifted out more than in past recessions.

One way to measure this is by estimating a Cobb-Douglas matching function and then looking at the Solow residual to get the matching function's "productivity." Stephen Williamson provides a graph of this residual for 2001 - 2010:


He doesn't show earlier recessions, but the shift for the current recession certainly looks much larger than the 2001 recession.

Posted by: Jon | August 29, 2010 at 09:46 PM

Post a comment

Comments are moderated and will not appear until the moderator has approved them.

If you have a TypeKey or TypePad account, please Sign in