Monday, August 2, 2010

Stop Worrying About Structural Unemployment

The economic blogosphere has suddenly become very concerned about the possibility that structural unemployment – resulting from a mismatch between the needs of employers and the capabilities of available job-seekers – has increased in the US. Paul Krugman is worried; Brad Delong is convinced; it’s obvious to Tyler Cowen; The Economist presents a variety of opinions; and any number of other bloggers and fora have been discussing the topic.

One major source of this newfound concern is a post by Dave Altig of the Atlanta Fed, who has detected a shift in the relationship between job openings and unemployment – the Beveridge curve. While the shift is unmistakable in his chart (see below), I have looked more closely at the data, and I have come to the conclusion that it does not represent a major increase in structural unemployment. Rather, I believe it represents the normal dynamics of the business cycle in the context of an incipient recovery from a historically severe recession that, in some ways, has not quite ended.

First, let’s get a clear idea of what’s going on in the chart:

Consistent with a common practice by people (including me) who plot economic data, Dave Altig has drawn a linear regression line to represent the general relationship between job openings and unemployment. But we should not therefore assume that the true relationship is a linear one. If you ignore the regression line, you can see a distinctly curved pattern to the points. We should expect a curved pattern: a strictly linear relationship wouldn’t make sense, because it would mean that, if there were enough job openings, unemployment could go below zero, and if there were enough unemployment, job openings could go below zero. In practice, when one of the series gets very low, it becomes less responsive to the other series. Thus the pattern in 2009, where the unemployment rate keeps rising while job openings become nearly flat at a very low level, is exactly what one might expect. It’s certainly what I expected, having plotted curves like this before.

But the point labeled “2010 Q2” breaks the pattern. It appears that we’ve suddenly moved off the old Beveridge curve onto a new one that has yet to be traced and that promises to associate a significantly greater amount of unemployment with any given number of job openings. But have we, really?

To answer this question, we need to think about how job openings (as well as other factors) affect the number of unemployed workers. Take a look at the actual numbers: in 2010 Q2 there were about 15 million unemployed workers and just over 3 million job openings. If all 3 million job openings were filled, it would (other things equal) reduce unemployment to about 12 million. But if you were to plot that hypothetical point on the chart, it would still be above the old Beveridge curve. So even with what seems a rather optimistic assumption about the matching process, it was inevitable, given the appearance of a comparatively large number of job openings in Q2, that they produced a point that was off the old curve. That result has nothing to do with structural unemployment; it’s just because there are many more available workers than openings.

My assumption is not really as optimistic as it seems, though, because in fact job openings fill very quickly. The May (most recent) JOLTS report, for example, shows 3.2 million job openings but 4.5 million new hires – which implies that the average job opening gets filled in less than a month. What about all those employers complaining that they can’t find people with the right qualifications? Apparently they are a minority – or else they end up settling.

If we’re looking for evidence of an increase in structural unemployment, we need to compare the rate at which openings fill today to the rate at which they filled in the past. When was the last time that there were this many job openings? In November 2008, there were 3.2 million openings but only 4.1 million hires. So job openings are filling faster now than then. You might expect them to fill faster, since there are more unemployed people with whom to fill them (15 million vs. 11 million). Indeed, the fact that they fill only a little bit faster could be taken as evidence that some of the additional unemployment is structural. But these data don’t support the idea that there has been a dramatic shift, that the pool of the unemployed is a significantly worse match for the available job opportunities than it was a few years ago. To find a point where actual hiring was happening as quickly as it is today, you have to go back to August 2008, before the fall of Lehman, when there were 3.7 million job openings.

So if 4.5 million people (equivalent to 30% of the unemployed) find jobs in a given month, how come so many people are still unemployed? Because people are losing jobs almost as quickly. That’s what I meant when I said that the recession, in some ways, has not quite ended. While the average rate of job losses during the recent recession was not particularly severe, those job losses continued for a long time (as it was a long recession) and pushed more and more people into unemployment, while there was an unusual lack of new jobs to get them out of unemployment. The new jobs are finally starting to appear, but the job losses are continuing. When I declared last year that “job losses are not the problem,” it hadn’t occurred to me how long the job losses might last. By the standards of a recovery, job losses are the problem today.

Well, part of the problem. Notice that even after the recent jump, there are fewer job openings than there were at any time during the 2001 recession, and only about as many as there were at the depth of the 2003 “job recession” that lingered after the official recession had ended. Whether you measure in terms of job losses or job openings, the job market is still depressed. There’s plenty of reason to expect persistent cyclical unemployment. Structural unemployment, not so much.

DISCLOSURE: Through my investment and management role in a Treasury directional pooled investment vehicle and through my role as Chief Economist at Atlantic Asset Management, which generally manages fixed income portfolios for its clients, I have direct or indirect interests in various fixed income instruments, which may be impacted by the issues discussed herein. The views expressed herein are entirely my own opinions and may not represent the views of Atlantic Asset Management.


Anonymous said...

So, as a hobby economist, I actually followed this post and can see how silly it is to put a straight line on the graph.

That being said, you didn't mention the effect on the unemployment numbers of people like me, who have been unemployed and underemployed for eight years (since the 2002 tech bust). I spent most of last year looking in vain for work, and finally went back to school to retrain in a completely new field (for the second time in my life).

Where am I in these statistics? I haven't gotten unemployment benefits since 2002. I'm thinking some of the unemployment may be accounted for by the very long term unemployed renewing their search for work and starting to show up in the counts again.

Maybe a way of looking at the problem that could shed more light would be to include the underemployed and the ones who have given up in a "total unemployed" headcount, and see how that stacks up.

Sue Greer

David Pearson said...


Good piece. What do you think about the % of unemployed > 26wks? Also, the majority of job losses in this recession come from the service sector (vs. about 18% in the two previous recessions). Put these two facts together and they depict an economy losing less-cyclical jobs (service sector), with the unemployed taking much longer to find new jobs. Is this evidence of "structural" or "cyclical" unemployment? I don't know if it is a useful distinction. The data fit the hypothesis of spending and employment trends propped up by a "bubble", with both evaporating once the bubble pops. If you spend what you don't earn and hire whom you don't need, then when you stop, is it cyclical or structural? (BTW, I am referring to the broader credit bubble rather than the narrow housing one). Perhaps a better concept to consider is what "trend growth" looks like absent a bubble or consistent stimulus. The answer would tell us a lot about the nature of unemployment.

MW said...

Andy, what would cause you to start worrying about structural unemployment, i.e. what are the appropriate indicators to watch?

Anonymous said...

Solitary free-lancers occasionally work on speculation, completing specific projects and warehousing them, sometimes for decades, with intent to market when opportunity arises. When a hourly or annual wage opportunity arises, they may set aside their speculative work for a period of time with intent to return to the speculative work when the wage position ends.

These individuals may have no income during the time of their speculative work. Consequently, they may be officially classified as unemployed.

Nevertheless, when their speculative work is purchased, they have restructured the economy with their innovation.

Andy Harless said...


I guess you (during the times you were unemployed and not underemployed or out of the labor force) would be in the “structurally unemployed” category, and would already have been in that category before this year – thus an example in support of my view that structural unemployment has not risen a lot in the last couple of years. Incidentally, I know at least one other person who has been “sort of out of work” since 2002.

In the broader unemployment statistics, there are a lot of “discouraged workers” but not out of proportion with the number of officially unemployed (at least the last time I looked). What is unusual is the number of “part time for economic reasons” (a category in which I proudly count myself). This category might help explain why the number of job openings has been so low (as I said, after the recent jump, only up to where it was at the bottom of the last downswing), since employers who might otherwise have been hiring can just reactivate part-timers instead.


There are several possible ways of defining “structural” vs. “cyclical,” and the implications may be very different. Given my generally Keynesian orientation, I would like to think of cyclical unemployment as unemployment that is associated with downward pressure on nominal wages. So the structurally unemployed would be those who don’t exert downward pressure on wages because they are in categories where nobody is trying to hire (and where the threat of layoff for still-employed workers isn’t effective at keeping wages down, perhaps because there are so few left that their wages are now immaterial to the aggregate level of wages). I think there’s a case to be made that there is no such thing as structural unemployment – that is to say, that the downward pressure, or lack thereof, on wages, is not related to the characteristics of the unemployed. It’s typically argued that longer spells of unemployment are more likely to be structural, because long-term unemployed lose their skills, etc., so aren’t going to be hired. But it’s also the case that people with longer spells are likely to be more desperate and therefore more willing to accept lower wages, so in that respect, they may be less “structural” than others, in the sense in which I want to use the word.


Barring more detailed analysis of subsequent data, I suppose what would make me more concerned about structural unemployment would be if I saw job openings continuing to rise and/or unemployment claims falling without seeing declines in the unemployment rate.

Robert said...

Structural unemployment is real and is here to stay. Even those economists who re beginning to recognize the problem vastly underestimate the ultimate impact. It is not just about a "skill mismatch." It is ultimately going to be about a total lack of demand for skills (and workers). It is something unprecidented and it is caused by advancing technology. And ,no, it is NOT the same thing that has been going on for decades. Technology is moving faster and faster. In the next couple of decades will will see a staggering level of progress.

For an excellent overview of this problem, check out this book (available as a free PDF): The Lights in the Tunnel: Automation, Accelerating Technology and the Economy of the Future. (

If there were a textbook on this issue of technological unemployment and where it will lead, this book is it. I wish every economist would take a break from data analysis and READ THIS BOOK.

The author also has a blog at

yakenzu toby said...

There is lot of information and they are very innovative and informative. I have read the article very well and it seems to me awesome. agen bola, judi bola

ibcbet wap said...

This is often very good. We watch this in turn subject matter so we are astounded. We are attracted to one of these methods. Solitary appreciate going up on your guidelines, and treasure doing with this. Please keep modifying. They are remarkably valua. main bola tangkas