Advertisements

A Disjoint Labor Market and the Two Recessions

Peter Orszag, Barack Obama’s former OMB director, argues that structural factors – like labor market mismatch, globalization, and automation – are unlikely to explain much of the recent shift the the Beveridge Curve. That is, we have a historically low hiring rate given the number of job openings and unemployment.

While it would be foolish to discount cyclical dynamics, we ought to pay heed to certain structural inconveniences. Some economists like to classify recessions into letter shapes including: “V” (rapid fall, rapid recovery), “U” (fall, stagnate, recover), “W” (double-dip and recover), “L” (fall and stagnate). Europe, and much of the United States (with apologies to North Dakota) are somewhere in between the latter, crappier, two. Measured by income or unemployment, at least.

But job openings tell a fundamentally different story:

Image

What we see is an (approximately) L-shaped recovery for hires and a very V-shaped recovery for openings. Openings are showing a rapid recovery because they fell twice as hard. (So it’s unfair for Orszag to claim that openings are recovering much faster without noting the respective falls). While this data set does not go far back, we can see this tension of shape only in this recession. Previously, as theory and common sense would expect, both followed a similar path, if to different magnitudes. This time, one might say, there are two different recessions. (And as I tackle in a later post, openings should map very well to aggregate demand as a whole, suggesting a much stronger demand-side recovery than many suggest.)

This suggests the skill mismatch thesis has more support than Orszag believes. He makes an interesting point: the opening-hire ratio is abnormally high even for Retail Trade, which isn’t a particularly skill intensive sector. While that’s a fair characterization, retail is also a remarkably noisy data set:

Image

Now, if we wanted to draw any inference, there clearly is a negative trend since 2011. That’s important because this deterioration doesn’t go as far back as 2008 – or even 2009 – as would be suggested by a cyclical downturn. This is an important distinction from ferment in the labor market as a whole. It’s entirely fair to argue that this inference derives from too-noisy data: but it’s the only inference one can make. We certainly may not conclude that the opening-hire ratio in retail trade speaks against mismatch theories as a whole.

For example, there seems to be no problems for workers in “Food and Accommodative Services” another historically low-skill, low-wage market:

Image

In fact, we should note that not only the magnitude, but also the shape, of hirings and openings are in harmony – in 2001 as well as 2009. This tells us something has changed between then and now in the market as a whole that has not changed in the minimum wage market. A best guess might be skills.

Research and common sense suggests workers at the bottom crust of our labor market are slightly more persistent against globalization and automation than those in the middle. Technology and globalization tend to hollow out jobs open to middle-class, blue-collar builders and factory workers: jobs that made the 20th century American. On the other hand, it’s markedly more difficult to outsource your fry cook to China.

As I’ve noted before, many of the problems caused by sticky wages have likely evaporated. Aggregate demand is well short of where it should be, without front-loaded spending cuts, but it’s becoming harder to attribute as a primary cause. Paul Krugman recently cited evidence that industries that were hit the hardest are recovering the quickest, which is great evidence in favor of the cyclical downturn hypothesis.

However, while it’s certainly clear that there was an incredible aggregate demand shortfall between 2008 and 2012, it’s harder to argue that this deficit continues today. Let me be clear: I think smart stimulus today – the type that boosts supply, demand, and low-income welfare – is a good policy today. However, the fact that many of the less desirable shifts to which Orszag refers begin years after the trough informs me that structural adjustment may be an important component of the recovery.

Some on the right, at this point, throw their hands up and claim “we’ve done what we can”. I’m not so complacent. America lost jobs that aren’t coming back. A recession only made it easier to plow through the industrial hurdles of such a change. This is precisely the time to offer high-quality, technical education through free two-year colleges targeted at our broad middle class. This is precisely the time to roll back payroll taxes completely. This is precisely the time to expand the earned income credit.

There’s one very important argument in Orszag’s column that deserves more attention: internal recruiting. There’s some more evidence that this might be true:

Image

A healthy labor market “churns” a lot: a lot of people get fired, a lot get hired, and still others quit. But we’re not seeing a recovery in fires, hires, and quits. Fewer people are fired if they can be internally rehired to do something more important. There is an important problem with this argument, that is “layoffs and discharges” are down to their trough in a healthy market. Regardless, the problem with internal job markets is that only the employed benefit: and a rise thereof would suggest painful problems for the long-term unemployed. Economists know employers look for many “signals” that provide information about a candidate (in other words, its not the amazing education you get at Harvard that get you hired). One potentially important signal is employment itself. That’s a big problem.

Orszag notes that “over the past three years, the number of job openings has risen almost 50 percent, but actual hiring has gone up by less than 5 percent. Companies are advertising a lot more jobs, in other words, but not filling them.” That’s fair enough, but it’s somewhat tricky to look at the recovery without scrutinizing the downturn. The real problem is the L-shaped hires graph and V-shaped recovery graph. And if we don’t do something about it soon enough, jobs Americans never had will go overseas.

Advertisements
8 comments
  1. Ryan said:

    I think it’s broadly reasonable to interpret the Beveridge Curve shift as evidence of structural issues. It’s not clear, however, which structural issues are to blame. You suggest “skills” as the best guess, but that deserves some unpacking. Have there been widespread changes in technology that have rendered some skills permanently obsolete? We certainly have anecdotal evidence for this.

    Another alternative is that the reallocation machine is broken. This can be a skills story too. Many measures of reallocation have been in secular decline for a few decades–migration, worker flows, job flows, entry/exit of firms and/or establishments–so if there are geographic, industry, or other sources of skills mismatch, a broken reallocation machine could be partly to blame. We don’t know why it’s not working–it could be policies, or it could be demographics, or it could also be technology. But if it’s broken, we have a problem. Conceptually, reallocation is how we adjust not only to cyclical forces but also to disruptions caused by technology, trade, and immigration.

    There are, of course, other explanations–changes in reservation wages/voluntary unemployment and similar issues that we can’t talk about in polite company. And these are likely related to the issues above. But the point I’m trying to make is that a lot of secular stuff is going on, so the blogosphere’s intense focus on cyclical forces and angry dismissal of any mention of structural forces is overconfident and unproductive.

  2. The author’s discussion and comments as well fail to note that ‘job openings’ is a point-prevalence number while the number of hires is an incidence rate cumulated over the entire month (the number of days between last-month’s and the current-month’s point-prevalence counts). This is why the ‘hires’ curve is usually above the job-openings curve. All speculations about the ‘gap’ between curves and how this gap increases/decreases over time MUST start with recognition of this ‘statistical’ fact.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s