Why do some jobs disappear forever after recessions?

During the financial crisis of 2008, employment fell dramatically, as was expected. But in the economic recovery that followed, only certain jobs bounced back. A research paper by Wharton finance professor Nikolai Roussanov looks at this phenomenon and correlates it with technological adoption by companies during a down economy.

The paper, whose co-authors are Cornell University professor Mathieu Taschereau-Dumouchel and Wharton doctoral student Alex Kopytov, is titled “Short-Run Pain, Long-Run Gain? Recessions and Technological Transformation.” Roussanov recently sat down with Knowledge@Wharton to explain the scope and goal of the research as well as what they found.

An edited version of the conversation follows.

Knowledge@Wharton: Can you tell us about your research?

Nikolai Roussanov: In this paper we examine the role of technological transformation in the changing composition of the U.S. labor force over the last 20 years, but also focusing in particular on what happened during the Great Recession and in the years subsequently.

We all know that employment fell dramatically during the Great Recession. It was also slow to come back up following the recession, during the recovery. This, of course, is a well-known fact. What is perhaps less well understood is why the jobs that did return were primarily the higher-skill jobs, or higher-skill ‘cognitive’ jobs — meaning jobs that require a fairly high level of education and working with information technology or working creatively, and so on. It did not really help lower-skill, so-called ‘routine’ jobs — jobs on the factory assembly line, for example — recover as much.

What we wanted to understand is what is the role of the Great Recession itself in this job polarization. What we were trying to do was combine [that with] the logic of long-term technological transformation. We know that improvements in information technology, automation, robotics have shifted the demand for labor from low-skill, routine work to more higher-skill occupations. What is not very well understood is what is the role of recessions and economic downturns in driving this process.

Knowledge@Wharton: What did you find?

Roussanov: In combining this logic of long-term technological transformation with a fairly standard business cycle analysis, we find that, indeed, the model that is aimed at matching the long-term trend in the decline of the share of routine workers, and of long-term job polarization, can also account for the magnitude of this accelerated job polarization during the recession, and in particular during the Great Recession.

Knowledge@Wharton: You talk about this job polarization in the economy. Should we be concerned about that? Or is it just another step forward as the economy transitions to another technology?

Roussanov: This is a very important question. What is job polarization? So far I’ve talked about low-skill jobs disappearing, and higher-skill jobs increasing their share of total employment. When people talk about “job polarization,” what they really mean is the disappearance of the middle-class job, the hollowing-out of the middle of the wage distribution — so the jobs that traditionally were the mainstay of America, such as the factory assembly line jobs, the machine operator jobs. If we go back to the 1980s, let’s say, we had a lot of typists. And then computers came about, and those jobs disappeared.

What is happening is that, of course, there is a growing demand for high-skilled, educated workers, but also an increasing share of employment is attributable to lower-skill but non-routine workers, such as workers who cannot be easily replaced by machines or computers — for example, hairdressers, janitors and so on.

When we talk about “polarization,” this is exactly what we mean — the distribution of wages becomes more polarized. We have very high-skill workers, and we have very low-skill workers, but this middle is, in some sense, disappearing. Why this is important, and whether this is a good thing or a bad thing, it’s not an easy question to answer.

In our model — the way we set it up — this is a natural process. When new technology comes about, this technology is much more skill-intensive. The natural thing that happens is that there’s demand for high-skill workers, so the workers should go and get trained, become high-skill, acquire these skills that are in demand — and everything is how it should be.

The question is whether this is actually happening in reality to the same extent that our somewhat stylized model would predict. A lot of the concern that people have, both in academic circles and in policy circles, is that some of the skill transformation and skill acquisition is actually not happening.

We have declines in labor force participation that are not fully explained by people going back to school, say, to acquire skills, and then coming back. This is what our models say should be happening. Of course we know in reality that post-secondary enrollment did jump in the Great Recession, exactly when our model predicts that should have happened. But we also know there’s a lot of people who exited the labor force and did not go to get trained and acquire new skills. That is, of course, a source of concern because the longer these people stay out of the labor force, the harder it is going to be for them to re-enter, if they’re going to be able to at all.

Knowledge@Wharton: There are two things that I found quite interesting in your paper that I want to highlight for a minute. One is the finding that 88% of job losses in the so-called “routine” occupations — such as bank tellers, manufacturing plant jobs, and office clerks — happened during economic downturns, and this is a trend that has been going on since the mid-1980s. Interestingly, this was also around the same time when innovation and automation started to pick up. These two seem to be correlated. Are they?

Roussanov: This is exactly the main empirical fact that our model aims to explain or at least understand. We were not the ones who documented this fact, but this has become an important piece of information for macroeconomists to wrap our heads around — the fact that this job polarization process seems to be primarily happening during relatively short periods of time, which are recessions.

This phenomenon is really recent — the last 20 to 30 years — and it was particularly pronounced in the Great Recession, which of course was the largest recession in quite a long time. But also to some extent, it showed up in recessions in 1990-1991, and 2001. This is exactly at the heart of our approach, trying to understand how this long-run trend — and the decline of routine jobs and the polarization of jobs — is related to the cyclical fluctuations in the macroeconomy.

Typically macroeconomists study those two things — long-run trends and the cyclical fluctuations — quite separately. Our effort and our innovation is to bring these two things together. The logic of our model is fairly simple. What we argue is that for a business to take advantage of new technology is costly, given that it is operating in the old, less IT or automation-intensive technology.

You have to potentially stop production. You have to hire some consultants and other external workforce to help restructure the business. You have to invest in new equipment, and so on and so forth. So when are you likely to undertake that transformation? Well, more likely you’re going to do it when demand for your product is relatively low, and you don’t need to have all these workers producing. You can lay a lot of them off and restructure your production, in the hopes of the economy recovering.

So it’s natural that this process of restructuring from the old technology to the new technology will happen during economic downturns. And this is not by itself a new idea — and not necessarily our idea. What is important, though, in our approach is the understanding that when the newly restructured businesses actually come online, they will also need high-skill workers and fewer lower-skill workers.

So what are the workers to do — those who were laid off during this restructuring process? Well, our model says that they should be using this time to go and acquire skills. Obviously acquiring skills — ‘human capital investment’ as we call it in economics lingo — takes time, as well as being potentially somewhat costly. But the value of this time, the opportunity cost of studying, is much lower during a recession, when jobs are scarce and wages are low. So it’s a natural time for workers to go and acquire skills.

And as I said, we do observe that post-secondary enrollment, for example, spikes during recessions, and in particular this happened during the 2008 recession, so to some extent we see empirical evidence of this mechanism of our model operating, but there are also a lot of the workers who did get laid off who did not use this time to acquire new skills. This is outside of our model, and it is an important question to understand why.

Knowledge@Wharton: How do you apply your findings to the practical world? What does it mean to workers, to companies, to public policy?

Roussanov: This is, of course, important for all of those constituencies. Our model assumes that everybody is doing ‘the right thing,’ and that the firms know that a recession is the optimal time to restructure. Of course, it could be that if a recession is accompanied by tightened access to capital, they may find it difficult to do so. Although we do see that — for example, shipment of industrial robots spikes up sharply in the aftermath of the Great Recession — it is consistent with this idea that firms are transitioning from the old technology to the new technology. And there is also evidence on employment, that firms that shed workers during the Great Recession were hiring workers that were higher-skilled subsequently.

So it’s probably the case that most firms know that this is the right thing to do and are already doing the right thing. The question is, do all the workers realize that when a recession hits, it is no longer a completely transitory phenomenon in terms of their job loss? The fact that if they do not acquire new skills while they are unemployed or even exit the labor force — temporarily, hopefully — [for training] they will potentially become unemployable subsequently when the economy recovers.

This is also an important lesson for policy-makers who may want to direct attention at the problem of how do we get people the skills that they need to operate in this new economy, and also help those who are perhaps too old or are unable to acquire those skills and are maybe just left behind permanently.

Knowledge@Wharton: Can you tell us what sets your research apart from other work in this area?

Roussanov: A lot of macroeconomic research treats these long-run growth trends very separately, very differently from cyclical fluctuations. We argue that it is actually quite important to bring the two together, because as we have seen in recent decades, there is an important interaction between the long-run trend and the job polarization process and the cyclical fluctuations.

The fact that the Great Recession was a very large, dramatic economic downturn, it actually accelerated this technological transformation. But at the same time, the fact that this recession occurred during a fairly late phase of this technological transformation process also amplified the recession itself. This is actually what our model shows — when recession strikes during a period of active restructuring, by amplifying the firm’s incentives to undertake the restructuring at that time, it also amplifies the drop in output, which means a more dramatic, deeper recession.

Knowledge@Wharton: How will you follow up your research?

Roussanov: An important question that our paper thus far does not address is what happens to workers who are not acquiring new skills after being laid off? What explains that? And what to do about it?

The decline in labor force participation was quite dramatic and cannot be all attributed to schooling — people going to, let’s say college or some sort of training program. And even though today the unemployment rate is very low, we also know that there is a large chunk of the population that is out of the labor force after having exited the labor force during, potentially, the downturn — or soon after. And it’s not clear that this exit from the labor force was completely voluntary in the sense that these people just decided that they didn’t want to work anymore.

Perhaps some of it is driven by the lack of ability to acquire new skills and lack of access, maybe, to the training options. But it also could be that not being employed is not as painful as it used to be. There is evidence by a group of macroeconomists showing that in particular, younger males now spend a lot more time playing video games and watching Netflix and so on. The quality of leisure that they enjoy has increased dramatically, relative to, say, 20 or 30 years ago. So maybe they are not as incentivized as they were before to go out and acquire skills and get these new jobs.