The novel coronavirus pandemic is certainly not good for the labor market. Recent weeks have seen unemployment claims surge to record levels as businesses and entire industries shutter in order to stop the spread of the novel coronavirus. As a result, the economy has plummeted, with the Dow Jones Industrial Average and S&P down more than
from their February highs.
While social distancing measures may be temporary , this economic downturn’s effect on the labor market will have long-lasting effects. In a joint post with his colleagues, Mark Muro, a senior fellow and policy director at the Brookings Institution’s Metropolitan Policy Program, recently wrote, “any coronavirus- related recession is likely to bring about a spike in labor-replacing automation. ”
Economic downturns, he argues, bring about increased levels of automation, which is already an existential threat to many jobs. And a coronavirus recession , due to its breadth and scale, could cause even more automation.
This interview has been lightly edited for clarity and brevity.
It feels counterintuitive to me that automation would spike in an economic downturn, because automation is expensive, at least in the short term. Why do you expect the rate of automation to increase?
It might seem that in hard times, in a downturn, human labor would be cheaper and therefore automation would go down. But in fact, it’s the opposite. What happens is that, because of the crisis of the bottom line and a crash in revenue, humans become relatively more expensive compared to automation.
So a firm that might have been thinking about automating is under a whole lot of pressure to do that, especially in the first two years of a new downturn. And that’s what a lot of research over the last few recessions has shown. People are relatively more expensive, including with their benefits. Meanwhile, you can restructure your business using new technology that increases productivity. So the typical move is to replace less-skilled workers with a fewer number of more-skilled workers or retain higher-skilled workers but then to bring in new technology.
The other thing is that new technology, meaning automation but also enterprise software, is no longer nearly as expensive as it was a decade ago, say around the financial crisis. The cost curve has been declining. We think there’s a lot of new high-quality technology on the shelves waiting for deployment, including a more “turnkey” standard offering, often from the big tech companies. So I think a lot has happened to sharpen what was already a highly visible dynamic in the last turndown, the financial crisis.
And machines don’t get sick or stay home when there’s a pandemic.
Right. Viruses may be transferred from one to another, but it’s not the same kind of virus.
So what exactly has the research shown?
Nir Jaimovich and Henry Siu
(found that, in three recessions over the last (years, percent of job loss took place in routine highly automatable occupations. And that was essentially all of the jobs lost in the crises. So, these crises have historically inordinately been visited on workers whose work actually was automatable.
Another finding by Brad Hershbein and Lisa Kahn of the University of Rochester looked at something like million online job postings and found that firms in the hardest-hit metros were steadily replacing workers who performed automatable routine tasks with this mix of tech, yes, but also more skilled workers. So there’s a kind of sorting that occurs: you get more automation but also upskilling, both of which will be tough for already jittery workers.
What makes this economic downturn different than the last recession?
The thing that is different this time is that all that I ‘ ve said about already less educated, lower-skilled workers is beginning to apply more to middle-skill and even higher-skill professional and white-collar work, which may become more susceptible given the improvement of things like AI. Our recent research has shown that AI is disproportionally utilized in white collar , middle management, or upper management areas. It’s also used by line workers and in administrative functions. So it could be that all of this use of telecommuting technologies and communication gear may be pointing to the readiness for more wholesale reorganization of offices that may well put more pressure on not just workers in more routine, traditionally automatable occupations, but more professional ones, I think, are very real possibility. I think the big takeaway here is that downturns drive more, not less, dislocation through automation. And all bets are off about how that will work through this cycle, where the event is huge and there may be more ready to go automation, AI, or remote work platforms.
Rani Molla What jobs were automated after the last recession? Mark Muro
The initial event had a lot to do with finance, but it was tied up with a substantial crisis in the auto and manufacturing sector. Those were kind of the poster children of the financial crisis, and I think we do have a different picture of the most affected sectors this time. Rani Molla
So what will be most affected this time? Mark Muro
Food service and accommodations have already been under a lot of pressure, including through the kind of kiosk ordering. But also the middle-skill administrative positions and offices have been under a lot of pressure. And then, of course, manufacturing has been at the forefront for years. For those who think that manufacturing is as automated as it can be, we still see very high exposures there, and I think the kind of suite of AI susceptible middle management occupations – whether it’s bookkeeping, financial analysis, even things like software development – may see pressure. Rani Molla
Can you spell out what it means to be susceptible to automation or AI?
Given existing technologies, a total of about 51 million jobs in multiple categories and industries and occupations could be replaced by machines. Now, that doesn’t mean they will. But that gives us a sense of the size of the exposed jobs. It’s not a huge share of the whole economy, but it’s a substantial number of people obviously. That means their work is relatively predictable – or what’s called routine – and therefore susceptible to replacement, either by robotics or office software, for instance.
Which jobs are safe?
Health, education, government – those will be under relatively less pressure in some ways, at least of immediate layoffs. But we may also see moves to remote learning, for instance, in education changing the mix of workers or government under fiscal pressure, needing to bring in new technology.
I think this episode seems broader. It will likely touch all kinds of occupations, but there’s no doubt that direct face-to-face interactions for personal health, for instance, will probably be pretty safe. But there will be the need to reduce some of them, so I think that becomes difficult too.
Your research has also found that automation will increase existing inequality in the United States. How so?
Because the exposure to automation is heavily shaped by the industries in which particular occupations are situated, we have all kinds of variation across the board. Because the gender or ethnic mix of a particular occupation or industry varies, we see extreme variation there. You have clear vulnerability of younger workers who may be more concentrated in food service or accommodations – jobs that have always been a driver of exposure to automation and that, clearly, is in the forefront under social distancing.
Men are a bit more exposed because of their involvement in the fulfillment and trucking sector, which is showing hiring now because of the huge shift to ecommerce. But it’s also a highly automatable area. So we look for a lot of flux in that area. Then Latinos and black Americans are also in heavily exposed industries and occupations, whether it’s again food service or mid-level office work. So this plays out across all of these occupational mixes. The pandemic pressure then will have its own patterns, but these exposure levels are pretty well known. This doesn’t mean that the jobs will necessarily go away, but it’s a mapping of the shape of exposure and where the pressure may be.
(The fact that this is a global event means it could have an even broader impact than a
(typical recession) . Is that right?
The scale of the event is largely going to multiply or compound the overall tendency for automation to surge during downturns. We know the cyclical pattern is a historical fact. The sheer scale of this event is going to exacerbate that. You have to think that what can be automated, likely will be now, given that we’ve had a lot of inertia through the good times. The good times aren’t the time when the pressure is applied. Now, the pressure will be felt by business organizations. And with more technology having arrived and being on its way, I think we will see more pressure, not just in the standard types of automation that we’ve seen in the past, whether robotics and factories or kiosk-ordering in restaurants. Retail with cashierless checkouts, such as the Amazon Go stores, may spread. Then I do think we will see more and greater use of AI in the middle class and professional workplace in offices. I think that we’re going to see much more automation than we have even in recent downturns. Rani Molla
What will the work landscape look like after this?
The potential scale of this event isn’t just going to bring an end to the plentiful supply of jobs we’ve had. It’s also going to bring, because of this automation link, a new round of much more structural change again, in both what the demand for skills is and what the labor market looks like.
So we’re not just losing a job-rich decade, we’re likely diving back into a period of tech-driven structural change . That’s going add complications for workers, and it’s going to really, really ratchet up the anxiety that I think people feel. Because it’s not like there’s going to be a return to the same normal. There’s likely going to be the insertion of new technological platforms that will change and really alter what normal is.
GIPHY App Key not set. Please check settings