Chicago Booth hosted a conversation in February as part of its 2024 lineup of Economic Outlook events. In a discussion moderated by journalist Tim Phillips, Booth’s Randall S. Kroszner and Chad Syverson and New York University’s Julia Lane analyzed the labor-market transition underway. The following is an edited and abridged excerpt.

Tim Phillips: When we discuss artificial intelligence, it represents a bag of all sorts of different technologies. Do we really know what we mean by AI at the moment, and can we easily capture, measure, and estimate its impact?

Julia Lane: Those numbers, I think one might call them bollocks. They are not robust. These claims that are being made are from feelings or interviews with business managers, which kind of feed off each other. So there’s a lot that needs to be done.

Chad Syverson: Julia raises a very fair concern about direct measurement of AI. And in terms of productivity, it isn’t enough to measure AI investments themselves; related investments and intangibles have to be part of the equation too.

But I think there is a data-driven case to be made that maybe we’re starting to see an emergence from the productivity growth slowdown that’s been going on for 15-plus years. Almost every economy, regardless of income level or geographic location, has seen this slowdown, and with it, stalled economic growth.

We’ve all been looking for signs of an emergence from the slowdown, and it might be happening—although we’re far from knowing that the productivity growth slowdown is over.

Phillips: Is it too early to be able to think about the macroeconomic impacts of AI, for example, increased demand for investment?

Randall S. Kroszner: Certainly we’ve seen a lot of excitement in the stock market over a number of companies that have been either producing or using AI technologies. That suggests there might be a lot of investment going into this area that could affect many industries.

A few years ago, there was a lot of discussion of so-called secular stagnation, which was a way of talking about the productivity slowdown. Under the secular stagnation hypothesis, you’d see interest rates fall because demand for investment would not be very high. But if you’ve got a lot of excitement about investment in AI, that could lead to a substantial increase in investment demand over the next few years. All other things being equal, that would lead to interest rates moving up.

This is a big debate right now. After central banks start to normalize interest rates once inflation comes down, where will interest rates end up?

In this scenario, rates will end up, in the long run, closer to what they were before the financial crisis rather than at the very low levels that we’ve seen recently. The level of interest rates, in turn, has an impact on thinking about the valuations of startups that might be coming into the AI space. It’s important to work out these scenarios, particularly if you’re thinking about investment in this area.

Phillips: There is a debate at the moment about the impact of AI at that micro level in the workplace and whether it is going to replace tasks, augment workers, or supplant jobs entirely. What’s the research telling you?

Syverson: Is this IT 2.0, where we’re going to see even more skill-biased technological change and further increases in inequality? Or might we see an actual compression in the earnings distribution?

It’s early, and we only have a few good case studies. But those so far seem to indicate it’s the latter, that AI technologies, at least in the applications where they’ve been used, help workers at the lower end of the skill distribution and bring up productivity most among those workers. Time will tell.

Phillips: How would you even begin to measure the productivity effects of something like generative AI?

Lane: The challenge is that AI is neither an industry nor a scientific field. We don’t have a statistical infrastructure that deals with it. We’re not producing physical goods or services as much as ideas. Growth is not occurring through investments in capital, labor, energy, material, or services, but in how they’re being put together.

If you think ideas are embodied in people, you need to trace people. See who’s doing AI work at universities. You can track them through research grants—not just them but the graduate students, postdocs, undergraduates, and so on who we know have value because we are seeing them flow into the labor market and get paid $750,000 a year plus stock options.

We created something about 15 years ago that captures the human-resource and finance information that is spent on the grants at those universities. You can track those people into the wage records of the companies in the private sector. You can identify which of those companies are touched by AI or any of the other critical and emerging technologies, such as quantum computing or synthetic biology.

It’s a technology-agnostic approach that says, “OK, what we’re really looking at is not just manufacturing industries or service industries but what we’re calling the industry of ideas.” That’s how you capture what’s going on, what happens to the other workers in those companies, and what skill needs there are.

“Any kind of general-purpose technology is going to have effects all over, and some of those are going to be larger than others, and some of those are going to be extreme.”

— Chad Syverson

Phillips: Who will be the winners and losers?

Kroszner: It’s early to speak definitively about that, but we might be able to draw some insights from big technology revolutions from the past. In the 1910s and ’20s, there was an enormous revolution in terms of the electrification of plants and the development of small machines, and then, of course, the application of the innovations in auto manufacturers. There was a lot of debate in that period about what would happen to workers. The concern initially was that low-skilled jobs would be eliminated by these machines, but obviously that wasn’t the case.

You could make a more optimistic case now that AI is not going to be eliminating those sorts of jobs. It could be complementary to many types of jobs and tasks, simply making people more productive, or asking people to do things that are a bit different than what they were doing.

In the early part of the 20th century, most countries had a six-day workweek. Then we moved to a five-day workweek. We seem to be on the verge of a four-day workweek, and if you’re superoptimistic, maybe it’ll be down to three. But it still seems that at least for the foreseeable future, there’s still going to be a lot of work that people will have to do.

Phillips: One way work has changed over the past decade is the rise of gig work, nontraditional work structures. How do you see AI affecting that?

Syverson: We have seen the way communications technology alone has changed where work takes place. Flexibility for workers in terms of their location is tied in part to the gig economy. I’m not sure how AI is going to either accelerate that further or reverse some of the changes we’ve seen over the past years.

Who wins from remote work? Is it the workers who get to stay at home because they don’t want to commute anymore, even if there’s a cost to the companies, maybe, in productivity?

Or can the companies actually force the workers to come in?

The late economist Ronald Coase held that wages will adjust to make whatever the most efficient thing is happen. If it’s really efficient and the productivity gains to companies are from having the workers all together in an office, they’re going to have to compensate the workers to come in.

If the productivity gains aren’t that large, certainly relative to the cost of commuting for workers, the workers will be able to stay at home. They’re not going to get compensated for the gains in terms of wages, but they’ll get compensated for it in terms of having more flexible schedules and not having to commute, perhaps.

Whatever way the technology pushes us, we’ve got this price out there that we call the wage that can move around to compensate the parties, and that the parties can share among themselves to let the most productive thing happen in the long run.

The Screen Actors Guild strike is an example of that. The two sides [SAG-AFTRA and the Alliance of Motion Picture and Television Producers] came to an agreement to share those gains in a particular way, at least for the time—though the fact that the strike happened and took a while to resolve indicates there are still some bargaining frictions out there. That’s what I would expect more of as we see the diffusion of the technology into other areas.

Recommended Reading

Phillips: Is this a new economy where different skills are going to be demanded?

Lane: Education provides the skills for particular workers. But one of the things we’re expecting to see in the data is the increased importance of different types of certifications.

We’ve always needed a flexible labor market, and now it’s much easier to acquire different types of skills. Capturing certifications that are going to make sense to companies will be a really important activity. There will be a lot more interest in understanding what the high-quality signals are, what the low-quality signals are, and what skills workers can acquire.

Phillips: Sometimes some labor markets in Europe, the Middle East, and Africa are criticized for being too rigid. How do you think that the impact of AI, if it demands flexible labor markets, will play out in these countries?

Kroszner: We can’t live in the past or try to keep everything exactly as it was, preserving all jobs and all tasks. Jobs and tasks are definitely going to change. AI will be likely to drive greater flexibility in those markets—that is, the technological innovation may erode those rigidities.

Twenty-five years ago, for example, it was difficult to obtain a landline in most of Africa, with long waiting lists and much lower penetration than in developed countries. Precisely because of that constraint, however, cellular technology spread rapidly. Africa leapfrogged most of the developed world, with much faster diffusion of mobile telecommunications. This quick adoption then facilitated a much more rapid diffusion of mobile phone banking in Africa than in developed countries. The rest of the world is still trying to catch up! This example gives me a glimmer of hope that applications of AI may allow Africa to overcome some of the rigidities in the system and benefit.

Phillips: Do you foresee radical changes in particular industries?

Syverson: I can’t tell you which industry it’s going to be in, but I think it will happen.

Thirty years ago, over half of the revenue in the travel-agent industry was from airline ticket commissions. In the United States, you’d buy an airline ticket from a travel agent and there’d be a $50 commission on there—and commensurate amounts in other places. Over the course of six years, the airlines figured out they could sell tickets directly to consumers, and consumers trained themselves to buy tickets directly from airlines. Half of the travel-agency industry’s revenue disappeared.

Most industries did not see half their revenue decline because of information technology. But for that one particular industry, there was a radical shift. Any kind of general-purpose technology is going to have effects all over, and some of those are going to be larger than others, and some of those are going to be extreme.

Phillips: Policy makers are under pressure to act from all sides. What should they do?

Lane: Build high-quality data. Think about where those data sources are going to come from and where you’re going to have trust in the data. Understanding and building a labor-market information system that is local, actionable, and timely is going to be critical, and we don’t have that statistical infrastructure right now. We need to rethink the data system and push toward faster, better data—otherwise, the policies are going to be bad.

Kroszner: You may want to provide some support in transition for workers. Best to allow technology to evolve, but have a support mechanism for workers and households in the transition. Trying to control a rapidly evolving technology can generate unintended consequences that may actually make groups that policy makers are trying to protect worse off.

Syverson: Protect the worker, not the job. Try to keep the jobs flexible, and then help the workers who are hurt when technology changes the nature of jobs. Education and retraining is part of that. It’s hard to predict the future; we all know that. You really want to have a system that offers options and can react to the realities that might happen.

Kroszner: Make sure a safety net is there for the transition because we don’t know exactly what the consequences are going to be. But I would be wary of jumping very rapidly into regulating an area we really don’t quite understand yet.

Randall S. Kroszner is the Norman R. Bobins Professor of Economics at Chicago Booth. Julia Lane is a professor at the NYU Wagner Graduate School of Public Service and an NYU Provostial Fellow for Innovation Analytics. Chad Syverson is the George C. Tiao Distinguished Service Professor of Economics at Booth.

More from Chicago Booth Review

More from Chicago Booth

Your Privacy
We want to demonstrate our commitment to your privacy. Please review Chicago Booth's privacy notice, which provides information explaining how and why we collect particular information when you visit our website.