“Equitable Growth in Conversation” is a recurring series where we talk with economists and other academics to help us better understand whether and how economic inequality affects economic growth and stability.

In this installment, Equitable Growth Economist Kate Bahn and Director of Markets and Competition Policy Michael Kades talk with Ioana Marinescu, professor of economics at the University of Pennsylvania School of Social Policy & Practice, and Herbert Hovenkamp, the James G. Dinan University Professor at the University of Pennsylvania Law School and Wharton School. The conversation focused on how anticompetitive mergers can decrease wages and how antitrust policy can help address this problem. Professors Marinescu and Hovenkamp have written an article, “Anticompetitive Mergers in Labor Markets,” which is forthcoming in publication in the Indiana Law Review.

[Editor’s note: This conversation took place on August 2, 2018.]

Michael Kades: My first question is for Professor Marinescu. Let’s start with some basics. What do you mean by labor market concentration?

Ioana Marinescu: If you think about the product market, and you’re talking about some market where some goods or services are sold—for example, you might think about the mobile telecom market, where there are just a few companies that sell most of that product or service—then that means that it’s a highly concentrated market. And that’s what antitrust is very much about.

Now, another way of looking at this is to look at the labor market. In that case, we can also talk about labor market concentration in a similar way, except that this time we’re not looking at firms potentially increasing prices for the products they are selling, but rather at firms decreasing the price that they pay for inputs—meaning the inputs that they purchase in order to produce. And one key input that I’m interested in is labor. Firms are buying labor services in order to produce with it. And so, therefore, you can calculate labor market concentration based on available jobs.

So, the question becomes, in a given labor market—let’s say the market for baristas—how many companies are currently hiring in that particular market? There might be many jobs, let’s say many jobs at Starbucks, but if they are all jobs at Starbucks, then that could still be a very concentrated market.

Kate Bahn: How do mergers affect labor market outcomes, and what do you think the broader economic consequences would be of that?

Marinescu: When we think about mergers, traditionally in antitrust, one of the tools that is being used to measure concentration is the Herfindahl-Hirschman Index [which measures concentration based on the number of firms in a market and their market shares], and, in fact, the horizontal merger guidelines—which the government follows in determining whether a merger is anticompetitive—explicitly say that this can also be used for the input market, including the labor market.

And there are specific thresholds in concentration that are low, medium, and high. All of them are used in order to determine whether a merger will likely yield anticompetitive price increases or, in the case I’m interested in, anticompetitive wage decreases. And, in fact, using data on wage postings from a very large online website called CareerBuilder.com, we show that the greater the labor market concentration, the lower the wages that companies post on this website.

Because of that reason, it is very likely that a merger that significantly increases labor market concentration will put downward pressure on wages. And what are the broader consequences? Well, obviously the immediate consequence is that wages in the affected markets will be lower, but if you look at the broader time trend, there’s been a wave of mergers and acquisitions since the 1980s, so it is possible that this phenomenon has contributed to stagnating wages.

There’s other research by Efraim Benmelech (Northwestern University – Kellog School of Mangement), Nittai Bergman (Tel Aviv University), and Hyunseob Kim (Cornell University – Samuel Curitis Johnson Graduate School of Management) showing that labor market concentration in manufacturing has, in fact, increased over time, and that the effects of labor market concentration on wages very much depend on context such as unionization, so that in a very unionized labor market, labor market concentration increases don’t have much effect on wages at all.

Therefore, it is possible, if you look at the increase in labor market concentration fueled by mergers and acquisitions, and if you combine that with the decline in unionization, that those phenomena together could contribute to explaining the wage stagnation that we’ve seen for 80 percent of the population since 1980.

This phenomenon could be one piece of the puzzle toward explaining the increase in inequality since the highest-wage people have seen wage growth, but as I said, 80 percent of U.S. workers have seen no wage growth since 1980. This phenomenon of greater concentration could be part of the reason why this has happened, although the exact extent of the contribution is very much something that we hope further research will tell us more about.

Kades: Professor Hovenkamp, in “Anticompetitive Megers in Labor Markets,” there is both a discussion of the economic evidence on concentration in labor markets and a description of how mergers that suppress wages can violate the antitrust laws.

And the article lays out the case in a very almost matter-of-fact way that these are mergers that can violate the antitrust laws. But up at the front, the article explains that no court—as far as we are aware—has ever condemned a merger because of its anticompetitive effects in the labor market.

Why do you think that, so far, antitrust law hasn’t focused on this particular type of competitive harm?

Herbert Hovenkamp: I think there are a couple of reasons. One of them is historical, which is that at the time that the Sherman Antitrust Act was passed in 1890 and all the way through the New Deal, the concern about labor was really with labor cartels and high wages rather than low wages. The first 15 indictments under the Sherman Act after it was passed were against labor unions, not against producers.

Eugene Debs is a good example. Debs went to prison in 1895 for organizing the Pullman railcars strike, but he was convicted under the Sherman Act of an antitrust violation. So, the United States went through a very, very long period in which we were thinking of the laborers as the monopolists. It wasn’t until 1948 that we even focused antitrust attention on markets in which firms buy, and even then, it’s only been a tiny, tiny percentage of antitrust cases across the map—not just merger cases, but across the map. We’ve focused on selling rather than buying, and labor just kind of dropped off.

I think there’s also an economic premise that Ioana’s work has shown to be incorrect, and that is antitrust traditionally assumed a far greater degree of labor mobility than, in fact, exists. Economists thought, “Well, laborers, they don’t have major brick and mortar assets, they’re selling their own skills which are as portable as they are, and so they can move around.” There was this thinking that labor markets are big, and as a result, we don’t think monopsony is a particularly big problem if people can avoid low wages in one area by simply relocating to another.

And the fact is that the evidence speaks strongly to the contrary. Laborers are not particularly mobile, they are very vulnerable to wage suppression, and as a result, we get the effects that we see now. But you know, economists are very late to the party in this case.

Kades: Given the empirical development, Professor Hovenkamp, what would you recommend to the antitrust agencies in terms of how they should be thinking about this issue going forward?

Hovenkamp: I think they should start out with the guidelines, which they have. I think the guidelines are going to need some revision for a number of reasons, only one of which is labor. Then they should look at both sides of markets rather than focusing on one side, which means the first thing they have to do is after they’ve drawn a product market, in order to test for anticompetitive consequences in the product market, they’re going to have to define relevant labor markets.

And that means definitions in terms of specialty and also in terms of geography. That’s what we do with product markets now. And once they define those markets, then they will compute the Herfindahl concentration, just as Ioana just suggested a few minutes ago, and if those concentration levels are worrisome, then they can decide whether or not they want to pursue that particular merger. I don’t think the methodology is going to be very different. Some of the factual evidence is going to be different, but the methodology is pretty much in place.

Kades: Is it fair to say you think that the agencies need to make concerns about labor monopsony more important?

Hovenkamp: Yes they should, and they shouldn’t assume that labor market boundaries track product market boundaries. I mean, that’s one thing we cited in our paper, and it seems pretty clear, right? Labor markets follow labor specialties and labor mobility. They may coincide to some extent with product market boundaries, but they may not.

And so, you can’t just do a workup on the product side and then assume you’ve gotten all the work done. If you’ve got a special class of employees, like computer engineers, those engineers might work for firms that don’t compete with each other at all on the product side, and that means that that market will end up having different boundaries than the product market has for those same firms.

Bahn: Professor Hovenkamp, I want to follow up on that a little bit. Does the antitrust scholarship give us any insight as to how this can be carried out and applied to labor markets and how you would define one of these labor markets? For example, your paper talks about Intuit Inc. and eBay Inc. having very different product markets but having the same market for workers, so how could antitrust agencies actually think through this problem?

Hovenkamp: Well, the merger law says where the effect of a merger may be to substantially lessen competition, number one, in any section of the country, and number two, in any line of commerce. Section of the country means a geographic market, and line of commerce means a product market, and so, under the guideline’s approach, what you would do is you would start with a very small grouping of sales—people that obviously do the same thing in the same area—and you would ask whether that’s a market.

And you would ask that by hypothesizing a wage decrease and asking how many people would attempt successfully to exit from that market in response to that decrease. If so many would exit that it would be sufficient to keep the merger—to keep prices at the premerger level—then you would know that your market is too narrow, and you’d have to draw a little bigger market.

But you would repeat that formulation until you find a market such that if wages were suppressed, the laborers would not be in a position to do anything about it but accept the lower wages, and then you would look at the impact of the merger in that market.

Marinescu: Can I follow up on that?

Bahn: I would love that. Go ahead.

Marinescu: In practice, it is important to understand that just the fact that sometimes people do leave jobs doesn’t mean that the market is defined too narrowly. In any kind of job, there are some people who are more mobile and will leave. The practice from antitrust in the product market is to ask about a critical threshold. How many people will leave? That’s the key question, not that no one leaves if the wage is lower. We require that not enough people leave, so that on balance, the company gains.

So, here’s the way to think about this. The company lowers wages for those workers that it can keep, and that’s an increase in profit. But some workers—typically some, usually not zero—will leave. The question is, how many and is it worth doing given that some will leave?

If it’s a lot leaving, then it’s not worth doing, just like Herb was saying, but if there’s enough people who stick around because they have no better opportunities, even at that lower wage, then it can be worth doing for the firm. And that’s where we can use prior research in economics to inform us about that threshold in terms of how sensitive people are to the wage. When they get a wage cut, how likely are they to quit a job?

And so, what we have seen across a number of studies is that, generally speaking, people are fairly immobile in the sense that there are enough people who will stay even if you cut wages. It’s quite striking that even in markets such as Amazon Mechanical Turk—a crowdsourcing jobs site for temporary “gig” employment—there’s a recent paper that shows in principle it’s very easy for workers to choose a job online that pays better wages, and even employers who pay better wages don’t attract a lot more workers than those who pay lower wages, showing that even in a market that in principle has very little obstacles to workers moving around, you don’t see big reactions from workers to wages being lower.

So, generally speaking, you know, the evidence is building up that workers often are not very reactive to changes in wages, and therefore in many concrete cases, it might turn out to be profitable for companies to lower wages, even though they lose a few people on balance, it might be a profitable operation and therefore mergers do have the potential to significantly suppress wages.

Bahn: Are you talking about the previous research that estimates labor supply elasticity?

Marinescu: Exactly.

Bahn: If you had an estimation of labor supply elasticity in an industry, you can get a sense of what we think the effect may be?

Marinescu: That’s exactly right. But what’s even more striking is that the labor supply elasticity for even the individual firm is very low. In an extreme case, you might consider that even an individual firm is already a market in itself, and therefore almost by definition any merger is bad.

I’m not necessarily going to claim that, but the degree of low labor supply elasticity that we see at the individual firm is not inconsistent with some firms constituting a market on their own. And that’s something that people have failed to appreciate until recent times, and I think it’s very exciting to see antitrust being brought together with labor economics. We have a lot to talk about and bring insights from both sides to better understand how market structure affects workers’ outcomes.

Bahn: This is a good transition to my next question for you, Professor Marinescu. The original theory of monopsony was developed by the economist Joan Robinson in the early 1930s and has often been used to describe previous kinds of of employment that seem less realistic now such as a geographically isolated company mining town. It’s kind of the prototypical example. Then Alan Manning pioneered this concept of dynamic monopsony based on the Burdett-Mortensen job search model, where we can do some of these elasticity estimations that you talk about, which expanded the definition of monopsony to include search frictions that give employers wage-setting power.

So, why do you think that monopsony was relegated to a sort of secondary field in economics and then why has it risen into prominence so recently?

Marinescu: The whole history of this is so very interesting, and in some way dovetails with my own personal history. I was a Ph.D. student at the London School of Economics, where Alan Manning is a professor, and already back then, almost 20 years ago, on the wall there was this book, Monopsony in Action, that Alan had written. Frankly, at the time, I wasn’t too interested in that. I thought it was cool, but is this really important?

I think part of what was happening in that intellectual era of the time was just a general belief that labor markets are pretty close to competitive equilibrium, and there’s not too many frictions. We all knew even back then that, yes, of course there are some frictions. The question is whether all in all it’s close enough to the competitive equilibrium.The belief tended to be that it was. And more generally, I think in economics 20 years ago, there was the belief that generally all markets are, you know, more or less close to the competitive equilibrium.

Since then a lot has changed. For example, there’s been the rise of behavioral economics and even behavioral finance that talks about how workers, consumers, and investors don’t behave in the ways that are predicted by traditional economics. We’ve also seen the Great Recession erode our confidence in the perfection of markets. And then there’s wage stagnation, which I mentioned earlier, whereby 80 percent of U.S. workers have seen no wage increase since 1980. All of these factors together bring us economists to question the degree to which perfectly competitive labor markets are a good description of reality.

Nowadays, there’s a lot more interest in understanding the degree to which maybe it’s not a great description, and we’re going to look for evidence about why it’s not a great description. So right now there’s a huge effort going in that direction. For myself, my team, and also a large and growing number of colleagues and young people, it’s very exciting to be part of this research agenda.

I also think that when it comes to labor economics, some of the data that is very peculiar to understanding details about the labor market might have been missing. I’ve been working with all this online job vacancy data. There is extremely detailed information about where exactly a job is located and what exact specialty the job has to a degree that has never been known before with such precision. So, we are able—thanks to far better, more systematic data—to see a lot more about this granularity that helps us understand what a labor market is, how it functions, and how wages are determined there.

You also asked me about modeling assumptions, and how these have perhaps changed. I think that practically speaking the classic monopsony model is perhaps still the most useful for antitrust because it can allow us to relate the Herfindahl-Hirschman Index, which is widely used, to monopsony power. And so, that kind of constitutes a somewhat logical set of tools that facilitates our understanding of what’s going on.

Now, are the assumptions from the model exactly sticking to reality? It’s still a model, and I will say that right now it’s something that is at the frontier of research to understand to what extent the empirical regularities that we observe in terms of low labor supply elasticity, quite a bit of labor market concentration, and the fact that such concentration seems to suppress wages—how exactly is that happening at the micro level? What kinds of strategies are firms and workers using?

That’s something that we don’t know as much about. This is an emerging and exciting area for research, to understand the more detailed underpinning of this phenomenon that we can observe in the data. So, I think that there’s a lot more work to do to understand what kind of model is right, and no model will be exactly correct, but which one strikes the best balance between helping our understanding, being simple enough but capturing sufficient amounts of the reality to also be useful.

Kades: Professor Marinescu, to follow up on that last question. As an antitrust lawyer, I’ve heard economists complain that we lawyers tend to oversimplify empirical work. We just want simple answers. If one study supports our case, then it is the gold standard, and if it doesn’t, then we like to poke holes in it.

So, I thought it would be interesting, from an antitrust lawyer’s perspective, to ask you if the Federal Trade Commission or the U.S. Department of Justice or state attorneys general were to come to you and say: “Look, we are really interested in the work that’s being done here. What we’re trying to really assess is how comfortable can we be or how robust is the research showing that labor supply elasticity is low, that concentration’s high, and then, most importantly, that high concentration tends to lead to the suppression of wages?” How would you respond?

Marinescu: Obviously, the evidence has to be gathered on a case-by-case basis. So, the question is, do we have enough evidence to make sure information gathering is worthwhile? And there I would say the answer is certainly yes. And the reason is that the literature has shown evidence across many markets that monopsony power is prevalent and that’s also what theory predicts. You know, the theory of imperfect competition.

It is evident in my own work. But it’s also evident in other people’s work that I was mentioning earlier in terms of the labor supply elasticity in online labor markets, which in principle should be so much more fluid, yet we see low labor supply elasticity and therefore the opportunity for online employers to suppress wages in a market where workers should be so mobile. The fact that this evidence exists strongly suggests that this monopsony issue is a real issue.

We are sure that it exists.But is it important in every case? That’s something that has to be determined in each specific case, and I will say that as economists working on this issue, we are still in the process of collecting evidence on the exact size of the impact of monopsony power on wages. That is a very interesting endeavor fraught with some uncertainty, in part because context matters.

So, for example, I mentioned before that my colleagues have shown that concentration is not decreasing wages when unionization is high. That’s one example of context where you say, “Well, yes, in theory, we have reasons to believe concentration will decrease wages, but if there are unions that are strong, then it doesn’t seem to happen.”

Another context that’s more relevant to legal proceedings and that Herb and I are discussing in our paper are things like antipoaching agreements. We can measure the Herfindahl-Hirschman Index from data on job vacancies, but that doesn’t tell us if there is an antipoaching agreement, and if that’s the case, then the actual concentration in the market that workers are facing is greater than we can measure in our job vacancy postings because, essentially, the employers in that market are not really actively competing with each other because they have agreed not to hire each other’s workers.

Similar effects are derived from noncompetition agreements, which also limit the opportunities for workers to go and work elsewhere in that labor market. These are important context issues to understand that I believe would be natural in a legal proceeding to bring whatever evidence we have to bear about these various context factors.

To sum up, I think the evidence for monopsony power is big enough to make it worthwhile to investigate in legal proceedings, and then, of course, each case is different and the relevant evidence has to be gathered.

Hovenkamp: Let me add just a couple of things to that. First of all, I think it’s very important to distinguish questions about the quality of the evidence from questions about the volume. The quality of the evidence I believe here is very high. It’s as high as it is for the exercise of monopoly power in product markets. And that makes it certainly worth pursuing.

What we don’t have yet is a big volume of empirical testing of important questions. For example, in product mergers we’ve got decades of government-supported and private retrospective testing of mergers to see what the impact has been in the past on product markets. We need to do more of that with respect to labor markets to find out what was the impact of a particular merger or series of mergers on wages.

There’s also a behavioral issue that has to be confronted because it makes measurement in some cases more difficult, and that is laborers have utility functions while product-producing firms have cost functions, which makes labor output behave a little bit idiosyncratically. For example, when wages are driven very, very low, workers may actually work more because they have to maintain substinence. That contradicts the usual assumption that when prices are pushed down, a firm produces less. On the other end, when wages are driven very, very high, workers may, in fact, work less because they value leisure more than they value the additional money they get from extra work.

These are things we don’t have varied, well-developed theories about yet insofar as merger policy is concerned, and that’s going to take more work. But I think we’ve got a very good basis for getting started with this.

Marinescu: I want to add something more to this point, which is that another particularity of labor markets compared to product markets that makes any level of concentration potentially more damaging in terms of suppressing wages is that this is a two-sided market. By this I mean when you’re trying to buy a can of beer, the can of beer cannot say no to you. So, as a consumer, you always have a choice of buying products that are available for sale on the market.

In the labor market, however, as a worker, you can’t just say, “Oh, I like this job, OK, I’m taking it.” Instead, the employer must also agree to that. And so, this in principle curtails workers’ opportunities even more and is a reason why there are reasons to believe that the Herfindahl Index, as computed for the labor market, might have a different meaning as compared to the product market.

But this is something that we still have to investigate to understand quantitatively how important it is. It’s another very important consideration that workers cannot simply go around and choose whatever jobs they want—they have to be approved by employers.

Kades: I’m going to change back to legal issues. Professor Hovenkamp, as you’re well aware, currently there is a debate in antitrust law over whether the consumer welfare standard—which some argue means antitrust law is focused only on consumer harm—is the appropriate foundation for how we think about antitrust law and when we think conduct should be considered anticompetitive.

So, critics of that standard say that the consumer welfare standard excludes the very types of harm that your article with Professor Marinescu identified. Now, obviously, you have eloquently defended the consumer welfare standard and its flexibility, and you disagree with this criticism. I thought it would be helpful to let you explain how the labor monopsony harm fits within the consumer welfare standard.

Hovenkamp: First of all, the very fact that we call it the “consumer welfare standard” tells us mainly that that name was developed at a time when antitrust was completely preoccupied about sales in product markets. That is, if we had been equally concerned about monopoly and monopsony, I think we would have given it a different name.

And there’s nothing in the early literature on consumer welfare that suggests much concern with a buy-side monopoly rather than a sell-side monopoly.

More importantly, and to the point, I think it’s very important to identify the consumer welfare standard as about output—not as about price as such, but about output. The consumer welfare standard we want is one that encourages the maximum amount of output that is consistent with competition. Just as we don’t want firms to encourage higher prices by reducing product output, we don’t want them to encourage lower wages by reducing labor market output.

That is to say, in competitive equilibrium on both sides, if both the product market is competitive and the labor market is competitive, then you get maximum output on both sides and any kind of output restriction on either side causes harm, although only consumer harm emerges in the first instance on the sell side. Now, it turns out that the suppression of wages can lead to higher product prices if the firm has market power in the market in which it sells, because in order to suppress wages it has to consume less labor, and as it consumes less labor it’s going to produce less output. So, if a firm has market power in the market that it sells to at the same time that it has market power in the market in which it buys labor, then wage suppression also leads to higher product prices.

The more problematic area is the firm that has market power in the market where it consumes wages but is a perfect competitor in the market in which it sells, because that firm will produce less as it suppresses wages, but it will not necessarily show up in higher product prices.

I think the best way to deal with this is to think of the word “consumer” as being really a term of art that refers to vertically related firms or trading partners with the firm whose conduct is in question. And any kind of output decrease that is an anticompetitive output decrease on either side counts as a harm, although of course technically only those on the sell side are harms to consumers.

Finally, I would just add the fairly obvious point, which is that we are all consumers. Even laborers are consumers. So, “consumer welfare” broadly defined means that suppressed wages also lead to suppressed buying power for laborers, and that harms consumer welfare as well.

Kades: As you probably both know, one of the things Equitable Growth as an institution is interested in is promoting and funding research that increases our knowledge, that leads to better better policy solutions. We’re always on the lookout for the big, important questions or trends that we need more research on. We like to ask our subjects in our In Conversation series what’s one research area that you think could use a lot more attention?

Hovenkamp: Well, actually, Ioana and I have been discussing something off and on. We haven’t really decided what we’re going to do about it yet, but that is noncompetition agreements, which are spreading like wildfire and out of the historical range where they usually occurred.

That is, noncompetition agreements were usually made with employees that had a lot of specialized, firm-specific skills or firm-specific information such as customer lists or trade secrets, so that their migration to a different employer would be costly. There’d be a free rider problem, but just this week there was litigation reported involving Jimmy John’s.

Well, Jimmy John’s hires mainly unskilled and low-skilled laborers, and furthermore, they’ve got noncompetition agreements with practically everybody. If you drive a car delivering a Jimmy John’s sandwich, there’s a noncompetition agreement in your employment contract.

Those agreements serve as very significant limitations on worker mobility. And, of course, the key to keeping labor markets competitive is the right of workers or the power of workers to be able to move from one job to another, and so noncompetition agreements can create these little pockets where employees lack the ability to migrate from one employer to another. That can give an employer a lot of power to suppress wages without being concerned that its employees will ditch them for someone else.

Marinescu: Yes, I think I already mentioned a number of areas. One of them is to better understand the details of how both company strategy and workers’ reactions play out in order to have an increase in concentration yield wage suppression. So, what exactly is going on on the ground and that it should be informed by economic theory, I think, are important areas of research, both because they would inform us about the structure of the labor market and how wages are determined—which, for economists, is always an exciting thing—but also practically speaking, because when we’re talking about a case, we often talk about the particulars of that market.

And so, having some blueprint for the behaviors that lead to certain outcomes based on economic theory and on statistical evidence would probably also be quite useful for the practice of antitrust.

Kades: Great! On behalf of Washington Center for Equitable Growth, I want to thank both of you for sitting down with us. We really enjoyed it. Your work is really interesting, and it’s great to see some interdisciplinary expertise brought to bear on this issue. Thank you for being so accessible.

Bahn: Yes, thank you very much.

Marinescu: Thank you!

Hovenkamp: Thank you for having us.