You are reading Economic Forces, a free weekly newsletter on economics, especially price theory, without the politics. You can support our newsletter by signing up here:
In July 2021, President Biden issued an “Executive Order on Promoting Competition in the American Economy.” In many ways, this was step one of the political response to growing concerns about possible failings in the U.S. economy: falling productivity growth, declining business start-ups and dynamism, a falling share of national income going to labor, rising market concentration, and more.
The popular narrative—well, popular among economists—ties all of these ailments to issues of competition and rising market power, especially from the “superstar” firms, as Autor et al. refer to them. The most shocking figure on the market power problem comes from a 2020 QJE paper by Jan De Loecker, Jan Eeckhout, and Gabriel Unger. Fun note: this paper was the “Editor’s Choice” in the same issue as the superstar paper linked above.
Their first figure shows average markups in the U.S. have risen from around 1.2 in 1980 to 1.6 by 2016. The markup is defined to be the ratio of price over marginal cost. If all of our measurements and estimates were correct and the economy was perfectly competitive, markups would be equal to one, and we would have our usual price equals marginal cost condition. If their figure is true, the last 40 years have seen a massive decline in competition.
In a future newsletter, I will go into the debate around their estimates and others in the literature. Don’t forget to subscribe, so you don’t miss that newsletter.
In today’s newsletter, I want to think more about the theoretical side of markups. Why do we care about them? How do we interpret them?
Why do Markups Matter?
Despite lots of talk about “market power,” it is not something we can directly measure. It’s more like consumer preferences or human capital: it’s a conceptual tool to clarify our thinking, not something we measure.
In the Stone Age of industrial organization (i.e., up to the 1970s), people would resort to things like market concentration as proxies for market power. We see this connection between concentration and power in sentences like “Google controls over 90 percent of search traffic.” The word “control” is highly misleading. As Thomas Sowell has pointed out:
Antitrust proponents have scored a verbal coup by constantly terming such percentages the "share" of the market "controlled" by certain firms, as if they were discussing prospective behavior rather than retrospective numbers.
If people are free to choose and most happen to choose one seller, that’s not control in any meaningful sense. That’s not power. This is one of many problems with using concentration as a measure of power, as I explained in a previous post.
Instead of concentration, when a modern economist wants to measure market power for a seller, they look at the ratio of price over marginal cost: the markup. How much is the price marked up above costs? For a buyer with market power (“monopsony”), the proper measure is the markdown. How much is the price marked down from marginal value to the buyer? For my discussion today, the two are equivalent, so I will talk about markups.
Markups can be thought of as profits on the last unit sold (if we ignore that the seller needed to lower the price to sell one more unit). Higher markups lead to higher profits which partially captures the idea of market power. Profits show they have power! In practice, due to data limitations, we are always measuring something like average variable cost, not marginal cost. In that case, we don’t have a clear way to distinguish the average profit rate from the marginal profit rate, i.e., the markup. So I will be a little fast and loose with the distinction between markups and profits. Forgive me. This is a newsletter.
While markups are a drastic improvement over concentration measures as a measure of market power, they are not perfect. First, they are difficult to measure. They require good measures of price and marginal cost; neither is generally available, as I said above. In contrast, concentration measures are relatively easy.
Second, even if you have perfect measures of marginal costs and marginal revenue, we cannot conclude the markup is bad for consumers, as the word “power” suggests. In fact, as I’ve said before, in the 101 monopoly model, the inefficiency comes because the monopolist is not paid enough. We would fix the inefficiency by subsidizing the monopolist. It’s kind of weird that the optimal policy is to subsidize the side with power.
If we see a firm’s markup rising, there are lots of possible reasons: some good, some bad. If they improve their product, the demand may shift out, and the markup could rise.1 That’s good for consumers. If their competitors’ products decline in quality, their markup may rise. That's bad for consumers. As I’ve explained before, we can’t do a full welfare analysis just by looking at one seller. We need to consider the market of sellers.
Now suppose I see firms exiting the market and there remains a single firm now with a high markup. Is that good or bad?
Antitrust, again back in the Stone Age, used to think of driving out competitors as harmful to competition. It may be. If the remaining firm somehow cut off a vital input for other firms, that is like making their product worse. But if the surviving firm simply was able to outcompete the rest and runs out the competitors, who cares if markups rise? That’s a welfare improvement for the consumers.
Everything up to this point is mostly standard. Now I’m going to push a weird idea.
Markups as Residuals
I’m going to argue that without more investigation into what caused the markups to rise, we should think of markups as residuals that we haven’t really explained. Steve Berry, in a keynote to the FTC, also emphasized that markups are residuals. They are a return to a firm, but we, the economists, do not know what for.
First, what do I mean by return? Let’s think of the world of perfect competition without markups. Everyone involved in the production process is earning a return for their efforts. For simplicity, we will think of their return as purely monetary. A business makes a certain amount of total revenue, and that gets divided up among people who contributed. But how does it get divided up? This is the big question in value theory. Around 1900, economists were obsessing over it.
We should think of all returns as returns to property rights. Drink! I mentioned property rights! People earn a reward as a return for some asset that they own: their labor, their capital, their knowledge, etc.
In the standard firm model, the firm owner hires workers and rents some machines. The revenue generated is split between the workers, the owners of capital, and the firm owner. In the standard model, since anyone can set up a firm and the ability to turn capital and labor into widgets is public knowledge, the asset “knows how to run a firm” is not scarce and so the firm owner earns no profits. In practice, the firm often owns the capital too. That owner gets both returns and earns the return which the accountants may call profit. We don’t need to take a stand on whether we call that profits; it’s a return to some assets.
Now let’s imagine the firm owner discovers some new idea that allows them to make an even better product than their competitors. The firm earns a return from that superior technology. If they could sell that technology, we would more easily see that the money they made came from the sale of the idea, not from running the business. They would again earn zero return to actually operating the firm.
One fundamental problem is that you can’t buy and sell ideas like you can buy and sell widgets, so it makes sense for the person with the idea to run the business. Those two assets (knowledge and firm ownership) are bundled together in practice. Wait for my paper on this in the next… few years?
So we see this business owner earning “excess” profit on the margin. They have a markup. Why is the markup not competed away so that price is driven down to marginal cost?
In this case, we, again as the economists studying the firm, do not realize that this firm has a different production technology than all the other firms in the industry, so they are earning a return above and beyond what we think they “should,” according to the estimate of production functions in this industry. It is a residual that we cannot explain. In this case, its benign.
Consider another case. Now, instead of better technology, the firm gets an exclusive privilege from the government to run their business. No one can enter. Now we understand why price isn’t driven to marginal cost. It’s charging a markup and earning a profit, but that profit is a return to the property right “allowed to exist in the market.” In the usual way, if the government grants more of those property rights, it becomes less scarce, and the return falls.
Again, the economist looks at this industry and doesn’t see why this markup is not being competed down. They do not realize that entry is restricted and the production function doesn’t just involve labor and capital but a government license. There is a residual return left unexplained.
My contention is that for any market where we find markups, if we search hard enough, we can find the property right that the markup is a return to. That does not mean that the markup is good. It could be a return to something we don’t like, such as a government privilege, or to an asset that hurts competitors. But simply measuring the markup, even with great data on costs and residual demand curves, does not tell us what is going on.
Micro’s Solow Residuals
While I’m calling it a residual, I don’t mean it is worthless. Residuals have proven quite useful within economics. If you’ve heard of a residual in economics, it is the Solow residual.
The way the Solow residual works is that we look at a bunch of countries over time and we estimate the relationship between inputs (like capital and labor) and outputs (like real GDP). Some of the output is attributed to labor, some to capital, and the rest gets swallowed up by the residual. The residual captures that the economist doesn’t know the true relationship between inputs and outputs and we need a line on the accounting to capture our own error.
But the same is true for firms!
Return to the firm with capital and labor. Suppose the economist isn’t properly accounting for the opportunity cost of the capital since the firm owns the capital. The firm will have some surplus not accounted for that may be labeled profits or markups over its cost, which the economist thinks is just labor costs. But there is a markup because we don’t have the right model. There is a residual gap that we call profits or markups, as shown in the figure below.
We found a markup but it’s actually a return to capital that we aren’t properly accounting for. While no one would be so dumb as to ignore capital costs, the principle is more general. The markup is a return to something we are leaving out of the model.
We may imagine that we are explaining things by pointing to markups, but that’s like explaining growth by pointing to the Solow residual. Yes…. and No. That’s not really an explanation of what is happening in the market. Maybe we should incorporate human capital. Maybe we should incorporate some measure of ideas.
The same is true when we find markups. We should look more into what is happening. What’s causing them? It is an invitation to dig deeper! Finding the markup is the start of the investigation, just as we should look more into what is driving growth when we see that the Solow residual is large. It tells us that something weird is going on here and production isn’t as simple as Y= K^1/3 x L^2/3.
There is evidence that what people called markups were actually just businesses with different production technologies than we realized. As I’ve heard John Haltiwanger say, different businesses are increasingly doing business differently. But as I warned, the empirics are saved for an upcoming newsletter.
We cannot necessarily say that an increase in demand generates an increase in markup. We need to look at elasticities. In practice, the distinction between a shift and a rotation that could raise markups doesn’t matter that much, since we are basically always measuring something like average willingness to pay or average variable cost, not marginal cost.
Oh this is good.