Price theory. RIP?

On Tyler Cowen's new book and price theory

Tyler Cowen has a new book, The Marginal Revolution: Rise and Decline, and the Pending AI Revolution. The book is much more wide-ranging, but at one point, he writes a eulogy for price theory. Maybe not dead exactly, but more like a distinguished elder statesman that the profession has stopped listening to.

Price theory, he argues,

has moved to being a niche interest in economics. To some economists, especially from a few decades ago, that may sound almost contradictory, almost like saying “economics has become a niche field within economics.” Well, that is a bit true as well.

He quotes Steve Levitt, who retired from the University of Chicago at 57, on price theory: “I gotta say that the Chicago price theory really has lost.” And “I think it is essentially lost to posterity at this point.”

Not so fast. There’s a big difference between mostly dead and all dead. Mostly dead is slightly alive.

Sure, he’s probably right about the academic market. As he points out, theoretical papers in the top journals fell by 30% percentage points from 1963 to 2011. This has been documented for decades now and applies to theory more broadly than just price theory. He describes how graduate programs increasingly recruit math and computer science majors, on the theory that “the economics you can figure out along the way, or for some topics you may not need to know much of it at all.” Kevin Murphy no longer teaches the Chicago PhD price theory course. Kevin also no longer teaches the Chicago Price Theory Summer Camp has trained hundreds of PhD students over 14 years, including being foundational for my graduate time. (The 2026 edition has Garicano, Glaeser, Hurst, and Mulligan on the faculty, so the rearguard is fighting.) Even the framing of needing this type of camp, Cowen describes as “a rearguard action.”

In some sense, this is good for me personally. Product differentiation and all that. But I’m not so negative. Cowen is focused on one part of the production process: publishing formal research. Yes, it’s the high-status part, but it’s only one part.

Reasoning vs. technique

Cowen defines price theory as “the view that the basic intuitive economic concepts, as would be taught in intermediate microeconomics, are highly useful and for advanced problems too.” A hypothesis should be “intelligible in terms of microeconomic concepts that you can hold in your mind and understand.” You should be able to explain it to a smart non-economist.

That’s what we’ve been doing at this newsletter for six years. And what Cowen describes is real; the profession has moved on.

It’s important to note that he’s measuring the market share of price theory as a research technique, and finding it has fallen. Fair enough. The credibility revolution raised the bar for publishable empirical work. Machine learning generates predictions from 360,000 factors. Structural estimation recovers parameters from computational models. Mechanism design proves theorems. In all of these, the technique itself is a large part of the contribution. Price theory doesn’t work that way. You can’t build a career around “I thought clearly about the problem using supply and demand.” I would be much more high status if one could. But the profession rewards new techniques, it always has, and broad, basic price theory isn’t one in 2026.

Instead, price theory plays a different, hidden role in research.

Price theory is the reasoning that tells you whether the technique was pointed at the right question and whether the answer makes sense. It’s upstream of the identification strategy, upstream of the structural model, upstream of the theorem, heck even upstream of data collection. It’s the discipline that hears “corporate profits rose during inflation” and immediately asks, relative to what, as a share of what, and is that consistent with the standard model, or does it require a special story?

The profession has gotten extraordinarily good at technique. The tools are more powerful than they were 30 years ago. The results are more carefully identified. But every one of those results still needs someone asking whether the mechanism is plausible, whether the magnitudes are realistic, whether the finding generalizes or is specific to one context, and what it means for policy. Those questions are not answered by running the technique again with better data. They require economic reasoning. Price theory.

The profession does reward this reasoning, just not on its own. Kevin Murphy is the archetype, sure. Someone who understood price theory deeply enough to extend it into genuinely new territory on addiction, human capital, inequality. The combination of deep reasoning plus a frontier contribution gets you tenure at a top department. But you still see that today. Top people are combining data and theory to ask economic questions. Yes, lots of people ask non-economic questions but people still care about prices. In any seminar I’ve been to, the reasoning matters a lot.

And that reasoning is useful even when it doesn’t produce a paper. Most of the price-theoretic work happening in the world isn’t published. It’s an economist reading a paper and thinking “that can’t be right, because...” It’s someone writing a newsletter explaining why a popular argument about inflation is incoherent. It’s going about your day thinking through causality, prices, what happens at the margin. The academic market doesn’t reward this directly. It rewards the downstream output. But the downstream output is worse without it.

Cowen looked at the downstream output and concluded that price theory is in decline. That’s fine, but we can’t forget the upstream reasoning. And—shocking for a guy with a newsletter—I think the broader reasoning matters a lot. Unfortunately, I do believe the reasoning has declined too but not as much as looking at research would suggest.

What reasoning produces

Let’s talk about COVID inflation again. Someday I will have a new example, but today is not that day.

Consider what happens when people actually need to understand the world in real time. People come up with all sorts of theories. We got the greedflation, we got sellers’ inflation, all sorts of stuff made up ad hoc.

Price theory has a clear framework for this, and it doesn’t require some machine learning or a structural demand model or a novel identification strategy. When marginal costs increase, profit-maximizing firms reduce their markups because they can only pass along part of the cost increase. Rising markups are the signature of demand-side pressure. If you see profits rising alongside prices, the most natural explanation in the standard model is that demand shifted out. The greedflation argument got the causation exactly backward.

Josh wrote a whole newsletter called “Price Theory as an Antidote,” making exactly this point. Price theory disciplines one’s thinking, and the greedflation episode is his prime example. As Josh put it, if rising markups caused the rise in inflation, “one would need to explain why firms, across the board, suddenly and simultaneously increased markups.” They couldn’t, because the standard price-setting model says rising marginal costs compress markups, not expand them. If markups are expanding, demand is doing the work. That’s a price theory question.

We understood this pretty quickly using basic reasoning, before the formal research caught up. But the initial price theory reaction was basically right. As Christopher Conlon’s recent piece argues , “prices rising faster than costs is a generic feature of markets with market power and does not, by itself, indicate a change in the nature of competition.”

None of the credibility revolution’s tools resolved this debate. What resolved it was economic reasoning. Someone asked whether the mechanism made sense, and it didn’t.

Scott Sumner, in his response to Cowen’s book, makes a related point from a different angle. He describes how smart scientists routinely say foolish things about economics, like insisting that greedy firms won’t pass on lower input costs to consumers. Sumner’s example: tell a room of high-IQ scientists that reducing fees on property developers will lower home prices, and a significant proportion will roll their eyes. “These developers are greedy, and they won’t pass on lower costs to consumers.” But as Sumner puts it, they are greedy, “which is precisely why they’ll pass on lower input costs to consumers.” If the profit-maximizing price falls when input costs fall, a profit-maximizing firm charges less. It’s the symmetry. Economics, Sumner argues, is “extremely easy but also quite difficult.” The models are simple. The intuitions are hard to hold onto without them.

360,000 factors

Cowen is particularly worried about machine learning. He describes a paper that built a model with 360,000 factors to predict stock returns, reducing pricing errors by 54.8 percent compared to the classic six-factor model. He asks: “What kinds of intuitions do you think possibly can be supported by those 360,000 factors?”

He asks the question and walks past it. But it’s the right question.

When you have 360,000 factors, you can’t reason about the model using the model. You need something outside it. A framework for asking whether the factors are capturing real economic mechanisms or statistical noise. Whether the out-of-sample performance will hold. Whether the returns are compensation for risk or artifacts of the sample.

Those are not questions that more data answers. They require economic reasoning about what’s generating the patterns in the first place.

“Marginalism will not die,” Cowen writes, “but we will automate it, and in the process drain marginalism away from the human intuitions of most economists.” But automating the technique doesn’t automate the reasoning about whether the technique’s output makes sense. It increases the volume of output that needs reasoning applied to it.

Lynne Kiesling, in her response to Cowen, makes this argument with price theory’s own tools. AI, she argues, is a shock to the relative prices inside the academic knowledge economy. It lowers the cost of routine cognitive labor: literature search, coding, specification exploration, data cleaning. Those tasks are becoming cheap. What becomes relatively more scarce? Judgment. The ability to decide what’s worth studying, which mechanism is first-order, which assumptions are plausible, which results travel across contexts.

Kiesling’s point is that AI doesn’t weaken the case for economic reasoning. It raises the return to it. If execution becomes cheap, the binding constraint is knowing what to execute and whether the result means anything. That’s a price-theoretic argument about the profession itself, and it points in exactly the opposite direction from Cowen’s conclusion.

Cowen proves his own point.

I actually think that Cowen’s book is evidence against this thesis.

The first chapter is full of price-theoretic reasoning. He explains why under some Chinese compensation schemes, drivers who hit pedestrians face a perverse incentive. The fine for killing a pedestrian can run $30,000 to $50,000, but lifetime disability payments for a surviving victim can reach hundreds of thousands. At the margin, the financial penalty falls if the victim dies. He explains why homeless people cluster in high-amenity cities, since they don’t pay rent, and the high prices signal good amenities rather than costs to them. He analyzes abortion clinic access through the lens of marginal valuations. He examines the Obamacare subsidy debate by asking what a large quantity response implies about how much people actually valued the insurance.

All of this is exactly the “explain to a well-educated non-economist” reasoning he says is in retreat. He does it throughout the book he wrote to announce its decline. Sure, he and I are not the journal-centric heart of the profession, but we exist.

And then in chapter 2, he explains why price theory is declining, using price theory. The mechanism, he argues, is competition. “Once you reach a certain level of microeconomic prowess, it is hard to ‘understand marginalism better than the other person.’ So competition moves into other fields of endeavor.” When a margin is exhausted, competition shifts to other margins. The same framework that explains why firms diversify explains why economists stopped differentiating on price theory.

He applies competitive market logic to explain the decline of competitive market logic. For an economist, the reasoning is so deeply embedded it operates even when you’re writing its obituary. We can’t lose that and need to protect it in education going forward, but it is not lost yet.

The real repricing

As a research technique, price theory’s market share has fallen. Cowen documents this honestly.

But as economic reasoning, the upstream discipline that tells you whether a question is well-posed and an answer makes sense, the demand has grown. More empirical results, more ML predictions, more computational models, all producing outputs that someone needs to check against basic economic logic. Does the mechanism work? Does the sign make sense? What’s happening on the other margins?

Kiesling sees AI pushing in the same direction whereby cheapening execution, raising the return to reasoning. Sumner sees the continued difficulty of the reasoning itself, the “seeing around corners” that even smart people fail at without the model.

The question I’d focus on, and the question this newsletter exists to answer, is not whether price theory is a competitive research technique. It is one input but not the showy part. I’d focus on whether the world still needs people who can hear a claim about the economy and ask whether it makes sense.

It does.