Steve Berry on his path to Yale and the origins of the BLP model
Beginning in the 1980’s, economists in the field of Industrial Organization revolutionized economic measurement by taking core concepts from theory to application. One of the early leaders in this movement was Steve Berry, economics professor at Yale University and faculty director of the Tobin Center.
Berry read Isaac Asimov’s famous science fiction Foundation trilogy as a kid, and he credits it with influencing him to study economics (Planet Money has previously covered how 'Foundation' has inspired economists). Berry did his PhD at Wisconsin after a brief time working in a savings and loan bank, where he was relegated to analysis rather than management. He felt the assignment validated his identity as a researcher, and if that weren’t confirmation enough, his job market paper on competition in the airline industry won The Econometric Society’s Frish Medal. Berry joined the faculty at Yale in 1988. In 2025, he was appointed as Sterling Professor of Economics, the highest honor that can be bestowed on a Yale professor.
Limits of early supply and demand models
At Wisconsin and Yale, Berry found himself surrounded by impressive econometricians, but his interest was in using data to model the impact of policy decisions. He knew supply and demand was a “tremendously powerful” tool, but it lacked enough specificity for the questions Berry wanted to answer.
In an interview with the economist Scott Cunningham about his impact on the field of Industrial Organization (IO) Berry talked about the limits of basic supply and demand models taught in an introductory economics class.
In the interview, Berry noted that the undergraduate demand and supply model is an amazing tool. If you know the separate demand and supply components of the system, you can predict the effects of policies that have not yet been tried, like tariffs or a carbon tax.
A classic question of empirical economics is how to learn about demand and supply from data. This is hard because the model has a feature called “simultaneity.” You can’t say that price causes quantity or that quantity causes price; they are jointly determined in the system.
A great achievement of empirical economics in the 20th-century was figuring out how to deal with this in the context of simple models of demand and supply. However, the freshman undergraduate model overlooks key features of reality, like imperfect competition and the fact that products within a market are often highly differentiated from one another. The market doesn’t just produce “cars,” it produces many different kinds of cars.
The Berry, Levinsohn, and Pakes (BLP) Model
When Jim Levinsohn presented a paper on the auto trade at Yale, Berry, and his colleague Ariel Pakes saw the potential to combine Levinsohn’s work with their own to create a powerful tool that could address the simultaneous impacts of a policy change.
One of the best tools for measuring demand for differentiated products at the time was the McFadden “discrete choice” model, which did not address how prices were determined except through a fixed constant that represented the unobservable elements of market demand.
By combining their joint expertise, they felt they could crack the problem. They could extend the discrete choice framework to incorporate a rich supply side featuring imperfect competition, coming much closer to a model that reflects reality. They showed how to deal with simultaneity in this more complex setting. Their data set included 20 years of auto prices, sales and characteristics such as size, fuel efficiency and number of doors.
For example, “we wanted to know: if you change the price of one car, what would happen to the sales of that car and similar cars?” That kind of prediction greatly affects how an economist would evaluate the effects of different policies. Their new model came to be known as the “BLP model.”
One model, endless applications for policy analysis
The ideas in the BLP model did not just apply to the auto industry. Many markets feature some combination of differentiated products and possibly imperfect competition. This includes, for example, hospitals, universities, and housing.
The paper written on the auto market has over 8,000 citations today. The tool has been extended in multiple dimensions, including by the original authors, and it is used across a variety of fields, from merger analysis in antitrust cases to the analysis of health policy.
“It's a little frightening, by the way, when you create a tool or crank, because people will crank it,” Berry said in his interview with Cunningham. “They'll crank it on the right idea and they'll crank it even when it's not such a great idea.”
The revolution in Industrial Organization had many contributors. Berry, Levinsohn and Pakes are certainly among them.
Read more from Berry’s students and colleagues about his impact on the field of industrial organization.