PodBrowser
a16z Crypto

Why markets fail — and how to fix them (ft. Nobel economist Alvin Roth)

Tuesday, 14 July 2026 · 4 min read · Listen to the episode ↗

Nobel laureate Alvin Roth joins the show to explain why some markets fail even when supply and demand exist, and what it takes to fix them. He walks through the three conditions any marketplace must satisfy in sequence: sufficient thickness, congestion management, and trustworthiness.

Alvin Roth, Nobel laureate and Stanford professor, argues that market design differs from conventional economics because it asks how the world should work rather than how it does work. He identifies three sequential requirements for a well-functioning marketplace: it must first be made thick enough to generate matches, then congestion must be managed, and finally it must be made safe and trustworthy. When any of these properties is missing, markets fail in predictable ways.

Approximately 130,000 Americans experience kidney failure for the first time each year, roughly half a million are on dialysis, and fewer than 30,000 kidney transplants are performed annually, meaning most people who need a kidney die without one. Paying a kidney donor is illegal almost everywhere under the National Organ Transplant Act of 1984, with Iran as the sole exception. Kidney exchange emerged as a workaround within that legal constraint, and about one third of living donor kidney transplants in the US now occur through exchange. Pairwise exchanges must be performed simultaneously because a promise of a future kidney would constitute illegal valuable consideration, requiring four operating rooms at once. Non-directed donor chains remove that constraint by allowing each patient-donor pair to receive a kidney before giving one, making a broken chain non-catastrophic rather than a complete failure. Over 500 non-directed donors participated in the US last year. The legislation uses the phrase kidney paired donation specifically to keep the word exchange out of the description, and the clarifying amendment passed unanimously only after its sponsor, Congressman Charlie Norwood, died and colleagues named the act after him.

The deferred acceptance algorithm, introduced by Gale and Shapley, finds stable matchings where no participant has an incentive to re-contract afterward. Roth co-redesigned the National Resident Matching Program to incorporate couples seeking placement in the same location, a case where stable matchings can theoretically fail to exist. In a market with roughly 45,000 applicants, 35,000 jobs, and 2,000 couples, Roth and Elliot Peranson showed that instability is rare in practice. The medical match in the 1950s independently used a mechanism equivalent to deferred acceptance, and couples matching was not a concern then because almost all American medical graduates were men, compared to just over 50 percent women today.

Roth, Parag Pathak, and a colleague designed the New York City high school match after Mayor Bloomberg centralized authority over city schools. Under the old system, students could list only five schools, schools could ignore applicants who did not rank them first, and only three rounds of offers could run before August, leaving 30,000 of 90,000 applicants administratively assigned to schools they had not listed. After implementing deferred acceptance, unmatched students fell from roughly 30,000 to about 3,000. The de Blasio administration later eliminated the supplementary match that gave remaining unmatched students another chance at schools with empty spots, replacing it with waiting lists that have changed in procedure each year since.

Congestion, defined as not having enough time to evaluate or transact, prevents markets from scaling. A telephone-based deferred acceptance market for psychologists took approximately six minutes per rejection chain exchange, limiting participants to roughly ten transactions per hour and making a national labor market unworkable. Sidecar, an Uber competitor that asked users to choose between individual drivers, went out of business because the added friction caused congestion. Roth notes that Uber faces a congestion problem at San Francisco airport that is mathematically similar to deceased donor organ allocation, because rides, like organs, cannot wait and become unviable if not matched quickly. Airbnb resolved its sequencing problem by showing rooms only where availability existed rather than letting users queue through rejections.

Repugnance functions as a hard constraint on market design. Blood plasma is on the WHO list of essential medicines, yet the WHO and EU hold that supply must come from unpaid donors within each country, and no country has achieved self-sufficiency on that basis. The United States pays plasma donors and exports approximately 70 percent of the world's plasma supply. Germany permits prostitution but prohibits surrogacy and kidney exchange, while the US tends toward the opposite. Britain and Canada permit surrogacy but prohibit payment, resulting in fewer surrogates than needed. American surrogates are paid approximately 50,000 dollars per birth. Roth argues that bans require broad social support to be effective and that permeable bans create unprotected parties, including the children involved.

Roth argues that matching markets are fundamentally different from commodity markets because being chosen matters as much as choosing, and wages are personalized between each employer-employee pair, making prices alone insufficient to coordinate outcomes. He also distinguishes equilibrium from equilibration, warning that companies running online experiments on a small fraction of transactions risk invalid conclusions because once a treatment is applied universally, participants notice and change behavior, shifting the equilibrium itself.

This summary was generated from the episode transcript and can contain mistakes.