David Friedman - Dating Markets, Legal Systems, Bitcoin, and Automation

The interview opens with Dwarkesh Patel asking whether the dating market is efficient, noting that it is not a competitive market in the textbook sense because he cannot be substituted for another bearded twenty-year-old Indian man. Friedman answers that he has long been puzzled about why online dating does not work better. In principle, he says, dating is largely a sorting problem, and the sorting is hard because people are not interchangeable: potential partners also have different tastes.

Friedman describes two models from the marriage-market chapter of his price theory text. In the simple one, everyone is identical and the market sets an implicit price in the terms of marriage, such as how much housework the husband agrees to do and how far the wife gets to make decisions. Within that model he argues that legalizing polygyny has the opposite effect from what people expect. Because some men would bid for two wives, the price of a wife rises. Ordinary monogamous men still marry, but on worse terms, which means they must do more of the dishes. Women gain, since they get either a whole husband on better terms or half of a husband on terms good enough that they prefer him to all of someone else. Legalizing polyandry would produce the same effect in reverse.

The harder model is the sorting model, and Friedman admits that even his version is simple: everyone's value as a spouse varies, but it is the same to all potential partners. Reality is harder because a person's value differs from one suitor to another. He illustrates this with his own marriage. The woman he eventually judged to be a one-in-ten-thousand catch was not being pursued by anyone else at the time, because other people had odd tastes and did not recognize her value. He adds that they are still married, about 38 years on.

He then names search costs as the central issue. Economics largely ignores them because it lacks a good way to handle them, yet they are at the heart of dating. He had expected computer dating to help a great deal: each person could feed objective, measurable characteristics into a system, search a million profiles, filter to a subset and then interact within it. As far as he can tell, that has not happened. Marriage has not worked better since computers arrived, and the people he knows who tried computer dating did not find it very successful. He relays a story that OKCupid, one of the most successful services, got worse over time, having been designed by enthusiasts who did clever things and then taken over by businessmen who did less clever things. He flags a limit on this evidence: the people he talks to may simply differ from the mass market.

On efficiency, Friedman says the question is always efficient compared to what. The dating market is certainly inefficient compared with a perfectly wise, benevolent matchmaker, but no such person exists. A more interesting comparison is with a system in which parents choose the spouse. He recalls a long conversation on a flight from Bombay to Sydney with a woman from southern India whose husband, a physician, had been chosen by her parents. She had a veto, but the search was done by her parents, and she was happily married, while his own first marriage had broken up. He treats this as a tiny sample favoring her system, but says the real value was meeting an educated contemporary from a different set of institutions, since it is much harder to insist that our way is obviously right when a real person is defending the alternative. He concludes that he does not know whether modern Western courtship works better or worse than the traditional system, and notes that few things are further from the economist's rational-choice model than a man in love.

Patel suggests that arranged marriages may work better because the social structure makes people happier within the marriage, not because parents are better at choosing. Friedman reads this as possibly meaning that people blame themselves less when a match they did not choose goes badly, or that those cultures are simply better at keeping couples together and content. To separate quality of choice from cultural support, he proposes looking at people from India, at least southern India, who live in the U.S. His impression is that it is not unusual for a young adult in the tech industry to migrate, then go back to marry the wife his parents chose, or have the arrangement made here and live in America. That, he says, is as close as one can get to a controlled experiment.

He also describes a harder research design that computer-dating firms have not seriously attempted. Take all pairs who had a first date, work out from each side's self-reported characteristics whether they get a second date, and follow the pairs through to marriage and to how long the marriage lasts. That would yield statistical guidance on how to pair people. One could then compare couples produced by American courtship with couples paired by parents and see which came closer to what the computer recommended. He calls it a neat project and says it would be very hard to carry out.

Patel connects this to the bureaucrat God in Friedman's book Hidden Order, and Friedman says parents in those societies are in a sense that figure, with nearly enough authority, since their choice can be rejected but usually is not, though they are not perfectly wise. He explains the device as his attempt to make sense of economic efficiency, which raises two issues. The first is what counts as an improvement. There he follows Alfred Marshall rather than the conventional Pareto, Hicks and Kaldor approach, which he regards as a way of pretending not to make interpersonal comparisons while doing so. The second is efficiency compared to what, and his answer is the best outcome one can imagine by the chosen criterion, as produced by an omniscient, omnipotent, benevolent actor. Nothing will match that, but some things come close.