Can a costly diploma prove ability even if it teaches nothing?

Yes, if it is cheaper for able people to earn than for less able ones. That is the surprising core of Michael Spence's 1973 idea. Employers want to hire able workers, but ability is invisible, which is the same asymmetric information that wrecks a used-car market. Spence proposed that the two sides could get around the problem if one party sends a signal that reveals some piece of relevant information to the other.

In his job-market signalling model, workers send a signal about their ability by acquiring education credentials. The credential is informative because the employer believes it is positively correlated with greater ability and difficult for low-ability workers to obtain. Spence makes one key assumption: good-type employees pay less for one unit of education than bad-type employees. The cost he means is opportunity cost, a mix of monetary and other costs such as psychological costs, time and effort, not just tuition.

And here is the twist. Spence discovered that even if education did not contribute anything to an employee's productivity, it could still have value to both employer and employee. The diploma works the way a hurdle at a race does: not by making anyone faster, but by showing who could clear it. A signal is most credible when it is differentially costly, harder or more expensive for low-quality signalers to produce than for high-quality ones.

Spence worked this out in a 1972 dissertation titled "Market signalling", supervised by Kenneth Arrow and Thomas C. Schelling, and shared the 2001 Nobel Memorial Prize in Economic Sciences with George Akerlof and Joseph Stiglitz. One of his Harvard classes, a graduate economics course, once included Bill Gates and Steve Ballmer.