Last week I posted the following final exam question:

Some sociologists have argued that discrimination on the basis of educational credentials should be illegal.  What do the human capital and signaling models of education predict about the effect of such a law?

Lots of thoughtful responses in the commentsAnthony and Silas Barta are my favorites.  My answer:

1. In a pure human capital model, employers have perfect information about workers’ productivity, so education only pays because it makes workers more skilled.  Assuming the law correctly identifies “discrimination on the basis of educational credentials,” there would be no effect on behavior.  The law simply bans something that no employer does, so no student has an incentive to revise his educational plans.

In the real world, however, discrimination laws usually take different average group outcomes as prima facie evidence of discrimination.  Employers who disproportionately hired and promoted the better-educated would expose themselves to legal action.  To reduce their legal risks, employers would, all else equal, prefer uncredentialed workers.  Firms would scramble to find qualified workers with poor credentials – and often settle for subpar “token” drop-outs to show their good faith.  Without draconian enforcement, however, employers would continue to prefer educated workers while loudly denying it.  Still, the net effect is that students would reduce – or deliberately conceal – their educational investments.  In a pure human capital model, this is a clear social loss.

2. In a pure signaling model, in contrast, any ban on educational discrimination has a big effect.  Assuming the law correctly identifies “discrimination on the basis of educational credentials,” imperfect enforcement is the only remaining reason to go to school.  Employers will try to find legal signals of intelligence, conscientiousness, and conformity, and workers will strive to provide these alternate signals.

But in the real world, to repeat, discrimination laws usually take different average group outcomes as prima facie evidence of discrimination.  New signals that closely mimicked the result of educational signaling would themselves be legally suspect.  So once again, employers would go out of their way to hire less-educated workers to shield themselves from prosecution – while covertly preferring educated workers.

In the pure signaling story, the net social benefit of banning educational discrimination is unclear.  Since education is currently heavily subsidized, it is entirely possible that the alternate signals that emerge will rank workers at a lower social cost than the status quo.  Even if the ranking of workers is absolutely worse, we have to weigh inferior job match against the massive social savings of slashing wasteful “investment” in education.