Kudos to Matt Yglesias for making an important but infrequently discussed point that price discrimination can be socially optimal. However, I must point out that his statement that “price-discrimination in monopolist-dominated markets is socially optimal” is not correct. It would instead be correct to say price discrimination by a monopolist can be socially optimal.
One reason for this is because, as Matt points out, you never get perfect price discrimination in the real world. This means that airlines can’t know, and therefore can’t charge, each customers their exact valuation of a ticket, and instead must try and find some way to sort customers by willingness to pay. If the sorting is costly and wasteful then the price discrimination can be inefficient, i.e. not socially optimal.
The irony is that a very common example used in textbooks of inefficient price discrimination is in fact airlines (see, for example, Kwoka and White’s antitrust book). One way that airlines price discriminate is to offer tickets at a discount if the person is willing to stay over saturday night. Because businessmen are typically unwilling to stay longer than they need to and are also less price sensitive, this allows airlines to offer a discount only to the more price sensitive leisure segment of the market… thus price discrimination. The inefficiency arises because people staying longer than they otherwise would is a distortion, and the loss of welfare due to distortions like these can more than offset the welfare gains from the discount. This cost of separating the market can make price discrimination on net less socially optimal than not price discriminating in the first place.
There are other ways price discrimination by a monopolist can be socially suboptimal, but Matt’s point that they can also be socially optimal is the less understood point, so overall I can forgive him for not getting it quite right.
[UPDATE: Yglesias responds. I agree.]