Prostitution and Income

I was reading an article the other day on the economics of the commercial sex industry and the author said we really don't have an estimate of the income elasticity of demand for prostitution services. I've actually only seen one measure of any elasticity of supply for prostitution and it was Levitt and Venkatesh's street prostitution paper that found the price elasticity of supply to be 2.0 (even though they mistakenly write 0.5 in the paper (p. 19). That estimate means if the wage rate of street prostitutes in Chicago went up 10% (say from $50/hr to $55/hr) that the number of prostitution services performed would increase 20% (say from 100 to 120). They found this mostly occurring on the external margin because they used an interesting instrument for demand which was a holiday. On the 4th of July, there was an increase in demand, and it led to both higher wages and more women working on that day - particularly by women who normally didn't work as prostitutes. That is only valid if the holiday shifted demand but not supply, though you can imagine it shifting both and so the elasticity would not have been identified.

Finding instruments that only affect one part of a labor market but not both is very difficult, since you're at the victim of letting random world events do these things for you, and then hope you can identify these natural experiments as they occur. Finding things that shift the incomes of clients, but not the incomes of providers, is even more difficult. In the article I read, the economists said they couldn't find any estimates of the income elasticity of demand (ie, by what percent does demand increase when client incomes increase by 1%?). The only thing they reported was evidence that wealthier people wanted to wear more condoms, which is hardly the same thing. Of course, it's not easy to get these things, and at least that paper acknowledged caveats, but I was really surprised to read this paragraph from Edlund and Korn's 2002 "A Theory of Prostitution" as I think they should know better. How did this get past a referee? It's not a big deal, the paper is still I'm sure fine, but this little piece of evidence is still annoying.

Prostitution has seen a secular decline in rich countries. For instance, while the Kinsey study, conducted in 1938–47, concluded that about 70 percent of the American white male population will ultimately purchase sex from a prostitute, the incidence among men surveyed in the NHSLS was 18 percent (Kinsey et al. 1948; Sullivan and Simon 1998). One may also note that during the half century that separates the two studies, male contact with prostitutes ceased to be considered common and normal. Part of this is undoubtedly due to better and cheaper contraceptives, which have increased the supply of noncommercial, nonmarital sex (e.g., Posner 1992, p. 132). However, higher income levels may also have played a role in the shift away from prostitution.

The only thing I have a problem with is noting that Kinsey concluded 70 percent of all American white males in the late 30s and early-to-mid-40s purchased prostitution services, but based on a NORC study in 1992, only 18% say they did recently. Reading this, if you didn't know anything about the NORC study (NHSLS was handled by NORC) or Kinsey's original sex studies, you'd think, "Hm. Looks like prostitution is an inferior good. Higher incomes and falling prostitution. May not be causal, but it's correlated."

The problem is that everyone knows Kinsey's original studies suffered from serious self-selection bias. It was because of the severity of that selection bias that NORC even decided to do the "new" sex study, NHSLS, at all. The problem was that Kinsey did not randomly draw his interviewees from the population, but rather from a non-representative social network. This was the homosexual friendship network, prisons, and the participants who voluntarily came to his lectures. Thus he tended to over-sample people with certain types of characteristics which were correlated with their sexual behavior and expression, and so tended to find much more larger estimates of various sexual behaviors and attitudes than NORC later found. All of the estimates were skewed, in fact. He estimated that 10% of all American males were currently homosexual (actively) and an even higher percentage had ever been, but NORC found closer to 1-2%. Do we also conclude homosexuality has been falling in the population over time, or do we like NORC simply note that his study was flawed and proper statistical sampling frames reveal more accurate current estimates. I'm not sure we can even compare these at all.

I would've preferred at least some recognition of this problem from Edlund and Korn. Like I said, not a big deal, but it seems like such a gross error. And the "evidence" is so typical of economists to just find something that could plausibly support your story, even if it's not as rigorously checked out as the rest of your paper. If Kinsey's problems were really obscure, it'd be no big deal, but everyone knows about them, so it's weird.