Fetal Voters, Redux

This post by Amy and these two by Eugene caused me to go back and read more carefully the article that claimed legal abortion may have cost Democrats elections by shrinking their numbers. I commented on the article here, after skimming it and finding the argument superficially plausible. (In my defense, I did say that I hadn’t checked the numerical analysis carefully!) It turns out, on closer inspection, that there are a number of flaws.

First, a lot of the figures about who has had abortions come from survey data, which is notoriously unreliable on questions about personal and potentially embarrassing matters.

Second, the question asked (“As far as you know, has anyone close to you had an abortion?”) doesn’t capture the relevant information even if answered truthfully. In order for the answers to tell us anything about how many Democrats and Republicans are having abortions, we’d have to know people’s propensity to feel “close” to someone of the opposite party.

Third, although it’s certainly true that people tend to “inherit” their political views from their parents, it’s not a 100% correlation. But the author implicitly treats all unborn children of Democrats as guaranteed Democrats, and all unborn children of Republicans as guaranteed Republicans. Taking into account the fact some children do not adopt their parents’ politics would shrink the size of the claimed effect (though not eliminate it).

Fourth, although the author claims to have taken the relatively low voting rates of 18-24 year olds into account, I don’t think he has. Working backward from his numbers, I figure he must have assumed a 47% voting rate for that age group, whereas the actual number (according to this site) is only about 42%. Again, that would tend to shrink the size of the effect (though not eliminate it). The author might respond, however, that it only delays the effect, since eventually the “missing voters” would reach the age ranges with higher voter turnout.

I still think the author’s argument has prima facie plausibility, because the crucial premises (that Democrats have abortions at higher frequency than do Republicans, and that children have a tendency to adopt their parents’ politics) are believable and, if true, lead in fairly straightforward fashion to the conclusion that legal abortion should reduce the number of future Democrats more than the number of future Republicans. But it would take much better statistical analysis to prove that conclusion with any degree of certainty.

Also, I stand by the final point made in my previous post: the phenomenon, if real, does not indict Democrats any more than Republicans. If Democrats diminish their electoral chances by supporting abortion, it follows that Republicans diminish their electoral chances by opposing it.

UPDATE: Alex Tabarrok points out that various factors may offset the alleged reduction in births due to abortion. First, the availability of abortion may just change the timing of births. Second, the abortion option probably increases the frequency of sex and decreases the use of birth control. Both factors would, again, tend to shrink the size of the "missing voter" effect. Alex suggests that, in theory, the effect could even go in the opposite direction.