Ed's Odd Pigovian Argument

In today's Boston Globe, my Harvard colleague Ed Glaeser opines on the alternative minimum tax. In the process, he gives us this thought-provoking passage:

When parents decide to have kids, they are creating a massive benefit for their children. As much as parents may love their children, they are unlikely to reap all the benefits those children will offer during their lives. Economists often think that it makes sense to subsidize behavior that generates big "external" benefits for others: parenting seems like a particularly natural example of such behavior.

The pragmatist in me recoils at this argument. Ed is implicitly comparing the utility of having been born with the utility of never having been born. But since we do not observe those people who were never born, how can we possibly know their utility? Any theory that relies on things that are intrinsically unobservable (such as the utility of potential people who were never even conceived) seems suspect as a basis for public policy.

Update: Ed emails me some background: "If op-eds included citations, I should have cited Becker and Murphy's Journal of Law and Economics paper on the family and the state that makes this case quite explicitly."