It's not every day there's a post by Casey Mulligan that I agree with

But this is one. Casey Mulligan says that we should treat survey respondents like "mercenaries, not slaves" (using the old Milton Friedman line about the all volunteer force).

I'm interested in the criticism about whether this will make the survey less representative. If this is actually true, you'd think it would make low income households over-represented. From a statistical standpoint that's fine - we can always re-weight to maintain representativeness. But it's not necessarily a bad thing that we over-sample low income families. It makes estimates about those households more precise. Indeed, many surveys deliberately over-represent these households for that reason!

There is another option, of course, which is to have an eminent domain approach to peoples' time: make them take the survey, but then compensate them for it (which is basically what we did under the draft, which kind of throws a monkey wrench into the old Friedman phrase, doesn't it? but "mercenaries, not people who are approached with the eminent domain powers of the Congress" doesn't have quite the same ring to it) .