No snarl intended, Arnold. I’m a big fan of Cox and Alm, too. But just pointing to basic numbers only gets you so far. Cox and Alm can show that living standards are rising, but if you want to figure out why, you’ve at least got to compute some conditional changes in mean (e.g. change in the U.S. vs. North Korea). And once you’re willing to look at conditional means, there’s no reason to reject regression, which is, in essence, a sophisticated way to calculate mean values conditional on more than one thing.

For example, suppose you wanted to understand party identification in the U.S. Regressing party on income, church attendance, race, sex, and other plausible influences on party seems like the obvious approach to me. It’s not perfect, but it’s pretty good.