Arnold makes me want to be a little more careful about my position on health care. By way of background, I’ll admit that I get most of this from Robin Hanson. But I was so incredulous when I first started hearing his views, that I’ve canvassed every other health economist who crossed my path. (Maybe 5 total, but a pretty random sample). And I was surprised to learn that Robin was telling me the standard conclusion. Robin is only radical in taking the standard conclusion seriously, while other economists admit that health care doesn’t do much for health, then get back to obsessing about the uninsured.

So what is the standard conclusion?

1. The marginal benefit of health care is on average about zero. How big of a margin? Roughly the last 30% of dollars spent.

2. The total benefit of health care is small relative to total increase in life expectancy. We live decades longer than we used to, and generous accounting says that a couple of those years come from better health care.

For cites and details, check out the section “Medicine and Health” in Hanson’s “Showing that You Care.”

My tweak on this literature, you may recall, is that the small effects of health care are probably a blending of large positive effects (e.g. saving my twins) with large negative effects (e.g. killing my grandpa).

Perhaps now that I’ve been more careful, Arnold won’t disagree as much. But maybe greater clarity will just make my errors more obvious! In any case, Arnold’s latest says:

Co-blogger Bryan and I have been arguing over two issues, one substantive and one methodological.


Substantive issue: I say that health care probably is effective. He says that there is no evidence that it is effective on average.

This overstates my position, though perhaps the fault is mine. “No evidence” is almost always too strong. I do believe that the bulk of evidence says that marginal health care is ineffective on average, and that the total effect of health care is a small fraction of the gain in life expectancy.

Incidentally, Paul Krugman says that there is no evidence that U.S. health care is effective at the margin, meaning that the additional money that we spend on health care relative to other countries does not lead to measurably better health outcomes.

Yes, and I’ve said that Krugman is treading on thin ice. If people knew how small the benefit of health care was, the hand-wringing about the uninsured would fizzle.

Methodological issue: I say that looking at aggregate data on health care outcomes is lazy econometrics. There is no substitute, I argue, for careful disaggregated studies.

Now, in response to my post on Murphy and Topel’s article about the huge benefits of health care, Bryan is sticking to his substantive position, but reversing himself on methodology. That is, he remains skeptical that health care provides benefits. Now, he argues that Murphy and Topel are failing to control for other factors that affect health.

Arnold conflates two different things. The literature I’m promoting tries to figure out how much health care has improved health on average. I think that’s a very valuable project. Assuming Robin’s critique of Murphy and Topel is correct, however, they don’t do this. They give an aggregate number, but to reach it they beg the vital question about the effectiveness of health care. (That might not be such a severe sin, if there weren’t already a big literature out there that says the opposite!)

I am more than willing to concede the substantive point that Murphy and Topel have failed to provide conclusive evidence of the benefits of health care. However, I claim progress on the methodological point that the question needs to be settled by careful, disaggregated analysis rather than lazy macro statistics.

The problem with disaggregation is that people lose the forest for the trees. It would be good to know which treatments have the biggest bang per buck. But given limited research resources and attention spans, it is better to know how helpful we can typically expect health care to be. And the answer is: A lot less than most people think.