Evan Soltas
May 2, 2013

Misunderstanding Medicaid

I haven't written a personal blog post in a while. I am hoping I will get some time to do that this summer. I do have some things I think might be worth writing up. In the mean time, here are two unrelated comments on things people are getting wrong.

(1) A major health study is out from a randomized controlled trial in Oregon, which expanded Medicaid. Its results are feeding a new line of conservative criticism that the expansion creates no health benefits. (See, for example, this piece by Michael F. Cannon of Cato.)

Whatever one thinks of the Medicaid expansion, such a statement makes an egregious, basic error in interpreting statistical data. Let me quickly explain why. The questions presented in the Oregon Health Study are hypothesis testing problems. Basically, we might ask such a question as "Is the Medicaid expansion associated with, say, a decrease in the probability of a diagnosis of diabetes?"

But here's how studies test this question, more exactly. They create two separate hypotheses, a null hypothesis and an alternative hypothesis. The null, in our case, is "No, the Medicaid expansion is not associated with a significantly decrease in the probability." The alternative is that "Yes, it is associated."

What the Oregon study did was that it did not reject the null in favor of the alternative in several relevant instances. It's not that the study proved the null. It's not that it rejected the alternative. Sorry, that's just not how statistics work. It said that we do not have sufficient confidence in the improvement in health (which was there in the data) to say that it is statistically significant. This an important subtlety. Those who are saying otherwise don't know how to interpret statistical evidence properly, let alone interpret this study.

(2) Matt Yglesias has a new column out for Slate in which he tries to throw cold water on the idea of "disruptive innovation." I think Yglesias' basic argument -- that disruption has been overweighted as a way to better goods and services, and that we should look to other methods like incremental gains in quality -- might be wrong and refutable with some data I just happen to have read recently.

From a new NBER working paper by Daron Acemoglu et al.: "[T]he existing literature attributes as much as 70% or 80% of productivity growth in the United States to reallocation -- exit of less efficient and entry of more efficient firms."

I think this is pretty much the narrowest definition of disruption you can have. And it's a shockingly high share of all productivity gains. It's not clear to me that we spend enough time talking and writing and legislating about ways that we can encourage more corporate disruption.