Random Critical Analysis (or RCA, in honor of the dizzying array of acronyms the author there employs) has a blog post up in response to a recent New York Times article by Austin Frakt and Aaron Carroll in which they (Frakt and Carroll) reiterate the case famously made by Uwe Reinhardt that the astronomical costs of the US healthcare system can be attributed to the astronomically high prices found in the US healthcare system (rather than, say, higher utilization), which has always seemed a pretty straightforward and obvious case to me. RCA claims that this is wrong. The thrust of their (RCA’s) argument seems to be that, if you look at healthcare spending in relation to the average disposable household income of various countries, rather than, say, as a % of GDP, US spending is not out of line at all, and therefore there’s little to explain in the first place. This is not the only argument that RCA makes, but it underlies most of the post, and it’s the one I am going to address here, briefly. RCA’s full post is very thoroughly written, very interesting, and well worth reading (as are the previous posts that it summarizes and links back to).
RCA claims that using real household disposable income is the proper way to put US healthcare spending in context, for two reasons. First, it is an incredibly accurate predictor for how much per capita health expenditure there will be in any given country. Second, it most reflects how people would intuitively think about healthcare spending as “domestic opportunity cost (i.e., how much real consumption is foregone when buying a comparable bundle of health care in a given country as compared to the OECD average)”.
RCA provides plenty of statistical evidence that the first aspect of the household disposable income variable is very true, i.e. it is a great predictor of healthcare expenditures. First I will give you their slick graph demonstrating this:
and then I will show you my terrible graph, just as evidence that I was able to track down the data and methodology that they use for each axis here (and that task proved difficult for some of their other graphs), as I will later expand on that data:
As you can see, our graphs are basically the same, other than one being clearly labeled and nice to look at, and the other one not like that at all. There are a few other differences. RCA used 2010 dollars, while I use 2014 data in 2014 dollars, and maybe we somehow ended up with a different country or two included in our respective datasets. But it is basically the same, and demonstrates the same thing. Per capita disposable household income is a great predictor of a country’s per capita healthcare spending level. And when we look at things like this, the US is right in line.
The second aspect of thinking about this variable is, beyond its predictive power, what does it really tell us? Again, RCA claims that this is a good way to measure relative costs because, if I understand their argument correctly, if I have a lot of disposable income, then of course I’m going to spend more freely on healthcare. This makes sense of course, taken as an assertion of its own. However, as justification for using the per capita disposable household income variable, it falls short. The primary reason that it does is because per capita means average, and in any given country, income may be distributed radically unequally. Thus, the average American household might see little meaningful loss of consumption opportunity when buying healthcare, but this doesn’t preclude the possibility that even the vast majority of Americans might see a lot of lost opportunity.
We can step back now and ask what it means that there’s such a strong correlation between healthcare cost and per capita disposable household income. Well, if we want to take it to the very extreme, it could be that the healthcare industry is very profitable for the upper income brackets in certain countries and that there is every incentive to keep it this way. Since having a lot of money generally translates to having political power, or so we all assume, there is, beyond the desire, the actual means to keep healthcare costs high. Add in massive income inequality, and you have a correlation that (in this extreme speculative way of thinking) just tells you that in some countries, people are allowed to get rich off of healthcare and they follow that to the logical, market economy conclusion. To be clear, I am not saying that this all is definitely why we see a strong correlation between per capita disposable income and overall healthcare costs. I am saying that “Everybody in the country is equally well-off and so doesn’t mind that a stay in the hospital in their country costs as much as a house in another country,” is not an obviously correct assumption to make about why that correlation between per capita disposable income and healthcare expenditures exists.
While I can’t resolve this all, what I can do is offer a clearer picture of how these costs are actually felt by people in a given country, beyond a per capita measure. To do this, I start with the same graph as above, and the same data that it’s based on. But I add an extra layer to it, the Gini index, which is a common measure of domestic inequality. Specifically, I take the per capita household disposable income number, and divide it by the World Bank Gini score (on a scale of 1–100, where 1 means perfect equality, and 100 means perfect inequality.) In order to bring things back to scale, I multiply this by the average Gini score for all of the countries in the analysis. Now we have a really adjusted measure of per capita household disposable income, i.e., one that means something in the context of real people.
First, to demonstrate that this isn’t a very stupid idea, let’s look at the predictive power of using this as a variable. Here is a regression on per capita healthcare costs by the Gini-adjusted income measure, excluding the U.S.:
As a predictive model, it is very respectable. An R-squared of 0.82 isn’t quite as good as the 0.94 from RCA’s model, but it’s not as though we are outside of the actual world here. At any rate, the point of doing this isn’t to get a better prediction about the world, but rather a fuller interpretation of the world.
Now let’s add the US in and do our graph with a regression line. Again, none of the countries are labeled here. This is just down-and-dirty. I apologize. The US is circled in red, if that helps:
As can be seen, we are quite a bit away from our regression line. What does it mean? Well, it means that relative to the typical person in the typical country, the typical American is indeed facing absurdly high healthcare costs.
And it’s absolutely worth figuring out why, and fixing it. My guess is that, at this point, we can get costs down somewhat by doing things like streamlining the insurance process and lightening the administrative load (and if you look carefully, this is supported by both RCA and Reinhardt’s analysis). In the big picture, though, we’re probably better off focusing more on getting the Gini coefficient down, to get us more in line with where we should be. We could do both at the same time with a progressively financed single payer system.