We all know that health care expenditures as a percentage of GDP (per capita) has increased greatly over the last few decades and that this phenomenon has been observed in the US and foreign countries alike. Recent research has tried to understand the determinants of this increase and one common explanation is that health care is a luxury good: that is, as incomes rise people demand more and more of it. In some sense, this might make rising health care expenditures less ominous: we spend more only because it is an expression of our preferences.
Empirical evidence linking incomes to health generally supports the luxury good hypothesis and is based on establishing correlations between the two in micro and aggregate data. However, in a recent working paper, Daron Acemoglu, Amy Finkelstein, and Matthew Notowidigdo argue that this evidence may be misleading for two reasons. First, simple correlations do not capture other unobserved factors associated with income that might affect health. Second, such models do not distinguish between/account for partial and general equilibrium effects: for example, rising demand for health care generated by income may increase spending both through increased local demand, but also through supply side changes in medical technology or practices that respond to changes in demand. In addition, rising incomes and demand may lead to changes in the politics around health care and health services. In either case, it is important to understand both partial and general equilibrium t truly characterize the relationship between income and health.
Acemoglu, Finkelstein and Notowidigo try to get around both of these issues by utilizing shocks to oil prices. The basic idea of their paper is the following:
1) Look at a bunch of smaller areas which may or may not have pre-existing oil industries.
2) Changes in world oil prices, which are not driven by small industry in any single area will affect localities with oil industries differently than those without them. Thus, these two types of areas will experience different "shocks" to income. (Thus, the effect of income on health care demand is identified by the interaction between pre-existing oil industries and world oil price shocks). The next step is to look at the association between predicted income from oil price shocks and measures of health care demand.
3) Establish that general equilibrium effects occur at the level of localities and that it is unlikely that changes in local demand have equilibrium effects on larger regions (such as nations or the world).
The authors findings strongly suggest that health care is NOT a luxury good and that rising incomes likely cannot explain an important portion of the rise in health care expenditures.
Neat paper on an interesting area of research, and definitely worth reading.
4 comments:
Definitely an important topic but its a bit of a weird identification strategy, in my opinion.
You'd have to be convinced that wealth determined by transitory oil price shocks is a decent measure of permanent income, which is probably what we are truly interested in...
thats a good point -- and i'm not at all convinced that transitory oil price shocks translate into permanent income. maybe the question of whether health is a luxury good is just one of those mysteries that we can never know about.
Now that I think about it, it's not implausible for these shocks to translate into permanent income. I was just watching "Friday Night Lights" and thought back to the book upon which the book is based (a true story). The book spends a great deal of time discussing West Texas towns during the oil boom years. Not only did this boom last for a while, but people's spending habits changed greatly, in ways that speak to (expectations about) increases in permanent income. So, if this were the case, or even if people the shock to imply a permanent shift in income, we could probably still learn a great deal about the substantive question from this paper.
However, this discussion hits at another potential issue in the paper: west Texas towns were sensitive to price shocks but towns in, say, Alabama, were not. Differential trends in health care usage or other factors impacting health care usage could have biased the results. The authors go through some trouble to show this was not the case, but the spectre of differential trends is always out there.
Post a Comment