Ambulance Chasing Econometrics

A recent post in Slate asks whether IMF loans cause Tuberculosis. The article refers to a recent study published in an on-line journal, apparently peer-reviewed journal. The author of the Slate post, Sydney Spiesel, is a physician at Yale, not an economist, so I suppose I should not be too surprised at the lack of skepticism in the article. Still it makes one wonder about the uses physicians make of statistics. If a Yale Professor of Medicine accepts this silliness, I am not sure I want to take his prescriptions for any illness I have.

One should read the published article, International Monetary Fund Programs and Tuberculosis Outcomes in Post-Communist Countries, which is available on-line here.

The basic premise of the study is that IMF programs result in reduced government spending, which leads to reduced spending on health, which leads to more TB. One might expect that the authors had tested this by looking to see if countries with IMF programs reduce spending on TB programs. That is, test the mechanism directly. Instead they regressed changes in mortality (over time across countries) from TB on the length of time a country operates under an IMF program. How hard is it to see the problem with this exercise?

Suppose we ran a regression of mortality from accidents on the amount of time people spend in ambulances. Does anyone doubt that we would find that ambulances cause accidents. This would especially be so since people who never get in an accident never find themselves in an ambulance. Would Dr. Spiesel tell us not to use ambulances to stay healthy? Well, countries that did not experience economic crises do not need to go to the IMF (for example, Slovenia, in the data sample used by the authors).

What the authors of the article and the physician miss is that admission to IMF programs is not a random occurrence. Countries go to the IMF only when things get so bad that accepting the pain from the program seems better than accepting a collapse of the currency or of the State (sometimes both, remember Suharto and Indonesia?). But the factors that have caused the economic crisis may also be causing the health crisis. In that case the correlation with the IMF program is spurious.

One good thing about this paper is that the data they used is available online. One can examine the data and see what is going on. You can see, for example, that countries with stronger state capacity that have joined the EU (e.g., Czech Republic, Hungary) had short IMF programs early in transition and that weak states have had long stays with the IMF.