Did the IMF Cause TB Deaths in Eastern Europe?

♠ Posted by Emmanuel in , at 10/25/2008 03:26:00 PM
This post is a timely one for very unfortunate reasons. As you probably know, Hungary, Ukraine, and Belarus have entered discussions with the IMF for lender of last resort funding. One of the points the IMF is emphasizing with this round of crises is that it will be sparing with the use of politically unpopular conditionalities. (I hope so.) Think of these conditionalities as similar to loan covenants. For instance, you may lend money to an errant relative provided that he cuts spending on alcohol and looks for gainful employment. Like other lenders, the IMF is interested in getting paid back. Accordingly, it has used conditionalities that it believes are conducive to placing a borrowing country in a position to make repayments and improve its financial outlook.

Unlike many other lenders, however, the IMF's status as lender of last resort allows it further discretion than most over a country's economic policies. Among other things, conditionalities have included targets for budget deficits, current account deficits, and inflation. A common assertion made by (usually left-leaning) critics of the IMF is that its conditionalities include strict limits on national spending that have resulted in cuts in health and education expenditures. Insofar as these cuts have resulted in adverse social effects, the IMF is implicated.

Something that recently caught my attention was an academic study concerning how IMF lending during the post-Soviet era is said to have resulted in increased difficulties with tuberculosis in Eastern European countries through the mechanism described above. The chart to the left is taken from this study of Eastern European countries. The IMF has famously been faulted for exacerbating crises by ignoring the social effects of its policies--policies which may have instead worsened matters via instability-inducing social dislocations. If true, this study would serve as more ammunition for the IMF's legion of critics (admittedly including myself on occasion) at a time when it is about to swing into motion in a big way.

The IMF offers many points of contention as to why this particular study has doubtful conclusions. In quoting several studies with the IMF imprimatur for counterarguments, I think few critics will be convinced. However, one thing that strikes me as curious is that the authors do not appear to have incorporated the most important study variable of all: spending on health. A Millennium Development Goal that Eastern Europe is having difficulty meeting is halting and reversing the incidence, prevalence, and death rates associated with tuberculosis by 2015 (see target 6.9). Accordingly, the study's authors are considering if this goal has been hindered by IMF lending conditionalities. The following line of argumentation must hold:

[IMF lending in Eastern European countries][led to cuts in public health expenditures],[resulting in increased tuberculosis problems].

Notice the three brackets I have placed. What the IMF points out is that the middle part of the argument has gone missing in the study cited, leading to an argument in this form:

[IMF lending in Eastern European countries][resulted in increased tuberculosis problems].

Notice how omitting the middle bracket makes the argumentation uncertain. Some might even say it sounds like a conspiracy theory. In the specification of their regression model, the authors have used the presence of IMF lending as an independent variable together with control variables such as democratization, occurrence of military or ethnic conflict, urbanization, population dependency ratios, and population education levels. (Control variables account for possible alternative causes of tuberculosis problems.) Each in turn, they use tuberculosis incidence, prevalence, and mortality rates as the dependent variable. As you can read in their study, "IMF program participation was associated with increases in tuberculosis incidence, prevalence, and mortality by 13.9%, 13.3%, and 16.6%, respectively."

Firing back at the IMF, David Stuckler, one of the study's authors, cites a whole list of variables they also incorporated in post-hoc tests. However, throwing in the kitchen sink does not address the study's Achilles heel which the IMF pointed to above: changes in health expenditures are not accounted for. Although "government spending as a percentage of GDP" is considered, this clearly isn't identical to changes in health expenditures during the period in which IMF lending was present.

Thus returning to the question posed in the title, I cannot offer a definitive answer here based on this study. The process by which IMF lending led to increased problems with tuberculosis is not given adequate exposition. OTOH, while the IMF points out a fundamental flaw with the paper, it doesn't offer conclusive proof that it is not complicit here, either. Another of the study's authors, Sanjay Basu, makes a pretty strong statement in a Huffington Post interview:
Question: Are you able to say what the increased mortality rates mean and what the increased incidence rates mean in terms of overall numbers?

Basu: In a statistical sense, you can say how many statistical deaths have resulted. In terms of deaths, it was on the order of tens of thousands. In terms of incidence of new cases, it was on the order of hundreds of thousands. Of course, these are all jumbled in reality, but in a statistical sense, that is part of the effect due to the IMF programs.
Associating IMF programs with thousands of deaths from tuberculosis is no laughing matter. To make such an assertion stick requires that a study be nearly unimpeachable. At this point in time, I am unconvinced that it is for reasons given above. That a main variable of interest is omitted doesn't inspire confidence. To their credit, the authors move in the preliminary direction of using a path analysis or structural equation model (SEM) instead of a regression model. The advantage of the likes of SEM over regression is that the former can handle different levels of variables. In plain English, instead of just considering two of the brackets above as the authors do with regression, SEM can consider all three brackets simultaneously. Again, however, what's in the middle bracket is important here--public health spending, not government spending as a percentage of GDP or any other variable.

If the authors are serious about fending off what are legitimate criticisms, then they should earnestly work towards addressing them to convince us that methodological issues are ironed out.