Selection Bias and the Output Costs of IMF Programs

Questions over the role of the IMF in the economic development and adjustment in developing countries have been the topic of intensive research and debate in recent years. Although most studies find that participation in an IMF program helps facilitate balance of payments adjustment, research in this area almost uniformly finds that growth is reduced at the same time (e.g. Bordo and Schwartz, 2000; Przeworski and Vreeland, 2000). In this paper we emphasize that the evaluation of the benefits and costs of participating in IMF-sponsored stabilization programs is complicated by the fact that countries typically enter into an agreement with the IMF only when facing dire economic problems. We argue that the sample selection bias is mainly responsible for the common perception that real output growth declines because countries choose to participate in IMF programs. This article uses four recently developed "matching" statistical methods (e.g. Heckman et al., 1997 and 1998; Rubin and Thomas, 1992; and others), based on the "selection on observables" bias, to estimate the growth effects of IMF program participation. In contrast with the extant literature, none of the matching method results (nearest neighbor, strata, radius and regression-adjusted) find an adverse growth effect. Rather, there is some evidence of a positive impulse to economic growth when countries entering IMF programs are compared to the appropriate counter-factual (i.e. non-participating countries with similar characteristics).