IMF Programs and Growth: A Source-Informed Robustness Reanalysis

Abstract

This article reassesses the meta-analytic evidence on the effect of International Monetary Fund programs on economic growth. The point of departure is the influential meta-analysis by Balima and Sokolova, which assembles 994 estimates from 36 studies and reports a positive average effect with substantial heterogeneity. The reanalysis presented here imposes four stricter requirements. It treats the study, rather than the reported estimate, as the primary inferential unit; it uses a source-informed classification of causal credibility; it models within-study dependence through study aggregation, correlated-effects sensitivity, multilevel CR2 inference, and robust variance estimation; and it evaluates publication-bias sensitivity through Egger, PET, PEESE, WAAP-like top-precision analysis, p-curve diagnostics, trim-and-fill, and exploratory selection models. The central result is not that IMF programs reduce growth everywhere, nor that the true effect is exactly zero. The result is narrower and stronger: the positive average effect in the aggregate literature is not robust once dependence, publication selection, heterogeneity, and credibility of identification are treated as first-order concerns. In the most defensible specifications, the average effect is statistically indistinguishable from zero, while equivalence to a substantively negligible effect is only partially supported and depends on the chosen equivalence bound.