IMF Program Suspensions: Theoretical Issues in Model Specification

Abstract

Recent criticisms of the International Monetary Fund focus on its inattention to the role of political institutions in shaping the ability of states to meet the conditions required in austerity programs. To what extent are these criticisms, which I term the information hypothesis, valid? As an initial pass at this question, I assess the Fund’s response to the approach of elections in countries under its programs. The conventional wisdom is that elections lead to program suspension, though this raises the question of whether the IMF responds proactively to the approach of elections by altering conditionality. The problem with using conventional discrete choice models to answer this question is that they force the variance across cases to be a constant. The variance surrounding program suspensions, however, is a proxy for the countries “margin of error” between economic performance and program suspension. As a result, models that force the variance to be constant across cases limit what we can learn about the Fund’s response to political shocks. Using a heteroskedastic probit model, I find that the approach of elections alters the variance surrounding program suspensions, providing another piece of evidence to counter the information hypothesis.