Currency Crises, Capital Controls, and Selection Bias

Michael Hutchison
Department of Economics University of California, Santa Cruz

Abstract

Whether or not legal and administrative controls on international payments (capital controls) are effective in insulating economies from currency and balance of payments turmoil is an important but unresolved policy issue. Theoretical and empirical literatures, and countless policy experiences, suggest that capital controls may or may not be effective depending on a host of economic, administrative and political factors. Unfortunately, this is not very helpful to economic policymakers seeking practical benchmarks in guiding their decisions on whether to restrict international payments in the hopes of warding off currency attacks.

This paper contributes to this literature by focusing on why capital controls are imposed— Are capital controls associated with greater vulnerability to currency crises mainly because of the special characteristics and “self selection” of countries that choose to restrict international payments? That is, countries with macroeconomic imbalances, financial weaknesses, political instability, and/or institutional problems may choose to impose capital controls in order to avoid difficult economic reforms or to avoid capital outflows that may trigger a crisis. Conversely, countries with sound macroeconomic and political environments and more robust financial systems and institutions are not only less likely to experience crises, but also less likely to enact capital controls.

We address this issue by employing the matching and propensity score methodology that was developed precisely for estimation problems associated with sample selection bias. We use “nearest neighbor”, “radius,” and “stratification” matching methods, as well as a “regression-adjusted” matching estimator—all methods designed to account for selection on observables bias— in a panel investigation of 69 developing countries over 1975-97. We develop a model of capital controls—identifying those economic and political characteristics that lead governments to impose restrictions—and match countries with similar predicted propensities to impose controls but with different experiences. These groups are tested for differences in the frequency of currency crises. All of our results suggest that, even after controlling for sample selection bias, capital restrictions are associated with a greater likelihood of currency crises. These results are robust to changes in methodology, equation specification, and so on. The point estimates suggest that countries with capital controls are 5-11 percent more likely to experience a currency crisis in any given year than countries that do not—even though both groups have almost identical ex ante likelihoods of imposing capital controls (i.e. very similar economic and political characteristics).

Seminar Paper