Reserves and currency crises: the role of the exchange rate regime

ABSTRACT

Although it is widely accepted that more reserves are required to support a peg relative to a more flexible currency, there is lack of empirical evidence motivating this phenomenon. This paper shows evidence that international reserves help countries decrease the probability of a currency crisis. However, the alleviating effect of reserves on the likelihood of currency crises is lower under pegged exchange rate regimes. This result implies that countries with pegged regimes need to accumulate more reserves to avoid currency crises, relative to their peers with more flexible exchange rate regimes.

KEYWORDS:

• Currency crisis
• early warning systems
• international reserves
• pegged exchange rate

JEL CLASSIFICATION:

• F31
• F33
• F41
• O24
• G01; G00

Acknowledgments

The views expressed in this paper are those of the author and do not necessarily represent the views of the IMF, its Executive Board, or IMF management. I thank Ata Can Bertay, Eda Emil and Temel Taşkn for insightful comments and suggestions. All errors are my own.

Disclosure statement

No potential conflict of interest was reported by the author.

Notes

¹ Also see, Krugman (1979), Ghosh et al. (2002) and Jeanne and Ranciere (2006), among others.

² For instance, Eichengreen, Rose, and Wyplosz (1995), Frankel and Rose (1996), Kaminsky, Lizondo, and Reinhart (1998), Kaminsky and Reinhart (1999) and Milesi-Ferretti and Razin (2000), Bussière, Saxena, and Tovar (2012) and Mendoza and Terrones (2012).

³ Results are robust to excluding these control variables. I do not report the coefficient estimates of the control variables throughout the paper, but illustrate those in Appendix A.

⁴ To mitigate the reverse causality problem, I use lagged values of these variables following the standard practice in the literature on the predictors of currency crises (e.g. Milesi-Ferretti and Razin 2000). Moreover, the definition of crisis episodes alleviates the reverse causality concern further. In the dataset, there is no case with a continuation of the same currency crisis for 2 or more consecutive years, as mentioned in Section 2.1. To be more specific, in the sample, the closest crisis events within a country has a 5-year gap as noted earlier. Thus, it is not very likely for a crisis at year $t$ to affect, for instance, the arrangement or reserves at year $t-1$. Moreover, I run a Hausman test for model selection as discussed in Appendix B.

⁵ Throughout regressions, the joint effect from reserves and the interaction between reserves and the dummy for pegged regimes is always different than zero (${\alpha }_1+{\alpha }_2<0$) at the $5\mathrm{\%}$ significance level.

⁶ However, as shown in the Appendix C, there does not exist such a large difference in terms of reserve holdings across the two regimes.