International reserve holdings: interest rates matter!

Abstract

We argue that the literature on optimal international reserve holdings in an era of high-capital mobility fails to find interest rates is a strongly significant factor because of the endogeneity of interest rates and reserves under fixed exchange rate regimes. Using two-stage least squares we control this and regain statistical significance for interest rates.

Acknowledgements

We would like to thank Eric J. Brunner for his insightful comments and suggestions. Financial support came in part from the Quinnipiac University Research Fund.

Notes

¹ Bahmani-Oskooee and Brown (2002) place the buffer stock model in its proper context in the relevant literature while also providing an excellent review of the literature in general.

² The measure of exchange rate flexibility is computed by using the SD of the innovation to the percentage change in the nominal effective exchange rate. Openness is captured by capital and trade flows.

³ Sebastian Edwards (1985) found that interest rates were significant. Flood and Marion, however, emphasize that their results are for the modern era of high-capital mobility. Since Edwards (1985) was prior to the high-capital mobility era, Flood and Marion's results remain the ones we must address.

⁴ Edwards (1983) and Bahmani-Oskooee (1987) show that the degree of exchange rate flexibility plays an important role in determining the demand for international reserves as well as the adjustment between the actual and the desired level.

⁵ The relationship still holds if the fixed regime does not enjoy perfect credibility or occasionally sterilizes. Perfect noncredibility or perpetual sterilization implies that ‘fix’ is a misclassification. Anything less than that adds noise to the empirical observation of the relationship but does not eliminate the endogenous relationship itself.

⁶ The sample period considered in FM is 1988–1997, while the one considered in this study is 1987–2000. We use the same 36 countries (listed in the data appendix) as in Flood and Marion.

⁷ As in Flood and Marion, the estimated coefficient is actually positive but statistically insignificant when the level of imports is used as the scale variable.

⁸ The exchange rate regime classification follows Levy-Yeyati and Sturzenger (2005).

⁹ For two of the four cases, the coefficient is not statistically different from −0.25 at the 5% confidence level, while for the other two this holds at the 10% level.

¹⁰ Further details on the shadow fundamental rate are in Flood and Marion (1999).