Comments and Questions

This paper examines the impact of interest rate changes in the presence of capital controls. It presents a model that indicates that increased interest rates may either appreciate, or depreciate the currency, conditional on the level of indebtedness. The topic is important and interesting. But for the paper to be viable, both further theoretical and empirical development is imperative. Addressing the theoretical concerns listed below would serve to deepen the insights of the paper. It is vital to address the empirical identification questions to clarify the meaning of the results, and to allow assessment of the robustness of the findings. Principal comments: 1. The theoretical framework of the paper is simple, with an interest income equation subject to the current account balance. Specifications are static, rather than intertemporal, (there is no explanation of this modeling choice – but in the absence of non-linearity in adjustment costs, this is unlikely of great significance, so is not taken further here). The effect of changes in the interest rate are derived straightforwardly by means of total differentials for the case of no leakages, and with leakages (imperfectly or incompletely defined):a. Implicit in the derivations presented, is the assumption that only dE and di are non-zero (i.e. there is adjustment only in the exchange rate and the interest rate). Specifically, for the non-leakage case the assumption is that dD=0, so that there is no adjustment of the debt position. For the leakage case, the dD=0 assumption is supplemented by dEe=0 (i.e. no adjustment in expectations of the exchange rate), no adjustment in capital control enforcement so that dt0=0 and dt1=0, and that there is no alteration in the risk premium, dp=0. Given the sudden stop scenario being modeled, seems an especially striking limitation.b. Perhaps this might be appropriate if the focus is purely on an instantaneous effect (though I still doubt it). Yet typically we are not – and in empirical application the authors employ the 2009-15 period. Over this length of time, debt, risk premia, monitoring of capital controls, exchange rate expectations are all subject to substantial potential change. The model does not reflect this possibility – though it easily could be made to reflect these changes. The consequences appear more severe for the leakage than the no leakage case.c. While the paper introduces “leakage,” this is not clearly enough defined. The associated assumptions being made require clarification. The paper avoids all mention of the literature on capital flight, the channels though this occurs, what implications follow for the efficacy of capital controls, and what implications this carries for the modeling exercise. 2. The theoretical models of sections 2 and 3 present much greater depth and information than the empirical specification (15) of section 4. After having gone through the derivation, why not exploit the predictions of the theory empirically? There is no explanation as to why this is not done. At the very least, theory and empirical specification are not as closely linked as is desirable. 3. Estimation is by VECM. The authors record two cointegrating vectors present in the data. This requires just or over identification for estimation. But no information is provided as to what the identifying restrictions are, making interpretation of the results difficult. Clear specification of what long run equilibrium relationships serve for identification purposes, and the associated parameter values are essential, to allow an assessment of the coherence of the estimation results.