A Tale of Two Exchange Rates: South Africa's Dual-Rate Experiment

Abstract

This paper investigates South Africa’s dual exchange rate experiment during the period 1985–1995, wherein the commercial rand rate is set by current account transactions while the financial rand rate is set by capital account transactions. Tests show that the dual-rate system was highly effective in segmenting FX markets for the commercial vs. financial rands, thereby providing a ‘controlled lab’ setting to study key exchange rate questions. Our analysis reveals striking differences in empirical features of the commercial and financial rates. Although the financial rate follows a random walk process, the commercial rate does not and hence, is predictable. The financial rand rate exhibits a high, time-varying volatility occurring in clusters while the commercial rate shows a relatively low. The purchasing power parity (PPP) tends to hold for the commercial rate but fails to hold for the financial rate. Similarly, tests show that the commercial rate is linked to macroeconomic fundamentals but the financial rand rate is disconnected from the fundamentals. Finally, fundamental-based models outperform the random walk benchmark in predicting the commercial rate but fail to do so in predicting the financial rate. Overall, our findings suggest that the rising dominance of international investment over trade may be responsible for the well-known behavior of recent flexible exchange rates such as the random walk, failure of PPP, and a tenuous link with fundamental economic variables.