Detecting currency manipulation: An application of a state-space model with Markov switching

Abstract

Following an idea of Milton Friedman's “plucking model,” we propose to use a state-space model with Markov switching as an auxiliary tool for detecting currency manipulation. Without imposing any a priori restrictions, our model tests if fluctuations of a country's exchange rate are symmetric or if there exists a time-varying support level or resistance level of exchange rate. Using weekly and monthly data of countries on the “monitoring list” of the US Treasury as of April 2017, we find that exchange rates of China, South Korea, Switzerland, and Taiwan rarely fall below their time-varying trends, but are plucked upward from time to time by transitory shocks, suggesting that the FX (foreign exchange) authorities of these countries may have been intervening more actively against appreciation shocks. Our sub-sample analysis reveals that our model accurately captures the period of Switzerland's minimum exchange rate policy with probability of one and Japan's exchange rate rarely falls below its trend after implementing Abenomics. We discuss the difficulties of detecting FX intervention along with the relative advantage of our approach.