Currency Effects on International Equity Portfolios: Evidence from Taiwan, Hong Kong, China and Singapore

This study uses multivariate adaptive regression spline (MARS) threshold modeling method to examine the possible nonlinear and interactive relationship between the currency and stock markets. Four countries including Taiwan, Hong Kong, China, and Singapore, that are within the investment universe of the Greater China funds, are analyzed with both linear Vector autoregressive (VAR) model and the nonlinear MARS_VAR model. In contrast to the VAR model, MARS_VAR model has the capability of extracting more information from the residual series estimated by the VAR model and estimating accurate threshold values at which the conditional effects are turned on. The MARS_VAR result shows that lagged currency returns and lagged stock returns interactively affect present stock’s performance. The MARS_VAR result also shows that there exists a two-way mapping relation between the two markets. The threshold values of extreme conditions provide the portfolio managers with tools of managing currency risk of the international equity portfolios. Using the conditions and threshold values estimated by the MARS_VAR model, one can generate trade signals for capital withdrawal purpose in order to avoid the potentially extremely negative returns on the stock market(s). With a data sample covering the period of March 2003 to August 2013, first 10 years of 2,345 daily observations are used to analyze and estimate the threshold values by MARS_VAR, and the most recent 1 year of 290 observations are used to check the out-of-sample performance of the model estimation. The in-sample and out-of-sample simulations on portfolio performance both show a better risk-adjusted return profile than their corresponding passively-managed benchmark portfolios due to the reduced portfolio volatility. The major contribution of this search is that the more dynamic relationships between the currency and stock markets are uncovered with a nonlinear MARS_VAR model, in terms of their lead-lag effect, the persistence of the effects, and how the relationship is nonlinear and interactive. This research also uses the data from a post-Asian financial crisis period to discuss the day-to-day relationship between the currency and stock markets instead of focusing only on event studies.