The dependence structure and portfolio risk of Malaysia's foreign exchange rates : the Bayesian GARCH–EVT–copula model

Purpose: This paper investigates the dependence structure and market risk of the currency exchange rate portfolio from the Malaysian ringgit perspective. Design/methodology/approach: The marginal return of the five major exchange rates series, i.e. United States dollar (USD), Japanese yen (JPY), Singapore dollar (SGD), Thai baht (THB) and Chinese Yuan Renminbi (CNY) are modelled by the Bayesian generalized autoregressive conditional heteroskedasticity (GARCH) (1,1) model with Student's t innovations. In addition, five different copulas, such as Gumbel, Clayton, Frank, Gaussian and Student's t, are applied for modelling the joint distribution for examining the dependence structure of the five currencies. Moreover, the portfolio risk is measured by Value at Risk (VaR) that considers the extreme events through the extreme value theory (EVT). Findings: The finding shows that Gumbel and Student's t are the best-fitted Archimedean and elliptical copulas, for the five currencies. The dependence structure is asymmetric and heavy tailed. Research limitations/implications: The findings of this paper have important implications for diversification decision and hedging problems for investors who involving in foreign currencies. The authors found that the portfolio is diversified with the consideration of extreme events. Therefore, investors who are holding an individual currency with VaR higher than the portfolio may consider adding other currencies used in this paper for hedging. Originality/value: This is the first paper estimating VaR of a currency exchange rate portfolio using a combination of Bayesian GARCH model, EVT and copula theory. Moreover, the VaR of the currency exchange rate portfolio can be used as a benchmark of the currency exchange market risk.