DYNAMIC RELATIONSHIP ANALYSIS OF SINGAPORE FOREIGN EXCHANGE RATES

Abstract

Eight series of spot rates in terms of Singapore Dollars are analysed using multiple time series approach proposed by Tiao and Box (1981). The inter-region and within region structure of the exchange rates are investigated using the sample cross correlation matrices. The results show that the European series are leading the Asia Pacific series. Although such analysis contributes to our understanding of the dynamic relationships among the series, the forecasts generated from this model are further examined. A few other simpler univariate models such as the random walk and exponential weighted moving average are used to generate forecasts for comparison with those from multivariate technique. In addition, we also formulated a regression model between the exchange rate return and interest rate differential based on the conventional wisdom that when interest rate of one country increases (decreases), the currency of that country will appreciate (depreciate). The one-step ahead forecasts for each technique are computed and in terms of the precision, the forecasts from the multiple time series technique fails to outperform those from the naive mean method. However, in terms of directional forecasts, the multivariate technique outperforms all the techniques discussed in this exercise and this is confirmed by applying the nonparametric Binomial Test on the frequencies of correct directional forecasts. The forecasting results show that complicated modelling techniques do not necessarily produce precise forecasts. Nevertheless, the large amount of effort invested in the multiple times series model still yield some information; although not extensively useful it is after all better than those who conclude that foreign exchange rates are totally unpredictable.