On Predictability in the Taiwan Foreign Exchange Market

• Main Authors: Yun-Shan Dai, 戴韻珊
• Other Authors: Wei-Ming Lee
• Format: Others
• Language: en_US
• Published: 2010
• Online Access: http://ndltd.ncl.edu.tw/handle/51345474358941821969

博士 === 國立中正大學 === 國際經濟所 === 98 === Numerous empirical studies examine the efficiency of the foreign exchange markets by answering two basic questions: first, whether exchange rate returns follow a martingale difference (MD) process and hence are serially uncorrelated; second, whether technical trading rules generate significantly positive returns in the foreign exchange markets. Most of the studies on predictability of asset returns rely on a variety of variance ratio (VR) tests. However, such tests fail to capture nonlinear predictability in asset returns as pointed out by Kuan and Lee (2004, SNDE). In view of this, in addition to the wild bootstrapping VR test proposed by Kim (2006, Economics Letters), in the chapter 2, we also employ the MD tests of Kuan and Lee (2004) and Escanciano and Velasco (2006, Journal of Econometrics) to examine if daily and weekly returns in the Taiwan foreign exchange markets are predictable. Our empirical results reveal that daily returns are serially correlated and hence linearly predictable, in contrast with the results of Lee, Pan, and Liu (2001, Journal of International Financial Markets, Institutions and Money). As for weekly returns, they are not linearly predictable but may be nonlinearly predictable as suggested by the MD tests. In view of the results from chapter 2, we also examine the predictability of various technical trading rules for the Taiwan foreign exchange markets in chapter 3. In addition to spot exchange rates, we also adopt daily and weekly data on the 10-day, 30-day, 60-day, and 90-day forward exchange rates. In order to avoid the data snooping bias, we apply the reality check (RC) test of White (2000, Econometrica) and the superior predictive ability (SPA) test of Hansen (2005, JBES) to examine if the buy-and-hold strategy is outperformed by the best technical trading rules. Our results reveal that the returns in Taiwan’s spot and forward markets can be predicted by the best trading rules, regardless of the frequency of data. In particular, it is found that the best trading rule can generate significant profits even when the transaction costs (TC) are taken into account. Our studies confirm that the returns in the Taiwan foreign exchange markets are predicted, that is, the Taiwan foreign exchange markets are inefficient. In chapter 4, in order to analyze whether the inefficiency of the Taiwan foreign exchange markets is derived from official interventions, we adopt the individual VR test of Lo and Mackinlay (1988, Review of Financial Studies) and the joint VR test of Chow and Denning (1993, Journal of Econometrics), and Kim''s (2006) wild bootstrapping VR test again. We divide the full sample period into each year, each quarter, each month for investigate the influence of official interventions. In addition to VR test, we also use technical analysis to reexamine the sub-samples. Our empirical results find that the inefficiency of Taiwan foreign exchange markets may be influenced by official interventions.