The Periodic Return and Volatility Pattern of the Intraday Currency Futures

• Main Authors: Shih-feng Chien, 簡士峰
• Other Authors: Ming-shu Hua
• Format: Others
• Language: zh-TW
• Published: 2003
• Online Access: http://ndltd.ncl.edu.tw/handle/27060155503892523154

碩士 === 國立暨南國際大學 === 國際企業學系 === 91 === Varies periodic patterns of return and volatility, such as overnight effect, time-of-the-day effect, and day-of-the-week effect, of current futures contracts are found in literatures. Most studies applied linear regression or GARCH model with seasonal dummy variables to capture return and volatility patterns. However, autocorrelation is found to be varying according to government macroeconomic announcement during the day time, which makes traditional modeling method a bias approach to periodic pattern researches, and time-vary parameters model is needed. Using periodic ARMA, we arm to verify all documented periodic patterns and further discover possible seasonal formations in currency futures market. The samples are Deutsche Mark futures, British Pound futures, Japanese Yen futures traded in 1998 and 1999 in Chicago Mercantile Exchange. We found no significant intraday return pattern, but overnight effect of significant negative return are appeared in all three contracts. Intraday volatility pattern in more details are verified in the research: volatility pattern shows U-shaped cross the day in general, and we further find market are more violate in the morning then in the afternoon. Beside, the volatility is expected to increase during government macroeconomic announcements periods. And British Pound futures also show a volatility increase during British market closing. The result as well shows day-of-the-week effect does exist, and government macroeconomic announcements are partly contributed to this phenomenon. Evident of inventory control effect reveals in our data. Surprisingly, three currency futures show similar intraweek patterns after controlling for the announcements effect and possible inventory control effect while they are dramatically distinct originally. Interestingly, according to the results in the periodic ARMA model, the coefficients of autocorrelation are larger in the afternoon relative to that in the morning, and are smaller during the government macroeconomic announcements periods then during the non-announcement periods in average.