| Summary: | 碩士 === 中原大學 === 國際貿易研究所 === 94 === ****This study uses the probability distribution techniques to explore the intraday effect and weekday effect of the 10-minuate high frequency returns for the Dow Jones Industrial Average (DJIA) and NASDAQ composite indices. We find that both DJIA and NASDAQ can form a U-shaped pattern by using absolute intraday returns. However, the U-shape from the DJIA index slightly declines at the closing time, while the U-shape from the NASDAQ index still goes up at the closing time. It implies that most investors of DJIA finish their last trading strategy before ten minutes of the closing. Moreover, the width of the Gaussian distribution in NASDAQ is always wider than in DJIA. This result proves that the trading market in NASDAQ is more risky than in DJIA.
Later, as we re-group total intraday returns into the opening, lunch and closing three subgroups, we find that no matter DJIA or NASDAQ index, the lowest center and widest width of the Gaussian distribution occur at the opening trading period. This means that the overnight information is cumulated until next morning, so the opening trading period is the most risky for a whole trading day.
Furthermore, it is more likely to cause the sign of intraday returns at one specific trading period to be swamped by the sign of intraday returns at the other specific trading period when we only discuss the weekday effect without considering the intraday effect. If we only consider the weekday effect, the weekend effect seems disappearing. However, if we add the intraday effect and re-classify total intraday returns according to the trading time and the weekday, we will find that the weekend effect still exists, but the occurrence of the weekend effect may be advanced or postponed.
Besides, using the Gaussian function to refit the intraday returns, we can find that no matter DJIA or NASDAQ, Thursday has the widest width, while Monday and Tuesday have the narrowest width. Meanwhile, the lunch trading period on Tuesday has the highest height. As a result, we infer that the intraday effect will affect the weekday effect in intraday returns. Similarly, using the log-normal distribution to refit the intraday volatilities, we find that Monday and Tuesday have the lowest average volatility and the lowest peak, while Thursday has the highest average volatility and the highest peak. However, the intraday effect is not significant via observing distributions of intraday volatilities.
|
|---|
