Research On Seasonality Anomalies: Based On High Frequency Stock Market Data

The emergence of stock market anomalies presents a significant challenge to the classical capital asset pricing theory,which is based on the efficient market hypothesis.The increasing accessibility of financial highfrequency data acquisition and processing methods has enabled the detection of intraday high-frequency market anomalies.One prominent anomaly in financial markets is seasonality,which exhibits cyclical and regular changes in asset prices and has become a hot topic in academic and investment circles.Thus,this dissertation focuses on seasonality,using intraday high-frequency stock market trading data and following an empirical asset pricing research paradigm,to investigate the pre-holiday cross-sectional seasonality of individual stocks and the overnight intraday seasonal reversal effect of factor Beta cross-sectional pricing power.First,this dissertation investigates the presence of pre-holiday crosssectional seasonality for individual stocks.This effect is characterized by the observation that stocks with exceptional return performance in the past pre-holiday window tend to exhibit the same trend in the future pre-holiday window.This dissertation employs Fama-Mac Beth regressions to validate the presence of notable pre-holiday cross-sectional seasonality,along with a post-holiday seasonal reversal.Moreover,by constructing a long-short portfolio based on this anomaly,this dissertation finds that the portfolio not only generates a significantly positive risk-adjusted return,but also significantly enhances the efficient frontier of the mainstream multi-factor pricing model.To test the robustness of the anomaly,cross-sectional and time-series robustness tests were performed,indicating that the anomaly remains significant in recent years and is not significantly influenced by other calendar effects,external environment,or holiday attributes.Second,this dissertation examines the intraday high-frequency performance characteristics of the pre-holiday cross-sectional seasonality.To begin,this dissertation decomposes the pre-holiday stock returns into pre-holiday overnight return components and pre-holiday intraday return components,revealing that the pre-holiday cross-sectional seasonal effect primarily occurs during the intraday trading session.By analyzing highfrequency intraday trade-by-trade stock data,this dissertation further decomposes the intraday trading session into opening,mid-day,and closing sessions and discovers that the effect mainly persists from the opening session until the closing session.Furthermore,this dissertation demonstrates that neither other stock return cross-sectional seasonality nor firm characteristics with predictive power for stock returns can completely explain the pre-holiday cross-sectional seasonality.Lastly,using international stock market trading data,this dissertation confirms the consistency of the overall characteristics and the intraday-specific variability of the pre-holiday cross-sectional seasonality in international stock markets.Third,this dissertation examines the overnight intraday seasonal reversal effect of the cross-sectional pricing power of factor Beta.Based on the theory of multi-factor pricing models,other factor Beta and market factor Beta equally reflect systemic risk.However,this dissertation reveals that the relationship between factor Beta and daily stock returns is insignificant.Instead,after decomposing daily returns into overnight returns and intraday returns,this dissertation identifies a significant seasonal reversal effect of the cross-sectional pricing power of factor Beta in the overnight intra-day period.The effect is prevalent across industry stocks and varies depending on the external market state.Furthermore,this seasonal reversal effect remains significant after undergoing robustness tests,such as modifying the sample filter,adjusting the sample interval,altering the estimation method,and considering other factor models.The innovations of this dissertation are noteworthy.Firstly,it contributes to the study of seasonality by discovering a new stock market anomaly,pre-holiday cross-sectional seasonality,which adds to the existing knowledge of seasonality research in both time series and crosssectional dimensions.Secondly,it provides a clear and comprehensive analysis of the high-frequency intra-day variation characteristics of preholiday cross-sectional seasonality at the individual stock level.This analysis helps to understand the pre-holiday cross-sectional anomalies from low to high frequency,pinpoint the specific time periods when these anomalies occur,and identify the driving factors behind them.Thirdly,this dissertation expands the research horizon of previous literature on singlefactor Beta pricing by accurately analyzing the overnight intraday seasonal reversal of the cross-sectional pricing power of factor Beta in a multi-factor pricing model.This finding extends the research scope of risk-return relationships,and sheds light on the dynamics of stock market anomalies.