中国股市的截面和时序因子模型

THE CROSS-SECTION AND TIME-SERIES FACTOR MODELS IN CHINA

蔡嘉旭

11930285

硕士

数学

王赫

2021-05-20

2021-06-10

南方科技大学

深圳

本文采用假设检验方法构建了两类统计检验量——GRS统计量和Hotelling’s T-square统计量，系统地对比了截面因子模型和时序因子模型对资产回报的定价能力。时序因子模型描述了风险与回报的关系，一直被广泛应用于解释资产的回报。然而Fama和French在2020年针对美国股市提出了基于其五因子模型的截面因子模型，并发现截面因子模型表现全面优于传统的时序因子模型。本文在中国的股票市场中对比了时序因子模型和截面因子模型的表现，寻找类似的发现。其中时序因子模型采用Liu等人在2019年提出的中国版四因子（CH-4）模型，截面因子模型基于前述模型采用Fama和French在2020年提出的方法构建。在构建因子模型前，本文先采用Fama-MacBeth回归进行检验，找出中国市场存在显著风险溢价的因子。Fama-MacBeth回归是一种针对面板数据的两步回归方法，能够降低残差因在截面上的相关性所产生的计算误差，得到更为准确的标准误。结果显示，在本研究样本中采用CH-4模型是合理的。本文的资产定价实证结果显示两类模型各有优劣，甚至不同的统计量给出了完全不同的结论。总体而言，采用截面因子的时序模型稍优于传统的时序模型，而新型的截面模型要优于上述两类时序模型。而分类异象的实证结果显示，截面因子模型在解释收益率模式较明显的异象，特别是异常换手率异象时，表现明显优于传统的时序模型。

In this paper, two kinds of statistical tests, GRS statistics and Hotelling's t-square statistics, are constructed by using hypothesis testing method, and the pricing power of cross-sectional factor model and time-series factor model on asset returns are systematically compared.The time series factor model describes the relationship between risk and return, and has been widely used to explain the return of assets. However, Fama and French propose a cross-section factor model corresponding to their five-factor model for the US stock market in 2020, and find that the cross-section factor model perform better than the traditional time-series factor model. This study wants to compare the performance of the time-series factor model and the cross-section factor model in the Chinese stock market, and looks for similar results. The time-series factor model in my study is based on the four-factor (CH-4) model proposed by Liu et al. (2019). The cross-section factor model corresponds to the CH-4 model and constructed by the method of Fama and French (2020).Before constructing the factor model, Fama-Macbeth regression is used to find out the factors that have significant risk premium in the Chinese stock market. Fama-Macbeth regression is a two-step regression method for panel data, which can reduce the calculation error caused by the correlation of residual errors on the section and obtain more accurate standard error. The results show that it is reasonable to use the CH-4 model in the sample of this study.The empirical asset pricing results show that the two types of models have their own advantages and disadvantages, and even different statistics give completely opposite conclusions. In general, the time-series model using the cross-section factors is slightly better than the traditional time-series model, and the cross-section model is better than the above two types of time-series models. The empirical results of each anomaly show that the cross-sectional factor model performs better than the traditional time series model in explaining the anomaly with significant return patterns, especially the abnormal turnover anomaly.

资产定价模型 时间序列因子模型 截面因子模型 异象 中国股市

asset pricing model time-series factor models cross-section factor models anomalies Chinese stock market

中文

独立培养

南方科技大学

蔡嘉旭. 中国股市的截面和时序因子模型[D]. 深圳. 南方科技大学,2021.