A factor pricing model based on double moving average strategy

Existing asset pricing models, including Fama and French's five-factor model (FF5) and Liu et al.'s (2019) four-factor model (CH4) for the Chinese market, struggle to fully explain market anomalies. This paper addresses this limitation by incorporating behavioral finance principles and the heterogeneous market hypothesis, which posits that investors possess diverse investment time horizons. The study focuses on the Chinese market, characterized by a high proportion of individual investors who often employ technical analysis, including moving averages (MA), to guide their trading decisions. The authors hypothesize that MA strategies with varying time horizons can capture the behaviors of these heterogeneous investors and improve the explanatory power of existing asset pricing models. The primary research question is whether incorporating double MA factors, representing short-term and long-term investment horizons, can enhance the pricing power of the CH4 model in the Chinese market. The importance of this research lies in its potential to refine asset pricing models and provide a more accurate representation of the Chinese stock market's dynamics, considering the unique characteristics of its investor base.

The heterogeneous market hypothesis suggests that investors with different time horizons interact in non-linear ways, leading to specific price correlations at varying time intervals. Models like HAR-RV quantify this interaction through volatility measures. Behavioral finance highlights the underreaction or overreaction to new information, causing trading delays and varied frequencies. Technical analysis, particularly moving averages, has shown predictive power in various markets. Studies demonstrate the profitability of MA strategies in the US and other markets (Han et al., 2016; Xiang and Lu, 2018). The authors review studies that use moving averages to capture investor behavior with different term structures and find that MA strategies yield significant excess returns. The existing literature provides theoretical and empirical support for the use of MAs to capture investor heterogeneity and improve asset pricing models.

The study utilizes data from the Shanghai and Shenzhen Stock Exchanges from January 2000 to June 2020, obtained from WIND Info. The authors follow Liu et al.'s (2019) methodology to construct the four factors of the CH4 model: market factor (Mkt), size factor (SMB), value factor (VMG) based on the earnings-to-price ratio (EP), and sentiment factor (PMO). Unlike Liu et al. (2019), the authors include the bottom 30% of stocks by market capitalization. The double MA strategy is implemented by calculating short-term and long-term MAs for different periods (1, 3, 6, 9, 12, 18, and 24 months) using Equation (4). A trading signal (S_jt,L) is generated by dividing the short-term MA by the long-term MA (Equation 5). The SCL factor (Equation 6) is formed using a low minus high approach to account for the reversal effect observed in the Chinese market. Spanning regression is used to select non-redundant factors. The authors compare the performance of the four-factor model with an augmented six-factor model that includes the SCL(1,3) and SCL(1,12) factors. The GRS test evaluates the pricing power of these models. A bootstrap test is also used to evaluate the in-sample and out-of-sample performance. Finally, the study examines the state dependence of the models by analyzing different GDP growth rates and Shanghai Composite Index trends.

The study finds that MA strategies using a 1-month short-term MA and various long-term MAs (3, 6, 9, 12, 18, and 24 months) generate significant positive excess returns in the Chinese market, with the highest returns observed for SCL(1,3). Other term structures show insignificant returns. The inclusion of SCL(1,3) and SCL(1,12) factors significantly improves the pricing power of the four-factor model. The augmented six-factor model passes the GRS test at the 5% level, indicating that it can explain the 18 market anomalies selected for analysis. The average absolute alpha value decreases by approximately 50% compared to the four-factor model. The six-factor model exhibits a higher maximum squared Sharpe ratio than the four-factor model. Including the bottom 30% of stocks in the analysis does not compromise the model's results. The six-factor model also outperforms the four-factor model across different macroeconomic states defined by GDP growth rate and Shanghai Composite Index trends, demonstrating robustness. The findings suggest that short-term trading strategies are more prevalent and profitable in the Chinese market.

The findings support the hypothesis that incorporating double MA factors improves asset pricing models in the Chinese market. The significant excess returns from specific MA strategies highlight the impact of investor behavior and differing time horizons on asset prices. The superior performance of the six-factor model, as evidenced by the GRS test and Sharpe ratio, confirms the value of considering short-term and long-term investment perspectives. The inclusion of small-cap stocks suggests that these stocks are not immune to technical analysis and contributes to more comprehensive results. The robustness of the model across different macroeconomic conditions reinforces its applicability and practical value. Overall, the findings suggest that investors in the Chinese market, particularly retail investors, are sensitive to both short-term momentum and longer-term trends, and this heterogeneity can be effectively captured by using different moving averages with different term structures.

This study's primary contributions include demonstrating the effectiveness of double MA factors in capturing investor behavior with various term structures in the Chinese market, showing that including small-cap stocks enhances the results, and expanding the CH4 model to improve its pricing power. Future research could explore the state-dependent aspects of the MA factors in greater detail and investigate the interaction of MA factors with other sentiment indicators. Investigating the applicability of this model in other markets and exploring other technical indicators would provide further insight into their relevance for asset pricing.

The study's limitations include the use of data only from the Chinese market. Therefore, the generalizability of the findings to other markets may be limited. The reliance on historical data could limit the prediction of future market behavior. The choice of specific time horizons for the MA strategies might influence the results. Furthermore, the study relies on publicly available data from WIND Info, which might have some limitations. Finally, the data used is not publicly available due to commercial confidentiality.