Pham, Hoang Thach (2023) Stock Selection for Trading Strategies Based on Risk Factors: A Study of The Ho Chi Minh Stock Exchange. PhD thesis, Victoria University.

Abstract

This research focuses on stock selections for trading strategies based on cross-sections of stock returns and risk factors in the Ho Chi Minh Stock Exchange (HSX), an emerging market. To select the right stocks for trading strategies, this thesis is divided into three steps. First, I analyse different characteristics that affect stock returns with new ones: dynamic beta, Value-at-Risk (VaR), and conditional Value-at-Risk (CVaR); and traditional ones: static beta, firm size, firm value, momentum, and illiquidity. This study uses different regression models on sample panel data in the HSX. For dealing with heteroskedasticity, different robustness techniques, including the traditional Newey–West method (when the residuals are correlated across time) and new clustering techniques (when the residuals are correlated across firms or across time, or across both firms and time) (Millo, 2019; Petersen, 2009) are used to reduce biases in testing the effect of the above characteristics on stock returns. Second, I study different available risk factors such as market, size, value, momentum, profitability, investment, illiquidity, Value-at-Risk, and develop a new factor called conditional Value-at-Risk. The GRS test (Gibbons et al., 1989) is applied to determine the appropriate risk model for this market. The final part of this thesis tests stock selection strategies based on understanding factors that affect stock returns. The performance of strategies is evaluated by a non-parametric test using a t-test and a parametric test using alpha (the intercept of the selected risk factor model). If the alpha of a strategy is positive, this strategy outperforms the market (the return of the strategy is higher than that of the selected risk factor model), and investors can apply it for their trading. If this strategy underperforms the market (the return of the strategy is lower than that of the selected risk factor model), investors can reject it. If the alpha is zero, the return of a strategy is predicted by the market (the return of the strategy is indifferent from that of the selected risk factor model). The first study found that double clustering panel regressions are more appropriate than OLS, between-estimator, and Fama–MacBeth regressions for coefficient estimations because both individual and time effects exist in the errors (Petersen, 2009; Sun et al., 2018). Furthermore, based on the Hausman test (Croissant & Millo, 2018), the fixed-effect models are preferred to the random-effect models. These models indicate that stock returns are positively and significantly correlated with momentum and dynamic beta but negatively correlated with firm size. The second study found that different combinations of risk factors can explain stock returns in the HSX. However, the GRS test shows that the three-factor model (containing the market, size, and investment factors) performs better than other multifactor models. The last study found that both long and arbitrage strategies earn positive returns and positive alphas. In particular, buying small-size stocks in the high Value-at-Risk (SHVaR portfolio) outperforms the market, and this strategy generates the highest return (approximately 2.38% monthly) compared to other strategies. This implies that the HSX is not efficient and recommended strategies in this thesis should help investors earn positive returns. The results of this thesis provide practical insights. First, policymakers can utilize the risk factors to evaluate the efficiency of the market. Second, investors can select stocks for their portfolios based on the correlation between stock returns and their characteristics. Last, investors can select the best risk model to evaluate if their investment strategies earn higher returns than that of the market required.

Search Google Scholar

Repository staff login