Search

In the era of intensified volatility of domestic and international capital markets, exploring effective factors and market information to construct appropriate investment portfolio strategies is of great significance to control risks and obtain stable and sustainable excess returns. In this study we select the performance reports of A-share listed stocks from the first quarter of 2018 to 2022 as research objects. Taking 1 to 12 weeks after the company's earnings announcement as the time window, we examine the post-earnings-announcement drift (PEAD) of stock prices, selecting the expected excess earnings factor, a proxy variable of market anomaly, and other 5 related market anomaly factors. The effective factors are screened and tested by information coefficient (IC), information ratio (IR) and double sorting methods. Considering that the quantitative stock selection is a classification task with low data volume, low frequency and high validity of eigenvalues, a multi-factor quantitative strategy based on a lightweight gradient boosting tree is used to construct a portfolio to predict stock returns. This approach is compared with traditional quantitative strategies (simple scoring method, single factor model based on expected surplus, IC value weighted multi-factor stock selection model) and quantitative strategies based on other machine learning models such as support vector regression (SVR), artificial neural network (ANN) and extreme gradient boosting (XGBoost). Empirical results show that in forecasting stock excess returns based on the PEAD effect in the A-share market, the portfolio constructed by the light gradient boosting machine (LGBM) multi-factor quantitative strategy achieves an average annual return of 21. 633% in the long-short combination portfolio, exceeding the benchmark annualized return of 20. 184%. The comprehensive analysis of various indicators reflect that the investment portfolio constructed by LGBM multi-factor quantitative strategy has excellent performance in the A-share market and more show significant improvement compared to other quantitative strategies while maintaining stability, which can better control portfolio risks and achieve higher excess returns. [ABSTRACT FROM AUTHOR]

Literature has found that regression of independent (nearly) nonstationary time series could be spurious. We incorporate this idea to examine whether significant regression could be treated as insignificant in some situations. To do so, we conjecture that significant regression could appear significant in some cases but it could become insignificant in some other cases. To check whether our conjecture could hold, we set up a model in which both dependent and independent variables Yt and Xt are the sum of two variables, say Y t = Y 1 , t + Y 2 , t and X t = X 1 , t + X 2 , t , in which X 2 , t and Y 2 , t are independent and (nearly) nonstationary AR(1) time series such that X 2 , t = α 1 X 2 , t − 1 + ε t and Y 2 , t = α 2 Y 2 , t − 1 + e t . Following this model-setup, we design some situations and the algorithm for our simulation to check whether our conjecture could hold. We find that on the one hand, our conjecture could hold that significant regression could appear significant in some cases when α1 and α2 are of different signs. On the other hand, our findings show that our conjecture does not hold and significant regression cannot be treated as insignificant when α1 and α2 are of the same signs. We note that as far as we know, our article is the first article to discover that significant regression can be treated as insignificant in some situations. Thus, the main contribution of our article is that our article is the first article to discover that significant regression can be treated as insignificant in some situations and remains significant in other situations. We believe that our discovery could be an anomaly in statistics. Our findings are useful for academics and practitioners in their data analysis in the way that if they find the regression is insignificant, they should investigate further whether their analysis falls into the problem studied in our article. [ABSTRACT FROM AUTHOR]

Inspired by Maslow's need hierarchy theory, we construct a new portfolio selection framework using the bi-level optimization technique in which the lower-level need relates to safety (low risk) while the upper-level need is concerned with self-actualization (high payoff). Specially, we consider a bi-level portfolio selection model using variance and conditional value-at-risk associated with the lower-level need and upper-level need, respectively. Accordingly, we propose a procedure to solve this bi-level optimization problem without the normal distribution assumption. Empirical study on the American stock market and U.K. stock market shows that our new model can determine optimal portfolios with moderate diversification. Out of sample performance also confirms that in our framework investors can obtain higher returns in a safe way compared to models in the traditional framework. [ABSTRACT FROM AUTHOR]

We propose and develop mean-variance-ratio (MVR) statistics for comparing the performance of prospects (e.g., investment portfolios, assets, etc.) after the effect of the background risk has been mitigated. We investigate the performance of the statistics in large and small samples and show that in the non-asymptotic framework, the MVR statistic produces a uniformly most powerful unbiased (UMPU) test. We discuss the applicability of the MVR test in the case of large samples and illustrate its superiority in the case of small samples by analyzing Korea and Singapore stock returns after the impact of the American stock returns (which we view as the background risk) has been deducted. We find, in particular, that when samples are small, the MVR statistic can detect differences in asset performances while the Sharpe ratio test, which is the mean-standard-deviation-ratio statistic, may not be able to do so. [ABSTRACT FROM AUTHOR]