The value premium and uncertainty: An approach by support vector regression algorithm

Khoa Bui Thanh; Tran Trong Huynh Article The value premium and uncertainty: An approach by support vector regression algorithm Cogent Economics & Finance Provided in Cooperation with: Taylor & Francis Group Suggested Citation: Khoa Bui Thanh; Tran Trong Huynh (2023) : The value premium and uncertainty: An approach by support vector regression algorithm, Cogent Economics & Finance, ISSN 2332-2039, Taylor & Francis, Abingdon, Vol. 11, Iss. 1, pp. 1-15, https://doi.org/10.1080/23322039.2023.2191459 This Version is available at: https://hdl.handle.net/10419/304029 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by/4.0/ Cogent Economics & Finance ISSN: (Print) (Online) Journal homepage: www.tandfonline.com/journals/oaef20 The value premium and uncertainty: An approach by support vector regression algorithm Bui Thanh Khoa & Tran Trong Huynh To cite this article: Bui Thanh Khoa & Tran Trong Huynh (2023) The value premium and uncertainty: An approach by support vector regression algorithm, Cogent Economics & Finance, 11:1, 2191459, DOI: 10.1080/23322039.2023.2191459 To link to this article: https://doi.org/10.1080/23322039.2023.2191459 © 2023 The Author(s). This open access article is distributed under a Creative Commons Attribution (CC-BY) 4.0 license. Published online: 19 Mar 2023. Submit your article to this journal Article views: 933 View related articles View Crossmark data Citing articles: 2 View citing articles Full Terms & Conditions of access and use can be found at https://www.tandfonline.com/action/journalInformation?journalCode=oaef20 FINANCIAL ECONOMICS | RESEARCH ARTICLE The value premium and uncertainty: An approach by support vector regression algorithm Bui Thanh Khoa 1 * and Tran Trong Huynh 2 Abstract: Risk premium plays an important role in stock investing. Experiments have shown that value stocks typically have a higher average return than growth stocks; however, this effect persists indefinitely, even disappearing in some stages. Some studies suggested high volatility in the series of returns, broken structures, market volatility, or the impact of financial crises. This study aimed to build the uncertainty index and control it in the regression analysis model to solve the limitations above. The empirical analysis in Ho Chi Minh Stock Exchange (HOSE) showed that a value premium exists, and value stocks have a higher average return than growth stocks due to the higher overall risk. Furthermore, this study combined the Support Vector Regression (SVR) algorithm with the risk premium theoretical framework for the forecasting model; consequently, it is the most efficient model. Subjects: Quantitative Finance; Statistics for Business, Finance & Economics; Machine Learning Bui Thanh Khoa ABOUT THE AUTHORS Bui Thanh Khoa earned his Master’s degree in Business Economics from the Université Toulouse 1 Capitole in France in 2012, and he will acquire his doctorate in Business Administration from the Ho Chi Minh City Open University in Vietnam in 2020. His works may be found in the SCOPUS and ISI databases. He is a member of the Advisory International Editorial Board of Jurnal the Messenger, an ISI system journal, as well as the editorial boards of Scopus-indexed journals such as Journal of System and Management Sciences; Advances in Operations Research; Journal of Logistics, Informatics and Service Science, as well as the International Journal of Technology Transfer and Commercialisation from Inderscience Publisher. His study interests are methodology, electronic commerce, organizational behavior, machine learning, and consumer behavior. He may be reached by email at [email protected]. Tran Trong Huynh is a lecturer at FPT University; he got a Master’s degree in Mathematics in 2013 at Ho Chi Minh City University of Education and Finance in 2020 at the University of Economics Ho Chi Minh City. His current research interests include finance, applied mathematics, data science, econometrics, and machine learning. He can be contacted at email: [email protected]. PUBLIC INTEREST STATEMENT Value premiums are one of the most controversial topics in the financial sector. Some experiments involving value complements have had mixed results. There are two interpretations of value premium: related to behavioral finance and risk premium. This study found that the value premium’s cause comes from risk arbitrage. Specifically, value portfolios have a higher overall risk. As a result, the rate of return is offset by a corresponding premium. Khoa & Huynh, Cogent Economics & Finance (2023), 11: 2191459 https://doi.org/10.1080/23322039.2023.2191459 Page 1 of 15 Received: 15 October 2022 Accepted: 10 March 2023 *Corresponding author: Bui Thanh Khoa, Faculty of Commerce and Tourism, Industrial University of Ho Chi Minh City, Ho Chi Minh City, Vietnam E-mail: [email protected] Reviewing editor: David McMillan, University of Stirling, Stirling, UNITED KINGDOM Additional information is available at the end of the article © 2023 The Author(s). This open access article is distributed under a Creative Commons Attribution (CC-BY) 4.0 license. Keywords: Value premium; uncertainty; risk premium; SVR; book value JEL Classifications: D81; E47; P45; G17 1. Introduction Researchers and investors often use the Book to market ratio (BM) to sort stocks into growth and value categories. Using data from the United States of America m (US market), E. Fama and French (1992); E. F. Fama and French (1993) documented that high BM earned higher average returns than low BM over the 28 years from July 1963 to June 1991. Some results are similar to Chan et al. (1991); Rosenberg et al. (1985); Stattman (1980). E. F. Fama and French (1993) also found that the value premium is not explained by the Capital Asset Pricing Model (CAPM). With the evidence from previous studies, Fama & French proposed a 3-factor model: market factor, size premium, and value premium. Research by Loughran (1997) showed that there is no value premium in large stocks, only in small stocks; however, this study only considers stocks in the US after 1963. E. F. Fama and French (1998) đconfirms the absence of a value premium in the US and some other markets. There is still debate about the interpretation of the value complement. Some researchers rely on behavioral finance, others on risk. Behavioral finance researchers argue that the overreaction of naive investors creates disparities in returns between portfolios. Specifically, they rely on the business’s past performance; for growth firms they overestimate the price level but underestimate the value business (Neves et al., 2021). Furthermore, these studies found that the performance of growth and value stocks varied between different periods around the global financial crisis. Furthermore, investor sentiment is important in growth and value stock returns. Some researchers interpreted risk premiums as risk compensation. Value stocks are riskier than growth stocks, which a higher expected return should offset—case studies such as E. F. Fama and French (1993); Qadan and Jacob (2022). N. F. Chen and Zhang (1998) used several risk characteristics related to firm characteristics, such as firm distress, financial risk, and the riskiness of future cash flows, to support the risk premium argument. Studies in non-US markets such as China also have conflicting results. Gang et al. (2019) did not find a valuable complement in the Fama-French 3-factor model. Clark and Qiao (2020) also gave similar results; however, they found that value premiums are related to macro policies. Meanwhile, Liu et al. (2019) proved that the value factor is important in the Fama-French 3-factor model. In Vietnam, Anh (2017) used data from 313 companies listed on the Ho Chi Minh Stock Exchange (HOSE) from October 2011 to October 2016 to evaluate the 3-factor model. The results showed that 6 out of 7 categories have a value effect for the linear regression method. However, the percentile regression results indicated that all categories have a value effect. The study also found that the value complements in the periphery percentiles are stronger than those in the central percentiles. The limitation of the study is that it has not considered the uncertainty affecting the model and has not evaluated the error in the forecasting model. Estimating the value premium by determining the difference in fertility rates of the value portfolio and the growth portfolio, according to E. F. Fama and French (2021), has many limitations; instead, they determine the value premium by comparing a portfolio against the market portfolio. This method works for several reasons: the market portfolio has always been the focus of pricing models; it indicates whether a category is in the growth or value category. E. F. Fama and French (2021) tested the US market and showed a diminishing valuecompensation effect. However, high volatility in monthly value premiums is an irrefutable cause of different hypotheses of the value premium at two periods. Several previous and subsequent studies have also demonstrated the effect of the uncertainty principle, such as Freyberger et al. Khoa & Huynh, Cogent Economics & Finance (2023), 11: 2191459 https://doi.org/10.1080/23322039.2023.2191459 Page 2 of 15 (2020); Gagliardini et al. (2016); Smith and Timmermann (2022). Therefore, this study constructed an uncertainty principle similar to the study of Ismailov and Rossi (2018); Rossi and Sekhposyan (2015). The advantage of this index is that when macroeconomic instability occurs, such as a financial crisis, oil price shock, etc., the uncertainty index will be high. Finally, this study considers the predictive model under the machine learning approach. Specifically, the value premium and the uncertainty index were applied to forecast portfolio returns. The algorithm is Support Vector Regression (SVR). The SVR is one of the most powerful algorithms widely applied in continuous output variable prediction. Many studies have shown its effectiveness in finance compared to traditional econometric models. Specifically, Zheng et al. (2021) used the SVR to forecast stock indexes in China from 1/2016 to 12/2020. The results show that, in general, SVR is more efficient than ANN (Artificial Neural Network) and RF (Random Forest) algorithms. Khoa and Huynh (2022a) have used SVR under the CAPM framework and obtained positive HOSE market results. Khoa and Huynh (2022b) used SVR in forecasting the movement of securities in the VN30 portfolio in the HOSE market, from which this study proposes a short-term trading method to obtain outstanding profits (Henrique et al., (2018); Huang et al., (2016); Pan et al., (2017); Y. Chen & Hao, (2017); Zhang et al., (2021)). The limited empirical study of value premiums and uncertainty is the main motivation for this research. This study focuses on three main objectives: ●To test the existence of value premium in the HOSE market according to the method of E. F. Fama and French (2021). ●To analyze the impact of uncertainty on the value premium. ●To use value premium and SVR algorithm to forecast portfolio return. 2. Theoretical framework 2.1. Value premium Early studies showed a gap between value stocks and growth stocks (Capaul et al., (1993); E. F. Fama & French, (1993); E. Fama & French, (1992)). In particular, value stocks appear to have higher average returns than growth stocks. This difference is called the value premium. There are two possible explanations for the value premium: behavioral finance and risk premium. Studies using behavioral finance suggested that market underperformance stems from the overwhelming majority of investors being too excited about growth companies. As a result, they overvalue growth firms and undervalue value firms (Lakonishok et al., (1994); Skinner & Sloan, (2002)). Jaffe et al. (2020) split the BM ratio into mispricing and growth option. This study identified that the mispricing component is stronger in explaining the value premium. Qadan and Jacob (2022) used data from 1965 to 2019 to show that value premiums are correlated and can be predicted by investors’ risk appetite. Specifically, investors’ appetite for risk translates into increased demand for value stocks over growth stocks. The other explanatory branch supports the efficient market hypothesis and arbitrage returns caused by risk arbitrage (Avramov & Chordia, (2006); E. F. Fama & French, (1998)). These studies suggested that higher-risk value stocks should be offset with a high expected return. The value premium is, therefore, the compensation for taking on more risk that the CAPM model has passed. Furthermore, this line of research also finds that value premiums are positively correlated with some forms of systematic risks, such as aggregate labor income (Jagannathan & Wang, (1996)), economic growth (Kirby, (2019); Koijen et al., (2017)), cash flow risk (Campbell & Vuolteenaho, (2004)), technological shocks (Berk et al., (1999)). Value premium relates to the firm’s characteristics, such as leverage (Doshi et al., (2019)) and growth firms (Ebrahim et al., (2014)). Angelidis et al. (2015) provided evidence that the dispersion of returns can serve as a variable indicative of the state of the economy since it is a reliable predictor of the value premium. Specifically, high volatility predicts a worsening economic situation and higher expected value premiums. Khoa & Huynh, Cogent Economics & Finance (2023), 11: 2191459 https://doi.org/10.1080/23322039.2023.2191459 Page 3 of 15 Most studies focus on the difference between the value portfolio and the growth portfolio to account for the value premium. The Fama-French 3-factor model used three factors in the pricing model, including the market factor, the factor related to the size and value premium (E. Fama & French, (1992)). Some empirical evidence revealed that the 3-factor model needs to be completed; E. F. Fama and French (2015) proposed a 5-factor model by adding two factors related to investment and profit. Following this line of research, E. F. Fama and French (2021) used the market portfolio as the basis because the author argues that the market portfolio is always central in most pricing models. Furthermore, for a particular portfolio, a comparison with the market portfolio indicated whether it is in the value or growth category; they defined it as the excess return value to the market portfolio (BM—BM M ), in which they used US data from June 1963 to June 2019 and did not include two samples. Research has found that the value premium tends to decrease from the early stage to the later period. Furthermore, the high volatility of the value premium is an irrefutable cause of the difference between the two periods. Subsequent work by Smith and Timmermann (2022) suggested that there exists a plucking structure that E. F. Fama and French (2021) have not considered. This study showed four breaks from 1950 to 2018 in the 6-factor model; these breaks occurred at some special events, such as the oil price shocks of the early 1970s, the change in the US Federal Reserve (FED) monetary policy regime, the collapse of the dot-com bubble, and the Global Financial Crisis (GFC). Previous studies have also noted a significant impact of variation in typically expected returns, such as Freyberger et al. (2020) or the effect of crises on traditional econometric models (Gagliardini et al., (2016)). Thus, the empirical evidence that the value complement remains a challenging problem for researchers. There exist two streams of explanation around the value premium, the behavioral asset and the risk premium. In some studies, the existence/disappearance of value premiums has been explained by causes of uncertainty principles such as economic fluctuations, crises, and even broken structures in the valuation models 2.2. Linear εsupport vector regression (linear ε-SVR) Support vector regression (SVR) is an algorithm to predict the continuous output. SVR is built on the idea of a Support Vector Machine (SVM) algorithm developed by Cortes and Vapnik (1995). SVM is a classification algorithm (supervised learning) in statistics. Assume a layered hyperplane perfectly separates the data. Each point in space will be assigned by −1 or 1. The SVM problem is finding a subclass H so that the margin between the two classes is maximum. If the subclass H hyperplane has the equation: wTxþb¼0 and without loss of generality, this result can assume that for the points closest to H, and have: wTxþb¼1 for the point to be subclassed one, and wTxþb¼  1 for the point to be subclassed −1. Then, the distance between the two classes is: margin ¼2 w. The SVM problem becomes: min w;b 1 2wTw s:t:yiwTxiþb  ��1 In some cases of complex distributed data, the method works poorly; the original data set is mapped to a more dimensional space where the classification is obvious, and the kernel function defines the mapping (Benkraiem & Zopounidis, (2021)). The technique used after the transformation is similar to the case where a perfectly classed hyperplane or the soft margin case exists. With the same idea as SVM, instead of locating an optimal classifier hyperplane, the SVR problem is to find a regression function f xð Þ ¼ wTxþb such that most of the observed points lie in the interval ε;ε½ �. This problem is called linear ε-SVR (Thomas et al., (2017)). Khoa & Huynh, Cogent Economics & Finance (2023), 11: 2191459 https://doi.org/10.1080/23322039.2023.2191459 Page 4 of 15 However, linear ε-SVR is very sensitive to outliers. Specifically, when outliers are present, ε will be larger, making the error based on the report higher. To overcome this, an offset parameter �i; �� i is added to limit the effect from outliers, allowing these outliers’ values to lie outside the boundary (Dhiman et al., (2019)). The ε-SVR problem becomes: min w 1 2kwk2þC∑ n 1 �iþ�� i  �subject to εþ�� i  ��yiwTxib�εþ� Where C > 0 is a regularization parameter, defining the trade-off relationship between the flatness of function f and the prediction errors. 3. Methodology 3.1. Data description The data are all stocks listed on Ho Chi Minh City Stock Exchange, period January 2012 to June 2022. At the end of 30 June 2022, there were 546 listed securities codes, including 403 stocks, two closed-end fund certificates, 9 ETF certificates, 128 covered warrants, and four bonds. The total volume of listed shares reached over 130 billion shares. This study kept only 403 stocks and grouped them into six categories by BM. Stocks with high BM are called value stocks (30%) and are grouped into two categories; stocks with low BM are called growth stocks access (30%). 2 portfolios, the remaining stocks (accounting for 40%) are grouped into two categories. The categories are sorted each year after the end of June (E. F. Fama & French, (2015)). The VN-index was analyzed to represent the market portfolio. This research built the uncertainty index I it based on the CAPM model; the idea is based on the previous studies of Ismailov and Rossi (2018); Rossi and Sekhposyan (2015). Accordingly, this study forecast error eit ¼pit Etpit ð Þj j, in whichEtpit ð Þ ¼ rft þβit rMt rft  �(the expected rate of return is calculated by the CAPM model with a beta coefficient estimated from over 36 months of data). Next, the error to [0,1] is normalized by using the formula: Eit ¼eit Min eit ð Þ Max eit ð Þ Min eit ð Þ A large value of E it implies that the observed rate of return is very different from the expected value. Specifically, the uncertainty index was constructed as follows: Iit ¼0;if0:25 �Eit �0:75 1;elsewhere � Table 1. Variable description Name Variable Description rft 1-Year Government Bond Yield rMt The return rate of the VN-index portfolios BMit Book value over the market value of portfolio i BMMt Book value over the market value of the market portfolio Iit Uncertainty index. I it = 0 implies that the portfolio is stable relative to the market. pit Weighted rate of return pit rmt Outstanding rate of return of the portfolio BMit BMMt Book-to-market ratios in excess of market Khoa & Huynh, Cogent Economics & Finance (2023), 11: 2191459 https://doi.org/10.1080/23322039.2023.2191459 Page 5 of 15 Some of the variables in the study are summarized in Table 1. 3.2. Data processing 3.2.1. Testing the relationship between the value premium and book-to-market ratios in excess of market The portfolio’s expected return always fluctuates due to different impacts from company-specific factors and macroeconomic variables (Khoa & Huynh, (2022c)). Measuring volatility (risk) is a challenge. Previous studies used lagged dividend-to-price (DP ratio) to predict and report the stock’s rate of return (Lettau & Ludvigson, 2005; Yin & Nie, (2021)). The logic of this argument is that the value of the stock is the present value of the expected stream of future dividends by the Gordon model (Gordon, (1959)). Fluctuations in expected return negatively impact the stock price. Because the present value of a stock is the present value of the expected future stream of dividends, the use of a dividend lag would be a poor proxy for the expected dividend. This can be seen in startups, always prioritizing growth overpaying dividends, or dividends are always negative. Furthermore, dividends can be affected by financial decisions more than the stock’s book value. For the above reasons, this study uses the lag of the BM outperforming the market ratio, BM—BM M , to forecast the outperforming market rate, R-R M . The regression Eq.1 has the form: Model1ð Þ :pit rMt ¼αiþβiBMit1BMMt1 ð Þþεit (1) Where: αi: the intercept coefficients; βi: the slopes; εit: error terms The hypothesis was performed to test H 0 : “The value complement does not exist”; the assumption is that the regression coefficients are constants over time. 3.2.2. The value premium and uncertainty index Assuming constant regression coefficients is a challenge. There are many reasons to reject this assumption, such as an increase in the supply of stocks, even shocks in the economy, such as the financial crisis, the oil price crisis, and the COVID-19 pandemic, can create a broken structure (E. F. Fama & French, (2021)). The regression equation with dummy variable I t as Eq.2 was established: Model2ð Þ :pit rMt ¼β0iþβ1iIit þβ2iBMit1BMMt1 ð Þþβ3iIit BMit1BMMt1 ð Þþεit (2) Where I it = 1 if volatility in the portfolio’s return to the market is high and I it = 0 otherwise. 3.2.3. The value premium and support vector regression The SVR algorithm is one of the most powerful output prediction algorithms in machine learning and is widely used in finance (Ma et al., (2021)). The greatest difficulty in using these algorithms is using the appropriate inputs. Fortunately, the theoretical framework of value complements helps in this regard. A useful theoretical framework combined with a powerful prediction algorithm is expected to bring positive results. Model 3 used the SVR algorithm based on Eq.2, and the parameters include the radial kernel function, cost = 1, and gamma = 0.5. Model3ð Þ :pit rMt ¼f Iit;Mit1;MMt1 ð Þ (3) This study divided the research data into two sets: training and testing at the ratio of 7:3 to evaluate the model. Specifically, the period from January 2012 to December 2017 is used for training, and the remaining period is used for testing. The evaluation criterion is a deviation (Ouerhani et al., (2022)). Khoa & Huynh, Cogent Economics & Finance (2023), 11: 2191459 https://doi.org/10.1080/23322039.2023.2191459 Page 6 of 15 Deviation ¼Y^ Y ������ Where: Y: observed values, ^ Y:predicted values 3.2.4. F test for predictive models Statistically, the mean of deviation cannot conclude whether one model is more efficient. In other words, this study needs a formal test to increase the reliability of the conclusion. This research uses the null hypothesis H 0 : “The forecast deviation in the models is the same,” and analysis of variance (ANOVA) was used for this test. The following assumptions are (1) normal distribution, (2) homogeneous variance, and (3) independent observations. 4. Results In Figure 1, at the end of the last trading session of June 2022, the VN-Index reached more than 1,197 points, down 7.36% compared to May 2022, equivalent to a decrease of more than 20% compared to the end of 2021. Stock market liquidity in June, compared to May, the average trading volume and value reached over VND 14,529 billion and 547.70 million shares, respectively, down 2.8% in value and up 1.38% in average volume. In the second quarter of 2022, the average trading value of shares reached more than 17,113 billion VND, with the average trading volume reaching more than 589.15 million shares; respectively, down 20.02% in value and 18.33% in average volume over the same period in 2021. By the end of June 2022, on HOSE, there are 42 enterprises with a market capitalization of more than 1 billion USD, of which three enterprises have a capitalization of over 10 billion USD, including Joint Stock Commercial Bank for Foreign Trade of Vietnam (VCB), Vinhomes Joint Stock Company (VHM) and Vingroup Corporation (VIC). From 1/2012 to 6/2022, the VN-index and BM M tend to increase, as shown in Figure 1. However, at the beginning of 2018, end of 2019, and end of 2022, the market volatility in these periods is very large. The HOSE market increased by 48% in 2017. The VN-index is ranked among the indexes with the most impressive gains globally. The “miracle story” of securities is expected by market participants to be continued in 2018. Therefore, assessment reports on the market outlook also lean towards possible VN-index, which continued double-digit growth to reach 1,120 and then even 1,250 points. Vietnam has become the stock market with the strongest increase in the world, ahead of Brazil, Russia, and Argentina, and nearly three times the increase of the NASDAQ index (National Association of Securities Dealers Automated Quotation System). However, the “hot rise” Figure 1. The VN-index point (divided by 1000) and the BM M ratio for the market portfolio. Khoa & Huynh, Cogent Economics & Finance (2023), 11: 2191459 https://doi.org/10.1080/23322039.2023.2191459 Page 7 of 15 Gang, J., Qian, Z., & Chen, F. (2019). The Aumann-Serrano risk factor and asset pricing: Evidence from the Chinese A-share market. Quantitative Finance, 19(10), 1599–1608. https://doi.org/10.1080/14697688.2019. 1605187 Gordon, M. J. (1959). Dividends, earnings, and stock prices. The Review of Economics and Statistics, 41(2), 99–105. https://doi.org/10.2307/1927792 Henrique, B. M., Sobreiro, V. A., & Kimura, H. (2018). Stock price prediction using support vector regression on daily and up to the minute prices. The Journal of Finance and Data Science, 4(3), 183–201. https://doi. org/10.1016/j.jfds.2018.04.003 Huang, H., Wei, X., & Zhou, Y. (2016). A sparse method for least squares twin support vector regression. Neurocomputing, 211, 150–158. https://doi.org/10. 1016/j.neucom.2015.12.133 Ismailov, A., & Rossi, B. (2018). Uncertainty and deviations from uncovered interest rate parity. Journal of International Money and Finance, 88, 242–259. https://doi.org/10.1016/j.jimonfin.2017.07.012 Jaffe, J. F., Jindra, J., Pedersen, D. J., & Voetmann, T. (2020). Can mispricing explain the value premium? Financial Management, 49(3), 615–633. https://doi. org/10.1111/fima.12272 Jagannathan, R., & Wang, Z. (1996). The conditional CAPM and the cross-section of expected returns. The Journal of Finance, 51(1), 3–53. https://doi.org/10. 1111/j.1540-6261.1996.tb05201.x Khoa, B. T., & Huynh, T. T. (2022a, 21-22 January, 2022). Factors affecting forecast accuracy of individual stocks: SVR algorithm under CAPM framework. Paper presented at the 2022 International Conference for Advancement in Technology (ICONAT). Goa, India. https://doi.org/10.1109/ICONAT53423.2022.9725916 Khoa, B. T., & Huynh, T. T. (2022b). Forecasting stock price movement direction by machine learning algorithm. International Journal of Electrical and Computer Engineering (IJECE), 12(6), 6625. https://doi.org/10. 11591/ijece.v12i6.pp6625-6634 Khoa, B. T., & Huynh, T. T. (2022c. 23-25 March, 2022). The impact of company characteristics and macro factors on stock prices: Machine learning approach. Paper presented at the 2022 International Conference on Decision Aid Sciences and Applications (DASA). https://doi.org/10.1109/DASA54658.2022.9764989 Khoa, B. T., Huynh, T. T., & Gong, D. (2021). Is it possible to earn abnormal return in an inefficient market? An approach based on machine learning in stock trading. Computational Intelligence and Neuroscience, 2021, 1–14. https://doi.org/10.1155/2021/2917577 Kirby, C. (2019). The value premium and expected business conditions. Finance Research Letters, 30, 360–366. https://doi.org/10.1016/j.frl.2018.10.022 Koijen, R. S., Lustig, H., & Van Nieuwerburgh, S. (2017). The cross-section and time series of stock and bond returns. Journal of Monetary Economics, 88, 50–69. https://doi.org/10.1016/j.jmoneco.2017.05.006 Kudar, A., & Sayılgan, G. (2021). A new sustainable dividend policy and valuation model: Decreasing growth rate model. Managerial and Decision Economics, 42(6), 1638–1642. https://doi.org/10. 1002/mde.3332 Lakonishok, J., Shleifer, A., & Vishny, R. W. (1994). Contrarian investment, extrapolation, and risk. The Journal of Finance, 49(5), 1541–1578. https://doi.org/ 10.1111/j.1540-6261.1994.tb04772.x Lettau, M., & Ludvigson, S. C. (2005). Expected returns and expected dividend growth. Journal of Financial Economics, 76(3), 583–626. https://doi.org/10.1016/j.jfi neco.2004.05.008 Liu, J., Stambaugh, R. F., & Yuan, Y. (2019). Size and value in China. Journal of Financial Economics, 134(1), 48–69. https://doi.org/10.1016/j.jfineco.2019.03.008 Loughran, T. (1997). Book-to-market across firm size, exchange, and seasonality: Is there an effect? The Journal of Financial and Quantitative Analysis, 32(3), 249–268. https://doi.org/10.2307/2331199 Ma, Y., Han, R., & Wang, W. (2021). Portfolio optimization with return prediction using deep learning and machine learning. Expert Systems with Applications, 165, 113973. https://doi.org/10.1016/j.eswa.2020.113973 Neves, M. E., Pinto, M. A., de Assunção Fernandes, C. M., & Vieira, E. F. S. (2021). Value and growth stock returns: International evidence (JES). International Journal of Accounting & Information Management, 29(5), 698–733. https://doi.org/10.1108/IJAIM-05-2021-0097 Ouerhani, N., Loehr, B., Rizzotti-Kaddouri, A., Santo De Pinho, D., Limat, A., & Schinderholz, P. (2022). Datadriven thermal deviation prediction in turning machine-tool-a comparative analysis of machine learning algorithms. Procedia computer science, 200, 185–193. https://doi.org/10.1016/j.procs.2022.01.217 Pan, Y., Xiao, Z., Wang, X., & Yang, D. (2017). A multiple support vector machine approach to stock index forecasting with mixed frequency sampling. Knowledge-Based Systems, 122, 90–102. https://doi. org/10.1016/j.knosys.2017.01.033 Qadan, M., & Jacob, M. (2022). The value premium and investors’ appetite for risk. International Review of Economics & Finance, 82, 194–219. https://doi.org/10. 1016/j.iref.2022.06.014 Rosenberg, B., Reid, K., & Lanstein, R. (1985). Efficient capital markets: II. Persuasive Evidence of Market Inefficiency, 11(3), 9–16. https://doi.org/10.3905/jpm.1985.409007 Rossi, B., & Sekhposyan, T. (2015). Macroeconomic uncertainty indices based on nowcast and forecast error distributions. The American Economic Review, 105(5), 650–655. https://doi.org/10.1257/aer.p20151124 Skinner, D. J., & Sloan, R. G. (2002). Earnings surprises, growth expectations, and stock returns or don’t let an earnings torpedo sink your portfolio. Review of Accounting Studies, 7(2), 289–312. https://doi.org/10. 1023/A:1020294523516 Smith, S. C., & Timmermann, A. (2022). Have risk premia vanished? Journal of Financial Economics, 145(2), 553–576. https://doi.org/10.1016/j.jfineco.2021.08.019 Stambaugh, R. F. (1999). Predictive regressions. Journal of Financial Economics, 54(3), 375–421. https://doi.org/ 10.1016/S0304-405X(99)00041-0 Stattman, D. (1980). Book values and stock returns. The Chicago MBA: A Journal of Selected Papers, 4(1), 25–45. Thomas, S., Pillai, G. N., & Pal, K. (2017). Prediction of peak ground acceleration using ϵ-SVR, ν-SVR and Ls-SVR algorithm. Geomatics, Natural Hazards and Risk, 8(2), 177–193. https://doi.org/10.1080/19475705.2016. 1176604 Varela, O. (2022). Duration-adjusted betas. Review of Financial Economics, 40(2), 168–173. https://doi.org/ 10.1002/rfe.1144 Yin, L., & Nie, J. (2021). Adjusted dividend-price ratios and stock return predictability: Evidence from China. International Review of Financial Analysis, 73, 101618. https://doi.org/10.1016/j.irfa.2020.101618 Zhang, J., Li, L., & Chen, W. (2021). Predicting stock price using two-stage machine learning techniques. Computational Economics, 57(4), 1237–1261. https://doi.org/10.1007/ s10614-020-10013-5 Zheng, J., Wang, Y., Li, S., & Chen, H. (2021). The stock index prediction based on SVR model with bat optimization algorithm. Algorithms, 14(10), 299. https:// doi.org/10.3390/a14100299 Khoa & Huynh, Cogent Economics & Finance (2023), 11: 2191459 https://doi.org/10.1080/23322039.2023.2191459 Page 14 of 15 © 2023 The Author(s). This open access article is distributed under a Creative Commons Attribution (CC-BY) 4.0 license. You are free to: Share — copy and redistribute the material in any medium or format. Adapt — remix, transform, and build upon the material for any purpose, even commercially. The licensor cannot revoke these freedoms as long as you follow the license terms. Under the following terms: Attribution — You must give appropriate credit, provide a link to the license, and indicate if changes were made. You may do so in any reasonable manner, but not in any way that suggests the licensor endorses you or your use. No additional restrictions You may not apply legal terms or technological measures that legally restrict others from doing anything the license permits. Cogent Economics & Finance (ISSN: 2332-2039) is published by Cogent OA, part of Taylor & Francis Group. Publishing with Cogent OA ensures: • Immediate, universal access to your article on publication • High visibility and discoverability via the Cogent OA website as well as Taylor & Francis Online • Download and citation statistics for your article • Rapid online publication • Input from, and dialog with, expert editors and editorial boards • Retention of full copyright of your article • Guaranteed legacy preservation of your article • Discounts and waivers for authors in developing regions Submit your manuscript to a Cogent OA journal at www.CogentOA.com Khoa & Huynh, Cogent Economics & Finance (2023), 11: 2191459 https://doi.org/10.1080/23322039.2023.2191459 Page 15 of 15