Due to the influential works of Fama and French (1992, 1993), the value premium often refers to the book-to-market effect although there are alternative valuation measures. Recent evidence indicates that the book-to-market effect has weakened over time or even disappeared in the U.S. stock market. The literature suggests that, even in the Korean stock market, the value premium measured by an individual valuation ratio, including book-to-market, may significantly fluctuate over time, and a value factor constructed based solely on the book-to-market ratio may not capture the true value premium. This study aims to verify the predictability of value premiums based on various valuation measures, not just the book-to-market equity ratio, and to identify which value predictor is consistently useful in the Korean stock market over a long period from 1987 to 2023. We further attempt to investigate why the most useful valuation measure has a better predictive performance than others. To this end, we consider six valuation measures, including the book-to-market ratio, retained earnings-to-market ratio, contributed capital-to-market ratio, earnings-toprice ratio, cash flow-to-price ratio, and dividend-to-price ratio. The main findings are summarized as follows. First, except the contributed capital-to-market ratio, all alternative value measures can significantly predict the cross-section of stock returns when book-to-market is controlled for. The most significant is retained earnings-to-market, followed by dividend-to-price, and these two measures predict value premiums more strongly than the book-to-market ratio. The results are consistent with Ball et al. (2020) in that retained earnings-to-market is a strong predictor of the value premium whereas contributed capital-to-market is not, but controlling for the retained earnings-to-market ratio does not eliminate the predictive power of book-to-market in Korea. Second, we construct value factors based on each of the six valuation measures to compare their historical performance and find that the HML factor, which represents the book-to-market effect, has a Sharpe ratio of 0.35 in the period prior to July 2005 and 0.83 in the period after July 2005, indicating a stronger recent performance unlike in the U.S. market. However, when compared to alternative value factors based on useful valuation measures, the HML factor performs worst before July 2005 and best after July 2005, which implies that the book-to-market effect substantially varies over time, as in the U.S. market. On the other hand, the retained earnings-to-market factor has the highest Sharpe ratio of 0.64 throughout the sample period and almost equal Sharpe ratios during the periods before and after July 2005. In addition, the retained earnings-to-market factor has higher correlations with the earnings-to-price, cash flow-to-price, and dividend-to-price factors than the HML factor. Therefore, retained earnings-tomarket is the strongest and most reliable predictor of the true value premium in any period. The HML factor incorporates not only the useful information contained in retained earnings-to-market but also the information in contributed capital-to-market, which has little to do with the value premium, and as a result, it performs worse in periods when it is highly affected by contributed capital-to-market. In this sense, the HML factor may be a relatively unstable indicator of the true value premium. Third, in spanning regressions, the retained earnings-to-market factor subsumes all other value factors, including HML, and has additional information that the other value factors do not have. The next most valuable value factor is the one based on the dividend-to-price ratio. The remaining value factors and the HML factor do not seem to capture all of the information from each other and their priority is not clear. As a result, the value factor that is most useful as an asset pricing factor seems to be the retained earnings-to-market factor. Finally, considering that retained earnings represent the difference between accumulated earnings and accumulated dividends over the firm’s history, we investigate whether accumulated earnings or accumulated dividends over recent years are more responsible for predicting the cross-section of expected returns and find that accumulated dividends, rather than accumulated earnings, primarily drive the predictive power of retained earnings-to-market in Korea. Moreover, we find that retained earnings-to-market can strongly predict the growth in earnings over two to three years. In contrast, the book-to-market, cash flow-to-price, dividend-toprice ratios do not predict earnings growth. Therefore, the outperformance of retained earnings-to-market as a predictor of the value premium may be related to useful information about future earnings growth. Although we show that the book-to-market effect has recently become stronger in Korea, unlike in the U.S., our findings are consistent with recent evidence in the U.S. market in that the book-to-market effect may not always represent a true value premium. Our findings consistently show that the retained earnings-to-market ratio is the best predictor of the value premium among well-known and easily observable valuation measures. Moreover, our results of spanning tests suggest that it may be desired to construct a value factor based on retained earnings-to-market rather than book-to-market in asset pricing models.
