Investor attention, firm-specific characteristic, and momentum: A case of the Korean stock market

Abstract

This study examines the sources of negative momentum profits by combining investor attention and the properties of common and firm-specific factors. We choose the Korean stock market as a good case to characterize the negative momentum profits identified in Asia. In both portfolio and stock analyses, a method is devised to generate return data involving the property of each common and firm-specific factor within stock groups by investor attention. This study found significant negative momentum profits within the stock group with high investor attention. This momentum effect is highly dependent on the reversed performance of the past loser portfolio, not the continued performance of the past winner portfolio, and this reversal is strongly attributable to the properties of firm-specific factors, and not those of common factors. These results are robustly consistent regardless of changes in empirical design and the consideration of influence factors, market dynamics, and other stock markets.

Introduction

Momentum and reversal effects have attracted interest from many researchers aiming to develop practical investment strategies using past return information.¹ Research on cognitive psychology has determined that attention in human brain activity is based on limited resources (Kahneman, 1973; Pashler and Johnston, 1998), and behavioral studies report that investor attention is a significant factor affecting investment decisions (Barber and Odean, 2008; Corwin and Coughenour, 2008; Peng and Penguin, 2009). From the perspective of behavioral models, investor attention based on past information leads to either an overreaction-driven bias or an underreaction-driven bias toward future investment performance (Daniel et al., 1998; Barberis et al., 1998).² Hou et al. (2009) report evidence supporting the dual role of investor attention, such as overreaction-driven price momentum and underreaction-driven earnings momentum according to investor attention. These studies suggest that investor attention is an important source to explain positive momentum profits.

Asian stock markets, including that of Korea, rarely exhibit the existence of positive profits from the momentum strategy in contrast to developed markets (Chui et al., 2000, 2010). Chui et al. (2000) explain negative momentum profits using the cultural differences between civil law and common law, and emphasize that positive momentum profit is not verified after the opening to foreign investors of Asian stock markets such as Japan, Taiwan, Korea, and Indonesia.³ Chui et al. (2010) also attributed this to another cultural difference between the collectivism of Asia countries and the individualism of Western countries. On the other hand, Du et al. (2009) present an alternative explanation for the negative momentum profits in the Taiwanese stock market. They found that the more frequent down market states relative to the U.S. stock market can be a significant factor, rather than the cultural difference proposed by Chui et al. (2000).⁴ Studies that investigate the momentum strategy in the Korean stock market report the existence of negative momentum profits (Chae and Eom, 2009; Lee and Cho, 2014). Kim and Byun (2011) also report the existence of negative momentum profits through a review of the previous studies focusing on the Korean stock market since the 1980s. These findings contrast with those of prior studies that generally support the existence of positive momentum profits in developed markets, including the U.S. stock market. This contradiction necessitates further research to explore the sources of the negative momentum profits, especially in Asian stock markets. The Korean stock market is a good case to empirically investigate the sources of the negative momentum profit. Thus, this study utilizes data on the Korean stock market. In addition, as compared to traditional momentum by Jegadeesh and Titman (1993), Blitz, Huij and Martens (2011) argue that positive momentum profits are created due to the properties of firm-specific factors. They present evidence that the momentum portfolio constructed using residual returns after removing the properties of common factors through the pricing models has more significant positive profits in the future periods than that of the traditional momentum, which is called the residual momentum. Using residual momentum by CAPM, Chaves (2016) reports evidence supporting the existence of positive momentum profits in the Japanese stock market, which is well known as a representative stock market that does not support the existence of positive momentum profits.

This study investigates the sources of the negative momentum profits identified in the Korean stock market, based on the perspectives of residual momentum by Blitz et al. (2011) and investor attention by Hou et al. (2009). For this research purpose, we employ the following empirical design. First, stock groups are classified according to investor attention, and then the momentum effect within each stock group is empirically investigated. This study employs trading volume as a proxy of investor attention that is prevalently utilized in the previous studies⁵, that is, trading volume turnover. The study data is separated into a stock group with high investor attention (high turnover stock group) and a stock group with low investor attention (low turnover stock group), after which the momentum effect is examined within each stock group. Also, in order to discriminate whether momentum is related to the overconfidence of investor attention, the performance patterns according to prolonging the holding period are examined. Second, traditional momentum (Jegadeesh and Titman, 1993) and residual momentum (Blitz et al., 2011) are used together to investigate the effect of firm-specific factors on momentum profit. For this purpose, three types of return data are generated: original returns, common returns that include only the properties of common factors, and residual returns that remove the properties of common factors. In this study, the residual returns are defined as firm-specific returns featuring only the properties of firm-specific factors.⁶ This study employs a method that combines random matrix theory (RMT) and singular value decomposition (SVD) in order to decompose the property of each common and firm-specific factor from the original returns that include the properties of all factors. Unlike Blitz et al. (2011), our method offers the advantage of overcoming the problem of residual returns being dependent on the selection of the pricing model; i.e., it is a model-free method. Third, the arbitrage portfolio suggested by Lo and MacKinlay (1990) is utilized to identify the momentum effect. Based on the average holding period returns of all stocks within each stock group in the prior period, stocks are categorized in the winner and loser portfolios, i.e., stocks with a greater holding period return than the average holding period returns of all stocks are classified in the winner portfolio, and otherwise, in the loser portfolio. The arbitrage portfolio is constructed under the condition of buying winners and selling losers. Using the performance of the arbitrage portfolio in the future period, the hypothesis on the momentum effect is statistically evaluated.⁷ If the arbitrage portfolio in the future period has significant positive performance, this is evidence supporting the hypothesis of positive momentum profit. The existence of negative momentum profit is determined by the significant negative performance of the arbitrage portfolio. Fourth, the cross-sectional regression of Fama and MacBeth (1973) is employed to examine the momentum effect from the perspective of individual stocks. This method offers the advantage of confirming the robustness of the portfolio analysis from a complementary viewpoint. We investigate whether past holding period returns of stocks belonging in winner and loser portfolios significantly explain the changes of future holding period returns of stocks in short- and long-term periods. In addition, factor premiums on each factor risk from the three types of return data are examined using the two-step regression method. Using time-series regression in the past period, the factor risks of stocks are estimated using factor returns with each property of the three types of return data. Next, using the cross-sectional regression in the future period, the factor premium is estimated using three types of factor risk. Fifth, the relationship between investor overreaction and future performance is measured for each of the winner and loser portfolios. Based on the hypothesis of overconfidence bias suggested by Daniel et al. (1998), the performance patterns of winner and loser portfolios by means of extending the holding period from 1-month to 36-month are investigated within each stock group. For further exploration, the strength of investor overreaction is quantified by using the two measures of cross-sectional mispricing index (MI) by Eom (2018) and information discreteness (ID) by Da et al. (2014).

Finally, the following robust test on the previous results is conducted. The first is changes in the criterion for classifying stock groups. This study utilizes firm size and book-to-market equity ratio, which were found in previous studies to significantly affect the momentum effect. The second is the effects of ownership structure on the momentum effect. We utilize two equity ratios of minority shareholders and foreign shareholders based on Chui et al. (2000). The third is changes in the method used to form the portfolios. We employ the quintile portfolio and its zero-cost portfolio between the extreme winner portfolio and the extreme loser portfolio. The fourth is changes in the roll-sampling method on the sub-period in the whole period. This study utilizes the overlapping holding period method, similar to Jegadeesh and Titman (1993), compared to the non-overlapping holding period method applied to the main results of this study. The fifth is to consider the up and down market dynamics on the momentum effect. This is based on Copper, Gutierrez and Hameed (2004), Du et al. (2009), and Asem and Tian (2010). The sixth is the investigation for whether results identified in the Korean stock market are unique. This study utilizes data in the Japanese stock market, which is known to show evidence supporting the negative momentum profit. Under the aforementioned empirical design, this study robustly investigates the sources of the negative momentum profits identified in the Korean stock market by means of the combination between stock groups by investor attention and return data types by common and firm-specific factors. We expect our results to give new insight into the possible sources that characterize negative momentum profits.

The main results are summarized as follows. First, the momentum profits from the 12 M/1 M strategy with a 12-month formation period and a 1-month holding period are negative in all arbitrage portfolios, regardless of turnover stock groups. However, significant results are only obtained for the high turnover stock group. The negative momentum profits are due to the reversed performance of the past loser portfolio with high positive profit in the future period, not the continued performance of the past winner portfolio. Especially, this performance is attributable to the properties of firm-specific factors, rather than those of common factors. Second, the results on extending the holding period from 1-month and 36-month also show the negative momentum profits in all arbitrage portfolios. However, in terms of statistical significance, the high turnover stock group has significant performance until the 12-month holding period. The performance of the arbitrage portfolio exhibits a downward trend during the extension of the holding periods, and is dominantly dependent on the reversal of the loser portfolio, and not on the winner portfolio. That is, the loser portfolio has a significant upward trend in performance, but the winner portfolio does not have significant performance. On the other hand, when comparing performance according to the type of return data, the negative momentum profits from the original returns are definitely attributable to firm-specific factors rather than common factors. Third, the cross-sectional regression on stocks shows that stock performance in the future holding period has a significant negative relationship with the performance of stocks that belong in the past loser portfolio, and this result is affected by the properties of firm-specific factors. Meanwhile, the characteristics of factor premiums on the factor risks of the three-type return data differ between winner and loser portfolios. Specifically, stocks that belong to the past loser portfolio have a significant positive factor premium on the risk of firm-specific factors. However, stocks belonging to the past winner portfolio do not show a significant factor premium for any factor risks. Finally, the robustness of the results is verified in the changes of empirical design, such as the criterion used to classify the stock groups, the method used to form the portfolio, roll-sampling method, the consideration of ownership structure, market dynamics, and the Japanese stock market. The robust test results are similar to the previous results. Therefore, our findings suggest that the negative momentum profit identified in the Korean stock market is found within the high turnover stock group with high investor attention. Within the high turnover stock group, the reversed performance of the past loser portfolio (not the continued performance of the past winner portfolio) leads to the existence of negative momentum profit and the properties of firm-specific factors (not common factors) drive the performance of the past loser portfolio.

The study is organized as follows. Chapter 2 describes the data and methods used in the testing process. Chapter 3 shows the results of the portfolio and stock analyses according to the research objectives, and Chapter 4 presents the robust test results. The final chapter summarizes the main findings and presents the study implications.