The Winner–Loser Effect in Japanese Stock Returns

Abstract

This study examines the winner–loser effect using stocks listed on the Tokyo Stock Exchange (TSE) from 1975 to 1997. We uncover significant return reversals dominating the Japanese markets, especially over shorter periods such as 1 month. No momentum effect is observed, however. The 1- month return reversal remains significant even after adjusting for firm characteristics or risk. While the 1-month return reversal is not related to industry classification, it is partially a result of higher future returns to loser stocks with low trading volume. Our results show that investor overreaction may be a possible explanation for the 1-month return reversal in Japan.

The original article first appeared in the Japan and the World Economy 16: 471–485, 2004. A newly written addendum has been added to this book chapter.

Addendum

This addendum has been newly written by Toshifumi Tokunaga for this book chapter.

1.1 Ten Years After “The Winner–Loser Effect in Japanese Stock Returns”: A Review and Recalculation

Around 2001, when we wrote this chapter, the short-term momentum effect was investigated intensively, while the discussion of the return reversal effect was temporarily settled in the U.S. At that time, our paper presented evidence that the short-term momentum effect did not exist in the Japanese stock market.

Discussion regarding the absence of the short-term momentum effect in the Japanese stock market was already underway even in the U.S. Daniel et al. (1998) proposed a model that simultaneously explains the return reversal effect and the short-term momentum effect based on a behavioral approach. They specifically referred to the absence of short term momentum in Japan. The main contribution of our paper is to confirm the absence of the short-term momentum effect using long-term time-series data in a way similar to the standard approach adopted by previous studies in the U.S.

We examine the possible explanations to the short term momentum which has been proposed by recent studies since our paper appeared. The key question is whether short term momentum is really absent in Japan using a longer period of data. A brief review and the results of recalculation follow below.

1.1.1 Discussion After Our Paper

There are two different approaches in the subsequent discussions of the long-term return reversal (De Bondt and Thaler 1985) and short-term momentum (Jegadeesh and Titman 1993).

The first is a rational approach. Fama and French (1996) who found that although the long-term reversal effect can be explained by their three-factor model, the short-term momentum effect cannot be explained (Grundy and Martin 2001; Korajczyk and Sadka 2004; Muga and Santamaria 2007).

The second is a behavioral approach, which is represented by the following three studies (Daniel et al. 1998; Barberis et al. 1998; Hong and Stein 1999).

Since then, researchers have presented several studies focusing on the discovery of overlooked factors and improvement of the model structure by adding a liquidity measure to Fama and French’s three factors (Lee and Swaminathan 2000; Avramov et al. 2006; Sadka 2006), switching a model structure under the market conditions (Chordia and Shivakumar 2002; Cooper et al. 2004; Bhojraj and Swaminathan 2006), and using proxy variables to represent investor behavior (Grinblatt and Han 2005; Hvidkjaer 2006; Chui et al. 2010).

1.1.2 A Revisit to the Momentum Effect in Japan

Though De Bondt and Thaler (1985) and Jegadeesh and Titman (1993) were exposed to many criticisms at their time of publication, they later reviewed the robustness of their results (De Bondt and Thaler 1987; Jegadeesh and Titman 2001).

The Fama–French alphas for momentum portfolio returns and market states in Japan from 1977 to 2005 are reported in Table 21.7, which is reconstructed from the results in Tokunaga (2008a, b, c, 2009). This table updates the data used in our paper 8 years ago. The results (“ALL”) show that the significant short-term momentum effect still does not exist in the Japanese stock market.

Table 21.7 Fama–French alphas for momentum portfolio returns and market states in Japan (1977–2005)

However, once the momentum portfolio returns are classified according to market states, a very interesting result appears. When the stock prices for the past 36 months are trending upward, the risk-adjusted returns of the subsequent momentum portfolios are positive and statistically significant. These results are consistent with those found in the U.S. market (Cooper et al. 2004). On the other hand, when the stock prices for the past 36 months are trending downward, the risk-adjusted returns of the subsequent momentum portfolios are negative, but statistically insignificant. These results are consistent with those found in the U.S. market (Cooper et al. 2004), and are consistent with the viewpoint that momentum returns are not significant.

If the result as a whole is considered, why does the momentum effect appear in the U.S., but not in Japan? There is one interesting number. Although NUP/NALL, the ratio of the number of observations, in the U.S. (Table I in Cooper et al. (2004)) is approximately 85 %, it falls to approximately 65 % in Japan. This result might be a clue to solve the lack of momentum effect in Japan. Thus, trifles are possibly the cause. In order to make this idea less controversial, many empirical studies addressing the Japanese stock market would need to be published from now on.