Feb. 2007, Vol.3, No.2 (Serial No.21) Journal of Modern Accounting and Auditing, ISSN1548-6583, USA Medium-term and Long-term Momentum and Contrarian Effects on China during 1994-2004 DU Xing-qiang, NIE Zhi-ping (School of Management, Xiamen University, Xiamen 361005, China) Abstract: We examine momentum and contrarian effects in China s stock market during 1994-2004 and find that no medium momentum effects exist. Meanwhile, contrarian strategy works effectively over the horizon of 18-36 months. 1-month gap for avoiding bid-ask bounce and lead-lag effect make no considerable change to our empirical results. Transaction costs seem to have no significant impact on contrarian strategies profit. Key words: medium-term momentum; long-term momentum; contrarian effects; transaction costs 1. Introduction Return continuation or return reversal has long been debated for their existence causing people to adopt momentum strategy or contrarian strategy. These two strategies work as two faces of one mirror. If stock price changes is larger than zero (this phenomenon is also referred as price continuation), then momentum strategy should be used. Otherwise, contrarian strategy outperforms momentum strategy. Generally speaking, in developed countries, momentum strategy has been confirmed during intermediate period while contrarian strategy is discovered efficient during long-term period and short-term period. However, no consistent evidence about medium-term momentum strategy has been found in emerging markets. Since China belongs to one of the largest emerging market, its stock price features and related strategies may draw many international investors attention. Allowing for past researches lack consistent results, we decide to put our efforts solely on medium-term momentum strategy in China and expand the time horizon as long as possible. 2. Literature Review De Bondt & Thaler (1985) documented return reversal over long horizon and in fact found contrarian strategy could bring impressive abnormal return in the long run. Meanwhile, Jegadeesh & Titman (1993) documented momentum strategy over medium-term horizon. They found this strategy could outperform the market with even 1% abnormal return in 3-12 months. Besides, they had extended their criteria for selecting past winners and losers to earn announcements. Later researchers extended momentum researches by using different criteria, comparing these strategies return internationally, adopting different methodology and discovering the This paper is supported by the National Natural Science Foundation of China (NSFC, No.70302012), the New Century Excellent Talents (NCET-04-0596) and the Fok Ying Tung Education Foundation (No.101087). DU Xing-qiang (1974- ), professor of School of Management, Xiamen University; research field: accounting and financial research in capital market. NIE Zhi-ping (1979- ), h.d. candidate of School of Management, Xiamen University; research field: accounting and financial research in capital market. 63
sources of momentum strategy. Rouwenhorst (1998) found that an internationally diversified portfolio that invested in medium-term winners and sold past medium-term losers and earned approximately 1% per month. We should note that the samples of this research are 12 European countries. When Rouwenhorst (1999) focused on individual stock returns from 20 emerging countries, momentum strategy seemed to work unsuccessful in some countries. The author considered that this phenomenon resulted from imprecise measurement of return premiums in individual countries caused by small number of stocks in some countries and high volatility of returns. Fama and French (1996) analyzed return reversal and return continuation and concluded that the profit of long-horizon contrarian strategy could be explained by three-factor model, but the same model failed to account for the profit of medium-horizon momentum. With the unsolved momentum effect, Carhart (1997) proposed a four-factor model. Furthermore, he discovered that although momentum phenomenon did exist, transaction cost might offset the abnormal return brought by momentum strategy. ZHANG Ren-ji, ZHU ing-fang and WANG Huai-fang (1998) found that return discontinuation for winners and return continuation for losers. SHEN Yi-feng and WU Shi-nong (1999) discovered that neither winners nor losers earned abnormal return compared to the market. WANG Yong-hong and ZHAO Xue-jun (2001) examined the profitability of 1-12 months momentum strategies and 1-3 years contrarian strategies over 1993-2000 interval in China s stock market. They found prominent contrarian strategies profit but no evident momentum strategies profit. ZHOU Lin-jie (2002) documented the profitability of momentum strategies over 1995-2000 interval in China s stock market. The research discovered that the formation and holding period is adversely related to returns of momentum strategies. Since selling short was prohibited in China at present, solely selling winners could not make profit. WU Shi-nong and WU Chao-peng (2003) tested momentum strategy with 342 listed firms during 1997-2002 in China s Shanghai stock market. They discovered that buying winners with high profits and selling losers with low profits in previous 6 months could earn evident abnormal return in the following one year. WANG (2004) examined the returns of momentum strategies between July of 1994 and December of 2000 and found losers in previous 6-12 months turned out to be winners in next 6-12 months. This result showed that contrarian strategies work well in China s stock market. Furthermore, the contrarian strategies profit could be well explained by Fama & French s three-factor model. 3. Returns of Medium-term Momentum Strategies: An Overall Results Our research focuses on price momentum strategies. One momentum strategy (we also call J-month/K-month strategy) is similarly constructed as in Jegadeesh & Titman (1993). For example, (3,12) strategy is constructed as follows: at the beginning of each month t the securities are ranked in ascending order on the basis of their return in the past 3 months. Based on these rankings, ten decile portfolios are formed that equally weight the stock contained in each decile. The top decile portfolio is called the losers decile and the bottom decile is called the winners decile. In each month t, the strategy buys the winner portfolio and sells the loser portfolio, holding this position for 12 months. Meanwhile, the strategy closes out the position initiated in month t-12. Therefore, under this strategy we revise the weights on 1/12 of the securities in the entire portfolio in any given month and carry over the rest from the previous month. The profit of the strategies is calculated for a series of portfolios that were rebalanced monthly to maintain equal weights. Our work documents the returns of the zero-cost portfolio strategies over 1994-2004. These portfolios consist 64
of all listed firms traded on the SHSE and SZSE. In order to exclude the impact of IOs, we require 6 months transaction record before strategy formation periods start. Allowance for all the above data comes from CSMAR database. Because bid-ask bounce can attenuate the continuation effect, we also show the average returns if the portfolio formation period is one month delayed related to ranking month. From Table 1, we can easily find that no strategies earn non-zero return when formation period or holding period is short (e.g. J or k belongs to 3-12 months). Returns of the losers portfolios for (3,24) and (3,30) strategies and the winners portfolios for (3,12), (3,24), (3,30), (6,9), (9,6) strategies are significant at 10% level or higher. All these returns are between 1.5% and 2.24%, quite substantial. However, when we turn our attention to winners minus losers portfolios, no medium-term strategies earn significant return. From this result, we conclude that intermediate momentum effect does not exist in China during 1994-2004. Formation period Table 1 Returns of zero-cost portfolios Holding period(k) (J) ortfolio 3 6 9 12 18 24 30 36 3 1 0.0168 0.0057 0.0090 0.0123 0.0148 0.0206 * 0.0224 * 0.0173 10 0.0154 0.0128 0.0145 0.0170 * 0.0150 0.0204 * 0.0199 * 0.0145 10-1 -0.0014 0.0072 0.0055 0.0047 0.0002-0.0002-0.0024-0.0027 6 1 0.0047 0.0052 0.0100 0.0112 0.0156 0.0200 * 0.0181 0.0175 10 0.0148 0.0144 0.0164 * 0.0135 0.0156 0.0175 0.0145 0.0120 10-1 0.0101 0.0091 0.0064 0.0023 0.0000-0.0025-0.0036-0.0056 9 1 0.0039 0.0072 0.0115 0.0127 0.0180 0.0196 0.0192 0.0218 10 0.0147 0.0157 * 0.0126 0.0129 0.0167 0.0149 0.0144 0.0141 10-1 0.0108 0.0085 0.0011 0.0002-0.0013-0.0047-0.0048-0.0078 12 1 0.0077 0.0098 0.0133 0.0149 0.0186 0.0157 0.0201 0.0269 10 0.0149 0.0110 0.0117 0.0130 0.0137 0.0100 0.0114 0.0144 10-1 0.0072 0.0013-0.0017-0.0018-0.0049-0.0056-0.0087-0.0125 18 1 0.0081 0.0104 0.0157 0.0158 0.0122 0.0155 0.0280 * 0.0286 10 0.0111 0.0124 0.0154 0.0133 0.0080 0.0069 0.0137 0.0088 10-1 0.0031 0.0020-0.0002-0.0026-0.0042-0.0086-0.0143 * -0.0198 * 24 1 0.0123 0.0136 0.0161 0.0119 0.0137 0.0234 0.0308 * 0.0414 * 10 0.0150 0.0123 0.0118 0.0085 0.0047 0.0083 0.0097 0.0126 10-1 0.0027-0.0013-0.0044-0.0034-0.0089-0.0150 * -0.0211 ** -0.0288 ** 30 1 0.0125 0.0097 0.0117 0.0114 0.0198 0.0258 0.0433 ** 0.0460 10 0.0093 0.0059 0.0068 0.0043 0.0049 0.0042 0.0120 0.0114 10-1 -0.0032-0.0038-0.0049-0.0071-0.0149 * -0.0216 ** -0.0313 ** -0.0347 * 36 1 0.0054 0.0074 0.0116 0.0163 0.0212 0.0358 * 0.0460 * 0.0476 10 0.0053 0.0024 0.0041 0.0045 0.0016 0.0065 0.0118 0.0024 10-1 0.0000-0.0049-0.0075-0.0118-0.0196 * -0.0293 ** -0.0342 ** -0.0452 * Notes: (1) At the beginning of each month t, the securities are ranked in ascending order on previous J-month performance. The top decile portfolio is called the losers decile and the bottom decile is called the winners decile. In each month t, the strategy buys the winner portfolio and sells the loser portfolio, holding this position for K months. The table gives the average monthly rebalanced returns on these portfolios for the period 1994 to 2004. The portfolios are formed immediately after ranking. T-stat is the average return divided by its standard error. T-stat is not provided directly in the table because otherwise the length of the table will occupy too much space. (2) * Denotes significance at the 10% level or higher. 65
(3) ** Denotes significance at the 5% level or higher. Table 1 also demonstrates to us other return patterns. The 4 4 square consists of formation period j or holding period k equal to 18, 24, 30, and 36 respectively. The longer the time horizon (denoting j or k) is, the higher the returns are and the more returns are significant. The winners minus losers portfolio returns for (18,30), (18,36), (24,24), (30,18), (30,36), (36,18), (36,36) are significant at 10% level or higher, while (24,30), (24,36), (30,24), (30,30), (36,24), (36,30) are significant at 5% level or higher. All these returns are below zero, from -4.52% to 1.43%. These results show that contrarian strategies can make profit over long-term horizon. Another interesting pattern is symmetrical strategies exhibiting similar return results. For example, strategy (24,30) results in -2.11%, while strategy (30,24) results in -2.16%. Both returns are significant at 5% level or higher. Usually, the returns for winners minus losers portfolios mainly come from the loser portfolio. But, it doesn t mean we can make profit from direct investing on the loser portfolio since returns for loser portfolios are less significant. To avoid bid-ask bounce and lead-lag effects, we postponed holding period k one month later and have related results shown in Table 2. Significant returns for shorter ranking or holding periods (eg. J or k belongs to 3-12 months) in Table 2 are a little smaller than their counterparts in Table 1, while significant for both longer ranking and holding periods (e.g. J and k belonging to 18-36 months) in Table 2 are a little higher than their counterparts in Table 1. However, these overall results reported in Table 2 are quite similar in Table 1. Most strategies with significant return are the same. Table 2 Formation period Returns of zero-cost portfolios with 1 month skip between j and k periods Holding period(k), skip 1 month (J) ortfolio 3 6 9 12 18 24 30 36 3 1 0.0122 0.0027 0.0086 0.0123 0.0141 0.0209 * 0.0222 * 0.0158 10 0.0153 0.0113 0.0141 0.0153 * 0.0143 0.0196 * 0.0193 * 0.0121 10-1 0.0031 0.0085 * 0.0055 0.0029 0.0003-0.0013-0.0030-0.0037 6 1 0.0023 0.0045 0.0097 0.0137 0.0166 0.0192 * 0.0192 0.0199 10 0.0129 0.0131 0.0134 0.0142 0.0160 * 0.0154 0.0165 0.0129 10-1 0.0105 0.0086 0.0037 0.0005-0.0005-0.0038-0.0028-0.0070 9 1 0.0042 0.0065 0.0123 0.0148 0.0186 0.0204 * 0.0178 0.0251 10 0.0145 0.0122 0.0123 0.0131 0.0157 0.0140 0.0117 0.0169 10-1 0.0104 0.0057 0.0000-0.0017-0.0029-0.0064-0.0061-0.0082 12 1 0.0073 0.0099 0.0140 0.0177 0.0173 0.0165 0.0216 0.0290 10 0.0122 0.0102 0.0109 0.0141 0.0115 0.0120 0.0118 0.0146 10-1 0.0049 0.0003-0.0030-0.0036-0.0058-0.0045-0.0099-0.0144 * 18 1 0.0090 0.0115 0.0166 0.0166 0.0121 0.0169 0.0291 * 0.0319 10 0.0101 0.0127 0.0141 0.0117 0.0089 0.0068 0.0133 0.0118 10-1 0.0011 0.0012-0.0025-0.0048-0.0032-0.0101-0.0158 * -0.0201 * 24 1 0.0133 0.0121 0.0167 0.0135 0.0144 0.0245 0.0324 * 0.0462 * 10 0.0138 0.0095 0.0097 0.0104 0.0034 0.0074 0.0111 0.0158 10-1 0.0005-0.0027-0.0070-0.0031-0.0110-0.0171 ** -0.0213 ** -0.0304 ** 30 1 0.0131 0.0077 0.0106 0.0141 0.0203 0.0277 0.0462 ** 0.0344 10 0.0068 0.0061 0.0037 0.0044 0.0028 0.0051 0.0138 0.0000 10-1 -0.0063-0.0016-0.0069-0.0097-0.0175 * -0.0226 ** -0.0324 ** -0.0344 * 36 1 0.0059 0.0071 0.0126 0.0195 0.0222 0.0385 * 0.0348 0.0525 * 10 0.0034 0.0007 0.0047 0.0045 0.0014 0.0080 0.0011 0.0058 10-1 -0.0025-0.0064-0.0079-0.0150-0.0208 * -0.0306 ** -0.0336 * -0.0468 * Notes: (1) At the beginning of each month t, the securities are ranked in ascending order on previous J-month performance. The 66
top decile portfolio is called the losers decile and the bottom decile is called the winners decile. In each month t, the strategy buys the winner portfolio and sells the loser portfolio, holding this position for K months. The table gives the average monthly rebalanced returns on these portfolios for the period 1994 to 2004. The portfolio formation occurs one month after the ranking takes place. T-stat is the average return divided by its standard error. T-stat is not provided directly in the table because otherwise the length of the table will occupy too much space. (2) * Denotes significance at the 10% level or higher. (3) ** Denotes significance at the 5% level or higher. 4. Returns Allowing for Transaction Costs Few of previous studies on China s momentum and contrarian strategies have considered transaction cost. Whether these strategies with significant returns can still make profits is a critical question to investors in reality. In our country, transaction costs mainly consist of two parts: commission fee and stamp tax. Usually, they are calculated as certain percent of transaction amount. Furthermore, these taxes are levied upon both buying amount and selling amount. Therefore, returns allowing for transaction costs should be calculated in Formula (1). In contrast, returns without considering transaction costs should be calculated in Formula (2). (1 c) i1 (1 + c) (1 + c) i0 i0 = R tc (1) i 0 i 1 i 0 = With Formula (1) and (2), we can deduct Formula (3). R (2) 1 R tc = (1 + R)(1 c) /(1 + c) + (3) Where, i1 is the total amount earned through selling stocks; i0 is the total amount expensed for buying stocks; R tc is the one-year buy-and-hold return, including transaction costs; R is the one-year buy-and-hold return, excluding transaction costs; c transaction costs, expressed as a fraction of the transaction amount. Table 3 compares the returns without considering commission and the returns allowing for it. For winners or losers portfolios, commission reduces the returns and related t-stats significantly. For example, there are nine significant returns without considering commission for the losers portfolios, while there are only 5, 2, zero, zero, zero significant returns when commission rate are 0.1%, 0.3%, 0.5%, 0.7%, 0.9% respectively. However, when it comes to the winners minus loser portfolios, commission only makes returns diminish marginally and all related t-stats keep unchanged. In the case of strategy (36, 36), the corresponding returns declined slightly, from -0.0452 to -0.0444. Other strategies are in the similar conditions. It seems that commission does not have serious impact on returns for contrarian strategies over the horizons of 18-36. 67
Table 3 Returns of some strategies allowing for commission ortfolios Strategies Commission rate (%) 0 0.1 0.3 0.5 0.7 0.9 1 (3,24) 0.0206 0.0186 0.0145 0.0105 0.0065 0.0024 (1.8639) * (1.6832) * (1.3207) (0.9567) (0.5913) (0.2244) (3,30) 0.0224 0.0203 0.0163 0.0122 0.0082 0.0041 (1.8566) * (1.6904) * (1.3571) (1.0224) (0.6863) (0.3489) (6,24) 0.0200 0.0180 0.0139 0.0099 0.0058 0.0018 (1.7454) * (1.5709) (1.2209) (0.8695) (0.5166) (0.1623) (18,30) 0.0280 0.0259 0.0219 0.0178 0.0137 0.0097 (1.7743) * (1.6474) (1.3930) (1.1375) (0.8810) (0.6234) (24,30) 0.0308 0.0287 0.0246 0.0205 0.0164 0.0124 (1.6844) * (1.5748) (1.3550) (1.1342) (0.9126) (0.6900) (24,36) 0.0414 0.0394 0.0352 0.0311 0.0270 0.0229 (1.7671) * (1.6817) (1.5104) (1.3385) (1.1658) (0.9925) (36,24) 0.0358 0.0338 0.0296 0.0255 0.0214 0.0174 (1.8408) * (1.7380) * (1.5317) (1.3247) (1.1168) (0.9080) (36,30) 0.0460 0.0439 0.0398 0.0356 0.0315 0.0274 (1.9412) * (1.8568) * (1.6874) * (1.5172) (1.3465) (1.1750) (30,30) 0.0433 0.0412 0.0371 0.0329 0.0288 0.0247 (2.0294) ** (1.9356) * (1.7475) * (1.5586) (1.3689) (1.1785) 10 (3,12) 0.0170 0.0150 0.0109 0.0069 0.0029-0.0011 (1.8539) * (1.6360) (1.1988) (0.7599) (0.3193) (-0.1232) (3,24) 0.0204 0.0184 0.0143 0.0103 0.0062 0.0022 (1.9094) * (1.7222) * (1.3467) (0.9697) (0.5912) (0.2111) (3,30) 0.0199 0.0179 0.0138 0.0098 0.0058 0.0017 (1.7496) * (1.5739) (1.2213) (0.8673) (0.5119) (0.1551) (6,9) 0.0164 0.0143 0.0103 0.0063 0.0022-0.0018 (1.7907) * (1.5719) (1.1328) (0.6919) (0.2493) (-0.1951) (9,6) 0.0157 0.0137 0.0096 0.0056 0.0016-0.0024 (1.6629) * (1.4507) (1.0250) (0.5976) (0.1685) (-0.2623) 10-1 (18,30) -0.0143-0.0142-0.0142-0.0141-0.0141-0.0140 (-1.7554) * (-1.7554) * (-1.7554) * (-1.7554) * (-1.7554) * (-1.7554) * (18,36) -0.0198-0.0198-0.0197-0.0196-0.0195-0.0194 (-1.9450) * (-1.9450) * (-1.9450) * (-1.9450) * (-1.9450) * (-1.9450) * (24,24) -0.0150-0.0150-0.0150-0.0149-0.0148-0.0148 (-1.8393) * (-1.8393) * (-1.8393) * (-1.8393) * (-1.8393) * (-1.8393) * (30,18) -0.0149-0.0149-0.0148-0.0148-0.0147-0.0147 (-1.7061) * (-1.7061) * (-1.7061) * (-1.7061) * (-1.7061) * (-1.7061) * (30,36) -0.0347-0.0346-0.0345-0.0343-0.0342-0.0341 (-1.9504) * (-1.9504) * (-1.9504) * (-1.9504) * (-1.9504) * (-1.9504) * (36,18) -0.0196-0.0196-0.0195-0.0194-0.0193-0.0193 (-1.7788) * (-1.7788) * (-1.7788) * (-1.7788) * (-1.7788) * (-1.7788) * (36,36) -0.0452-0.0451-0.0450-0.0448-0.0446-0.0444 (-1.8186) * (-1.8186) * (-1.8186) * (-1.8186) * (-1.8186) * (-1.8186) * (24,30) -0.0211-0.0211-0.0210-0.0209-0.0208-0.0207 (-2.1360) ** (-2.1360) ** (-2.1360) ** (-2.1360) ** (-2.1360) ** (-2.1360) ** (24,36) -0.0288-0.0288-0.0286-0.0285-0.0284-0.0283 (-2.2190) ** (-2.2190) ** (-2.2190) ** (-2.2190) ** (-2.2190) ** (-2.2190) ** (30,24) -0.0216-0.0216-0.0215-0.0214-0.0213-0.0212 (-2.0806) ** (-2.0806) ** (-2.0806) ** (-2.0806) ** (-2.0806) ** (-2.0806) ** (30,30) -0.0313-0.0312-0.0311-0.0310-0.0309-0.0308 (-2.3433) ** (-2.3433) ** (-2.3433) ** (-2.3433) ** (-2.3433) ** (-2.3433) ** (36,24) -0.0293-0.0293-0.0291-0.0290-0.0289-0.0288 (-2.2439) ** (-2.2439) ** (-2.2439) ** (-2.2439) ** (-2.2439) ** (-2.2439) ** (36,30) -0.0342-0.0342-0.0340-0.0339-0.0338-0.0336 (-2.0513) ** (-2.0513) ** (-2.0513) ** (-2.0513) ** (-2.0513) ** (-2.0513) ** Notes: (1) This table picks out portfolios with significant return in Table 1 and compares them with returns under the same 68
strategies but allowing for different levels of commission. That is, the commission rate varies from zero to 0.9%. T-stat given in the parentheses is the average return divided by its standard error. (2) * Denotes significance at the 10% level or higher. (3) ** Denotes significance at the 5% level or higher. 5. Conclusion We follow the strategy construction method in Jegadeesh & Titman (1993) and Rouwenhorst (1998) to examine medium-term and long-term momentum and contrarian effects in China s stock market during 1994-2004. We find that no momentum strategy can make significant profit over the horizon of 3-36 months, meanwhile, contrarian strategy work effectively over the horizon of 18-36 months. 1-month gap for avoiding bid-ask bounce and lead-lag effect make no considerable difference for our empirical results. At last, transaction costs have no significant impact on contrarian strategies profit. References: [1] Carhart, Mark M.. On ersistence in Mutual Fund erformance. Journal of Finance, 1997(52): 57-82. [2] WANG Chang-yun. Relative Strength Strategies in China s Stock Market: 1994-2000. acific-basin Finance Journal 2004(12): 159-177. [3] De Bondt, Werner F. M., Thaler, Richard H.. Does the Stock Market Overreact. Journal of Finance, 1985(40): 793-805. [4] Fama, Eugene F., Kenneth R. French. Multifactor Explanations of Asset ricing Anomalies. Journal of Finance, 1996(51): 55-84. [5] Jegadeesh, Narasimhan, Sheridan Titman. Returns to Buying Winners and Selling Losers: Implications for Stock Market Efficiency. Journal of Finance, 1993(48): 65-91. [6] ZHOU Lin-jie. Research on the rofitability of Momentum Strategy Applied in China s Stock Market. World Economy, 2002(8): 60-64. [7] ZHANG Ren-ji, ZHU ing-fang, WANG Huai-fang. Empirical Test on Shanghai Stock Market Overreaction. Economic Research, 1998(5), 58-64. [8] Rouwenhorst, K. Geert. International Momentum Strategies. Journal of Finance, 1998(53): 267-284. [9] Rouwenhorst, K. Geert. Local Return Factors and Turnover in Emerging Markets. Journal of Finance, 1999(54): 1439-1464. [10] WU Shi-nong, WU Chao-peng. Empirical Research on rice Momentum Strategy and Earnings Momentum Strategy Applied in China s Stock Market. Economic Science, 2003(4), 54-61. [11] SHEN Yi-feng, WU Shi-nong. Does Our Stock Market Overreact? Economic Research, 1999(2): 21-26. [12] WANG Yong-hong, ZHAO Xue-jun. Emperical Analysis on Momentum Strategy and Contrarian Strategy Applied in China s Stock Market. Economic Research, 2001(6): 56-62. (Edited by Ivy Xie and Shirley Hu) 69