Momentum and seasonality in Chinese stock markets

Tags: stock returns, China, stock market, strategies, Jegadeesh, trading strategies, profitability, seasonal pattern, holding periods, contrarian strategies, effects, stock, Fama and French, momentum strategies, weighted portfolio, contrarian, portfolio, portfolios, Shenzhen Stock Exchange, Journal of Finance, beta coefficient, N. Jegadeesh, historical returns, momentum strategy, University of Queensland, seasonal patterns, Griffith University, annual returns, Queensland University of Technology, trading strategy, intermediate horizon, formation, p1 p2 p3
Content: Journal of Money, Investment and Banking ISSN 1450-288X Issue 17 (2010) © EuroJournals Publishing, Inc. 2010 http://www.eurojournals.com/JMIB.htm Momentum and Seasonality in Chinese stock markets Bin Li Griffith Business School, Griffith University, Brisbane, QLD 4111, Australia E-mail: [email protected] Tel: +61-7- 3735 7117 Judy Qiu UQ Business School, the University of Queensland, Australia Yanhui Wu School of Economics and Finance, Queensland University of Technology, Australia Abstract In this paper, we follow Jegadeesh and Titman's (1993, Journal of Finance) approach to examine 25 momentum/contrarian trading strategies using monthly stock returns in China for the period from 1994 to 2007. Our results suggest that there is no momentum profitability in any of the 25 strategies. In contrast, there is some evidence of reversal effects where the past winners become losers and past losers become winners afterward. The contrarian profit is statistically significant for the strategies using short formation and holding periods, especially for the formation periods of 1 to 3 months and the holding periods of 1 to 3 months. The contrarian strategies can generate about 12% per annum on average. Moreover, we follow Heston and Sadka (2008, Journal of Financial Economics) to investigate where there is any seasonal pattern in the cross-sectional variation of average stock returns in our momentum/contrarian strategies. There is no evidence of any seasonal pattern, and the results are robust to different formation and holding periods. Keywords: Momentum, Market Efficiency, Seasonality, emerging market JEL Classification: G14, G15 1. Introduction The finance literature continues to debate whether the market is efficient. There are three forms of market efficiency: strong-form efficient, semi-strong form efficient and weak-form efficient. In testing weak-form efficiency, researchers examine whether the market fully reflects information contained in the past. Up to date, there is no overwhelming consensus on this issue. There are many anomalies identified in historical stock returns such as the momentum effect, which has caught much attention in the finance research. DeBondt and Thaler (1985, 1987) find that stock prices overreact to the past information, which suggests that contrarian strategies (which is buying past losers and selling past winners) can achieve abnormal returns. They find that stocks performed poorly over the previous 3 to 5 years are more likely to perform well over the next 3 to 5 years. Jegadeesh (1990) and Lehmann (1990) also document stock return reversals in short formation and holding periods. Later studies such as Lakonishok, Shleifer and Vishny (1994, 1997) and Schiereck, De Bondt, and Weber (1999) confirm
Journal of Money, Investment and Banking - Issue 17 (2010)
25
that the profitability of the contrarian strategy using different data. Fama and French (1996) argue that this phenomenon is due to the value premium because the previous poor performers are more likely to become value stocks over the next 3 to 5 years, thus achieving higher returns. Jegadeesh and Titman (1993, 2001) find return continuation over intermediate horizons (3 to 12 months). The stocks performed better from the previous 6 months are more likely to be one of the better performers over the next 6 months. It is this intermediate horizon pricing anomaly that is the most intriguing of the momentum anomalies. Chan, Jegadeesh and Lakonishok (1996) extend the analysis to include both price momentum effects and earnings momentum effects. Although it was initially argued that price momentum may be driven by earnings momentum. Chan, Jegadeesh and Lakonishok find that the earnings and price momentum effects are quite separate. Rouwenhorst (1998) finds momentum profits are significantly positive in 12 countries. Grundy and Martin (2001) provide further analysis on the risk-adjusted momentum effect. They use the Fama and French three-factor model and show that the momentum effect remains even after adjusting for risk. Kang, Liu and Ni (2002) find evidence of abnormal returns for short-horizon contrarian and intermediate-horizon momentum strategies in China using weekly data for the period of 1993 to 2000. They suggest that investors' overreaction to firm-specific information is the main cause of abnormal return for short-term contrarian but not intermediate-horizon momentum profits. Wang and Chin (2004) report that return momentum and mean reversal are driven by past trading volume over intermediate horizons in China's stock market. They suggest that market characteristics such as prohibition on short-sales and the dominance of individual investors in China's market may explain this phenomenon. Recent work by Shumway and Wu (2006) suggests that disposition effect may drive stock price momentum in China's market. The momentum effects are further deepened by the presence of seasonality. The common seasonal patterns that have been observed and addressed in the finance literature are monthly, daily and intra-day seasonality patterns. Monthly seasonality in stock returns was first documented by Rozeff and Kinney (1976). DeBondt and Thaler (1985) show that the momentum returns only occur in January which could potentially be explained as a piece of evidence of seasonal pattern, even not explicitly stated in their paper. As discussed by Jegadeesh and Titman (1993) and Grinblatt and Moskowitz (2000), the momentum return usually continues throughout February to November, increases in December, but changes to a strong reversal in January. Contrary to the existing finding of either monotonic reversal or continuation in stock returns, Heston and Sadka (2007) investigate seasonal patterns in the cross-section of expected returns on the U.S. stocks. They form a variety of seasonal strategies and find that the stocks returns exhibit an annual pattern and this pattern lasts up to 20 years. This effect is larger and longer than that of any previous studies. The purpose of this study is to examine the existence of momentum effects in the Chinese stock markets. We follow Jegadeesh and Titman's (1993) approach to explore 25 momentum/contrarian trading strategies using monthly stock returns in China for the period from 1994 to 2007. There is no evidence of the momentum profitability in any of the 25 strategies. By contrast, we find some reversal effects where the past winners become losers and past losers become winners afterward. The contrarian profit is statistically significant for the strategies using short formation and holding periods, especially for the formation periods of 1 to 3 months and the holding periods of 1 to 3 months. The contrarian strategies can generate about 12% per annum on average. However, there is no evidence of the strategies using longer formation and holding periods. Furhtermore, we follow Heston and Sadka (2008) to examine where there is any seasonal pattern in the cross-sectional variation of average stock returns in our momentum/contrarian strategies. The results suggest that there is not any seasonal pattern, and the results are robust to different formation and holding periods. This paper contributes to the existing literature in the following ways. Firstly, this paper fills the gap in the literature by investigating the existence of momentum in Chinese stock returns. Most of the previous research is done in the U.S. setting, and the study on the Chinese stock markets is relatively scarce. Given the scale and prospect of the Chinese markets, it is imperative to extend the thin literature on this issue. Secondary, despite the fact that the momentum issue has been a well-
26
Journal of Money, Investment and Banking - Issue 17 (2010)
documented feature of stock returns, the analysis on the seasonal pattern in momentum profit is quite limited. Our study provides more recent evidence using Chinese stock returns. The structure of the paper is as follows: we begin with a brief discussion of our data in Section 2. The methodology is presented in Section 3. We then test the existence of the momentum effects in Chinese stock returns followed by examining the seasonality in Section 4. Section 5 concludes the paper and provides some thoughts for future work.
2. Data We obtain monthly stock price data, which is adjusted for dividend distribution and capital adjustment for all the "A" shares listed in Shanghai Stock Exchange and Shenzhen Stock Exchange from the period of January 1994 to December 2007. We also obtain Shanghai and Shenzhen composite price indices and market value. All the data are from DataStream. We construct value-weighted market return index based on these two composite indices for the period after February 1, 1997, but using equal-weighted market index for the period before February 1, 1997 because the market value data are not available before that time on DataStream and the market value of both exchanges are very close before that period. In constructing risk-adjusted returns, we only include firms listed for at least four years into our portfolios, and firms must have 36 observations in the past 48 months to compute their betas.
3. Methodology 3.1. Momentum Trading Strategies If there are short-term autocorrelations in stock returns, we may find some profitable strategies by selecting stocks based on their past returns and forming portfolios and holding these portfolios for some period. In this paper, we first employ the trading strategies used by Jegadeesh and Titman (1993). We form portfolios based on the J-month lagged stock returns and hold them for the following K months where J and K take the value of 1, 3, 6, 9, and 12. There are 25 strategies in total. For the illustrating purpose, we provide an example of the 3-1 strategy (where J=3 and K=1). The procedures are as follows: (1) At the beginning of each month t, we compute the average monthly stock returns over the past 3 months (t-1 to t-3) for each "A" stock listed on Shanghai Stock Exchange and Shenzhen Stock Exchange. (2) We then rank these stocks into 10 deciles in ascending orders. The top decile (Decile 10) portfolio contains the top 10% of stocks on the basis of their average returns in the past 3 months. It is then called the "winner" portfolio. The bottom decile (Decile 1) is called the "loser" portfolio. (3) We form an equal-weighted portfolio in each decile for each month t, and calculate the average returns for each decile portfolio at t. We further calculate the time-series average of each decile portfolio's returns. The portfolios are rebalanced monthly. The overlapping holding period is used because it can increase the power of our test (Jegadeesh and Titman, 1993). We also consider the test without overlapping holding periods. The momentum or reversal pattern may be due to the fact that firms with similar risk selected into the same portfolio and the risk factor(s) determines the cross-sectional difference of these portfolios, thus explaining the pattern. In light of this, we consider the risk-adjusted stock returns. In the capital asset pricing model (CAPM) framework, the market risk premium is the only factor in explaining stock returns1. First, we calculate the beta for each firm at time t, 1 Fama and French (1993) shows that CAPM has no power to explain the cross section of average returns on assets
Journal of Money, Investment and Banking - Issue 17 (2010)
27
ri,t = ci,t + r i,t m,t + i,t ,
(1)
based on the past 48 monthly returns. A stock must have at least 36 observations in the past 48 monthly
returns to selected into the portfolios.Second, we subtract the product of beta with market return from
stock return to obtain the risk-adjusted returns. These returns then substitute the raw return in
generating 25 momentum strategies.
3.2. Seasonality in Stock Returns
3.2.1 Simple Cross-Sectional Stock Return Regression
Unlike most previous studies where the winner and loser stocks are sorted based on their historical
returns over a multi-month formation period, we follow Heston and Sadka's (2007) method, in which a
single month return is used in order to exploit the one-month effect.
The cross-sectional simple regression (Fama and MacBeth (1973), and Fama and French
(1992)) is as follow:
ri,t = k.t + r k,t i,t -k + i,t
(2)
where ri,t is the monthly return of stock i at time t and the variable ri,t-k is the lagged return at time t-k.
The beta coefficient measures the relation of time t return to its lagged return. The value of k ranges
from 1 to 12 and then increased by 12 (months) interval afterward until the value reaches 60. The time
period covered is from 1991 to 2006. For each value k, we run the regression for every month and then
compute the average beta coefficient.
3.2.2 Multiple Cross-Sectional Stock Return Regressions
To examine the incremental effect of stock returns, we follow Heston and Sadka's (2007) multiple
cross-sectional regression model in which they extend Jegadeesh's (1990) model by adding longer lags:
12
4
ri,t = k.t +
r + k,t i,t -k
r + 12k i,t -12k
i,t
(3)
k =1
k =2
3.2.3 Periodic Momentum In this section, we propose a new trading strategy that examines the seasonality of momentum strategies in the Chinese stock markets. Every month we form 10 decile portfolios (the number of stocks in each portfolio are equal) based on their past performance. We use three different methods to determine the range of the past stock returns. The first method uses all the past returns in the interval to compute the past average return, which is the same as the method used in Section 3.1. The second one uses only the past returns lagged
assorted by size and book-to-market premium. It would have been desirable to use multifactor pricing models such as the Fama and French three-factor model. However, the data to construct the other two factors is not available. Therefore we based our analysis only on the CAPM framework.
28
Journal of Money, Investment and Banking - Issue 17 (2010)
Table 1: Momentum Strategies Using Raw Returns
Panel A. Portfolio Formed Based On the Past Average 1 Month Return
K/Portfolios
p1
p2
p3
p4
1
0.0229
0.0219
0.0242
0.0191
(0.1059) (0.1004) (0.1036) (0.0958)
3
0.0237
0.0209
0.0225
0.0203
(0.0632) (0.0604) (0.0616) (0.0609)
6
0.0207
0.0205
0.0214
0.0208
(0.0485) (0.0481) (0.0484) (0.0480)
9
0.0195
0.0202
0.0208
0.0202
(0.0413) (0.0414) (0.0420) (0.0418)
12
0.0187
0.0185
0.0192
0.0185
(0.0373) (0.0367) (0.0376) (0.0367)
Panel B. Portfolio Formed Based On the Past Average 3 Month Return
K/Portfolios
p1
p2
p3
p4
1
0.0258
0.0217
0.0214
0.0189
(0.1014) (0.0985) (0.0993) (0.0974)
3
0.0221
0.0208
0.0211
0.0202
(0.0632) (0.0603) (0.0621) (0.0615)
6
0.0193
0.0200
0.0208
0.0198
(0.0476) (0.0465) (0.0490) (0.0469)
9
0.0187
0.0194
0.0201
0.0200
(0.0403) (0.0403) (0.0411) (0.0407)
12
0.0176
0.0183
0.0188
0.0187
(0.0363) (0.0368) (0.0372) (0.0363)
Panel C. Portfolio Formed Based On the Past Average 6 Month Return
K/Portfolios
p1
p2
p3
p4
1
0.0200
0.0203
0.0220
0.0204
(0.1046) (0.0999) (0.1001) (0.1003)
3
0.0188
0.0196
0.0213
0.0202
(0.0625) (0.0604) (0.0630) (0.0620)
6
0.0177
0.0194
0.0207
0.0212
(0.0466) (0.0466) (0.0478) (0.0484)
9
0.0173
0.0198
0.0204
0.0202
(0.0396) (0.0407) (0.0413) (0.0419)
12
0.0174
0.0192
0.0193
0.0188
(0.0365) (0.0373) (0.0367) (0.0369)
Panel D. Portfolio Formed Based On the Past Average 9 Month Return
K/Portfolios
p1
p2
p3
p4
1
0.0201
0.0201
0.0200
0.0224
(0.1040) (0.0997) (0.1005) (0.1036)
3
0.0185
0.0193
0.0207
0.0219
p5 0.0191 (0.1013) 0.0204 (0.0617) 0.0204 (0.0476) 0.0203 (0.0411) 0.0190 (0.0368) p5 0.0207 (0.1012) 0.0221 (0.0633) 0.0218 (0.0483) 0.0213 (0.0417) 0.0194 (0.0362) p5 0.0188 (0.1017) 0.0205 (0.0624) 0.0214 (0.0491) 0.0212 (0.0426) 0.0193 (0.0375) p5 0.0194 (0.1021) 0.0213
p6 0.0188 (0.0982) 0.0196 (0.0612) 0.0204 (0.0476) 0.0200 (0.0412) 0.0187 (0.0367) p6 0.0180 (0.1008) 0.0203 (0.0613) 0.0206 (0.0477) 0.0198 (0.0410) 0.0187 (0.0363) p6 0.0211 (0.1031) 0.0216 (0.0637) 0.0211 (0.0485) 0.0207 (0.0422) 0.0194 (0.0372) p6 0.0234 (0.1081) 0.0235
p7 0.0173 (0.0966) 0.0198 (0.0609) 0.0206 (0.0478) 0.0201 (0.0407) 0.0188 (0.0358) p7 0.0159 (0.0989) 0.0194 (0.0622) 0.0201 (0.0477) 0.0203 (0.0417) 0.0194 (0.0372) p7 0.0212 (0.1030) 0.0229 (0.0630) 0.0226 (0.0494) 0.0222 (0.0423) 0.0205 (0.0372) p7 0.0188 (0.1011) 0.0208
p8 0.0172 (0.0983) 0.0187 (0.0612) 0.0194 (0.0463) 0.0196 (0.0400) 0.0189 (0.0359) p8 0.0170 (0.1025) 0.0191 (0.0618) 0.0209 (0.0480) 0.0204 (0.0415) 0.0192 (0.0369) p8 0.0174 (0.1005) 0.0199 (0.0628) 0.0207 (0.0480) 0.0199 (0.0408) 0.0185 (0.0360) p8 0.0184 (0.1029) 0.0204
p9 0.0162 (0.0975) 0.0183 (0.0608) 0.0198 (0.0462) 0.0193 (0.0395) 0.0187 (0.0361) p9 0.0170 (0.1040) 0.0184 (0.0607) 0.0209 (0.0484) 0.0201 (0.0410) 0.0190 (0.0365) p9 0.0158 (0.0968) 0.0199 (0.0605) 0.0206 (0.0473) 0.0198 (0.0400) 0.0185 (0.0359) p9 0.0150 (0.0957) 0.0184
p10 0.0128 (0.0977) 0.0171 (0.0616) 0.0188 (0.0454) 0.0187 (0.0396) 0.0181 (0.0359) p10 0.0143 (0.0989) 0.0181 (0.0599) 0.0193 (0.0455) 0.0193 (0.0404) 0.0181 (0.0361) p10 0.0137 (0.0957) 0.0182 (0.0586) 0.0193 (0.0450) 0.0185 (0.0390) 0.0166 (0.0354) p10 0.0158 (0.0934) 0.0191
p10-p1 -0.0101 (0.0115) -0.0066 (0.0071) -0.0019 (0.0054) -0.0008 (0.0047) -0.0006 (0.0043) p10-p1 -0.0115 (0.0113) -0.0041 (0.0070) 0.0000 (0.0054) 0.0006 (0.0047) 0.0005 (0.0043) p10-p1 -0.0062 (0.0113) -0.0006 (0.0069) 0.0016 (0.0053) 0.0012 (0.0046) -0.0008 (0.0042) p10-p1 -0.0044 (0.0112) 0.0005
Journal of Money, Investment and Banking - Issue 17 (2010)
29
(0.0626) (0.0606) (0.0622) (0.0631) (0.0631) (0.0663) (0.0624) (0.0632) (0.0583) (0.0596) (0.0070)
6
0.0178
0.0196
0.0210
0.0219 0.0213 0.0224 0.0217
0.0207
0.0191 0.0192 0.0014
(0.0473) (0.0464) (0.0486) (0.0489) (0.0489) (0.0506) (0.0489) (0.0478) (0.0460) (0.0446) (0.0053)
9
0.0188
0.0202
0.0207
0.0212 0.0213 0.0213 0.0211
0.0198
0.0181 0.0177 -0.0011
(0.0415) (0.0409) (0.0417) (0.0422) (0.0419) (0.0427) (0.0418) (0.0400) (0.0400) (0.0387) (0.0047)
12
0.0186
0.0195
0.0195
0.0195 0.0195 0.0196 0.0194
0.0189
0.0171 0.0159 -0.0026
(0.0380) (0.0374) (0.0373) (0.0369) (0.0370) (0.0377) (0.0368) (0.0360) (0.0354) (0.0353) (0.0043)
Panel E. Portfolio Formed Based On the Past Average 12 Month Return
K/Portfolios
p1
p2
p3
p4
p5
p6
p7
p8
p9
p10
p10-p1
1
0.0192
0.0205
0.0193
0.0226 0.0221 0.0213 0.0194
0.0163
0.0142 0.0171 -0.0021
(0.1051) (0.1029) (0.1013) (0.1052) (0.1052) (0.1042) (0.0996) (0.1008) (0.0965) (0.0928) (0.0112)
3
0.0184
0.0205
0.0211
0.0223 0.0225 0.0218 0.0219
0.0185
0.0172 0.0187 0.0003
(0.0635) (0.0619) (0.0642) (0.0643) (0.0645) (0.0628) (0.0637) (0.0611) (0.0590) (0.0576) (0.0069)
6
0.0192
0.0206
0.0210
0.0221 0.0224 0.0222 0.0225
0.0187
0.0168 0.0182 -0.0009
(0.0496) (0.0478) (0.0487) (0.0495) (0.0488) (0.0479) (0.0499) (0.0472) (0.0456) (0.0436) (0.0054)
9
0.0204
0.0209
0.0206
0.0210 0.0215 0.0210 0.0221
0.0185
0.0163 0.0172 -0.0032
(0.0442) (0.0419) (0.0417) (0.0422) (0.0415) (0.0404) (0.0425) (0.0408) (0.0388) (0.0387) (0.0048)
12
0.0202
0.0202
0.0195
0.0197 0.0199 0.0193 0.0204
0.0172
0.0157 0.0150 -0.0052
(0.0406) (0.0388) (0.0369) (0.0370) (0.0368) (0.0353) (0.0377) (0.0360) (0.0345) (0.0351) (0.0045)
Notes: The momentum strategy portfolios are formed based on the past average 1, 3, 6, 9 and 12-month return and the portfolios are
then held for K months. The values of K for different strategies are indicated in the first column. The stocks are ranked in
ascending orders based on the past average 1, 3, 6, 9 and 12-month return and an equally weighted portfolio of stocks in the
lowest past return decile is the loser portfolio (p1) and an equally weighted portfolio of stocks in the highest return decile is the
winner portfolio (p10). (p10-p1) represents the strategy of buying loser and selling winner. The average monthly returns of these
portfolios are presented corresponding to each K and portfolio in this table, while the standard deviation of each return is
indicated below the return. Panels A, B, C, D and E contain the returns for different strategy portfolios formed based on the past
average 1, 3, 6, 9 and 12-month return, respectively.
30
Journal of Money, Investment and Banking - Issue 17 (2010)
Table 2: Momentum Strategies Using Risk Adjusted Returns
Panel A. Portfolio Formed Based On the Past Average 1 Month Return
K/Portfolios
p1
p2
p3
p4
1
0.0040 0.0084 0.0065 0.0036
(0.0545) (0.0532) (0.0512) (0.0474)
3
0.0047 0.0050 0.0049 0.0038
(0.0318) (0.0278) (0.0271) (0.0288)
6
0.0022 0.0032 0.0033 0.0029
(0.0181) (0.0161) (0.0164) (0.0169)
9
0.0014 0.0022 0.0027 0.0028
(0.0128) (0.0123) (0.0117) (0.0126)
12
0.0006 0.0012 0.0022 0.0017
(0.0115) (0.0103) (0.0108) (0.0106)
Panel B. Portfolio Formed Based On the Past Average 3 Month Return
K/Portfolios
p1
p2
p3
p4
1
0.0116 0.0091 0.0058 0.0028
(0.0594) (0.0515) (0.0525) (0.0535)
3
0.0061 0.0048 0.0036 0.0041
(0.0342) (0.0302) (0.0277) (0.0301)
6
0.0022 0.0033 0.0022 0.0028
(0.0173) (0.0162) (0.0168) (0.0183)
9
0.0010 0.0022 0.0024 0.0024
(0.0136) (0.0123) (0.0124) (0.0130)
12
0.0002 0.0021 0.0016 0.0018
(0.0119) (0.0110) (0.0111) (0.0109)
Panel C. Portfolio Formed Based On the Past Average 6 Month Return
K/Portfolios
p1
p2
p3
p4
1
0.0065 0.0077 0.0052 0.0044
(0.0619) (0.0523) (0.0547) (0.0501)
3
0.0043 0.0030 0.0028 0.0030
(0.0313) (0.0296) (0.0287) (0.0283)
6
0.0012 0.0013 0.0028 0.0025
(0.0175) (0.0165) (0.0169) (0.0164)
9
0.0009 0.0017 0.0030 0.0027
(0.0132) (0.0124) (0.0140) (0.0126)
12
0.0011 0.0014 0.0023 0.0022
(0.0114) (0.0113) (0.0124) (0.0113)
p5 0.0037 (0.0480) 0.0035 (0.0273) 0.0030 (0.0171) 0.0031 (0.0127) 0.0023 (0.0109) p5 0.0030 (0.0487) 0.0020 (0.0281) 0.0027 (0.0179) 0.0026 (0.0130) 0.0022 (0.0111) p5 0.0032 (0.0492) 0.0035 (0.0302) 0.0035 (0.0187) 0.0031 (0.0137) 0.0025 (0.0119)
p6 0.0026 (0.0454) 0.0023 (0.0267) 0.0027 (0.0169) 0.0030 (0.0124) 0.0025 (0.0104) p6 0.0037 (0.0472) 0.0033 (0.0274) 0.0031 (0.0177) 0.0032 (0.0124) 0.0025 (0.0109) p6 0.0022 (0.0446) 0.0024 (0.0273) 0.0021 (0.0183) 0.0026 (0.0131) 0.0017 (0.0106)
p7 0.0009 (0.0459) 0.0005 (0.0266) 0.0022 (0.0176) 0.0025 (0.0130) 0.0017 (0.0105) p7 -0.0015 (0.0459) -0.0006 (0.0258) 0.0020 (0.0169) 0.0023 (0.0119) 0.0019 (0.0104) p7 0.0000 (0.0450) 0.0002 (0.0267) 0.0019 (0.0182) 0.0019 (0.0123) 0.0013 (0.0100)
p8 -0.0022 (0.0427) -0.0021 (0.0256) 0.0002 (0.0171) 0.0015 (0.0119) 0.0013 (0.0092) p8 -0.0037 (0.0418) -0.0002 (0.0265) 0.0016 (0.0183) 0.0020 (0.0119) 0.0015 (0.0098) p8 -0.0020 (0.0427) 0.0001 (0.0266) 0.0016 (0.0180) 0.0017 (0.0119) 0.0009 (0.0096)
p9 -0.0022 (0.0418) -0.0017 (0.0248) 0.0003 (0.0178) 0.0012 (0.0118) 0.0013 (0.0096) p9 -0.0058 (0.0394) -0.0045 (0.0239) 0.0000 (0.0180) 0.0012 (0.0107) 0.0011 (0.0088) p9 -0.0046 (0.0361) -0.0019 (0.0249) -0.0001 (0.0164) 0.0006 (0.0104) 0.0001 (0.0081)
p10 -0.0074 (0.0458) -0.0045 (0.0268) -0.0013 (0.0179) 0.0005 (0.0111) 0.0007 (0.0098) p10 -0.0060 (0.0392) -0.0043 (0.0235) -0.0015 (0.0175) 0.0005 (0.0105) 0.0000 (0.0088) p10 -0.0062 (0.0424) -0.0028 (0.0261) 0.0005 (0.0166) 0.0014 (0.0087) -0.0001 (0.0082)
p10-p1 -0.0115 (0.0062) -0.0092 (0.0036) -0.0035 (0.0023) -0.0010 (0.0015) 0.0001 (0.0014) p10-p1 -0.0176 (0.0062) -0.0103 (0.0037) -0.0037 (0.0022) -0.0004 (0.0016) -0.0002 (0.0014) p10-p1 -0.0127 (0.0067) -0.0071 (0.0036) -0.0007 (0.0022) 0.0006 (0.0015) -0.0012 (0.0013)
Journal of Money, Investment and Banking - Issue 17 (2010)
31
Panel D. Portfolio Formed Based On the Past Average 9 Month Return
K/Portfolios
p1
p2
p3
p4
p5
p6
p7
p8
p9
p10
p10-p1
1
0.0087 0.0037 0.0033 0.0049 0.0035 0.0008 0.0009 -0.0001 -0.0047 -0.0043 -0.0130
(0.0628) (0.0549) (0.0525) (0.0520) (0.0486) (0.0486) (0.0422) (0.0435) (0.0385) (0.0399) (0.0067)
3
0.0035 0.0012 0.0022 0.0026 0.0032 0.0012 0.0015 0.0012 -0.0034 -0.0015 -0.0050
(0.0348) (0.0300) (0.0287) (0.0294) (0.0302) (0.0279) (0.0274) (0.0261) (0.0224) (0.0223) (0.0037)
6
0.0013 0.0015 0.0020 0.0030 0.0030 0.0021 0.0023 0.0010 -0.0002 0.0004 -0.0009
(0.0189) (0.0192) (0.0183) (0.0182) (0.0188) (0.0186) (0.0185) (0.0171) (0.0147) (0.0149) (0.0022)
9
0.0017 0.0023 0.0019 0.0029 0.0028 0.0024 0.0024 0.0015 -0.0005 0.0001 -0.0016
(0.0136) (0.0141) (0.0130) (0.0136) (0.0143) (0.0134) (0.0127) (0.0112) (0.0086) (0.0087) (0.0015)
12
0.0009 0.0014 0.0012 0.0022 0.0017 0.0014 0.0019 0.0012 -0.0004 -0.0010 -0.0018
(0.0116) (0.0120) (0.0113) (0.0116) (0.0119) (0.0111) (0.0108) (0.0092) (0.0076) (0.0071) (0.0013)
Panel E. Portfolio Formed Based On the Past Average 12 Month Return
K/Portfolios
p1
p2
p3
p4
p5
p6
p7
p8
p9
p10
p10-p1
1
0.0072 0.0048 0.0028 0.0040 0.0036 0.0014 0.0017 -0.0018 -0.0033 -0.0034 -0.0106
(0.0640) (0.0585) (0.0559) (0.0525) (0.0487) (0.0475) (0.0471) (0.0410) (0.0346) (0.0374) (0.0067)
3
0.0042 0.0028 0.0021 0.0033 0.0027 0.0016 0.0018 -0.0012 -0.0024 -0.0021 -0.0064
(0.0346) (0.0321) (0.0307) (0.0305) (0.0300) (0.0284) (0.0307) (0.0239) (0.0213) (0.0215) (0.0037)
6
0.0028 0.0027 0.0018 0.0021 0.0029 0.0026 0.0030 -0.0002 -0.0016 -0.0012 -0.0040
(0.0192) (0.0198) (0.0190) (0.0182) (0.0184) (0.0183) (0.0196) (0.0172) (0.0144) (0.0152) (0.0023)
9
0.0021 0.0027 0.0013 0.0017 0.0024 0.0023 0.0032 0.0007 -0.0007 -0.0009 -0.0030
(0.0138) (0.0139) (0.0137) (0.0132) (0.0134) (0.0129) (0.0139) (0.0115) (0.0089) (0.0091) (0.0016)
12
0.0013 0.0015 0.0004 0.0009 0.0015 0.0013 0.0019 0.0005 -0.0012 -0.0017 -0.0029
(0.0120) (0.0122) (0.0112) (0.0113) (0.0109) (0.0109) (0.0111) (0.0098) (0.0070) (0.0079) (0.0014) Notes: The momentum strategy portfolios are formed based on the past average 1, 3, 6, 9 and 12-month risk adjusted return and the
portfolios are held for K months. The values of K for different strategies are indicated in the first column. The returns are
adjusted for market risk. The stocks are ranked in ascending orders based on the past average 1, 3, 6, 9 and 12-month risk
adjusted return and an equally weighted portfolio of stocks in the lowest past return decile is the loser portfolio (p1) and an
equally weighted portfolio of stocks in the highest return decile is the winner portfolio (p10). (p10-p1) represents the strategy of
buying loser and selling winner. The average monthly returns of these portfolios are presented corresponding to each K and
portfolio in this table, while the standard deviation of each return is indicated below the return. Panels A, B, C, D and E contain
the returns for different strategy portfolios formed based on the past average 1, 3, 6, 9 and 12-month risk adjusted return,
respectively.
Journal of Money, Investment and Banking - Issue 17 (2010)
32
yearly such as lagged 12 years and lagged 24 years to determine the past average return. For comparison, the third one uses the past returns except those lagged yearly. We examine two different formation intervals. The first interval uses the Jegadeesh and Titman's (1993) horizon of 1 year. The second one uses longer interval, which is from 2 years to 4 years. DeBondt and Thaler (1985) use horizons of 2 to 5 years and Heston and Sadka (2007) also use much longer intervals such as 6-10 and 11-15 years. However, due to the short history of Chinese stock exchanges, we cannot explore such long intervals.
4. Empirical results 4.1 Momentum Strategies Using Raw Returns and Risk-adjusted Returns Table 1 reports the results of monthly raw returns for equally-weighted winner-loser portfolios as well as the difference between them. The portfolios are constructed on the basis of 5 formation periods (1month, 3-month, 6-month, 9-month and 12-month). The winner (loser) portfolio is the one with the highest return which is marked as p10 (p1) in the table. Each portfolio return is then computed for a certain holding period. For each formation period, we consider 5 different holding periods (1, 3 6, 9 and 12 months). In total there are 25 strategies considered and reported in each panel of Table 1. For a better demonstration, we label them as F-H strategy where F stands for formation period and H stands for Holding period. For example, the 1-3 strategy refers to the one where portfolios are formed based on the past one month raw return and held for the next 3 months. As shown in Table 1, all return differences between winners and losers for all strategies are not statistically significant. The result seems not to lend itself to explain the contrarian profit found in the previous literature. Nevertheless there are some interesting findings. For 1-H strategies (Panel A), all differences are negative indicating that past winners become losers and losers become winners. Some strategies generate economically significant profit. For example, the 1-1 strategy generates 1% return per month or average 12% per annum. In panel B, it shows a reversal effect when the portfolios are held for a short time period, and then such effect disappears if they are held for 6 months. When the holding period increases, portfolio returns tend to reverse back where winners are still winners and losers stay as losers. However, the longer the formation horizon, the shorter is the intermediate-term momentum effect. Table 2 details the momentum profit using risk adjusted returns. Our result is consistent with the finding of Kang, Liu and Ni (2002) in which they report statistically significant abnormal profits for some short-horizon contrarian strategies. The contrarian profit is statistically significant at the 5% level for the 1-1 and 1-3 strategies in Panel A. In Panel B, the contrarian profit is statistically significant for the 3-1 and 3-3 strategies. For 6-H strategies, the 6-3 strategy generates significant difference in the winner-loser portfolio at the 5% significant level and the 6-1 strategy generates the contrarian profit which is significant at the 10% level. For 9-H strategies in Panel D, only the 9-1 strategy shows the evidence of the contrarian profit at the 10% significant level. There is no sign of significant contrarian profits in Panel E even though the differences are all negative. 4.2. Simple and Multiple Cross-sectional Regressions of Returns Using the simple regression of Equation (2), the first two lags are all negative and significantly different from zero. The lag of 6 months also exhibits statistically significance with a positive estimate of 0.0329. All the remaining lags are insignificant and display no apparent regular pattern. In measuring the incremental effect of stock returns, we use multiple regression of Equation (3) and find a close pattern to what are shown in the simple regression. The t-statistics of the first two lags are significant and the possible long-term
33
Journal of Money, Investment and Banking - Issue 17 (2010)
Table 3: Simple and Multiple Cross-Sectional Regressions of Returns
Panel A. Simple Regression
Panel B. Multiple Regressions
Lag

t-statistics

t-statistics

t-statistics
1
-0.0261
(-1.78)
-0.0409
(-2.54)
-0.0855
(-3.05)
2
-0.0370
(-2.79)
-0.0517
(-2.81)
0.0028
(0.09)
3
0.0083
(0.71)
0.0144
(0.81)
0.0120
(0.66)
4
-0.0037
(-0.27)
-0.0208
(-0.75)
-0.0295
(-1.61)
5
-0.0054
(-0.47)
-0.0096
(-0.56)
0.0359
(1.75)
6
0.0392
(3.78)
0.0333
(1.81)
0.0366
(1.94)
7
0.0133
(1.07)
-0.0025
(-0.19)
-0.0039
(-0.16)
8
-0.0206
(-1.54)
-0.0372
(-1.75)
0.0384
(2.01)
9
0.0147
(1.49)
0.0084
(1.00)
0.0109
(0.47)
10
-0.0041
(-0.38)
-0.0030
(-0.15)
-0.0062
(-0.32)
11
0.0001
(0.02)
0.0079
(0.65)
0.0042
(0.19)
12
0.0141
(1.25)
0.0186
(1.68)
0.0119
(0.68)
24
-0.0101
(-0.70)
-0.0022
(-0.16)
-0.0262
(-1.19)
36
-0.0106
(-0.70)
0.0071
(0.75)
-0.0220
(-1.06)
48
-0.0009
(-0.07)
0.0128
(0.59)
60
0.0125
(0.76)
0.0213
(1.43)
72
-0.0123
(-0.78)
-0.0203
(-1.24)
84
0.0074
(0.34)
0.0004
(0.03)
96
0.0048
(0.24)
-0.0031
(-0.21)
108
-0.0230
(-1.23)
-0.0131
(-0.94)
120
0.0051
(0.47)
0.0095
(0.77)
Notes: Monthly simple cross-sectional regressions of the form ri,t = k.t + r k,t i,t -k + i,t are computed for each month t
and lag k, where ri,t is the monthly return of stock i at time t and the variable ri,t -k is the lagged return at time t-k.
The beta coefficient measures the relation of time t return to its lagged return. The value of k ranges from 1 to 12 and then increased by 12 (months) interval afterward until the value reaches 60. The time period covered is from 1991 to 2006. For each value k, we run the regression for every month and then compute the average beta coefficient. The time-series average k,t is presented in Panel A. Panel B computes multiple cross-sectional
regressions, including all past lags in the same regression. Two regression specifications are considered: including lags 1 through 12, 24, and 36; then adding each 12th lag through 120.
Journal of Money, Investment and Banking - Issue 17 (2010)
34
Table 4: Seasonality of Momentum Strategies Based on Past Performance
Panel A: Use Raw Returns
Strategy
p1
p2
p3
p4
p5
p6
p7
p8
p9
p10
p10-p1
Year 1
ALL
0.0190
0.0203
0.0191
0.0223
0.0219
0.0213
0.0193
0.0167
0.0143
0.0170 -0.0020
(0.1048) (0.1026) (0.1010) (0.1050) (0.1048) (0.1038) (0.0993) (0.1006) (0.0962) (0.0925) (0.0112)
ANN
0.0194
0.0169
0.0197
0.0172
0.0167
0.0191
0.0191
0.0222
0.0218
0.0194
0.0000
(0.1156) (0.1026) (0.1019) (0.0988) (0.0947) (0.0977) (0.0969) (0.0978) (0.0967) (0.0939) (0.0119)
NON
0.0200
0.0202
0.0193
0.0219
0.0227
0.0208
0.0170
0.0160
0.0169
0.0165 -0.0034
(0.1049) (0.1014) (0.1002) (0.1027) (0.1057) (0.1026) (0.1002) (0.0989) (0.1005) (0.0925) (0.0112)
Year 2-4
ALL
0.0322
0.0324
0.0302
0.0288
0.0194
0.0214
0.0209
0.0190
0.0145
0.0126 -0.0197
(0.1389) (0.1233) (0.1195) (0.1250) (0.0986) (0.1071) (0.1044) (0.0976) (0.0981) (0.1005) (0.0143)
ANN
0.0219
0.0275
0.0194
0.0230
0.0268
0.0242
0.0282
0.0198
0.0207
0.0163 -0.0056
(0.1071) (0.1070) (0.1097) (0.1038) (0.1087) (0.1154) (0.1289) (0.1011) (0.1064) (0.1070) (0.0126)
NON
0.0315
0.0348
0.0284
0.0239
0.0280
0.0204
0.0147
0.0172
0.0196
0.0126 -0.0189
(0.1242) (0.1437) (0.1150) (0.1089) (0.1204) (0.1014) (0.1048) (0.0969) (0.1048) (0.0927) (0.0129)
Panel B: Use Risk-Adjusted Returns
Strategy
p1
p2
p3
p4
p5
p6
p7
p8
p9
p10
p10-p1
Year 1
ALL
0.0072
0.0048
0.0028
0.0040
0.0036
0.0014
0.0017 -0.0018 -0.0033 -0.0034 -0.0106
(0.0640) (0.0585) (0.0559) (0.0525) (0.0487) (0.0475) (0.0471) (0.0410) (0.0346) (0.0374) (0.0067)
ANN
-0.0027 -0.0007
0.0023
0.0000
0.0012
0.0031
0.0011
0.0058
0.0042
0.0024
0.0051
(0.0493) (0.0491) (0.0473) (0.0471) (0.0463) (0.0426) (0.0474) (0.0476) (0.0432) (0.0410) (0.0058)
NON
0.0058
0.0048
0.0051
0.0043
0.0049
0.0009 -0.0006 -0.0006 -0.0037 -0.0042 -0.0100
(0.0645) (0.0582) (0.0561) (0.0530) (0.0490) (0.0469) (0.0440) (0.0411) (0.0376) (0.0371) (0.0068)
Year 2-4
ALL
0.0057
0.0000 -0.0016 -0.0031 -0.0020 -0.0031 -0.0032 -0.0033 -0.0041 -0.0046 -0.0103
(0.0658) (0.0642) (0.0614) (0.0607) (0.0552) (0.0539) (0.0535) (0.0481) (0.0434) (0.0359) (0.0081)
ANN
-0.0033 -0.0042 -0.0009 0.0000 -0.0039 -0.0016
0.0002
0.0014 -0.0042 -0.0030 0.0002
(0.0597) (0.0540) (0.0504) (0.0542) (0.0522) (0.0523) (0.0525) (0.0520) (0.0508) (0.0530) (0.0087)
NON
0.0059
0.0027 -0.0024 -0.0025 -0.0019 -0.0047 -0.0034 -0.0053 -0.0035 -0.0041 -0.0100
(0.0644) (0.0645) (0.0616) (0.0569) (0.0580) (0.0545) (0.0498) (0.0510) (0.0419) (0.0372) (0.0081)
Notes: Every month stocks are grouped into ten portfolios (with equal number of stocks in each portfolio) according to various categories based on past
performance. For example, the trading strategy that is formed based on past annual returns during years 2-4 ranks stocks according to their average returns
during the historical lags 12, 24, 36 and 48. The stocks in each portfolio are assigned equal weight, and the portfolios are rebalanced monthly. The average
monthly return difference between the highest past-performing decile and the lowest past-performing decile is then calculated for the period 1994 through
2007. The table below contains the results of this procedure, performed separately for 10 different groups. Panel A reports the results for raw returns while
Panel B presents the results with risk adjusted returns.
Journal of Money, Investment and Banking - Issue 17 (2010)
35
momentum effect is presented for 6 months lag. When we extend the lag length up to 120 months, the multiple regression still yields a significant t-statistic at the 5% level for the 1 month and 6 months lags. In addition, the 8 month lag shows a momentum effect, suggesting the evidence produced from the multiple regression can be an indication of a short term return reversal and an intermediate-term momentum.
4.3. Momentum Strategies Exploring Seasonality Pattern To explore the seasonality effect in momentum, we form the winner-loser portfolios based on their returns over the previous year and then hold them for 1 month. The seasonality is examined at annual and nonannual intervals. We also consider a longer interval of 2-4 years as evidenced in DeBondt and Thaler (1985, 1987) in which they find long-term price reversal. For robustness check, we further run the tests for a shorter formation interval of 1, 3, 6 and 9 months combined with holding periods of 3, 6, 9, and 12 months. The results remain similar. For the simplicity, we only report the one with 12 months formation horizon and 1 month holding period. The other test results are available upon request. The average return to a momentum strategy reported in Table 4 indicate that there is no clear pattern that whether the decile portfolio returns are monotonically increasing or decreasing. For both the year 1 and years 2-4 strategies, the winner-loser profits are negative. Although the effects are not statistically significant, the contrarian profit of 1.97% per month from the long-term strategy is economically significant. All the annual lags for the year 1 and years 2-4 strategies show no signs of seasonal effect. The similar results are found when we use risk-adjusted returns.
5. Conclusion In this paper, we follow Jegadeesh and Titman's (1993) approach to explore the 25 momentum/contrarian trading strategies using monthly stock returns in China for the period from 1994 to 2007. We do not find the momentum profitability in any of 25 strategies. By contrast, we find some reversal effects where the past winners become losers and past losers become winners afterward. The contrarian profit is statistically significant for the strategies using short formation and holding periods, especially for the formation period of 1 to 3 months and the holding periods of 1 to 3 months. The contrarian strategies can generate about 12% per annum on average. However, there is no evidence of the strategies using longer formation and holding periods. Moreover, we follow Heston and Sadka (2008) to examine where there is any seasonal pattern in the cross-sectional variation of average stock returns in our momentum/contrarian strategies. Our results suggest that there is no seasonal pattern. The results are robust to different formation and holding periods. A full understanding of the source of the risk-adjusted momentum profit is still an Open question. Numbers of analyses have been done attempting to explain the momentum effects. Conrad and Kaul (1998) argue that the cross-sectional dispersion in mean returns of individual stocks can be an important determining factor in generating momentum profits. Chan, Jegadeesh and Lakonishok (1996) suggest that firm-specific events also contribute to momentum as investors may either underreact to information or there is delayed over-reaction to information caused by a positive feedback trading. Kang, Liu and Ni (2002) suggest that investors' overreaction to firm-specific information is the main cause of abnormal return for short-term contrarian but not intermediate-term momentum profits. Shumway and Wu (2006) suggest that the disposition effect drives stock price momentum in China's market. Given these possible and diverse explanations, the cause of the reversal in Chinese stock returns is still an open question which remains to one of the future works.
Acknowledgement We are grateful to comments made by participants at the 15th global finance Conference in Hangzhou, China in May 2008.
36
Journal of Money, Investment and Banking - Issue 17 (2010)
References [1] Bouman, S., and B. Jacobsen, 2002. "The Halloween indicator, 'Sell in May and Go Away': Another puzzle", American Economic Review 92, pp.1618-1635. [2] Chan, K.C., N. Jegadeesh, and J. Lakonishok, 1996. "Momentum strategies", Journal of Finance 51, pp.1682-1713. [3] Conrad, J., and G Kaul, 1998. "An Anatomy of Trading Strategies", Review of Financial Studies 11, pp.489-519. [4] DeBondt, W.F.M., and R.H Thaler, 1985. "Does the market overreact?", Journal of Finance 40, pp.793-805. [5] DeBondt, W.F.M., and R.H Thaler, 1987. "Further evidence on investor overreaction and stock market seasonality", Journal of Finance 42, pp.557-581. [6] Fama, E., and K. French, 1992. "The cross-section of expected stocks returns", Journal of Finance 47, 427-465. [7] Fama, E., and K. French, 1996. "Multifactor explanations of asset pricing anomalies", Journal of Finance 51, pp.55-84. [8] Fama, E., and J. MacBeth, 1973. "Risk, return and equilibrium: Empirical tests", Journal of Political Economy 81, pp.607-636. [9] Grundy, B.D., and J.S. Martin, 2001. "Understanding the nature of the risks and the source of the rewards to momentum investing", Review of Financial Studies 14, pp.29-78. [10] Heston, Steven L., and Ronnie Sadka, 2008. "Seasonality in the cross-section of stock returns", Journal of Financial Economics 87, pp.418-445. [11] Jegadeesh, N., 1990. "Evidence of predictable behavior of security returns", Journal of Finance 45, pp.881-898. [12] Jegadeesh, N., and S. Titman, 1993. "Returns to buying winners and selling losers: Implications for stock market efficiency", Journal of Finance 48, pp.65-91. [13] Jegadeesh, N., and S. Titman, 2001, Profitability of Momentum Strategies: An Evaluation of alternative explanations, Journal of Finance 56, pp.699-720. [14] Kamstra, M.J., L.A. Kramer, and M.D. Levi, 2003. "Winter blues: Seasonal Affective Disorder (SAD) stock market returns", American Economic Review 93, pp.324-343. [15] Kang, J., M. Liu, and S.X. Ni, 2002. "Contrarian and momentum strategies in the China stock market:1993-2000", Pacific-Basin Finance Journal 10, pp.243-265. [16] Lakonishok, J., A. Shleifer, and R.W. Vishny, 1994. "Contrarian investment, extrapolation, and risk", Journal of Finance 49, pp.1541-1578. [17] Lehmann, B., 1990. "Fads, martingales and market efficiency", Quarterly Journal of Economics 105, pp.1-28. [18] Rouwenhorst, K. Geert, 1998. "International momentum strategies", Journal of Finance 53, pp.267-284. [19] Rozeff, M., and W Kinney, 1976. "capital market seasonality: The case of stock returns", Journal of Financial Economics 2, pp.379-402. [20] Shumway, T., and G. Wu, 2006. "Does disposition drive momentum?" University of Michigan, Working Paper. [21] Wang, C., and S. Chin, 2004. "Profitability of return and volume-based investment strategies in China's stock market", Pacific-Basin Finance Journal 12, pp.541-564.

File: momentum-and-seasonality-in-chinese-stock-markets.pdf
Published: Wed Nov 3 10:59:48 2010
Pages: 13
File size: 0.13 Mb


New Engineer, 28 pages, 1.23 Mb

, pages, 0 Mb
Copyright © 2018 doc.uments.com