- Email: [email protected]
Echo effects and the returns from 52-week high strategies
ARTICLE IN PRESS[m1G;November 10, 2015;10:13]
JID: FRL
Finance Research Letters 000 (2015) 1–9
Contents lists available at ScienceDirect
Finance Research Letters journal homepage: www.elsevier.com/locate/frl
Echo effects and the returns from 52-week high strategies An-Sing Chen∗, Wayne Yang National Chung Cheng University, Department of Finance, 168 University Road, Min-Hsiung, Chia-Yi 621, Taiwan, ROC
a r t i c l e
i n f o
Article history: Received 1 August 2015 Accepted 16 October 2015 Available online xxx JEL classification: G11 G12 G14 G17 Keywords: 52-Week high Momentum Skip-period Trading strategies Investment strategies Echo effect
a b s t r a c t Echo effects have been shown by the existing literature to influence the performance of conventional return-based momentum portfolios. This effect has yet to be confirmed for 52-week high momentum strategies. Our results show that the 52-week high strategy also manifests an echo effect. Increasing the skip period between the date of portfolio formation and the date of portfolio purchase 3–6 months significantly improves performance in nearly all cases analyzed. The results are robust to both in-sample and out-of-sample analyses. They are also robust to controlling for the effects on the risk of the portfolio from its return exposure to commonly used empirical return factors. © 2015 Elsevier Inc. All rights reserved.
1. Introduction For conventional Jegadeesh and Titman (1993) (JT henceforth) momentum portfolios, market states have been shown to be able to influence their performance (Cooper et al., 2004; Asem and Tian, 2010). The effects of business cycles on JT momentum returns have also been examined (Chordia and Shivakumarm, 2002; Griffin et al., 2003; Arshanapalli et al., 2006). Novy-Marx (2012) finds the presence of an echo effect in the conventional JT momentum portfolios; namely, increasing the length of the skip-period between ∗
Corresponding author. Tel.: +886 5 2720411x34201; fax: +886 5 2720818. E-mail address: fi[email protected] (A.-S. Chen).
http://dx.doi.org/10.1016/j.frl.2015.10.015 1544-6123/© 2015 Elsevier Inc. All rights reserved.
Please cite this article as: A.-S. Chen, W. Yang, Echo effects and the returns from 52-week high strategies, Finance Research Letters (2015), http://dx.doi.org/10.1016/j.frl.2015.10.015
JID: FRL
ARTICLE IN PRESS[m1G;November 10, 2015;10:13]
2
A.-S. Chen, W. Yang / Finance Research Letters 000 (2015) 1–9
the time of the momentum portfolio formation computations and the time of the actual purchase of the portfolio to 6 months significantly increases the profitability of the conventional JT momentum strategies. The echo effect is an important factor of the performance momentum strategies. However, the previous literature does not address how it affects the 52-week high strategy. An important objective of this study is to investigate if the 52-week high momentum strategy, as another cross-sectional momentum strategy, is distinct from the JT momentum strategy. Will changing the skip period also influence anchoring based momentum strategies such as the George and Hwang (2004) (GH henceforth) 52-week high momentum strategy? Exhaustive survey of the literature shows that none of the existing studies on the 52-week high momentum strategy has addressed this issue. This study finds that the 52-week high strategy also manifests the Novy-Marx style echo effect; increasing the skip period 3–6 months dramatically improves performance in nearly all cases analyzed. Additional robustness checks show that the performance improvements from adjusting the skip period are also robust to controlling for commonly used return risk factors. For investors, the results of this study suggest that waiting up 3–6 months before purchasing (short selling) stocks identified by the 52-week high strategy can significantly improve performance in the holding period. 2. Data and method The data cover the sample period from July 1963 through December 2013 and include all stocks in the Center for Research in Securities Prices (CRSP) universe. The benchmark 52-week high momentum strategy is implemented in this study using the following modification of method described in GH to allow for skip-period. At the end of each month, we rank stocks based on the ratio of the current price to the past 52-week high. We then identify the top 30 percentile (winners) and the bottom 30 percentile (losers) stocks, wait (skip) s months before establishing a zero-investment (self-financing) portfolio consisting of long positions in the top 30 percentile and short positions in the bottom 30 percentile stocks identified s months ago. Specifically, stocks are ranked based on
Pi,t−s , highi,t−s
where Pi,t−s is the price of stock i at the
end of month t − s and highi,t−s is the highest price of stock i during the 12-month period that ends on the last day of month t − s. We then hold the winner and loser portfolios for h months. The momentum profits are calculated as the difference between the equally weighted average returns of the winner and loser portfolios established h months ago. When s is equal to 1, the implementation becomes the standard GH 52-week high momentum strategy discussed in the literature. For this study, we analyze the effects of varying the skip-period s, from 0, 1, 2, to 9 months. We also analyze holding periods h of 1, 3, and 6 months. For clarity of exposition, in the remainder of this paper, we will denote the various 52-week high momentum strategies analyzed in this study using the (12, s, h) notation. The “twelve” refers to the fact that 52-weeks are approximately equal to 12 months, s denotes the number of months in the skip period, and h denotes the number of months in the holding period. The standard one month skip period GH 52week high momentum strategy with a holding period of six months, for example, is denoted as (12, 1, 6) using this notation. In the analyses, the dataset is divided into two parts. The first two-thirds of the dataset from July 1963 to December 1996 is denoted as in-sample. The remaining one-third of the dataset from January 1997 through December 2013 is denoted as the reserved out-of-sample. This 2/3 and 1/3 split is widely used in the forecasting literature and allows us to check whether skip-period s that gives the best performance in the in-sample period will continue to give the best performance in the reserved out-of-sample period. Evidence that s does not change dramatically from in-sample to out-of-sample will give us additional confidence as to the robustness of our findings with respect to s and to its generalizability to future data that is not part of our dataset. 3. Results 3.1. Profits from momentum strategies Table 1 reports the average monthly returns of winner, loser, and self-financing portfolios for the insample period from July 1963 through December 1996 for various 52-week high momentum investing Please cite this article as: A.-S. Chen, W. Yang, Echo effects and the returns from 52-week high strategies, Finance Research Letters (2015), http://dx.doi.org/10.1016/j.frl.2015.10.015
ARTICLE IN PRESS[m1G;November 10, 2015;10:13]
JID: FRL
A.-S. Chen, W. Yang / Finance Research Letters 000 (2015) 1–9
3
Table 1 Returns for 52-week high momentum strategies (July 1963 through December 1996). h 1
3
6
s
Winner
Loser
Winner−loser
Winner
Loser
Winner−loser
Winner
Loser
Winner−loser
0
1.06∗∗∗ (4.80) 0.38∗ (1.68) 0.5∗∗ (2.23) 0.51∗∗ (2.27) 0.65∗∗∗ (2.86) 0.66∗∗∗ (2.91) 0.69∗∗∗ (2.99) 0.68∗∗∗ (3.00) 0.71∗∗∗ (3.11) 0.64∗∗∗ (2.82) 0.13∗∗∗ (3.45)
1.28∗∗ (3.30) 0.00 (0.01) −0.06 (−0.16) −0.01 (−0.04) 0.08 (0.23) 0.14 (0.37) 0.14 (0.40) 0.19 (0.53) 0.22 (0.61) 0.29 (0.83) −0.02 (−0.54)
−0.22 (−0.90) 0.38∗ (1.76) 0.56∗∗∗ (2.69) 0.52∗∗ (2.56) 0.57∗∗ (2.58) 0.53∗∗ (2.45) 0.54∗∗∗ (2.59) 0.49∗∗ (2.39) 0.49∗∗ (2.44) 0.35∗ (1.76) 0.15∗∗ (2.37)
0.71∗∗∗ (4.90) 0.52∗∗∗ (4.19) 0.58∗∗∗ (4.70) 0.61∗∗∗ (4.92) 0.68∗∗∗ (4.61) 0.69∗∗∗ (4.65) 0.70∗∗∗ (4.73) 0.68∗∗∗ (4.65) 0.66∗∗∗ (4.52) 0.61∗∗∗ (4.22) 0.09∗∗∗ (5.66)
0.33 (1.31) −0.03 (−0.16) −0.02 (−0.09) 0.02 (0.11) 0.05 (0.23) 0.09 (0.38) 0.12 (0.50) 0.17 (0.72) 0.22 (0.95) 0.30 (1.31) 0.05∗∗∗ (4.17)
0.38∗∗ (2.62) 0.55∗∗∗ (5.21) 0.6∗∗∗ (5.77) 0.59∗∗∗ (5.71) 0.63∗∗∗ (4.60) 0.60∗∗∗ (4.45) 0.58∗∗∗ (4.45) 0.51∗∗∗ (4.02) 0.44∗∗∗ (3.53) 0.30∗∗ (2.43) 0.04 (1.41)
0.69∗∗∗ (6.46) 0.63∗∗∗ (7.09) 0.66∗∗∗ (7.53) 0.68∗∗∗ (7.74) 0.69∗∗∗ (6.30) 0.68∗∗∗ (6.28) 0.66∗∗∗ (6.12) 0.61∗∗∗ (5.74) 0.57∗∗∗ (5.39) 0.53∗∗∗ (5.08) 0.05∗∗∗ (5.38)
0.18 (1.02) 0.05 (0.39) 0.08 (0.57) 0.11 (0.84) 0.13 (0.73) 0.17 (0.99) 0.23 (1.33) 0.30∗ (1.77) 0.36∗∗ (2.16) 0.42∗∗ (2.54) 0.06∗∗∗ (7.88)
0.51∗∗∗ (5.26) 0.57∗∗∗ (8.50) 0.59∗∗∗ (8.79) 0.57∗∗∗ (8.60) 0.56∗∗∗ (5.96) 0.51∗∗∗ (5.53) 0.43∗∗∗ (4.81) 0.31∗∗∗ (3.53) 0.21∗∗ (2.42) 0.11 (1.28) −0.01 (−0.52)
1 2 3 4 5 6 7 8 9 3−1
This table reports the average monthly portfolio returns from July 1963 through December 1996 for four different 52-week high momentum investing strategies. s denotes the number of months in the skip period, and h denotes the number of months in the holding period. The sample includes all stocks on CRSP; t-statistics are in parentheses. ∗ , ∗∗ , ∗∗∗ denote significance at the 10%, 5% and 1% levels, respectively.
strategies using a range of skip-period s and holding period h. For portfolios with skip-period of one and higher, the results show that all of the self-financing portfolios and the winner portfolios produce significantly positive returns. For loser portfolios, none of the returns are significantly different from zero. The bottom row shows that the difference between the three-month skip-period (12, 3, h) and the one-month skip-period (12, 1, h) strategies is statistically significant for all three holding periods for the winner portfolios. Novy-Marx (2012) shows that the return-based JT momentum portfolios exhibit an echo effect, whereby increasing the length of the skip-period between the time of portfolio formation and the time of portfolio purchase, improves the performance of the momentum portfolios. For the return-based JT momentum portfolios studied by Novy-Marx (2012), a skip-period of 6 months generates the best improvement. The in-sample results of Table 1 show that the 52-week high strategy also exhibits an echo effect. However, the length of the skip-period for the echo effect in the 52-week high portfolios differs from the length of the skip-period for the echo effect found by Novy-Marx (2012) for the JT momentum portfolios. For the 52-week high portfolios, a skip-period of 3 months generates the best performance improvement. For the zero investment portfolios, the results in Table 1 show that increasing the skip period between the date of portfolio formation and the date of portfolio purchase to 3 months significantly improves performance in all cases. To summarize, the in-sample results shown in Table 1 give support to the notion that anchoring based momentum strategies such as the 52-week high strategy are distinct from the return-based JT momentum strategies, with unique (echo) characteristics. GH showed that the 52-week high strategy performs better than the JT momentum strategy under a variety of measures. To this date, out of the set of the more commonly analyzed momentum strategies discussed in the literature, the GH 52-week high strategy by most accounts remains one of the best performing. Our finding of an echo effect for the 52-week high portfolios is important in that it reveals a new way to squeeze even more performance out of the already high performing GH strategy; by simPlease cite this article as: A.-S. Chen, W. Yang, Echo effects and the returns from 52-week high strategies, Finance Research Letters (2015), http://dx.doi.org/10.1016/j.frl.2015.10.015
JID: FRL
ARTICLE IN PRESS[m1G;November 10, 2015;10:13]
4
A.-S. Chen, W. Yang / Finance Research Letters 000 (2015) 1–9
ply imposing a skip-period of 3 month between the date of portfolio formation and the date of portfolio purchase.1 The first row of Table 1 also shows the in-sample performance results of various zero-skip-period 52-week high portfolios. The established methodology for researching return-based momentum and anchoring-based 52-week high strategies with monthly data is to use the one-month skip-period as the default. Examples include Jegadeesh and Titman (2001), George and Hwang (2004), Novy-Marx (2012), and Goyal and Wahal (forthcoming), among others. The zero skip-period portfolios are usually excluded from the analysis. The reasons given in the literature are that the zero skip-period portfolios are likely to exhibit short-term reversals, microstructure distortions, and bid-ask bounce effects. Jegadeesh (1990) and Lehmann (1990), are examples that use the one-month skip-period to avoid the effects of short-term reversals. Griffin, Ji, and Martin (2002) in analyzing momentum investing and business cycle risk focus on results that skip a month between portfolio ranking and investment periods to avoid microstructure distortions. Chordia and Shivakumar (2002) use a one-month gap between the portfolio formation period and the holding period to allow for an implementable strategy and mitigate any bid-ask bounce effects.2 In addition, it can also be argued that it may not be realistically possible for an investor in the real world to actually purchase/sell the zero skip-period momentum portfolio, since this would entail instantaneous data processing and the purchase or selling of securities using the same prices that were used in computing the composition of the zero skip-period momentum portfolio in the first place.3 In short, the above reasons can cloud the results of these zero skip-period portfolios and, thus, detract from the analyses of the true nature of these strategies. The results in the first line of Table 1 for the zero skip-period portfolios support the arguments of these previous studies. For the zero investment portfolios, the results are consistent with the pattern of returns for those portfolios with skip-period of one and higher; for all holding periods, the zero skipperiod portfolios generate the lowest returns, the returns increases as the skip-period increases from one to three months. The zero-investment 52-week high portfolio results are also consistent with the results of Griffin et al. (2003) for return-based momentum strategies. Griffin et al. (2003) show that the profits of return-based momentum strategies without skipping a month are smaller than those skipping a month. On the other hand, the results for the one-sided winner (and loser) zero skip-period portfolios are inconsistent with this pattern. For these portfolios, the returns are larger than those of the corresponding one-month skip-period portfolios. These results, however, support the various arguments given by previous researchers that the zero-skip period portfolios may contain distortions and that methodologically it is generally more appropriate to focus on the portfolios with skip-period of one month or greater in the analyses. Overall, the results of Table 1 show evidence of an echo effect for the 52-week high portfolios; increasing the skip-period between the time of portfolio formation to the time of portfolio purchase up to three months significantly improves the performance of the 52-week high portfolios. 3.2. Out-of-sample results Table 2A reports the corresponding results for the out-of-sample period from January 1997 through December 2013. For this period, the 6-month holding period self-financing and winner portfolios produce significantly positive returns. For portfolios with skip-period of one and higher, all of the loser portfolio strategies produce negative returns but are not significantly different from zero. For the self-financing strategies, increasing the skip-period to three months produces statistically significant positive returns when the holding periods are three and six months, generating returns of 0.48% and 0.38% per month, re1 Other more recent papers that study the 52-week high strategy include Li and Yu (2012) on aggregate market returns, Liu et al. (2011) on international stock markets, and Sapp (2011) on mutual fund returns and cash flows, among others. These papers also do not check for the existence of the echo effect in the 52-week high portfolios. 2 More recently, Asem and Tian (2010) in studying market dynamics and momentum profits skip one month between the portfolio formation and holding periods to minimize microstructure-induced biases. Hong et al. (2015) in researching industry information on the 52-week high effect, skip one month between the portfolio forming month and holding period to control for the effect of bid-ask bounce. 3 The argument “forming the zero-skip momentum portfolio may not be possible in the real world” may be somewhat outdated, considering how much the speed of trading and dissemination of information has developed during the recent decades. It could be interesting to consider in future studies whether the echo effect has changed over time.
Please cite this article as: A.-S. Chen, W. Yang, Echo effects and the returns from 52-week high strategies, Finance Research Letters (2015), http://dx.doi.org/10.1016/j.frl.2015.10.015
ARTICLE IN PRESS[m1G;November 10, 2015;10:13]
JID: FRL
A.-S. Chen, W. Yang / Finance Research Letters 000 (2015) 1–9
5
Table 2A Returns for 52-week high momentum strategies (out-of-sample period). h 1
3
6
s
Winner
Loser
Winner − loser
Winner
Loser
Winner − loser
Winner
Loser
Winner − loser
0
0.98∗∗ (3.25) 0.05 (0.18) 0.11 (0.37) 0.21 (0.69) 0.18 (0.57) 0.32 (1.00) 0.30 (0.98) 0.29 (0.93) 0.21 (0.67) 0.22 (0.71) 0.16∗ (1.71)
1.25 (1.59) −0.01 (−0.01) 0 (−0.01) −0.06 (−0.10) −0.29 (−0.42) −0.28 (−0.42) −0.24 (−0.37) −0.12 (−0.19) −0.05 (−0.08) 0.00 (−0.01) −0.06 (−0.65)
−0.27 (−0.43) 0.06 (0.11) 0.12 (0.22) 0.28 (0.56) 0.47 (0.91) 0.60 (1.20) 0.55 (1.14) 0.41 (0.89) 0.26 (0.60) 0.23 (0.59) 0.22 (1.27)
0.33∗ (1.67) 0.12 (0.73) 0.2 (1.21) 0.3∗ (1.82) 0.26 (1.32) 0.29 (1.51) 0.27 (1.34) 0.23 (1.16) 0.22 (1.07) 0.20 (1.00) 0.18∗∗∗ (9.34)
0.18 (0.38) −0.14 (−0.38) −0.16 (−0.44) −0.18 (−0.49) −0.31 (−0.71) −0.26 (−0.61) −0.17 (−0.41) −0.07 (−0.17) −0.02 (−0.06) 0.03 (0.07) −0.04 (−1.31)
0.15 (0.44) 0.26 (1.00) 0.36 (1.42) 0.48∗ (1.93) 0.57∗ (1.82) 0.55∗ (1.85) 0.44 (1.51) 0.30 (1.08) 0.24 (0.92) 0.18 (0.71) 0.22∗∗∗ (5.28)
0.27∗ (1.88) 0.21∗ (1.85) 0.27∗∗ (2.37) 0.31∗∗∗ (2.69) 0.24∗ (1.71) 0.25∗ (1.75) 0.23 (1.56) 0.20 (1.35) 0.18 (1.20) 0.18 (1.19) 0.09∗∗∗ (8.18)
−0.01 (−0.03) −0.12 (−0.46) −0.11 (−0.42) −0.07 (−0.29) −0.11 (−0.33) −0.05 (−0.17) 0.00 (0.01) 0.08 (0.26) 0.13 (0.45) 0.16 (0.56) 0.05∗∗ (2.37)
0.28 (1.01) 0.34∗ (1.67) 0.38∗∗ (1.97) 0.38∗∗ (2.05) 0.35 (1.46) 0.30 (1.33) 0.22 (1.06) 0.12 (0.58) 0.04 (0.22) 0.01 (0.06) 0.05 (1.54)
1 2 3 4 5 6 7 8 9 3−1
This table reports the average monthly portfolio returns in the out-of-sample period from January 1997 through December 2013 for four different 52-week high momentum investing strategies. s denotes the number of months in the skip period, and h denotes the number of months in the holding period. The sample includes all stocks on CRSP; t-statistics are in parentheses. ∗ , ∗∗ , ∗∗∗ denote significance at the 10%, 5% and 1% levels, respectively.
spectively. When the holding period is reduced to one month, the three-month skip-period self-financing strategy yields positive returns, larger than corresponding two-month and one-month skip-period returns, but is not statistically significant. For all three holding periods, the conventional one skip-period (12, 1, h) strategies produce the worst self-financing returns for portfolios with skip-period of one and higher, generating returns of 0.06%, 0.26% and 0.34% per month when the holding periods are 1, 3, and 6 months, respectively. All of the winner portfolios show positive returns. All of the six-month holding period winner portfolios show statistically significant positive returns. For the winner portfolios, the three-month skip-period (12, 3, h) strategies produce statistically significant positive returns, for portfolios with skip-period of one and higher, generating returns of 0.30% and 0.31% per month when the holding periods are 3 and 6 months, respectively. The return of the one-month holding period (12, 3, h) winner portfolio, however, is not statistically significant. For each holding period, the worst performing winner portfolios are the conventional one-month skip-period (12, 1, h) portfolios, giving returns of 0.05%, 0.12% and 0.21% per month when the holding periods are 1, 3, and 6 months, respectively. For portfolios with skip-period of one month, only the six-month holding period (12, 1, 6) winner portfolio produces returns that are statistically significant. The three-month holding period and the one-month holding period winner portfolios (12,1,1) and (12,1,3) provide returns that are not statistically significant. The last row in the table shows that, for winner portfolios, the differences between the three-month skip-period (12, 3, h) and the conventional one-month skip-period (12, 1, h) strategies are statistically significant for all three holding periods analyzed. For the out-of-sample period, the main finding with respect to the differences in performance, between the conventional one-month skip period portfolios and the three-month skip period portfolios, are comparable to those of the in-sample period. Comparing the results of Table 1 with those of Table 2A, it can be seen that for both the in-sample and out-of-sample periods, the winner portfolios using Please cite this article as: A.-S. Chen, W. Yang, Echo effects and the returns from 52-week high strategies, Finance Research Letters (2015), http://dx.doi.org/10.1016/j.frl.2015.10.015
JID: FRL
ARTICLE IN PRESS[m1G;November 10, 2015;10:13]
6
A.-S. Chen, W. Yang / Finance Research Letters 000 (2015) 1–9
Table 2B Risk-adjusted returns for 52-week high momentum strategies (out-of-sample period). h 1
3
6
s
Winner
Loser
Winner − loser
Winner
Loser
Winner − loser
Winner
Loser
Winner − loser
0
0.45∗∗∗ (4.27) −0.59∗∗∗ (−5.37) −0.55∗∗∗ (−5.72) −0.45∗∗∗ (−4.70) −0.37∗∗∗ (−4.01) −0.23∗∗ (−2.46) −0.23∗∗ (−2.52) −0.25∗∗∗ (−2.70) −0.32∗∗∗ (−3.14) −0.31∗∗∗ (−3.16) 0.13 (1.35)
0.36 (0.78) −1.05∗∗ (−2.55) −1.03∗∗ (−2.58) −1.06∗∗ (−2.76) −1.12∗∗∗ (−2.78) −1.10∗∗∗ (−2.83) −1.03∗∗∗ (−2.76) −0.91∗∗ (−2.53) −0.83∗∗ (−2.48) −0.77∗∗ (−2.55) −0.03 (−0.38)
0.07 (0.13) 0.44 (0.95) 0.46 (1.06) 0.6 (1.39) 0.73∗ (1.67) 0.85∗∗ (2.01) 0.78∗ (1.94) 0.64∗ (1.66) 0.50 (1.39) 0.44 (1.42) 0.14 (0.84)
0.07 (0.45) −0.04 (−0.23) 0.04 (0.27) 0.14 (0.90) 0.00 (0.02) 0.04 (0.25) 0.02 (0.12) −0.02 (−0.09) −0.04 (−0.25) −0.06 (−0.33) 0.16∗∗∗ (8.35)
−0.27 (−0.63) −0.45 (−1.22) −0.46 (−1.29) −0.48 (−1.35) −0.73∗ (−1.84) −0.68∗ (−1.78) −0.59 (−1.59) −0.48 (−1.33) −0.43 (−1.20) −0.37 (−1.06) −0.05∗ (−1.66)
0.33 (0.99) 0.39 (1.51) 0.49∗ (1.93) 0.6∗∗ (2.44) 0.72∗∗ (2.38) 0.70∗∗ (2.44) 0.59∗∗ (2.15) 0.45∗ (1.69) 0.37 (1.46) 0.29 (1.24) 0.19∗∗∗ (4.63)
0.11 (0.87) 0.17 (1.48) 0.24∗∗ (2.01) 0.27∗∗ (2.34) 0.09 (0.71) 0.09 (0.72) 0.08 (0.57) 0.04 (0.32) 0.02 (0.15) 0.02 (0.14) 0.08∗∗∗ (6.94)
−0.30 (−0.86) −0.22 (−0.81) −0.21 (−0.79) −0.17 (−0.67) −0.38 (−1.20) −0.32 (−1.03) −0.26 (−0.88) −0.17 (−0.60) −0.12 (−0.41) −0.08 (−0.28) 0.03 (1.56)
0.39 (1.46) 0.38∗ (1.85) 0.43∗∗ (2.18) 0.43∗∗ (2.27) 0.45∗ (1.90) 0.39∗ (1.74) 0.31 (1.51) 0.20 (0.99) 0.12 (0.60) 0.08 (0.42) 0.03 (1.05)
1 2 3 4 5 6 7 8 9 3−1
This table reports the average risk-adjusted return of monthly portfolio in the out-of-sample period from January 1997 through December 2013 for three different 52-week high momentum investing strategies. s denotes the number of months in the skip period, and h denotes the number of months in the holding period. The risk-adjusted returns are computed following the same risk-adjustment procedure described in George and Hwang (2004) of hedging out the strategy’s Fama and French (1996) factor exposure. Specifically, under this procedure, the risk-adjusted return of a particular portfolio would be the intercept from a timeseries regression of the portfolio’s unadjusted returns on the contemporaneous Fama–French factors. The sample includes all stocks on CRSP; t-statistics are in parentheses. ∗ , ∗∗ , ∗∗∗ denote significance at the 10%, 5% and 1% level, respectively.
the three-month skip-period provide significantly higher returns than the conventional one-month skipperiod portfolios. This is true for all three holding periods analyzed. The computed results support our hypothesis concerning the ’ad-hoc’ nature of the conventional one-month skip-period of the 52-week high strategies being used in the current literature. The numbers in the tables show that increasing the skip-period 3–6 months can indeed produce significant improvements in the performance of the 52-week high strategies for both the in-sample and out-of-sample periods. For the winner portfolios, in particular, there appears a general pattern of increase in performance as the skip-period is increased from 1 to 3 months, regardless of the length of the holding period. 3.3. Risk-adjusted return analysis Table 2B presents the results of Table 2A using risk-adjusted returns.4 The risk-adjusted returns are estimated as the intercept from the following Fama and French (1996) three factors model regression:
R pit,os − R f t,os = αi + βi [Rmt,os − R f t,os ] + si (SMLt,os ) + hi (HMLt,os )
(1)
where Rpit, os is the monthly return of portfolio i at time t in the out-of-sample period, Rft, os is the one month Treasury Bill return at time t in the out-of-sample period, α i is the estimated regression intercept 4 We follow the same risk-adjustment procedure described in George and Hwang (2004) of hedging out the strategy’s Fama and French (1996) factor exposure. Specifically, using their procedure, the risk-adjusted return of a particular portfolio would be the intercept from a time-series regression of the portfolio’s unadjusted returns on the contemporaneous Fama–French factors.
Please cite this article as: A.-S. Chen, W. Yang, Echo effects and the returns from 52-week high strategies, Finance Research Letters (2015), http://dx.doi.org/10.1016/j.frl.2015.10.015
ARTICLE IN PRESS[m1G;November 10, 2015;10:13]
JID: FRL
A.-S. Chen, W. Yang / Finance Research Letters 000 (2015) 1–9
7
Table 2C Risk-adjusted returns for 52-week high momentum strategies (two-stage method). h 1
3
6
s
Winner
Loser
Winner − loser
Winner
Loser
Winner − loser
Winner
Loser
Winner − loser
0
0.42∗∗ (3.13) −0.63∗∗∗ (−4.884) −0.58∗∗∗ (−4.907) −0.47∗∗∗ (−4.288) −0.39∗∗∗ (−3.31) −0.25∗∗ (−2.11) −0.27∗∗ (−2.18) −0.28∗∗ (−2.22) −0.35∗∗∗ (−2.72) −0.33∗∗∗ (−2.63) 0.14 (1.537)
0.23 (0.45) −1.09∗∗ (−2.479) −1.08∗∗ (−2.551) −1.14∗∗ (−2.773) −1.26∗∗∗ (−2.92) −1.24∗∗∗ (−2.98) −1.19∗∗∗ (−2.96) −1.06∗∗∗ (−2.76) −0.98∗∗∗ (−2.73) −0.93∗∗∗ (−2.82) −0.06 (−0.739)
0.17 (0.30) 0.44 (.866) 0.49 (1.016) 0.65 (1.396) 0.85∗ (1.78) 0.97∗∗ (2.10) 0.90∗∗ (2.04) 0.77∗ (1.81) 0.61 (1.57) 0.58∗ (1.66) 0.19 (1.121)
0.12 (0.64) 0.03 (.193) 0.11 (.646) 0.21 (1.222) 0.03 (0.15) 0.06 (0.34) 0.04 (0.21) 0.00 (0.02) −0.01 (−0.07) −0.02 (−0.09) 0.16∗∗∗ (8.000)
−0.17 (−0.38) −0.25 (−0.663) −0.27 (−0.717) −0.28 (−0.771) −0.64 (−1.57) −0.58 (−1.49) −0.49 (−1.28) −0.39 (−1.04) −0.34 (−0.93) −0.29 (−0.82) −0.05∗ (−1.756)
0.26 (0.78) 0.27 (1.009) 0.36 (1.394) 0.47∗ (1.873) 0.65∗∗ (2.12) 0.63∗∗ (2.16) 0.51∗ (1.83) 0.37 (1.40) 0.31 (1.22) 0.26 (1.07) 0.19∗∗∗ (4.547)
0.14 (1.02) 0.18 (1.587) 0.24∗∗ (2.101) 0.28∗∗ (2.405) 0.11 (0.84) 0.12 (0.86) 0.10 (0.70) 0.07 (0.50) 0.05 (0.36) 0.05 (0.34) 0.08∗∗∗ (6.356)
−0.17 (−0.49) −0.14 (−0.504) −0.12 (−0.467) −0.09 (−0.333) −0.27 (−0.83) −0.21 (−0.67) −0.15 (−0.51) −0.08 (−0.26) −0.02 (−0.07) 0.01 (0.04) 0.03 (1.620)
0.29 (1.07) 0.3 (1.480) 0.35∗ (1.773) 0.34∗ (1.824) 0.36 (1.50) 0.30 (1.35) 0.23 (1.09) 0.13 (0.63) 0.05 (0.27) 0.02 (0.10) 0.02 (.826)
1 2 3 4 5 6 7 8 9 3−1
This table reports the average risk-adjusted return of monthly portfolio in the out-of-sample period from January 1997 through December 2013 for three different 52-week high momentum investing strategies. s denotes the number of months in the skip period, and h denotes the number of months in the holding period. The risk-adjusted returns are estimated from Fama–French three-factor model. The factor loadings of the Fama–French three factors are estimated from the in-sample period. The risk-adjusted returns in the out-of-sample period are then obtained from portfolio return minus expected return. The sample includes all stocks on CRSP; t-statistics are in parentheses. ∗ , ∗∗ , ∗∗∗ denote significance at the 10%, 5% and 1% level, respectively.
for portfolio i, Rmt, os is the monthly return of the market portfolio at time t in the out-of-sample period, SMLt, os is the size factor at time t in the out-of-sample period, and HMLt, os is the book to market factor at time t in the out-of-sample period. The results show that even after risk-adjustment, many of the portfolios remain statistically significant. For the three-month holding period, the two-month and three-month skip-period, self-financed portfolios provide positive risk-adjusted returns that are statistically significant at the 10% and 5% levels, respectively. For the six-month holding period, the two-month and three-month skip-period, winner portfolios and self-financed portfolios generate positive risk-adjusted returns that are statistically significant at the 5% level. For the three-month holding period, the three-month skip-period, winner portfolios and self-financed portfolios provide significantly higher returns than the conventional one-month skip-period portfolios, statistically significant at the 1% level. For the six-month holding period, the three-month skip period, winner portfolios give significantly higher returns than the conventional one-month skip-period portfolios, statistically significant at the 1% level. An alternative method of controlling for the effects of risk is to directly compute the risk-adjusted returns by subtracting the expected portfolio return computed from the Fama–French three factor model from the corresponding actual portfolio return. Table 2C presents the resulting risk adjusted returns using this method for the out-of-sample period. The numbers shown on Table 2C can be interpreted as the riskadjusted returns of the corresponding unadjusted returns of Table 2A. In detail, the risk adjustment process for Table 2C requires two steps. In the first step, we compute the factor loadings from the following Fama–French three factor model using in-sample data:
R pit,is − R f t,is = αi + βi [Rmt,is − R f t,is ] + si (SMLt,is ) + hi (HMLt,is )
(2)
where Rpit, is is the monthly return on the portfolio i at time t in the in-sample period, Rft, is is the one month Treasury Bill return at time t in the in-sample period, α i is the estimated regression intercept for Please cite this article as: A.-S. Chen, W. Yang, Echo effects and the returns from 52-week high strategies, Finance Research Letters (2015), http://dx.doi.org/10.1016/j.frl.2015.10.015
JID: FRL
ARTICLE IN PRESS[m1G;November 10, 2015;10:13]
8
A.-S. Chen, W. Yang / Finance Research Letters 000 (2015) 1–9
portfolio i computed using in-sample data, Rmt, is is the monthly return of the market portfolio at time t in the in-sample period, SMLt, is is the size factor at time t in the in-sample period, and HMLt, is is the book to market factor at time t in the in-sample period. In the second step, the expected return for portfolio i in the out-of-sample period is computed by combining the factor loadings computed in the first step from in-sample data with the out-of-sample values of the three Fama–French risk factors for each of the months in the out-of-sample period using the Fama–French three-factor model.
E[R] pii = R f t,os + βi [Rmt,os − R f t,os ] + si (SMLt,os ) + hi (HMLt,os )
(3)
where E[R]pii is the expect monthly return under the Fama–French three-factor model for portfolio i at time t in the out-of-sample period, Rft, os is the one month Treasury Bill return at time t in the out-ofsample period, Rmt, os is the monthly return of the market portfolio at time t in the out-of-sample period, SMLt, os is the size factor at time t in the out-of-sample period, and HMLt, os is the book to market factor at time t in the out-of-sample period. The risk-adjusted returns for each month in the out-of-sample period are then obtained by subtracting the expected portfolio return computed from Eq. (3) from the corresponding actual portfolio return.
Radj,it = R pit,os − E[R] pit
(4)
where Radj, it is the risk-adjusted monthly return of portfolio i at time t in the out-of-sample period, Rpit, os is the unadjusted monthly return of portfolio i at time t in the out-of-sample period, and E[R]pit is the expect monthly return under the Fama–French three-factor model for portfolio i at time t in the out-ofsample period. The overall pattern of the results shown in Table 2C are similar to those in Table 2B. The zero investment portfolios, for example, maintain the pattern of performance getting better as the length of the skip-period is increased. Moreover, both Tables 2C and 2B show statistical significance for nearly the same set of portfolios. In both tables the performance improvement, for example, between the three-month skip-period portfolios and the one-month skip-period portfolios was statistically significant for the threemonth holding period zero-investment portfolios, the three-month holding period winner portfolios, and the six-month holding period winner portfolios. In sum, the risk-adjusted returns of Table 2C confirm the results of Table 2B which was computed using a different method. The results in Tables 2B and 2C also show that risk-adjusted returns are negative for several portfolios in the one-month holding period (winner) column. One possible reason why the risk-adjusted returns for the zero-skip period portfolios are positive but negative for other one-month holding period portfolios relates to the previous argument that forming the zero-skip momentum portfolio may not be possible in the real world. That is, even though the speed of trading and dissemination of information has developed during the recent decades making it technically feasible to form the zero-skip momentum portfolio, in actual practice the vast majority of investors are still not doing it and; thus, these zero-skip portfolios are still not efficiently priced in the market, resulting in the positive risk-adjusted returns shown in our results. One practical implication of these results is that given many momentum-based products are long-only ETFs, the investor needs to be careful in investing in these products and attempt to determine whether or not the fund manager has the ability to establish the zero-skip momentum portfolio which has the positive risk-adjusted returns. 4. Conclusions This study is related to George and Hwang (2004) and Novy-Marx (2012). George and Hwang (2004) show that nearness to the 52-week high is positively associated with future stock returns. Novy-Marx (2012) shows the existence of an echo effect in momentum returns; increasing the length of the skipperiod between the time of portfolio formation and the time of portfolio purchase improves the performance of the momentum portfolios. Novy-Marx (2012) finds that a skip-period of 6 months generates the best improvement. In addition to George and Hwang (2004), other more recent papers that study the 52-week high strategy include Li and Yu (2012) on aggregate market returns, Liu et al. (2011) on international stock markets, and Sapp (2011) on mutual fund returns and cash flows, among others. None of these papers checks for the existence of the echo effect in the 52-week high portfolios. Please cite this article as: A.-S. Chen, W. Yang, Echo effects and the returns from 52-week high strategies, Finance Research Letters (2015), http://dx.doi.org/10.1016/j.frl.2015.10.015
JID: FRL
ARTICLE IN PRESS[m1G;November 10, 2015;10:13] A.-S. Chen, W. Yang / Finance Research Letters 000 (2015) 1–9
9
Our results show that the 52-week high strategy also exhibits an echo effect. However, the length of the skip-period for the echo effect in the 52-week high portfolios differs from the length of the skipperiod for the echo effect found by Novy-Marx (2012) in JT momentum portfolios. For the 52-week high portfolios, a skip-period of 3–6 months generates the best improvement; specifically, our results show that increasing the skip period between the date of portfolio formation and the date of portfolio purchase 3–6 months significantly improves performance in nearly all cases analyzed. The results give support to the notion that anchoring based momentum strategies such as the 52-week high strategy are distinct from return-based JT momentum strategies, with unique (echo) characteristics. References Arshanapalli, Bala, Fabozzi, FrankJ., Nelson, William, 2006. The value, size, and momentum spread during distressed economic periods. Financ. Res. Lett. 3, 244–252. Asem, Ebenezer, Tian, GloriaY., 2010. Market dynamics and momentum profits. J. Financ. Quant. Anal. 45, 1549–1562. Chordia, Tarun, Lakshmanan, Shivakumar, 2002. Momentum, business cycle, and time-varying expected returns. J. Financ. 57, 985– 1019. Cooper, Michael J., Gutierrez Jr, RobertoC., Hameed, Allaudeen, 2004. Market states and momentum. J. Financ. 59, 1345–1365. Fama, Eugene F., French, KennethR., 1996. Multifactor explanation of asset pricing anomalies. J. Financ. 51, 55–84. George, T., Hwang, C.Y., 2004. The 52-week high and momentum investing. J. Financ. 59, 2145–2176. Griffin, John M., Ji, Xiuqing, Spencer Martin, J., 2003. Momentum investing and business cycle risk: evidence from pole to pole. J. Financ. 58, 2515–2547. Goyal, Amit, Wahal, Sunil. Ismomentumanecho? J. Financ. Quant. Anal. (forthcoming). Hong, Xin, Jordan, BradfordD., Liu, MarkH., 2015. Industry information and the 52-week high effect. Pacific-Basin Financ. J. 32, 111– 130. Jegadeesh, Narasimhan, 1990. Evidence of predictable behavior of security returns. J. Financ. 45, 881–898. Jegadeesh, Narasimhan, Titman, Sheridan, 1993. Returns to buying winner’s and selling loser’s: implications for stock market efficiency. J. Financ. 48, 65–91. Jegadeesh, Narasimhan, Sheridan, Titman, 2001. Profitability of momentum strategies: an evaluation of alternative explanations. J. Financ. 56, 699–720. Lehmann, Bruce N., 1990. Fads, martingales, and market efficiency. Q. J. Econ. 105, 1–28. Li, Jun, Jianfeng, Yu, 2012. Investor attention, psychological anchors, and stock return predictability. J. Financ. Econ. 104, 401–419. Liu, Ming, Qianqiu, Liu, Tongshu, Ma, 2011. The 52-week high momentum strategy in international stock markets. J. Int. Money Financ. 30, 180–204. Novy-Marx, Robert, 2012, Is momentum really momentum? J. Financ. Econ. 103, 429–453. Sapp, Travis R.A., 2011. The 52-week high, momentum, and predicting mutual fund returns. Rev. Quant. Financ. Account. 37, 149–179.
Please cite this article as: A.-S. Chen, W. Yang, Echo effects and the returns from 52-week high strategies, Finance Research Letters (2015), http://dx.doi.org/10.1016/j.frl.2015.10.015
JID: FRL
Finance Research Letters 000 (2015) 1–9
Contents lists available at ScienceDirect
Finance Research Letters journal homepage: www.elsevier.com/locate/frl
Echo effects and the returns from 52-week high strategies An-Sing Chen∗, Wayne Yang National Chung Cheng University, Department of Finance, 168 University Road, Min-Hsiung, Chia-Yi 621, Taiwan, ROC
a r t i c l e
i n f o
Article history: Received 1 August 2015 Accepted 16 October 2015 Available online xxx JEL classification: G11 G12 G14 G17 Keywords: 52-Week high Momentum Skip-period Trading strategies Investment strategies Echo effect
a b s t r a c t Echo effects have been shown by the existing literature to influence the performance of conventional return-based momentum portfolios. This effect has yet to be confirmed for 52-week high momentum strategies. Our results show that the 52-week high strategy also manifests an echo effect. Increasing the skip period between the date of portfolio formation and the date of portfolio purchase 3–6 months significantly improves performance in nearly all cases analyzed. The results are robust to both in-sample and out-of-sample analyses. They are also robust to controlling for the effects on the risk of the portfolio from its return exposure to commonly used empirical return factors. © 2015 Elsevier Inc. All rights reserved.
1. Introduction For conventional Jegadeesh and Titman (1993) (JT henceforth) momentum portfolios, market states have been shown to be able to influence their performance (Cooper et al., 2004; Asem and Tian, 2010). The effects of business cycles on JT momentum returns have also been examined (Chordia and Shivakumarm, 2002; Griffin et al., 2003; Arshanapalli et al., 2006). Novy-Marx (2012) finds the presence of an echo effect in the conventional JT momentum portfolios; namely, increasing the length of the skip-period between ∗
Corresponding author. Tel.: +886 5 2720411x34201; fax: +886 5 2720818. E-mail address: fi[email protected] (A.-S. Chen).
http://dx.doi.org/10.1016/j.frl.2015.10.015 1544-6123/© 2015 Elsevier Inc. All rights reserved.
Please cite this article as: A.-S. Chen, W. Yang, Echo effects and the returns from 52-week high strategies, Finance Research Letters (2015), http://dx.doi.org/10.1016/j.frl.2015.10.015
JID: FRL
ARTICLE IN PRESS[m1G;November 10, 2015;10:13]
2
A.-S. Chen, W. Yang / Finance Research Letters 000 (2015) 1–9
the time of the momentum portfolio formation computations and the time of the actual purchase of the portfolio to 6 months significantly increases the profitability of the conventional JT momentum strategies. The echo effect is an important factor of the performance momentum strategies. However, the previous literature does not address how it affects the 52-week high strategy. An important objective of this study is to investigate if the 52-week high momentum strategy, as another cross-sectional momentum strategy, is distinct from the JT momentum strategy. Will changing the skip period also influence anchoring based momentum strategies such as the George and Hwang (2004) (GH henceforth) 52-week high momentum strategy? Exhaustive survey of the literature shows that none of the existing studies on the 52-week high momentum strategy has addressed this issue. This study finds that the 52-week high strategy also manifests the Novy-Marx style echo effect; increasing the skip period 3–6 months dramatically improves performance in nearly all cases analyzed. Additional robustness checks show that the performance improvements from adjusting the skip period are also robust to controlling for commonly used return risk factors. For investors, the results of this study suggest that waiting up 3–6 months before purchasing (short selling) stocks identified by the 52-week high strategy can significantly improve performance in the holding period. 2. Data and method The data cover the sample period from July 1963 through December 2013 and include all stocks in the Center for Research in Securities Prices (CRSP) universe. The benchmark 52-week high momentum strategy is implemented in this study using the following modification of method described in GH to allow for skip-period. At the end of each month, we rank stocks based on the ratio of the current price to the past 52-week high. We then identify the top 30 percentile (winners) and the bottom 30 percentile (losers) stocks, wait (skip) s months before establishing a zero-investment (self-financing) portfolio consisting of long positions in the top 30 percentile and short positions in the bottom 30 percentile stocks identified s months ago. Specifically, stocks are ranked based on
Pi,t−s , highi,t−s
where Pi,t−s is the price of stock i at the
end of month t − s and highi,t−s is the highest price of stock i during the 12-month period that ends on the last day of month t − s. We then hold the winner and loser portfolios for h months. The momentum profits are calculated as the difference between the equally weighted average returns of the winner and loser portfolios established h months ago. When s is equal to 1, the implementation becomes the standard GH 52-week high momentum strategy discussed in the literature. For this study, we analyze the effects of varying the skip-period s, from 0, 1, 2, to 9 months. We also analyze holding periods h of 1, 3, and 6 months. For clarity of exposition, in the remainder of this paper, we will denote the various 52-week high momentum strategies analyzed in this study using the (12, s, h) notation. The “twelve” refers to the fact that 52-weeks are approximately equal to 12 months, s denotes the number of months in the skip period, and h denotes the number of months in the holding period. The standard one month skip period GH 52week high momentum strategy with a holding period of six months, for example, is denoted as (12, 1, 6) using this notation. In the analyses, the dataset is divided into two parts. The first two-thirds of the dataset from July 1963 to December 1996 is denoted as in-sample. The remaining one-third of the dataset from January 1997 through December 2013 is denoted as the reserved out-of-sample. This 2/3 and 1/3 split is widely used in the forecasting literature and allows us to check whether skip-period s that gives the best performance in the in-sample period will continue to give the best performance in the reserved out-of-sample period. Evidence that s does not change dramatically from in-sample to out-of-sample will give us additional confidence as to the robustness of our findings with respect to s and to its generalizability to future data that is not part of our dataset. 3. Results 3.1. Profits from momentum strategies Table 1 reports the average monthly returns of winner, loser, and self-financing portfolios for the insample period from July 1963 through December 1996 for various 52-week high momentum investing Please cite this article as: A.-S. Chen, W. Yang, Echo effects and the returns from 52-week high strategies, Finance Research Letters (2015), http://dx.doi.org/10.1016/j.frl.2015.10.015
ARTICLE IN PRESS[m1G;November 10, 2015;10:13]
JID: FRL
A.-S. Chen, W. Yang / Finance Research Letters 000 (2015) 1–9
3
Table 1 Returns for 52-week high momentum strategies (July 1963 through December 1996). h 1
3
6
s
Winner
Loser
Winner−loser
Winner
Loser
Winner−loser
Winner
Loser
Winner−loser
0
1.06∗∗∗ (4.80) 0.38∗ (1.68) 0.5∗∗ (2.23) 0.51∗∗ (2.27) 0.65∗∗∗ (2.86) 0.66∗∗∗ (2.91) 0.69∗∗∗ (2.99) 0.68∗∗∗ (3.00) 0.71∗∗∗ (3.11) 0.64∗∗∗ (2.82) 0.13∗∗∗ (3.45)
1.28∗∗ (3.30) 0.00 (0.01) −0.06 (−0.16) −0.01 (−0.04) 0.08 (0.23) 0.14 (0.37) 0.14 (0.40) 0.19 (0.53) 0.22 (0.61) 0.29 (0.83) −0.02 (−0.54)
−0.22 (−0.90) 0.38∗ (1.76) 0.56∗∗∗ (2.69) 0.52∗∗ (2.56) 0.57∗∗ (2.58) 0.53∗∗ (2.45) 0.54∗∗∗ (2.59) 0.49∗∗ (2.39) 0.49∗∗ (2.44) 0.35∗ (1.76) 0.15∗∗ (2.37)
0.71∗∗∗ (4.90) 0.52∗∗∗ (4.19) 0.58∗∗∗ (4.70) 0.61∗∗∗ (4.92) 0.68∗∗∗ (4.61) 0.69∗∗∗ (4.65) 0.70∗∗∗ (4.73) 0.68∗∗∗ (4.65) 0.66∗∗∗ (4.52) 0.61∗∗∗ (4.22) 0.09∗∗∗ (5.66)
0.33 (1.31) −0.03 (−0.16) −0.02 (−0.09) 0.02 (0.11) 0.05 (0.23) 0.09 (0.38) 0.12 (0.50) 0.17 (0.72) 0.22 (0.95) 0.30 (1.31) 0.05∗∗∗ (4.17)
0.38∗∗ (2.62) 0.55∗∗∗ (5.21) 0.6∗∗∗ (5.77) 0.59∗∗∗ (5.71) 0.63∗∗∗ (4.60) 0.60∗∗∗ (4.45) 0.58∗∗∗ (4.45) 0.51∗∗∗ (4.02) 0.44∗∗∗ (3.53) 0.30∗∗ (2.43) 0.04 (1.41)
0.69∗∗∗ (6.46) 0.63∗∗∗ (7.09) 0.66∗∗∗ (7.53) 0.68∗∗∗ (7.74) 0.69∗∗∗ (6.30) 0.68∗∗∗ (6.28) 0.66∗∗∗ (6.12) 0.61∗∗∗ (5.74) 0.57∗∗∗ (5.39) 0.53∗∗∗ (5.08) 0.05∗∗∗ (5.38)
0.18 (1.02) 0.05 (0.39) 0.08 (0.57) 0.11 (0.84) 0.13 (0.73) 0.17 (0.99) 0.23 (1.33) 0.30∗ (1.77) 0.36∗∗ (2.16) 0.42∗∗ (2.54) 0.06∗∗∗ (7.88)
0.51∗∗∗ (5.26) 0.57∗∗∗ (8.50) 0.59∗∗∗ (8.79) 0.57∗∗∗ (8.60) 0.56∗∗∗ (5.96) 0.51∗∗∗ (5.53) 0.43∗∗∗ (4.81) 0.31∗∗∗ (3.53) 0.21∗∗ (2.42) 0.11 (1.28) −0.01 (−0.52)
1 2 3 4 5 6 7 8 9 3−1
This table reports the average monthly portfolio returns from July 1963 through December 1996 for four different 52-week high momentum investing strategies. s denotes the number of months in the skip period, and h denotes the number of months in the holding period. The sample includes all stocks on CRSP; t-statistics are in parentheses. ∗ , ∗∗ , ∗∗∗ denote significance at the 10%, 5% and 1% levels, respectively.
strategies using a range of skip-period s and holding period h. For portfolios with skip-period of one and higher, the results show that all of the self-financing portfolios and the winner portfolios produce significantly positive returns. For loser portfolios, none of the returns are significantly different from zero. The bottom row shows that the difference between the three-month skip-period (12, 3, h) and the one-month skip-period (12, 1, h) strategies is statistically significant for all three holding periods for the winner portfolios. Novy-Marx (2012) shows that the return-based JT momentum portfolios exhibit an echo effect, whereby increasing the length of the skip-period between the time of portfolio formation and the time of portfolio purchase, improves the performance of the momentum portfolios. For the return-based JT momentum portfolios studied by Novy-Marx (2012), a skip-period of 6 months generates the best improvement. The in-sample results of Table 1 show that the 52-week high strategy also exhibits an echo effect. However, the length of the skip-period for the echo effect in the 52-week high portfolios differs from the length of the skip-period for the echo effect found by Novy-Marx (2012) for the JT momentum portfolios. For the 52-week high portfolios, a skip-period of 3 months generates the best performance improvement. For the zero investment portfolios, the results in Table 1 show that increasing the skip period between the date of portfolio formation and the date of portfolio purchase to 3 months significantly improves performance in all cases. To summarize, the in-sample results shown in Table 1 give support to the notion that anchoring based momentum strategies such as the 52-week high strategy are distinct from the return-based JT momentum strategies, with unique (echo) characteristics. GH showed that the 52-week high strategy performs better than the JT momentum strategy under a variety of measures. To this date, out of the set of the more commonly analyzed momentum strategies discussed in the literature, the GH 52-week high strategy by most accounts remains one of the best performing. Our finding of an echo effect for the 52-week high portfolios is important in that it reveals a new way to squeeze even more performance out of the already high performing GH strategy; by simPlease cite this article as: A.-S. Chen, W. Yang, Echo effects and the returns from 52-week high strategies, Finance Research Letters (2015), http://dx.doi.org/10.1016/j.frl.2015.10.015
JID: FRL
ARTICLE IN PRESS[m1G;November 10, 2015;10:13]
4
A.-S. Chen, W. Yang / Finance Research Letters 000 (2015) 1–9
ply imposing a skip-period of 3 month between the date of portfolio formation and the date of portfolio purchase.1 The first row of Table 1 also shows the in-sample performance results of various zero-skip-period 52-week high portfolios. The established methodology for researching return-based momentum and anchoring-based 52-week high strategies with monthly data is to use the one-month skip-period as the default. Examples include Jegadeesh and Titman (2001), George and Hwang (2004), Novy-Marx (2012), and Goyal and Wahal (forthcoming), among others. The zero skip-period portfolios are usually excluded from the analysis. The reasons given in the literature are that the zero skip-period portfolios are likely to exhibit short-term reversals, microstructure distortions, and bid-ask bounce effects. Jegadeesh (1990) and Lehmann (1990), are examples that use the one-month skip-period to avoid the effects of short-term reversals. Griffin, Ji, and Martin (2002) in analyzing momentum investing and business cycle risk focus on results that skip a month between portfolio ranking and investment periods to avoid microstructure distortions. Chordia and Shivakumar (2002) use a one-month gap between the portfolio formation period and the holding period to allow for an implementable strategy and mitigate any bid-ask bounce effects.2 In addition, it can also be argued that it may not be realistically possible for an investor in the real world to actually purchase/sell the zero skip-period momentum portfolio, since this would entail instantaneous data processing and the purchase or selling of securities using the same prices that were used in computing the composition of the zero skip-period momentum portfolio in the first place.3 In short, the above reasons can cloud the results of these zero skip-period portfolios and, thus, detract from the analyses of the true nature of these strategies. The results in the first line of Table 1 for the zero skip-period portfolios support the arguments of these previous studies. For the zero investment portfolios, the results are consistent with the pattern of returns for those portfolios with skip-period of one and higher; for all holding periods, the zero skipperiod portfolios generate the lowest returns, the returns increases as the skip-period increases from one to three months. The zero-investment 52-week high portfolio results are also consistent with the results of Griffin et al. (2003) for return-based momentum strategies. Griffin et al. (2003) show that the profits of return-based momentum strategies without skipping a month are smaller than those skipping a month. On the other hand, the results for the one-sided winner (and loser) zero skip-period portfolios are inconsistent with this pattern. For these portfolios, the returns are larger than those of the corresponding one-month skip-period portfolios. These results, however, support the various arguments given by previous researchers that the zero-skip period portfolios may contain distortions and that methodologically it is generally more appropriate to focus on the portfolios with skip-period of one month or greater in the analyses. Overall, the results of Table 1 show evidence of an echo effect for the 52-week high portfolios; increasing the skip-period between the time of portfolio formation to the time of portfolio purchase up to three months significantly improves the performance of the 52-week high portfolios. 3.2. Out-of-sample results Table 2A reports the corresponding results for the out-of-sample period from January 1997 through December 2013. For this period, the 6-month holding period self-financing and winner portfolios produce significantly positive returns. For portfolios with skip-period of one and higher, all of the loser portfolio strategies produce negative returns but are not significantly different from zero. For the self-financing strategies, increasing the skip-period to three months produces statistically significant positive returns when the holding periods are three and six months, generating returns of 0.48% and 0.38% per month, re1 Other more recent papers that study the 52-week high strategy include Li and Yu (2012) on aggregate market returns, Liu et al. (2011) on international stock markets, and Sapp (2011) on mutual fund returns and cash flows, among others. These papers also do not check for the existence of the echo effect in the 52-week high portfolios. 2 More recently, Asem and Tian (2010) in studying market dynamics and momentum profits skip one month between the portfolio formation and holding periods to minimize microstructure-induced biases. Hong et al. (2015) in researching industry information on the 52-week high effect, skip one month between the portfolio forming month and holding period to control for the effect of bid-ask bounce. 3 The argument “forming the zero-skip momentum portfolio may not be possible in the real world” may be somewhat outdated, considering how much the speed of trading and dissemination of information has developed during the recent decades. It could be interesting to consider in future studies whether the echo effect has changed over time.
Please cite this article as: A.-S. Chen, W. Yang, Echo effects and the returns from 52-week high strategies, Finance Research Letters (2015), http://dx.doi.org/10.1016/j.frl.2015.10.015
ARTICLE IN PRESS[m1G;November 10, 2015;10:13]
JID: FRL
A.-S. Chen, W. Yang / Finance Research Letters 000 (2015) 1–9
5
Table 2A Returns for 52-week high momentum strategies (out-of-sample period). h 1
3
6
s
Winner
Loser
Winner − loser
Winner
Loser
Winner − loser
Winner
Loser
Winner − loser
0
0.98∗∗ (3.25) 0.05 (0.18) 0.11 (0.37) 0.21 (0.69) 0.18 (0.57) 0.32 (1.00) 0.30 (0.98) 0.29 (0.93) 0.21 (0.67) 0.22 (0.71) 0.16∗ (1.71)
1.25 (1.59) −0.01 (−0.01) 0 (−0.01) −0.06 (−0.10) −0.29 (−0.42) −0.28 (−0.42) −0.24 (−0.37) −0.12 (−0.19) −0.05 (−0.08) 0.00 (−0.01) −0.06 (−0.65)
−0.27 (−0.43) 0.06 (0.11) 0.12 (0.22) 0.28 (0.56) 0.47 (0.91) 0.60 (1.20) 0.55 (1.14) 0.41 (0.89) 0.26 (0.60) 0.23 (0.59) 0.22 (1.27)
0.33∗ (1.67) 0.12 (0.73) 0.2 (1.21) 0.3∗ (1.82) 0.26 (1.32) 0.29 (1.51) 0.27 (1.34) 0.23 (1.16) 0.22 (1.07) 0.20 (1.00) 0.18∗∗∗ (9.34)
0.18 (0.38) −0.14 (−0.38) −0.16 (−0.44) −0.18 (−0.49) −0.31 (−0.71) −0.26 (−0.61) −0.17 (−0.41) −0.07 (−0.17) −0.02 (−0.06) 0.03 (0.07) −0.04 (−1.31)
0.15 (0.44) 0.26 (1.00) 0.36 (1.42) 0.48∗ (1.93) 0.57∗ (1.82) 0.55∗ (1.85) 0.44 (1.51) 0.30 (1.08) 0.24 (0.92) 0.18 (0.71) 0.22∗∗∗ (5.28)
0.27∗ (1.88) 0.21∗ (1.85) 0.27∗∗ (2.37) 0.31∗∗∗ (2.69) 0.24∗ (1.71) 0.25∗ (1.75) 0.23 (1.56) 0.20 (1.35) 0.18 (1.20) 0.18 (1.19) 0.09∗∗∗ (8.18)
−0.01 (−0.03) −0.12 (−0.46) −0.11 (−0.42) −0.07 (−0.29) −0.11 (−0.33) −0.05 (−0.17) 0.00 (0.01) 0.08 (0.26) 0.13 (0.45) 0.16 (0.56) 0.05∗∗ (2.37)
0.28 (1.01) 0.34∗ (1.67) 0.38∗∗ (1.97) 0.38∗∗ (2.05) 0.35 (1.46) 0.30 (1.33) 0.22 (1.06) 0.12 (0.58) 0.04 (0.22) 0.01 (0.06) 0.05 (1.54)
1 2 3 4 5 6 7 8 9 3−1
This table reports the average monthly portfolio returns in the out-of-sample period from January 1997 through December 2013 for four different 52-week high momentum investing strategies. s denotes the number of months in the skip period, and h denotes the number of months in the holding period. The sample includes all stocks on CRSP; t-statistics are in parentheses. ∗ , ∗∗ , ∗∗∗ denote significance at the 10%, 5% and 1% levels, respectively.
spectively. When the holding period is reduced to one month, the three-month skip-period self-financing strategy yields positive returns, larger than corresponding two-month and one-month skip-period returns, but is not statistically significant. For all three holding periods, the conventional one skip-period (12, 1, h) strategies produce the worst self-financing returns for portfolios with skip-period of one and higher, generating returns of 0.06%, 0.26% and 0.34% per month when the holding periods are 1, 3, and 6 months, respectively. All of the winner portfolios show positive returns. All of the six-month holding period winner portfolios show statistically significant positive returns. For the winner portfolios, the three-month skip-period (12, 3, h) strategies produce statistically significant positive returns, for portfolios with skip-period of one and higher, generating returns of 0.30% and 0.31% per month when the holding periods are 3 and 6 months, respectively. The return of the one-month holding period (12, 3, h) winner portfolio, however, is not statistically significant. For each holding period, the worst performing winner portfolios are the conventional one-month skip-period (12, 1, h) portfolios, giving returns of 0.05%, 0.12% and 0.21% per month when the holding periods are 1, 3, and 6 months, respectively. For portfolios with skip-period of one month, only the six-month holding period (12, 1, 6) winner portfolio produces returns that are statistically significant. The three-month holding period and the one-month holding period winner portfolios (12,1,1) and (12,1,3) provide returns that are not statistically significant. The last row in the table shows that, for winner portfolios, the differences between the three-month skip-period (12, 3, h) and the conventional one-month skip-period (12, 1, h) strategies are statistically significant for all three holding periods analyzed. For the out-of-sample period, the main finding with respect to the differences in performance, between the conventional one-month skip period portfolios and the three-month skip period portfolios, are comparable to those of the in-sample period. Comparing the results of Table 1 with those of Table 2A, it can be seen that for both the in-sample and out-of-sample periods, the winner portfolios using Please cite this article as: A.-S. Chen, W. Yang, Echo effects and the returns from 52-week high strategies, Finance Research Letters (2015), http://dx.doi.org/10.1016/j.frl.2015.10.015
JID: FRL
ARTICLE IN PRESS[m1G;November 10, 2015;10:13]
6
A.-S. Chen, W. Yang / Finance Research Letters 000 (2015) 1–9
Table 2B Risk-adjusted returns for 52-week high momentum strategies (out-of-sample period). h 1
3
6
s
Winner
Loser
Winner − loser
Winner
Loser
Winner − loser
Winner
Loser
Winner − loser
0
0.45∗∗∗ (4.27) −0.59∗∗∗ (−5.37) −0.55∗∗∗ (−5.72) −0.45∗∗∗ (−4.70) −0.37∗∗∗ (−4.01) −0.23∗∗ (−2.46) −0.23∗∗ (−2.52) −0.25∗∗∗ (−2.70) −0.32∗∗∗ (−3.14) −0.31∗∗∗ (−3.16) 0.13 (1.35)
0.36 (0.78) −1.05∗∗ (−2.55) −1.03∗∗ (−2.58) −1.06∗∗ (−2.76) −1.12∗∗∗ (−2.78) −1.10∗∗∗ (−2.83) −1.03∗∗∗ (−2.76) −0.91∗∗ (−2.53) −0.83∗∗ (−2.48) −0.77∗∗ (−2.55) −0.03 (−0.38)
0.07 (0.13) 0.44 (0.95) 0.46 (1.06) 0.6 (1.39) 0.73∗ (1.67) 0.85∗∗ (2.01) 0.78∗ (1.94) 0.64∗ (1.66) 0.50 (1.39) 0.44 (1.42) 0.14 (0.84)
0.07 (0.45) −0.04 (−0.23) 0.04 (0.27) 0.14 (0.90) 0.00 (0.02) 0.04 (0.25) 0.02 (0.12) −0.02 (−0.09) −0.04 (−0.25) −0.06 (−0.33) 0.16∗∗∗ (8.35)
−0.27 (−0.63) −0.45 (−1.22) −0.46 (−1.29) −0.48 (−1.35) −0.73∗ (−1.84) −0.68∗ (−1.78) −0.59 (−1.59) −0.48 (−1.33) −0.43 (−1.20) −0.37 (−1.06) −0.05∗ (−1.66)
0.33 (0.99) 0.39 (1.51) 0.49∗ (1.93) 0.6∗∗ (2.44) 0.72∗∗ (2.38) 0.70∗∗ (2.44) 0.59∗∗ (2.15) 0.45∗ (1.69) 0.37 (1.46) 0.29 (1.24) 0.19∗∗∗ (4.63)
0.11 (0.87) 0.17 (1.48) 0.24∗∗ (2.01) 0.27∗∗ (2.34) 0.09 (0.71) 0.09 (0.72) 0.08 (0.57) 0.04 (0.32) 0.02 (0.15) 0.02 (0.14) 0.08∗∗∗ (6.94)
−0.30 (−0.86) −0.22 (−0.81) −0.21 (−0.79) −0.17 (−0.67) −0.38 (−1.20) −0.32 (−1.03) −0.26 (−0.88) −0.17 (−0.60) −0.12 (−0.41) −0.08 (−0.28) 0.03 (1.56)
0.39 (1.46) 0.38∗ (1.85) 0.43∗∗ (2.18) 0.43∗∗ (2.27) 0.45∗ (1.90) 0.39∗ (1.74) 0.31 (1.51) 0.20 (0.99) 0.12 (0.60) 0.08 (0.42) 0.03 (1.05)
1 2 3 4 5 6 7 8 9 3−1
This table reports the average risk-adjusted return of monthly portfolio in the out-of-sample period from January 1997 through December 2013 for three different 52-week high momentum investing strategies. s denotes the number of months in the skip period, and h denotes the number of months in the holding period. The risk-adjusted returns are computed following the same risk-adjustment procedure described in George and Hwang (2004) of hedging out the strategy’s Fama and French (1996) factor exposure. Specifically, under this procedure, the risk-adjusted return of a particular portfolio would be the intercept from a timeseries regression of the portfolio’s unadjusted returns on the contemporaneous Fama–French factors. The sample includes all stocks on CRSP; t-statistics are in parentheses. ∗ , ∗∗ , ∗∗∗ denote significance at the 10%, 5% and 1% level, respectively.
the three-month skip-period provide significantly higher returns than the conventional one-month skipperiod portfolios. This is true for all three holding periods analyzed. The computed results support our hypothesis concerning the ’ad-hoc’ nature of the conventional one-month skip-period of the 52-week high strategies being used in the current literature. The numbers in the tables show that increasing the skip-period 3–6 months can indeed produce significant improvements in the performance of the 52-week high strategies for both the in-sample and out-of-sample periods. For the winner portfolios, in particular, there appears a general pattern of increase in performance as the skip-period is increased from 1 to 3 months, regardless of the length of the holding period. 3.3. Risk-adjusted return analysis Table 2B presents the results of Table 2A using risk-adjusted returns.4 The risk-adjusted returns are estimated as the intercept from the following Fama and French (1996) three factors model regression:
R pit,os − R f t,os = αi + βi [Rmt,os − R f t,os ] + si (SMLt,os ) + hi (HMLt,os )
(1)
where Rpit, os is the monthly return of portfolio i at time t in the out-of-sample period, Rft, os is the one month Treasury Bill return at time t in the out-of-sample period, α i is the estimated regression intercept 4 We follow the same risk-adjustment procedure described in George and Hwang (2004) of hedging out the strategy’s Fama and French (1996) factor exposure. Specifically, using their procedure, the risk-adjusted return of a particular portfolio would be the intercept from a time-series regression of the portfolio’s unadjusted returns on the contemporaneous Fama–French factors.
Please cite this article as: A.-S. Chen, W. Yang, Echo effects and the returns from 52-week high strategies, Finance Research Letters (2015), http://dx.doi.org/10.1016/j.frl.2015.10.015
ARTICLE IN PRESS[m1G;November 10, 2015;10:13]
JID: FRL
A.-S. Chen, W. Yang / Finance Research Letters 000 (2015) 1–9
7
Table 2C Risk-adjusted returns for 52-week high momentum strategies (two-stage method). h 1
3
6
s
Winner
Loser
Winner − loser
Winner
Loser
Winner − loser
Winner
Loser
Winner − loser
0
0.42∗∗ (3.13) −0.63∗∗∗ (−4.884) −0.58∗∗∗ (−4.907) −0.47∗∗∗ (−4.288) −0.39∗∗∗ (−3.31) −0.25∗∗ (−2.11) −0.27∗∗ (−2.18) −0.28∗∗ (−2.22) −0.35∗∗∗ (−2.72) −0.33∗∗∗ (−2.63) 0.14 (1.537)
0.23 (0.45) −1.09∗∗ (−2.479) −1.08∗∗ (−2.551) −1.14∗∗ (−2.773) −1.26∗∗∗ (−2.92) −1.24∗∗∗ (−2.98) −1.19∗∗∗ (−2.96) −1.06∗∗∗ (−2.76) −0.98∗∗∗ (−2.73) −0.93∗∗∗ (−2.82) −0.06 (−0.739)
0.17 (0.30) 0.44 (.866) 0.49 (1.016) 0.65 (1.396) 0.85∗ (1.78) 0.97∗∗ (2.10) 0.90∗∗ (2.04) 0.77∗ (1.81) 0.61 (1.57) 0.58∗ (1.66) 0.19 (1.121)
0.12 (0.64) 0.03 (.193) 0.11 (.646) 0.21 (1.222) 0.03 (0.15) 0.06 (0.34) 0.04 (0.21) 0.00 (0.02) −0.01 (−0.07) −0.02 (−0.09) 0.16∗∗∗ (8.000)
−0.17 (−0.38) −0.25 (−0.663) −0.27 (−0.717) −0.28 (−0.771) −0.64 (−1.57) −0.58 (−1.49) −0.49 (−1.28) −0.39 (−1.04) −0.34 (−0.93) −0.29 (−0.82) −0.05∗ (−1.756)
0.26 (0.78) 0.27 (1.009) 0.36 (1.394) 0.47∗ (1.873) 0.65∗∗ (2.12) 0.63∗∗ (2.16) 0.51∗ (1.83) 0.37 (1.40) 0.31 (1.22) 0.26 (1.07) 0.19∗∗∗ (4.547)
0.14 (1.02) 0.18 (1.587) 0.24∗∗ (2.101) 0.28∗∗ (2.405) 0.11 (0.84) 0.12 (0.86) 0.10 (0.70) 0.07 (0.50) 0.05 (0.36) 0.05 (0.34) 0.08∗∗∗ (6.356)
−0.17 (−0.49) −0.14 (−0.504) −0.12 (−0.467) −0.09 (−0.333) −0.27 (−0.83) −0.21 (−0.67) −0.15 (−0.51) −0.08 (−0.26) −0.02 (−0.07) 0.01 (0.04) 0.03 (1.620)
0.29 (1.07) 0.3 (1.480) 0.35∗ (1.773) 0.34∗ (1.824) 0.36 (1.50) 0.30 (1.35) 0.23 (1.09) 0.13 (0.63) 0.05 (0.27) 0.02 (0.10) 0.02 (.826)
1 2 3 4 5 6 7 8 9 3−1
This table reports the average risk-adjusted return of monthly portfolio in the out-of-sample period from January 1997 through December 2013 for three different 52-week high momentum investing strategies. s denotes the number of months in the skip period, and h denotes the number of months in the holding period. The risk-adjusted returns are estimated from Fama–French three-factor model. The factor loadings of the Fama–French three factors are estimated from the in-sample period. The risk-adjusted returns in the out-of-sample period are then obtained from portfolio return minus expected return. The sample includes all stocks on CRSP; t-statistics are in parentheses. ∗ , ∗∗ , ∗∗∗ denote significance at the 10%, 5% and 1% level, respectively.
for portfolio i, Rmt, os is the monthly return of the market portfolio at time t in the out-of-sample period, SMLt, os is the size factor at time t in the out-of-sample period, and HMLt, os is the book to market factor at time t in the out-of-sample period. The results show that even after risk-adjustment, many of the portfolios remain statistically significant. For the three-month holding period, the two-month and three-month skip-period, self-financed portfolios provide positive risk-adjusted returns that are statistically significant at the 10% and 5% levels, respectively. For the six-month holding period, the two-month and three-month skip-period, winner portfolios and self-financed portfolios generate positive risk-adjusted returns that are statistically significant at the 5% level. For the three-month holding period, the three-month skip-period, winner portfolios and self-financed portfolios provide significantly higher returns than the conventional one-month skip-period portfolios, statistically significant at the 1% level. For the six-month holding period, the three-month skip period, winner portfolios give significantly higher returns than the conventional one-month skip-period portfolios, statistically significant at the 1% level. An alternative method of controlling for the effects of risk is to directly compute the risk-adjusted returns by subtracting the expected portfolio return computed from the Fama–French three factor model from the corresponding actual portfolio return. Table 2C presents the resulting risk adjusted returns using this method for the out-of-sample period. The numbers shown on Table 2C can be interpreted as the riskadjusted returns of the corresponding unadjusted returns of Table 2A. In detail, the risk adjustment process for Table 2C requires two steps. In the first step, we compute the factor loadings from the following Fama–French three factor model using in-sample data:
R pit,is − R f t,is = αi + βi [Rmt,is − R f t,is ] + si (SMLt,is ) + hi (HMLt,is )
(2)
where Rpit, is is the monthly return on the portfolio i at time t in the in-sample period, Rft, is is the one month Treasury Bill return at time t in the in-sample period, α i is the estimated regression intercept for Please cite this article as: A.-S. Chen, W. Yang, Echo effects and the returns from 52-week high strategies, Finance Research Letters (2015), http://dx.doi.org/10.1016/j.frl.2015.10.015
JID: FRL
ARTICLE IN PRESS[m1G;November 10, 2015;10:13]
8
A.-S. Chen, W. Yang / Finance Research Letters 000 (2015) 1–9
portfolio i computed using in-sample data, Rmt, is is the monthly return of the market portfolio at time t in the in-sample period, SMLt, is is the size factor at time t in the in-sample period, and HMLt, is is the book to market factor at time t in the in-sample period. In the second step, the expected return for portfolio i in the out-of-sample period is computed by combining the factor loadings computed in the first step from in-sample data with the out-of-sample values of the three Fama–French risk factors for each of the months in the out-of-sample period using the Fama–French three-factor model.
E[R] pii = R f t,os + βi [Rmt,os − R f t,os ] + si (SMLt,os ) + hi (HMLt,os )
(3)
where E[R]pii is the expect monthly return under the Fama–French three-factor model for portfolio i at time t in the out-of-sample period, Rft, os is the one month Treasury Bill return at time t in the out-ofsample period, Rmt, os is the monthly return of the market portfolio at time t in the out-of-sample period, SMLt, os is the size factor at time t in the out-of-sample period, and HMLt, os is the book to market factor at time t in the out-of-sample period. The risk-adjusted returns for each month in the out-of-sample period are then obtained by subtracting the expected portfolio return computed from Eq. (3) from the corresponding actual portfolio return.
Radj,it = R pit,os − E[R] pit
(4)
where Radj, it is the risk-adjusted monthly return of portfolio i at time t in the out-of-sample period, Rpit, os is the unadjusted monthly return of portfolio i at time t in the out-of-sample period, and E[R]pit is the expect monthly return under the Fama–French three-factor model for portfolio i at time t in the out-ofsample period. The overall pattern of the results shown in Table 2C are similar to those in Table 2B. The zero investment portfolios, for example, maintain the pattern of performance getting better as the length of the skip-period is increased. Moreover, both Tables 2C and 2B show statistical significance for nearly the same set of portfolios. In both tables the performance improvement, for example, between the three-month skip-period portfolios and the one-month skip-period portfolios was statistically significant for the threemonth holding period zero-investment portfolios, the three-month holding period winner portfolios, and the six-month holding period winner portfolios. In sum, the risk-adjusted returns of Table 2C confirm the results of Table 2B which was computed using a different method. The results in Tables 2B and 2C also show that risk-adjusted returns are negative for several portfolios in the one-month holding period (winner) column. One possible reason why the risk-adjusted returns for the zero-skip period portfolios are positive but negative for other one-month holding period portfolios relates to the previous argument that forming the zero-skip momentum portfolio may not be possible in the real world. That is, even though the speed of trading and dissemination of information has developed during the recent decades making it technically feasible to form the zero-skip momentum portfolio, in actual practice the vast majority of investors are still not doing it and; thus, these zero-skip portfolios are still not efficiently priced in the market, resulting in the positive risk-adjusted returns shown in our results. One practical implication of these results is that given many momentum-based products are long-only ETFs, the investor needs to be careful in investing in these products and attempt to determine whether or not the fund manager has the ability to establish the zero-skip momentum portfolio which has the positive risk-adjusted returns. 4. Conclusions This study is related to George and Hwang (2004) and Novy-Marx (2012). George and Hwang (2004) show that nearness to the 52-week high is positively associated with future stock returns. Novy-Marx (2012) shows the existence of an echo effect in momentum returns; increasing the length of the skipperiod between the time of portfolio formation and the time of portfolio purchase improves the performance of the momentum portfolios. Novy-Marx (2012) finds that a skip-period of 6 months generates the best improvement. In addition to George and Hwang (2004), other more recent papers that study the 52-week high strategy include Li and Yu (2012) on aggregate market returns, Liu et al. (2011) on international stock markets, and Sapp (2011) on mutual fund returns and cash flows, among others. None of these papers checks for the existence of the echo effect in the 52-week high portfolios. Please cite this article as: A.-S. Chen, W. Yang, Echo effects and the returns from 52-week high strategies, Finance Research Letters (2015), http://dx.doi.org/10.1016/j.frl.2015.10.015
JID: FRL
ARTICLE IN PRESS[m1G;November 10, 2015;10:13] A.-S. Chen, W. Yang / Finance Research Letters 000 (2015) 1–9
9
Our results show that the 52-week high strategy also exhibits an echo effect. However, the length of the skip-period for the echo effect in the 52-week high portfolios differs from the length of the skipperiod for the echo effect found by Novy-Marx (2012) in JT momentum portfolios. For the 52-week high portfolios, a skip-period of 3–6 months generates the best improvement; specifically, our results show that increasing the skip period between the date of portfolio formation and the date of portfolio purchase 3–6 months significantly improves performance in nearly all cases analyzed. The results give support to the notion that anchoring based momentum strategies such as the 52-week high strategy are distinct from return-based JT momentum strategies, with unique (echo) characteristics. References Arshanapalli, Bala, Fabozzi, FrankJ., Nelson, William, 2006. The value, size, and momentum spread during distressed economic periods. Financ. Res. Lett. 3, 244–252. Asem, Ebenezer, Tian, GloriaY., 2010. Market dynamics and momentum profits. J. Financ. Quant. Anal. 45, 1549–1562. Chordia, Tarun, Lakshmanan, Shivakumar, 2002. Momentum, business cycle, and time-varying expected returns. J. Financ. 57, 985– 1019. Cooper, Michael J., Gutierrez Jr, RobertoC., Hameed, Allaudeen, 2004. Market states and momentum. J. Financ. 59, 1345–1365. Fama, Eugene F., French, KennethR., 1996. Multifactor explanation of asset pricing anomalies. J. Financ. 51, 55–84. George, T., Hwang, C.Y., 2004. The 52-week high and momentum investing. J. Financ. 59, 2145–2176. Griffin, John M., Ji, Xiuqing, Spencer Martin, J., 2003. Momentum investing and business cycle risk: evidence from pole to pole. J. Financ. 58, 2515–2547. Goyal, Amit, Wahal, Sunil. Ismomentumanecho? J. Financ. Quant. Anal. (forthcoming). Hong, Xin, Jordan, BradfordD., Liu, MarkH., 2015. Industry information and the 52-week high effect. Pacific-Basin Financ. J. 32, 111– 130. Jegadeesh, Narasimhan, 1990. Evidence of predictable behavior of security returns. J. Financ. 45, 881–898. Jegadeesh, Narasimhan, Titman, Sheridan, 1993. Returns to buying winner’s and selling loser’s: implications for stock market efficiency. J. Financ. 48, 65–91. Jegadeesh, Narasimhan, Sheridan, Titman, 2001. Profitability of momentum strategies: an evaluation of alternative explanations. J. Financ. 56, 699–720. Lehmann, Bruce N., 1990. Fads, martingales, and market efficiency. Q. J. Econ. 105, 1–28. Li, Jun, Jianfeng, Yu, 2012. Investor attention, psychological anchors, and stock return predictability. J. Financ. Econ. 104, 401–419. Liu, Ming, Qianqiu, Liu, Tongshu, Ma, 2011. The 52-week high momentum strategy in international stock markets. J. Int. Money Financ. 30, 180–204. Novy-Marx, Robert, 2012, Is momentum really momentum? J. Financ. Econ. 103, 429–453. Sapp, Travis R.A., 2011. The 52-week high, momentum, and predicting mutual fund returns. Rev. Quant. Financ. Account. 37, 149–179.
Please cite this article as: A.-S. Chen, W. Yang, Echo effects and the returns from 52-week high strategies, Finance Research Letters (2015), http://dx.doi.org/10.1016/j.frl.2015.10.015