Académique Documents
Professionnel Documents
Culture Documents
Zi Ning*
Nicolas Gressis
September 2006
Abstract
1
I. INTRODUCTION
An extensive range of literature has documented that the stock returns are
predictable based on their past returns. Particularly, stock returns exhibit positive serial
Chan et al., 1999). Jegadeesh and Titman (1993, 2001) report that trading strategies of
buying past winners and selling past losers realize significant positive returns over the
markets (Rouwenhorst 1998; Griffin et al. 2002). For example, Rouwenhorst (1998)
months) return continuation from 1980 to 1995. A diversified portfolio of past medium-
term winners outperforms a portfolio of medium-term losers by more than 1 percent per
as the sources of momentum returns. To date, no measures of risk have been found that
found that a three-factor asset pricing model cannot explain the returns of the short-term
momentum but only the long-term reversal (Fama and French 1996; Grundy and Martin
2
Korajczyk & Sadka (2004) find that transaction costs, in the form of spreads and
price impacts of trades, reduce but do not fully eliminate the return persistence of past
winner stocks. Chordia and Shivakumar (2002) show that macroeconomic instruments
for measuring market conditions can explain a large portion of momentum profits. They
argue that inter-temporal variations in the macroeconomic factors, such as dividend yield,
default spread, term spread, and short-term interest rates, are the main sources of
momentum profits. However, Cooper et al. (2004) find that the macroeconomic
induced biases.
Additionally, Lee and Swaminathan (2000) show that trading volume plays a role
in the profits to momentum strategies. Grinblatt and Moskowitz (2003) conclude that tax
In recent years, several behavioral and cognitive biases theories have been
developed to jointly explain the short-run momentum in stock returns. Some claim that
momentum profits arise because of inherent biases in the way investors interpret
information (DeBondt & Thaler, 1985; Daniel et al.,1998; Hong & Stein, 1999).
Other authors claim that momentum in stock returns is related to the market’s
al., 1995; Chan et al., 1996, 1999). For instance, firms reporting unexpectedly high
earnings outperform firms reporting unexpectedly poor earnings. The market incorporates
the news in stock prices gradually, so prices exhibit predictable drifts. These drifts last for
3
up to a year (Chan et al., 1999). Barberis et al. (1998) also demonstrate that momentum
Contrary to the under-reaction theory, Daniel at al. (1998) report that the
momentum effect comes from the continuing overreaction of informed investors. When
the direction of the market is upwards, traders’ overconfidence is boosted. Their model
predicts that momentum profits are stronger following bull markets, which are attributed
Cooper et al. (2004) show that the profits from momentum strategies are tightly
linked to the state of the market. Overreactions become stronger following up markets
generating greater momentum in the short run. A momentum portfolio is profitable only
following periods of market gains, consistent with the overreaction models of Daniel et
Intuitively, momentum effects should become even stronger if the overall market
is overconfident, such as the unusual years of the burst of High-tech bubble. During that
period, investors in general (both informed and uninformed) should be overoptimistic and
over that period. Using data from 1982 to 2000, the findings of this study are consistent
Also, in previous studies, it is usually the case that stock behaviors are examined
this study categorizes the momentum strategies into those that are with and without
4
accelerating monthly returns within a quarter. Intuitively, momentum strategies with
accelerating monthly returns should convey more positive information for investors and
thus drive up the momentum returns. The evidence shows that this is exactly the case.
Using S&P 500 index as a benchmark, momentum strategies with accelerating monthly
returns are the obvious winners over the last sub-period, from 1996-2000. Yet, there is no
distinct difference in terms of returns among the four momentum strategies over the
period from 1982 to 1996 and S&P 500 index. It appears that an over-optimistic market
tends to drive up the momentum effects. Such findings indicate that psychological factor
description of data, sample, and methodology, the stock selection rules defined,
compared and contrasted. Section III documents the findings and analysis. Section IV
II. METHODOLOGY
RATIONALE
In the prior momentum studies, it is very common that all stocks are ranked into
deciles of stocks based on their past 3-month or 6-month rate of return compound return.
However, such ranking may not catch the essence of momentum in that stocks may
exhibit different degrees of momentum within the formation period. The traditional
momentum approach assumes that all stocks in the decile portfolio are homogeneous in
5
information. Intuitively, the patterns of momentum within the formation period should
Let’s illustrate above arguments graphically. Here, it is necessary to out that only
stocks with positive return in the past quarter are considered. Let the sequence of intra-
quarter security prices be P0, P1, P2 where P0 (P2) is the beginning (end) of quarter
price.
Figure 1. The possible patterns of monthly returns that produce the same quarterly return are illustrated.
Intuitively, patterns 2 and 3 do not reflect the momentum idea in terms of trend
continuation. Thus, although all three price patterns produce the same quarterly return,
investors are likely to have more preferences patters 1 over patterns 2 and 3 in that
pattern 1 denotes more positive information. In addition to positive ROR, the monthly
returns are all positive. Figure 2 provides a more detailed illustration of pattern 1.
6
Figure 2. Intra-formation period monthly return sub-patterns.
momentum investors. Thus, in figure 2, the first sub-pattern is likely more attractive than
the other sub-patterns. The investors will expect on average higher subsequent returns
from securities exhibiting this sub-pattern than from the others. The primary goal of this
paper is to find out if the portfolio consisting of stocks selected by such investment
strategy generates above the average returns in a bull market. An appealing feature of this
study is that it considers not only momentum upon formation period but also the intra-
quarterly changes of the price patterns within the formation period as applied to the
This study limits the analysis to winners alone, referring to those stocks with the
highest rate of return (ROR). The existing literature indicates that a larger share of the
abnormal returns (without trading costs) to the long/short strategy is due to the short
positions in past losers. Thus, before trading costs, winners-only investing strategy is
7
SAMPLE CONSTRUCTION
The stocks are selected and evaluated on a quarterly basis. The sample data comes
from the Center for Research in Security Prices (CRSP) database. Both the CRSP
monthly and quarterly returns files are used, which included all domestic, primary stocks
listed on the New York (NYSE), American (AMEX), and Nasdaq stock markets. All
stocks priced below $10 are excluded at the beginning of the holding period so as to
ensure that “the results are not driven primarily by small and illiquid stocks or by bid-ask
bounce” (Jegadeesh and Titman, 2001).The deletion of low-priced stocks also lower the
magnitude of the sample variability. The data extends from October 1982 to September
First, at the end of each quarter, all stocks are ranked in ascending order on the
basis of their compound returns in the past 3 months. Then, stocks are selected based on
the four selection rules describe below. The top ten stocks that meet the requirement for
each strategy are grouped into one of the four portfolios accordingly. Thus, a total of
forty stocks are selected in each quarter. The selection rules are not easily met
particularly for strategy D. Portfolios are rebalanced for each quarter. Very likely
another forty stocks are selected for the subsequent quarter. There are a total of 288
portfolios over the study period. Each portfolio is held for three months, following the
ranking quarter. The average mean of quarterly returns are calculated and reported for
8
STOCK SELECTION RULES
Four stock selection rules are discussed here, namely, strategies A, B, C and D.
A. Strategy A
According to this traditional momentum strategy, stocks are ranked from top to
bottom based on its past quarterly compound return. Mathematically, the strategy can
be expressed as follows
Max(1+Rt-1)(1+Rt-2)(1+Rt-3)
where Rt-i ( i=1,2,3) is the return on a stock in the past three months. Simply put,
portfolio A comprises the ten stocks with the largest ranking period returns.
B. Strategy B
For this specific strategy, inequality ratios imply that the chosen stock’s price
mathematically written as
Max(1+Rt-1)(1+Rt-2)(1+Rt-3)
subject to
1 + Rt − 1 1 + Rt − 2
Rt-1> 0, Rt-2> 0, Rt-3> 0 and > 1, >1
1 + Rt − 2 1 + Rt − 3
C. Strategy C
In this case, it is assumed that more recent return movements convey more
information than less recent ones. Hence, investors pay more attention to Rt-1 and Rt-2
9
than to Rt-3 and, who would prefer to select the stock that exhibits price acceleration
over the two most recent months. Mathematically, this strategy is shown as follows
Max(1+Rt-1)(1+Rt-2)(1+Rt-3)
subject to
1 + Rt − 1
Rt-1> 0, Rt-2> 0, Rt-3> 0 and >1
1 + Rt − 2
D. Strategy D
This strategy selects only stocks with increasing monthly price acceleration over the
past three months. It is the most difficult one to implement due to the strict selection
Max(1+Rt-1)(1+Rt-2)(1+Rt-3)
subject to
1 + R t − 1 1 + Rt − 2
Rt-1> 0, Rt-2> 0, Rt-3> 0 and > >1
1 + Rt − 2 1 + Rt − 3
After the first round of screening, stocks that meet the criteria of particular
strategy are put into the appropriate portfolio accordingly. It is found that the ROR for
some stocks are missing in the subsequent quarter. Thus, the prices and returns of the
delisted stocks are obtained from CRSP individually. Very likely, those companies have
been either merged by other companies or simply went bankrupt. For all strategies other
than Strategy A, if less than 10 stocks meet the requirements, a certain percentage of
money would be invested in 3-month U.S. treasury bills. This situation is most likely to
10
happen in strategy D. Lastly, the mean returns of strategy portfolios are calculated, in
which way comparisons can be made among the four momentum strategies.
III. RESULTS
RAW RETURNS
This section documents the returns of the strategy portfolio described in the
previous section. Table I reports the mean, or the subsequent “realized” quarterly returns
from following each of the four momentum strategies over the 72 periods. It is
emphasized that the decision on which stock to invest in is made every quarter based on
The basic assumption in all computations is that at the beginning of each quarter
studied the investor puts an equal amount of money, supposed $1 into each common
stock, under the assumption that all dividends are reinvested in the month paid.
Table I presents terminal value over the post-formation period, which shows the
evolution of wealth over the entire sample period and the sub-periods. The findings
suggest that almost all momentum strategies (with the exception of strategy C)
Particularly, the analysis is motivated by the fact that typical momentum strategy
with accelerating monthly returns (strategy B) prove to be very profitable over the 18-
year period. Supposed we put $1 at the beginning of holding period, we would have
received nearly $25 by the end of holding period, more than double of the returns from
11
Table I
The momentum strategy portfolios are formed based on 3-month lagged returns and held for 3 months. The
stocks are ranked in ascending order on the basis of 3-month lagged returns. The momentum strategy
portfolios are formed immediately after the lagged returns are measured for the purpose of portfolio
formation. The terminal value of $1 invested in portfolios for strategies A, B, C and D are presented in this
table. The sample period is October 1982 to September 2000.
Figure 3-9 visually illustrates the performance of momentum portfolios over the
18-year period, the two 9-year sub-periods and four 4.5-year sub-periods.
30
25
20
15
10
5
0
82
84
85
86
87
89
90
91
92
94
95
96
97
99
00
19
19
19
19
19
19
19
19
19
19
19
19
19
19
20
Time
Figure 3. The cumulative return for strategy A, B, C, D and S&P over the entire 18-year period.
12
Terminal Value of $1 Invested from 1982.10-1991.9
Terminal Value of $1 Invested
3.5
3
2.5
2
1.5
1
0.5
0
82
83
84
85
85
86
87
88
88
89
90
91
19
19
19
19
19
19
19
19
19
19
19
19
Time
Figure 4. The cumulative return for strategy A, B, C, D and S&P for the first 9-year sub-period.
14
12
10
8
6
4
2
0
91
92
92
93
93
94
94
95
95
96
96
97
97
98
98
99
19
19
19
19
19
19
19
19
19
19
19
19
19
19
19
19
Time
Figure 5. The cumulative return for strategy A, B, C, D and S&P for the second 9-year sub-period.
Figures 6-9 revisit the comparative wealth behavior of the momentum strategies
13
Terminal Value of $1 Invested From 1982.10-1987.3
3
Terminal Value of $1 Invested
2.5
1.5
0.5
0
82
83
83
84
84
85
85
86
86
19
19
19
19
19
19
19
19
19
Time
Figure 6. The cumulative return for strategy A, B, C, D and S&P for the first 4.5-year sub-period.
2
Terminal Value of $1 Invested
1.5
0.5
0
1987 1987 1988 1988 1989 1989 1990 1990
Time
Figure 7. The cumulative return for strategy A, B, C, D and S&P for the second 4.5-year sub-period.
14
Term inal Value of $1 Invested from 1991.10-1996.3
6
Terminal Value of $1 Invested
0
1991 1992 1993 1994 1994 1995
Time
Figure 8. The cumulative return for strategy A, B, C, D and S&P for the third 4.5-year sub-period.
5
Terminal Value of $1 Invested
0
1996 1996 1997 1997 1998 1998
Time
Figure 9. The cumulative return for strategy A, B, C, D and S&P for the fourth 4.5-year sub-period.
15
The sub-period evidence gives a better picture of the performances of momentum
over the entire sample period. However, over the first sub-period from 1982-1991, all
500 portfolio. Over the second sub-period from 1991-2000, strategy B and D are the clear
winners, which perform much better than strategy A, C and S&P 500 portfolio.
As the attention is drawn to the shorter time intervals, we find that it is over the
sub-period from 1996-2000 that has a significant influence on the terminal wealth of all
index.
Such findings are not coincident. In a no-load mutual fund study done by Akhbari
et al. (2006), similar portfolio construction strategies are employed. It is found that only
in the past few years of the 1990s, when the stock market bubble burst, did the
momentum strategy B clearly exhibit superior performance. Both evidences from the no-
load mutual funds and stocks confirm the hypothesis that momentum effects are at least
RISK-ADJUSTED RETURNS
The Jensen-alpha
The model that is being adopted to incorporate risk is the standard Sharpe-Lintner
Capital Asset Pricing Model (CAPM) (1967). The risk-adjusted returns are estimated as
16
R - F = α + β (M – F) + ε
where R is the return on the portfolios under consideration, M is the market index, F is
the risk-free rate, α is the excess stock return and ε is the residual rate of return. I
In evaluating the performance of the momentum strategies, the S&P 500 portfolio
is used as the benchmark. The related variables in this study are defined as: R is the
quarterly rate of return for portfolio A, B, C and D, F is the three-month U.S. Treasury
bill rate constructed from one-month bill rates, M is the S&P 500 portfolio rate of return.
The models are estimated by regressing the mean of returns for each holding
period for strategies A, B, C and D separately. Table IV shows the estimation results for
Table II
This table reports the risk-adjusted returns of momentum portfolios based on strategy A, B, C and D. This
table reports the intercepts from Jensen CAPM alpha. The sample period is October 1982 to September
2000. The t statistics are reported in parentheses.
level. For strategy B, it is highly significant at 5% level. All the βs are slightly negative,
17
Several reasons may explain the result. First, the stock market is extremely
volatile and heterogeneous than mutual funds. Second, stocks that exhibit momentum
may catch more attention from investors, creating more volatility. They are affected by
such short- term macro-economy variables such as interest rate, federal money dealing,
and fiscal policy that affect all securities as well as some internal indicators such as the
company’s profits and sales, day-by-day performance, and analyst report. As more
attention is being paid, investors may overreact to these factors. Lastly, the small sample
size might also be a reason. The strategy portfolios consisted of only 10 stocks out of
4,000 to 6,000 stocks in each quarter and the sample is not very typical; whereas, the
S&P 500 portfolio better represents the whole market performances considering its large
sample size. One major reason that limit us from constructing portfolios with more stocks
IV. CONCLUSION
Optimism is contagious. The late 1990s are certainly a period of over- optimism
in the US. General investors tend to overreact to positive information, such as stocks with
expectations.
18
The results indicate that the recent record of stock prices do project future prices
and produce generous profits over the 18-year period from 1982-2000. The more nuanced
patterns. The findings show that the year of 1996-2000 is a critical period that typical
momentum strategy performs best, when the whole market is experiencing “irrational
exuberance”.
aspect of momentum effect, which is consistent with the over-reaction models of Daniel
19
REFERENCES:
Barberis, Nicholas, Andrei Shleifer, and Robert Vishny, 1998, A model of investor
Bernard, Victor L., Jacob K. Thomas, and James Wahlen, 1995, Accounting-based stock
Chan, Louis K.C., Narasimhan Jegadeesh, and Josef Lakonishok, 1996, Momentum
Chan,Louis K.C., Narasimhan Jegadeesh, and Josef Lakonishok, 1999, The profitability
80-90.
Chordia, Tarun, and Lakshmanan Shivakumar, 2002, Momentum, business cycle, and
Cooper, Michael J., Roberto C. Gutierrez Jr.., and Allaudeen Hameed (2004). Market
Psychology and security market under- and overreactions, Journal of Finance 53,
1839-1886.
De Bondt, Werner F.M., and Richard Thaler, 1985, Does the stock market overreact?
20
Fama, Eugene F. & Kenneth R.French, 1996, Multifactor Explanations of Asset Pricing
Griffin, John M., Xiuqing Ji & J. Spencer Martin, 2002, Momentum Investing and
Business Cycle Risk: Evidence from Pole to Pole. Journal of Finance 58(6),
2515-2547.
Grinblatta, Mark & Tobias J. Moskowitz, 2004, Predicting Stock Price Movements from
Past Returns: the Role of Consistency and Tax-loss Selling, Journal of Financial
Grundy, Bruce D. & J. Spencer Martin, 2001, Understanding the Nature of the Risks and
14, 29-78.
2143-2184.
Jegadeesh, Narasimhan, and Sheridan Titman, 1993, Returns to buying winners and
48, 65-91.
720.
Jensen, Michael C., and George A. Benington, 1969, Random walks and technical
Jensen, Michael C., 1969, Risk, the pricing of capital assets, and the evaluation of
21
investment portfolios, Journal of Business 42, 167-247.
Korajczyk, Robert A. & Ronnie Sadka (2004). Are Momentum Profits Robust to Trading
Latane, Henry A., and Charles P. Jones, 1979, Standardized unexpected earnings 1971-
Lee, Charles M.C. & Bhaskaran Swaminathan, 2000, Price Momentum and Trading
Sharpe, William F., 1964, Capital asset prices: A theory of market equilibrium under
22