J-SCAD Asset Management
Project Summary





Objective
Method
Hypothesis
Results
Further Research
Objective
To determine if momentum is the best strategy for obtaining positive alpha in mid-cap
equities in U.S, can it be improved by using it in combination with other factors, or
perhaps it can be replaced by other factors that produce a better alpha than it can. Several
studies have documented the alpha of momentum strategies, such as several done by
Jegadeesh and Titman. Momentum strategists look at a strong price chart, rapid earnings
growth and recent positive changes in earnings growth forecast to make investment
decisions based on the philosophy that stocks that have already outperformed the market
over the past 12 months are likely to continue their winning ways, at least in the short
term. Limiting the stock picks to strong 12-month performers is crucial to momentum
strategies due to elimination of noise in the sample, and relative strength is the tool of
choice for gauging performance. Although momentum strategies can entail momentum
on periods as short as 3 months and longer, we chose a 12 month time period for our
project in order to focus on one momentum strategy and dig deeper into the analysis of
the results of this strategy. If we had more time we would have looked at momentum over
a 3 month and 6 month period to compare and contrast results of factors versus different
momentum time periods.
Method
Our objective is to find the best factors, through the best time frame, to fit each of the
following momentum categories. The factors that we initially estimated to be the best
were:
Category
Factor
Relative Strength
12 M Total Return L 1M
6 M Total Return
Earnings Momentum
3 yr EPS growth
1 yr EPS growth
Revenue and Rev growth
Revenue Growth Q/Q
3 y Revenue Growth/share
Return on Investments
ROE Growth
ROA Growth
Trading and Volume
% change in price y/y
Average Volume change * Average Price Change
Relative Strength: Measures how a stock has performed compared to the overall market
over a specified time period.
Earnings Momentum: Starts with fast earnings growth and positive earnings surprises. It
measures the increase in the per share growth rate from one reporting period to the next.
Revenue and Rev Growth: A fundamental category that looks for a high minimum sales
growth over a specific time period
Return on Investments and Equity: Momentum based on a company’s profitability
Trading and Volume: This momentum category rules out the most risky stocks that are
out of favor with the trading community
In our project, we mainly focused on mid-cap equities in the U.S. market (market cap =
$1B to $10B). Our total universe was narrowed to 2,322 stocks. We wanted to use the
mid-cap equity screen because we felt that 1) mid-cap stocks would have less likelihood
to have momentum caused by inclusion or exclusion from index funds (ex: S&P 500
inclusion can cause heavy buying causing artificial momentum) and 2) mid-cap stocks
are likely to have more pure momentum from strong fundamental factors compared to
large-cap stocks that can have momentum caused from the latest news which may or may
not really affect the fundamental or economic factors driving the results of the company.
One setback in using mid-cap equities could be that mid-cap stocks could be more
susceptible to artificial momentum caused by investors fraudulently bidding up stocks
causing small investors to chase the returns and produce artificial momentum.
In regards to time frame of our data, we used data from January 1985 to December 2005.
In regards to weighting, we looked at both equally weighted and value weighted
portfolios. For value-weighted, we scaled down our market-capped fractiles by 7% every
year we go back in time to account for inflation.
We then wanted to look at the momentum for mid-cap stocks and see if a momentum
strategy could be improved by adding additional factors (both momentum based and nonmomentum based) in the scoring of the stock. We chose to run three different scoring
scenarios to see what strategy would maximize alpha. Our first scenario was to score our
universe purely on one factor, momentum, and to score Quintile 1 as 5 and Quintile 5 as 5. Our second scenario was to run a hybrid momentum strategy giving the momentum
factor a very high weight (5) and a score of 1 to the other factors; we then ranked them as
5 for quintile 1 and a very low factor (-5) for quintile 5. We then included other factors
such as price to book, revision ratio, 3-year EPS growth, and SUE (surprise unexpected
earnings), which we found were good test factors when used individually, and give these
low weights of 1 and -1 for quintile 1 and quintile 5, respectively. Our third scenario was
to run the scoring test with weights based upon our subjective judgment of the factors.
In addition to these scenarios, we wanted to run univariate, bivariate, and multivariate
scoring screens to determine if there were factors that scored a better alpha than
momentum did.
Hypothesis
We hypothesized that alpha would be greater in scenario 2 when momentum was given
the highest weight and other factors were given relatively lower weights. We thought
that this would be an improvement over a pure momentum strategy because stocks with
highest momentum (quintile 1) would not purely win out in a screening, but it would
consider other factors attributing to momentum to make sure stocks with possible
artificial momentum were weeded out of the top quintile and stocks that should have
better results were removed from the lowest quintile. We also hypothesized that the
subjective judgment of factors would cause alpha to be slightly better in scenario 3 than
the hybrid momentum strategy of scenario 2 because we were able to weight factors we
thought were best more accurately into our scoring system. However, we were interested
to see if a more objective scoring system (scenario 2) was better than a subjective scoring
system (scenario 3).
Results
The main conclusion of our results is that momentum is a major contributing factor to
attaining high alpha in and beating the market if combined in a hybrid portfolio. Using
momentum as a sole factor will achieve high returns, but not as high as when combined
with other univariate factors.
In determining the factors we would use, we began by looking through the factors we
initially expected would be effective in determining momentum. We found through
univariate screening that there were three factors that best captured momentum, which
were 12 month total return, 3 year growth EPS, and 1 year price change, that had a good
alpha rating. We also found 3 other factors that also showed positive alpha results and
we could use later in our testing. Overall, the other factors that we used were 12-month
total return, 3-year EPS growth, 1-year price change, revision ratio, SUE and book value
to price.
Scenario 1: Momentum 5 v 0
We then began running three scenarios to determine which strategy would work best with
momentum. We started by running our pure-momentum strategy. Our test showed an
alpha of 0.87% per month for the pure momentum strategy with an equally weighted
platform. This was slightly higher than the 0.82% alpha for a value weighted portfolio.
Although beta was lower for quintile 1 than quintile 5, as was expected, the overall beta
for quintile 1 was higher than expected, at 1.07. However, we understood this by the fact
that a momentum strategy would have an expected higher return due to the fact that they
will normally have risks higher than a market portfolio.
Alpha & Beta (5 v 0)
1.50
1.00
0.50
Alpha
0.00
Beta
-0.50
1
2
3
4
5
6
Quintiles
Scenario 2 – Momentum 5 v 1
We then tested what we called a hybrid momentum strategy, which entailed using 12
month total return and scoring this with a +5 and -5 for quintile 1 and 5, respectively. We
then included five other factors that we found were good factors for alpha (two being
momentum based, but different from our main momentum factor, and 3 other
fundamental variables) and gave these factors a score of +1 and -1 for quintile 1 and 5,
respectively. This was done to see if we could make a modification to momentum that
could enhance the results that momentum brings.
Our results show that alpha is actually lower for the hybrid portfolio than for the original
momentum strategy and a lower overall return. It also has a smaller beta differential
(beta difference between quintile 1 and 5). However, there are other factors, like
information ratio, or the ratio of expected return to risk, as measured by standard
deviation, for the pure momentum strategies which is lower than the hybrid portfolio.
This ratio, which is a measure of portfolio performance, may prove that having
momentum may not be our sole determining factor is a good way of beating the market.
Alpha & Beta ( 5 v 1)
1.20
1.00
0.80
0.60
0.40
0.20
0.00
-0.20
-0.40
Alpha
Beta
1
2
3
4
5
Quintiles
Scenario 3 – Momentum 5 v 5
We then wanted to test the pure momentum strategy and the hybrid momentum strategy
against a scoring strategy that was more subjective to our judgment of the factors to test
the idea that momentum strategies are one of the best strategies for obtaining positive
alpha and can be improved only a small amount (as shown in our earlier test of
momentum versus hybrid momentum). We realized an alpha slightly higher than the
pure momentum portfolio but had more risk (based on beta). However, this portfolio
performed much better against the benchmark (about 24% in the hybrid portfolio versus
17% in the pure momentum portfolio).
Our scoring entailed the following factors:
Q1
Q5
12 month total return:
+5
-3
3 yr EPS growth
+5
-5
1 yr price momentum
+1
-1
Revision ratio
+5
-3
SUE
+5
-3
BV to P
+5
-3
Alpha vs Beta ( 5 v 5)
1.20
1.00
0.80
0.60
Alpha
0.40
Beta
0.20
0.00
1
2
3
4
5
Quntiles
Univariate Factors
Finally, we ran our model on the univariate factors we had chosen. Our results showed
that none of the factors, on its own, was able to capture the 0.93 alpha we got by the
hybrid, equally weighted portfolio. Although the 12-month total return factor was the
best univariate factor on its own, none of the factors was able to beat the benchmark at
24%, reach a 0.16 Sharpe ratio, and a 1.07 information ratio. See table below.
Univariate:BV to P
1.40
1.20
1.00
0.80
0.60
0.40
0.20
0.00
-0.20
Alpha
Beta
1
2
3
4
5
Quintiles
Factor Comparison (Quintile 1 less Quintile 5) => for instance, using Momentum 5 v 5,
Q1 has an alpha that is 0.93 greater than Q5)
Alpha
Momentum - 5 v 5
Equally Weighted
Value Weighted
Beta
Universe Information
Return
Ratio
%>
Bench
% > Up
Bench
%>Down Excess
Bench vs. Bench
0.93
0.87
(0.01)
(0.01)
0.92
0.85
1.07
1.02
20.24
16.67
23.46
17.90
14.44
14.44
0.92
0.85
0.82
(0.13)
0.69
0.56
11.90
12.96
10.00
0.66
Momentum - 5 v 0
Equally Weighted
Value Weighted
0.87
0.82
(0.15)
(0.14)
0.71
0.68
0.53
0.53
12.70
10.32
16.67
10.49
5.56
10.00
0.69
0.65
Univariate Factors
12 Mo TR
3 Yr Gr EPS
1 yr price chg
Revision Ratio
SUE
BV to P
0.87
0.15
0.81
0.65
0.25
0.55
(0.15)
0.22
(0.12)
0.02
0.14
(0.24)
0.72
0.31
0.68
0.67
0.35
0.41
0.54
0.35
0.51
0.79
0.46
0.51
13.10
4.55
11.51
11.90
6.12
3.57
17.28
7.41
15.43
14.20
11.73
(11.11)
5.56
(3.33)
4.44
7.78
(7.78)
30.00
0.69
0.34
0.66
0.67
0.37
0.36
Momentum - 5 v 1
Equally Weighted
Value Weighted
Table legend:
Momentum 5 v 5 (scenario 3, momentum with subjective factors) = Momentum given
score of +5 or -5 (for quintile 1 or 5, respectively), and other factors were scored
subjectively with a score up to +5 or -5 based on quintile 1 or 5, respectively(see
explanation above for subjective scoring weights)
Momentum 5 v 1 (scenario 2, hybrid momentum strategy) = Momentum given score of
+5 or -5 (for quintile 1 and 5, respectively), and other factors given a score of +1 or – 1
based on quintile 1 or 5, respectively
Momentum 5 v 0 (scenario 1, pure momentum strategy = Momentum given a score of
+5 or -5 for quintile 1 or 5, respectively with no other factor weightings
Heat map analysis
By running the heat map model, we find that in scenario 3 (Momentum 5 v 5), the
momentum strategy sends us a consistent signal: long quintile 1 and short quintile 5:
Year
1
2
Equal weighted
3
4
Value weighted
5
1
2
3
4
5
1985
145.2
138.1
132.4
111.5
124.8
140.2
135.9
132.7
116.9
129.1
1986
131.9
125.7
120.8
123.6
105.5
131.7
126.1
123.9
126.1
106.6
1987
105.1
104.9
96.1
100.2
95.5
105.1
104.0
96.3
105.6
97.1
1988
127.0
122.1
119.8
128.8
115.5
126.1
118.2
121.2
131.9
117.0
1989
141.2
133.6
125.7
113.8
114.8
146.1
132.5
126.6
120.4
117.1
1990
1991
95.7
168.2
90.8
145.4
95.1
139.4
89.5
144.7
83.6
129.7
96.4
163.2
89.2
143.4
91.2
136.6
94.0
141.7
80.5
130.2
1992
119.9
115.0
119.1
119.5
112.7
117.3
114.8
118.2
117.3
113.3
1993
132.3
123.1
118.1
122.2
116.1
127.1
121.7
118.6
121.0
117.4
1994
103.2
99.6
97.9
99.9
97.5
101.8
95.3
98.3
98.5
98.8
1995
138.5
132.2
126.0
125.7
123.2
135.6
124.4
131.1
126.6
127.5
1996
125.6
122.2
119.0
117.4
113.5
122.4
121.9
118.9
117.5
112.1
1997
127.5
127.7
123.7
124.4
119.0
128.8
133.6
123.2
122.7
117.5
1998
109.2
103.8
99.7
104.3
92.8
112.4
105.5
101.6
104.6
96.2
1999
148.5
120.5
112.0
123.6
116.0
160.4
125.1
118.0
124.2
110.9
2000
100.9
99.5
100.7
90.2
103.0
100.6
99.1
101.4
88.5
104.6
2001
93.5
97.4
98.5
108.3
106.0
91.8
94.1
99.4
101.4
100.7
2002
87.4
89.7
87.2
78.4
68.6
86.2
89.9
87.4
81.1
69.6
2003
145.7
144.3
146.9
143.7
154.4
144.2
142.8
146.2
141.3
152.5
2004
120.0
122.0
122.3
117.0
117.5
119.2
123.3
123.7
114.8
120.1
2005
114.9
114.4
109.3
106.3
102.7
117.8
117.1
113.1
108.3
105.3
One interesting finding in the momentum scenario is that while the portfolio of going
long quintile 1 and shorting quintile 5 in the equal-weighted portfolio gave a Sharpe ratio
higher than that of the benchmark SP50, the value-weighted portfolio resulted in a Sharpe
ratio lower than that of the benchmark.
Long-Short - EW
Long-Short - VW
Performance measure/
Long 1
Long 1,2
Long 1
Long 1,2
Benchmk
Summary Statistic
Short 5
Short 4,5
Short 5
Short 4,5
Portfolio
Annualized Average Return
15.0%
9.2%
9.6%
5.5%
13.6%
STD deviation of returns (annualized)
12.6%
7.6%
13.3%
7.5%
15.9%
1.19
1.21
0.72
0.73
85.1%
Sharpe Ratio
Market/
Another interesting finding is that in a bubble year like 1999, the strategy of longing
quintile 1 and shorting quintile 5 could not help us to beat the market. However, when the
bubble burst in the following years like 2001, such a strategy could help us suffer a less
loss than the market. Our understanding is: in a bubble year, almost all stock prices
increase, and shorting would reduce your returns while reducing your risks. But in a
bearish year like 2001, such a strategy could help save your money.
Further Research
Areas that we would like to research further are how differing time period momentum strategies compare to
other alpha strategies in a scoring system. We would also like to study if momentum results are affected by
company size.