VOLUME 37 NUMBER 4 www.iijpm.

com SUMMER 2011

The Voices of Influence | iijournals.com

 Momentum in Japan: The
Exception That Proves the Rule
CLIFFORD ASNESS

S
CLIFFORD ASNESS ince their power for choosing U.S. asset class, security type, and time (an eclectic
is the managing and stocks was documented in the early subset of work includes Asness, Liew, and Ste-
founding principal of AQR
to mid-1990s (Jegadeesh and Titman vens [1997], Fama and French [1998], Asness,
Capital Management, LLC,
in Greenwich, CT. [1993] and Asness [1994]), the success Moskowitz, and Pedersen [2010], De Groot,
[email protected] of momentum strategies has become one of Pang, and Swinkels [2010], Blitz and Van Vliet
the strongest empirical regularities in finance. [2008], and Okunev and White [2001]).
Momentum has joined size (Banz [1981]) and There is, however, one very notable
value (Rosenberg, Reid, and Lanstein [1985], exception. Quite a few authors have pointed
Fama and French [1992, 1993, 1996], and out that momentum is an empirical failure for
Lakonishok, Shleifer, and Vishny [1994]) as stock selection in Japan (Asness, Moskowitz,
one of the “big three” anomalies, or risk fac- and Pedersen [2010], Fama and French
tors, of modern investing. [2010], Griffin, Ji, and Martin [2003], and
Any empirically successful trading Rouwenhorst [1998]). In both academic and
strategy, meaning one that has produced practitioner circles this result causes quite a
significant positive returns in the past, may bit of angst as many worry about how large
be the result of one of three explanations: a blow this is to our overall confidence in
1) exposure to a priced risk, 2) some market momentum strategies.
inefficiency, or 3) simple data mining. While This article examines the failure of
journals are full of debate over the first two momentum in Japan and asks the very basic
possibilities, it is the third we are concerned question: Do we care? That is, how dam-
with here, in particular as it might apply to aging are the Japanese findings to a belief
momentum. We can never fully eliminate the that momentum strategies have a healthy ex
chance that any empirical result is caused by ante positive expected return outside and
random chance uncovered by data-mining inside Japan? Does this documented excep-
researchers. But this worry can be progres- tion prove the rule or expose it as f lawed?
sively minimized with successful out-of- Our central finding is that the results in
sample tests. Japan are no blow to believers in momentum,
Momentum and value have handily not even a glancing one. In fact, we argue
survived such tests. In the ensuing near two they are ultimately supportive. First, we
decades since they first came to the forefront document and confirm the basic finding—
of finance, both value and momentum strat- value and momentum work everywhere,
egies have shown consistent out-of-sample except for momentum in Japan. Next, we
success when examined across geography, show that given the success of both value and

SUMMER 2011 THE JOURNAL OF PORTFOLIO M ANAGEMENT

JPM-ASNESS.indd 67 7/13/11 9:55:48 AM

 momentum strategies around the world (we limit our- Moskowitz, and Pedersen [2010]. Each month we divide
selves to value- and momentum-based stock selection in each market’s stocks into three equal groups based on their
four major developed regions), the ex ante chance that value or momentum rankings. Momentum is defined
one strategy in one region has delivered poor results, as the past 12-month return on each security, skipping
akin to those for momentum in Japan, is quite high. In the most recent month’s return, which means to form
other words, the true p-value of the Japanese finding is a portfolio in January 2011 we use January–November
unimpressive.1 2010 returns, skipping December. Value is defined as
Furthermore, because value and momentum are book-to-market equity where book is updated with a
negatively correlated, we argue that it is difficult and 6-month lag to ensure the data are available for con-
unproductive to study one without including the other. struction, and market is the current price at the month’s
These two factors are best studied as a system. In this start.2,3 Security returns are then value weighted within
context, we find that an ex post Sharpe ratio optimizer each group. The spread in returns between the portfo-
that can invest in value and momentum would still invest lios representing the top and bottom third of securities
heavily in the Japanese momentum strategy, despite its captures the value and momentum premia within each
seemingly poor performance. Next, we explore fur- market and asset class.
ther the intuition behind the importance of the nega- We restrict our universe to the top 90% of market
tive correlation between value and momentum. Finally, capitalization in each of the four markets. This is a
using a three-factor model, we find that momentum fairly conservative large-capitalization restriction. For
in Japan is actually a success. The three-factor model example, in the U.S. for December 2010, we have 707
intercepts are significantly positive, and comparable to firms in our sample—midway between the large-cap
other countries—meaning net of value, size, and market Russell 1000 and the very large-cap S&P 500. We use
exposure—momentum strategies in Japan have added value weighting and a large-cap universe to ensure that
considerable return over this period. implementation drags on our gross-of-trading-cost
In sum, we find the univariate failure of momentum, results are small. In fact, our size restrictions would allow
which has occurred only in Japan, to be no blow to our the reasonable use of equal-weighted portfolios held
belief in the power of momentum strategies. The global long and short. We view the equal-weighted results as a
(and indeed within Japan) results for momentum, partic- robustness check, and they confirm (at higher individual
ularly when considered as a system with the well-known Sharpe ratios for each strategy) all the results presented in
value effect, are powerful and almost unscathed by the this article. Exhibit 1 reports the number of stocks, and
oft-noted weak results in Japan. various other statistics (all in USD), for the four regions
in our sample at the end of our sample period.
DATA DESCRIPTION We are left with eight key series that form the
core of this article: the returns to a large-capitalization
Factor construction and data sources follow those liquid dollar-balanced long–short value strategy (or
of Asness, Moskowitz, and Pedersen [2010]. Interna- factor) within each of the four regions, and the returns
tional equity returns are from DataStream and are aggre- to a similarly constructed momentum strategy (or factor)
gated across four regions: the U.S., the U.K., Europe within each of the four regions, all monthly from July
(excluding the U.K.), and Japan. We choose these four 1981 through December 2010. Note that these are all
regions to be economically meaningful (as opposed to,
say, having each non-U.K. European country enter
separately). The data cover the 29½-year period from EXHIBIT 1
July 1981 to December 2010, the longest period over Statistics for Sample (Year-end 2010)
which we have both value and momentum for all four
regions. More detail on the source of these returns and
their construction are provided by Asness, Moskowitz,
and Pedersen [2010].
We obtain the value and momentum portfolios
in each of these markets in the same manner as Asness,

MOMENTUM IN JAPAN : THE EXCEPTION THAT P ROVES THE RULE SUMMER 2011

JPM-ASNESS.indd 68 7/13/11 9:55:48 AM

 zero investment portfolios, so both return and excess As a final preliminary, in each of the regions we
return over cash are the same. look at a portfolio that invests 50% in that country’s
value long–short strategy and 50% in that country’s
THE BASIC EVIDENCE momentum long–short strategy, rebalanced to 50/50
monthly (note that momentum has a long history of
We start with the value strategy. Exhibit 2 details realizing higher volatility per dollar than value, there-
the results to the value long–short strategy in each of the fore, ex post, and perhaps predictably ex ante, this port-
four regions, and in the “All” region that puts equal dol- folio is actually slightly more momentum than value).
lars (rebalanced monthly) in each of the four. We report Each region, including Japan, turns in a statistically
the average annualized returns, annualized standard significant performance when we examine a portfolio
deviations, Sharpe ratios achieved, and the t-statistics of value and momentum; in fact, Japan’s performance is
(against a mean of zero) associated with those Sharpe considerably better than that of the U.S. The All region
ratios. is particularly strong because, of course, it is diversifying
Basically, a value strategy has worked around the by value and momentum and by region. In particular,
world. It is not statistically significant on its own in note that even though Japan has 50% of its dollars, and
the U.S. and is borderline in Europe,4 however, the All more than half its ex post risk in the 0.03 Sharpe ratio
region turns in a very strong 0.57 Sharpe ratio and 3.12 momentum strategy, its 50/50 portfolio Sharpe ratio is
t-statistic. Note that the Japan region turns in the stron- quite strong.
gest value results, even stronger than the more-diversi- Of course, in Japan, and each region, the 50/50
fied All region (the importance of this strong result for strategies are benefiting from the negative correlation of
value in Japan will be apparent later in the article). Now, value and momentum. These correlations are quite neg-
we will examine the momentum strategy. ative and are clearly driving much of the large benefits
All the momentum strategies are positive, but of including both value and momentum, as indicated by
momentum in Japan is far and away the weak sibling. It the greatly superior Sharpe ratios reported in Exhibit 4
achieves a Sharpe ratio of effectively zero for 29½ years.5 compared to those reported in Exhibits 2 and 3. Next,
The All region survives its exposure to weak Japan with in Exhibit 5, we list these negative correlations (calcu-
a Sharpe ratio of 0.38 and a t-statistic of 2.06 (note this lated using monthly returns) for each region and the
analysis includes, and the results survive, some very poor combined All region.
results for momentum in the spring of 2009).
Essentially, this f lat result in Japan is what many
have noted as the failure of the momentum strategy EXHIBIT 3
in Japan and what has caused much hand-wringing Momentum Strategy Results
in quantitative finance circles.6 Interestingly, value July 1981–December 2010
and momentum in the U.S., over the large-cap value-
weighted portfolios we study, are better but also unim-
pressive when viewed alone. They have not performed
as badly as momentum in Japan, however, and have not
attracted nearly the same attention.7

EXHIBIT 2 EXHIBIT 4
Value Strategy Results 50/50 Value/Momentum Strategy Results
July 1981–December 2010 July 1981–December 2010

SUMMER 2011 THE JOURNAL OF PORTFOLIO M ANAGEMENT

JPM-ASNESS.indd 69 7/13/11 9:55:49 AM

 EXHIBIT 5 Simulations Considering Value
Value/Momentum Correlation and Momentum as a System
July 1981–December 2010
As a forerunner to the rest of this article, which
views value and momentum as a system, we can ask
another question. We have noted that Japan’s value
strategy has historically delivered the highest Sharpe
ratio. Because value and momentum are negatively cor-
related, when one is ex post high, we should expect the
SIMULATIONS other to be ex post low. We next ask, in what fraction of
our simulations do the high and low Sharpe ratios occur
We start by asking a simple question. Because we within the same region? The answer is 30% of the time.
only observe ex post returns, what are the chances that If we replace our correlation matrix with a diagonal
the poor results for momentum in Japan are just bad matrix (preserving their different empirical volatilities),
luck? Also, can simulations shed light on our contention this probability falls to 15%. Given the negative cor-
that value and momentum are best viewed as a system? relations of value and momentum, it becomes far more
likely to see the high and low Sharpe ratios within one
Basic Simulations region, in simulation and in real life.
Finally, using our simulations one more time to
To begin, we fix the variance-covariance matrix of argue that value and momentum are a system, and given
the eight strategies (value and momentum in each of the that their negative correlations must be viewed together,
four regions) to match the ex post 29½-year historical we ask, how often, when any region’s value strategy
realization. Then, we fix the mean return for each of the realizes a Sharpe ratio of greater or equal to 0.73 ( Japan’s
eight strategies such that it yields an ex ante 0.35 Sharpe historical figure), do we see a momentum Sharpe ratio
ratio, which is the average realized Sharpe ratio across in that same region of less than or equal to 0.03? The
the eight strategies. We then draw 10,000 multivariate answer is 29% of the time. Conditional on the strong
normal runs for the eight strategies over 29½ years. We value returns in Japan, the almost nonexistent Japanese
make special use of two historical Sharpe ratios. The momentum returns are, again, very far from shocking.
historical Sharpe ratio of momentum in Japan is 0.03, The puzzle of why momentum is so weak in Japan can
the lowest we observed, and the historical Sharpe ratio be dismissed as random, as we did earlier in the article,
of value in Japan is 0.71, the highest we observed. or we can go further and rephrase it as completely unsur-
For a 0.35 Sharpe ratio strategy, viewed alone, to prising given the fact that value is so strong in Japan.9
realize ex post a 0.03 Sharpe ratio over 29½ years is a
−1.74 standard deviation event. Realizing less than or EX POST OPTIMAL PORTFOLIOS
equal to a 0.03 Sharpe ratio has a p-value of 4.1%. Even
viewed alone, ignoring the fact that it “has not worked Value and momentum’s highly negative correla-
in Japan” and focusing instead on the statistical signifi- tion suggests that these two factors are best studied as
cance of this failure, the Japanese momentum statistical a system. Toward that end, we now consider, in each
event is not particularly earthshaking.8 region, the optimal combination of value and momentum
The more relevant question is, assuming eight strat- strategies.
egies (i.e., value and momentum in all four regions and For each region, Exhibit 6 reports the Sharpe ratio
each with an ex ante Sharpe ratio of 0.35), what are the of the value strategy, the Sharpe ratio of putting 50%
chances that one or more would come in below a 0.03? into value and 50% into momentum rebalanced monthly,
Running this simulation, we find the answer is 26%, the weight put in value (the weight in momentum is
which is a p-value nobody would write a word about 100% minus the weight in value) by an optimizer tasked
(unless we count the present instance). In other words, with maximizing the realized Sharpe ratio (this is an
under the hypothesis that all eight strategies are equally in-sample exercise), and finally the Sharpe ratio of this
good, the Japanese results are not at all shocking. ex post optimal portfolio. The All region puts 25% of its

MOMENTUM IN JAPAN : THE EXCEPTION THAT P ROVES THE RULE SUMMER 2011

JPM-ASNESS.indd 70 7/13/11 9:55:49 AM

 EXHIBIT 6 failure? There the ex post optimal amount of value is
July 1981–December 2010 Ex Post Optimal Accounts 70%. The improvement in the Sharpe ratio is not the
of Value and Momentum more-than-double amount of the other countries in the
study, but it is nontrivial, moving from 0.71 to 0.88. Even
switching to the 50/50 portfolio barely nudges down–
from 0.71 to 0.65 the Sharpe ratio of that achieved by a
100% value allocation. The 50% value/50% momentum
allocation means well more than 50% of the ex post
volatility comes from momentum, because momentum
capital each month in each of the four regions’ ex post has realized considerably more ex post volatility value
optimal mix of value and momentum. per dollar in the region of the world where momentum
For clarity, we can elaborate using the U.S. as an delivers no return and where value is the strongest in the
example. The historical Sharpe ratio of a strategy that world. And yet, even in Japan, the 50/50 allocation is
goes long U.S. value stocks and shorts U.S. growth stocks, almost as good as an all value allocation. An optimizer
as described in the preceding data section, is 0.14. The with full foresight wants to put 30% into the “failed”
Sharpe ratio of a portfolio that invests half of its assets Japanese momentum strategy. Some failure!
each month into the value portfolio and the other half
into a similarly constructed momentum long–short port- FURTHER ANALYSIS OF CORRELATIONS
folio is 0.40. The weight an ex post optimizer chooses to
invest in value if not restricted to a 50/50 allocation, but Exhibit 7 shows the rolling 12-month return to
choosing the weights to maximize the realized Sharpe the Japanese value and momentum strategies. The nega-
ratio over the 1981–2010 period, is 54% (and implicitly tive correlation we observed earlier at the monthly level
46% in momentum). Finally, the Sharpe ratio of this ex clearly shows up at the annual level. Exhibit 8 shows
post optimal 54/46 portfolio is 0.41. the rolling 12-month return to the 50/50 portfolio of
For the U.S., the U.K., and Europe, the optimal Japanese value and momentum.
weight in value is between 54% and 55%. For these coun- The worst 12-month return for Japanese value is
tries, the Sharpe ratio of the optimal portfolio is generally −61%. When looking at these large numbers recall that
very close to that of the 50/50 portfolio and is more than they are additive spread returns, so a −50% followed by a
double the Sharpe ratio of a pure value portfolio. But −50% is −100%, not −75% like in compounded portfolios,
what about in Japan where momentum is a much heralded

EXHIBIT 7
Rolling 12-Month Value and Momentum in Japan

SUMMER 2011 THE JOURNAL OF PORTFOLIO M ANAGEMENT

JPM-ASNESS.indd 71 7/13/11 9:55:50 AM

 EXHIBIT 8
Rolling 12-Month 50/50 Portfolio of Value and Momentum in Japan

which generates some large numbers. The worst return The intuition for W2 is straightforward. We assume
for Japanese momentum, despite it having a lower average strategy 2 has a zero Sharpe ratio, so if it is also uncor-
return and higher volatility than value, is not as bad at related to the first positive Sharpe ratio asset, it is ignored
−49%. Of course, the real point is the 50/50 combination (held at zero weight). If it is positively correlated, it is held
strategy. The worst 12-month return for this portfolio is short as a hedging asset. Finally, if negatively correlated, it
a comparatively tiny −13%. Of course, this is simply an is held long, again as a hedging asset. Now, if ρ = −0.55,
illustration of the power of diversification and a −0.55 like it does for value and momentum in Japan, we get
correlation, but perhaps an edifying one. W2 = 35%. This basically reproduces our optimization
One way to view these results is that value has results, except for the zero assumption for SR2 versus the
been on a 29½-year tear in Japan. Japanese momentum whopping historical 0.03 for Japanese momentum, and
has been −0.55 correlated to value over this period. the fact that real-life value and momentum volatilities
For momentum to “only” make a tiny bit of money, are not perfectly equal. To repeat, the intuition is simple.
as opposed to losing quite a lot of money, means it is a A zero Sharpe ratio does not seem that impressive, until
valuable component in a diversified portfolio. we find that it is achieved with a −0.55 correlation to an
It is simple to show that if you have two investment asset with a positive Sharpe ratio. Then, it is very impres-
strategies with Sharpe ratios SR1 and SR2, and correla- sive indeed (obviously the package is more impressive the
tion ρ, the optimal weight in asset 1 is10 higher the positive Sharpe ratio and the more negative
the correlation) because it is an excellent hedge for a
SR1 ρ SR 2 strategy that already does well on its own.
W1=
(SR1 SR2 1 ρ)
SR 2)(1− The next section makes this idea, that momentum
has actually added a lot to the highly successful Japanese
Consider the simple case where SR1 is positive and value strategy, even more explicit.
SR2 is zero. Then the formula reduces to
THREE-FACTOR REGRESSION INTERCEPTS
1
W1=
(1− ρ) For quite a few years now, the standard way empir-
ical finance evaluates historical strategy performance is
−ρ
W 2 1 W1 = in the context of a factor model. The industry-standard
(1 − ρ) approach is the Fama–French three-factor model (Fama
and French [1993, 1996]), which regresses the historical

MOMENTUM IN JAPAN : THE EXCEPTION THAT P ROVES THE RULE SUMMER 2011

JPM-ASNESS.indd 72 7/13/11 9:55:50 AM

 EXHIBIT 9 chance that the much ballyhooed failure of momentum
Momentum Regressed on Three-Factor Model in Japan is just random noise.
of Market, Size, and Value, July 1981–December 2010 Furthermore, our defense of momentum in Japan
goes beyond simply citing statistical significance. For
the last 29½ years, a Sharpe ratio–maximizing Japanese
investor with access to the value strategy would actually
have been better off with a lot of momentum in his port-
folio. Finally, viewed as a system along with negatively
correlated value strategies and when using a version of the
Fama–French three-factor model, we can easily reject, at
standard significance levels, the hypothesis that momen-
tum’s success in Japan is the result of chance. Using the
three-factor model, momentum is not simply “not a
Note: The intercept is reported as an annualized percent.
failure” in Japan, but a strong empirical success.
Let us pause for a moment and note how at odds
excess returns in question on the excess returns of the
our argument is with the practical world of real-life
market, a long–short small-minus-large factor, and a
portfolio management. In practice, managers as well as
long–short value-minus-growth factor. The loadings
individual strategies that do not perform over, say, three
measure the strategy’s factor bets and the intercept mea-
to five years are at the least considered quite suspect and
sures the economic value of the strategy over making
often tossed on the scrap heap. Yet, here we argue in
the same factor bets with costless index funds. Exhibit 9
favor of a strategy that viewed alone has been inept for
presents the three-factor results for the momentum
almost 30 years. We argue for it based on the success of
strategy in each of the four regions regressed on market
quite similar strategies elsewhere, its power as a hedge
excess returns and on small-minus-large and value-
for another successful strategy, and an appreciation of the
minus-growth factors. We highlight the all-important
power of random chance. This does not mean that every
intercept t-statistics with a box.11,12
strategy or manager should be given a 30-year benefit of
The size of momentum’s intercept in Japan is eco-
the doubt. But it does illuminate how at odds real-world
nomically and statistically large (and is better than the
time horizons can be with time horizons that actually
intercept in the U.S. market). Quite simply, viewed
mathematically matter.
through the prism of modern finance, momentum has
To sum up simply, we show that when viewed
actually worked very well in Japan. Considerable perfor-
properly momentum works fine in Japan. Everyone
mance, both economically and statistically significant,
move along, nothing to worry about here.
has been generated by momentum net of the standard
three-factor adjustment.
ENDNOTES
CONCLUSION I would like to thank Dana M. D’Auria, Andrea
Frazzini, Jacques Friedman, Jared Kizer, Ronen Israel, John
If we hypothesize that momentum has similar pre- Liew, Toby Moskowitz, Fraser Murray, Lars Nielsen, Aldo
dictive power around the world, equal to the historical Pascarella, Lasse Pedersen, and Nathan Sosner for exception-
average of value and momentum over each of the four ally helpful comments, and Sarah Jiang for the same, and
regions we studied, then the ex post “failure” of Japanese superb research assistance.
1
momentum is not at all statistically impressive with a The p-value in this case is the probability we would
p-value of about 26%. Viewed as one of eight possibly find Japanese momentum to be as bad, or worse, than we do
efficacious strategies around the world, we cannot come if the true mean return for momentum in Japan was similar to
close to rejecting that momentum’s Sharpe ratio in Japan that found ex post in other countries; 5% or lower is a com-
is ex ante equal to momentum and value’s average Sharpe monly used level where something is called “significant.”
2
Asness and Frazzini [2011] studied the difference
ratio everywhere. In other words, there is a significant
between using updated market price to calculate book-to-
market ratios versus using a lagged price that matches the

SUMMER 2011 THE JOURNAL OF PORTFOLIO M ANAGEMENT

JPM-ASNESS.indd 73 7/13/11 9:55:51 AM

 timing of the book data. Versions of value that lag price, such Fama–French large-cap value strategy, which also looks weak
as the Fama–French HML, look more like 80% value and but positive over this period.
8
20% momentum than they do like pure value. Thus, using Similarly, the empirical (not simulation) t-statistic of the
the lagged value measures, it is harder to see the importance mean of Japanese momentum versus the equal-weighted average
of adding some momentum to value in Japan, because the of the other three momentum strategies is only –1.65, and
lagged value measures already add momentum to the strategy. Japanese momentum versus the equal-weighted average of the
The key simulation result in this article, that the true p-value other seven strategies (value and momentum) is only –1.34.
9
of momentum’s “failure” in Japan is not impressive, is not Of course, we can only show that random chance is a
affected by this choice. strong possibility, we cannot dismiss causality. For example,
3
We do lag the price by one day to avoid loading on Chui, Wei, and Titman [2000, 2010] offered explanations
the short-term contrarian strategy that is possibly an artifact for momentum’s weakness in Japan (and other parts of Asia)
of the bid–ask spread. Also, for the three-factor regressions that include momentum being weaker in countries with less
reported in Exhibit 9, we tested explicitly adding the one- individualism, and momentum being weaker in civil versus
month contrarian strategy to each regression and found the common law countries.
10
results essentially unchanged. This formula is for equal volatility strategies. It does not
4
Keep in mind, to be conservative, we use a very large-cap perfectly apply to our empirical results because the momentum
formulation where many anomalies are known to be weaker. portfolio is more volatile than the value portfolio, although
See Israel and Moskowitz [2011] for further discussion of these the intuition carries through quite well.
11
results, analyzing both value and momentum among different The value-minus-growth factor is the “value” factor
size stocks. of this article. The market factor is the cap-weighted return
5
For comparison with another study, the Fama and on our universe of stocks (the 90% of market cap we use
French [2010] Exhibit 1 presents the results for momentum for this article) over cash, and the small-minus-large factor
in Japan over only large-cap stocks (the analogue to what is constructed like momentum and value in this article but
we do in our study). They found a t-statistic of 0.32 over using market capitalization at the end of the prior month (and
239 months. That translates to a Sharpe ratio of 0.07 versus requiring each firm to have value and momentum factors).
our Sharpe ratio of 0.03. The time periods and construction Because we restrict our universe to large-capitalization stocks
methods are not precisely the same, but the similarity in result to begin with, our size factor is very weak.
12
is comforting. We are employing “local” versions of the three-
6
Results in Japan pre- and post-1990 are essentially the factor model. Fama and French [2010] discussed the differ-
same, making the dramatic change in the Japanese market ence between local and global models. An exception is that
from long-term bullish to bearish not a potential explanation our All region strategy in Exhibit 9 is analyzed using global
for momentum’s failure. versions of the RHS factors (equal-weighted averages of the
7
This is also perhaps due to value and momentum’s region factors); it is not just an average of other results in the
testability and success over a longer period in the U.S., and exhibit but a separate global test.
the fact that the Fama–French HML includes both smaller
stocks and a bit of the momentum strategy (see Asness and
REFERENCES
Frazzini [2011]) as compared to our value strategy that uses
up-to-date unlagged price. The Fama–French HML gener- Asness, C. “Variables That Explain Stock Returns.” Ph.D.
ates a 0.42 Sharpe ratio over this same period, but averages dissertation, University of Chicago, 1994.
value’s performance among large and small stocks. Among
large-cap stocks, such as those we use in this article (similar Asness, C., and A. Frazzini. “Lagging Value.” Working paper,
although not precisely the same cut), the Fama and French AQR Capital Management, 2011.
data actually yield a Sharpe ratio of only 0.10 for value over
the period, but 0.65 for small stocks (see the Ken French Asness, C., T. Moskowitz, and L. Pedersen. “Value and
website, which is our source for the data). In separate bivar- Momentum Everywhere.” Working paper, University of
iate regressions of this article’s U.S. value strategy on the Chicago and AQR Capital Management, 2010.
Fama–French large-cap and small-cap value strategies, we
obtain a 0.87 coefficient and a 17.7 t-statistic on the Fama– Asness, C., J. Liew, and R. Stevens. “Parallels between
French large-cap value strategy, and 0.05 and a 1.0 t-stat on the Cross-Sectional Predictability of Stock and Country
the Fama–French small-cap value strategy. In other words, Returns.” The Journal of Portfolio Management, Vol. 23, No. 3
as a reasonableness check, our results look very much like the (1997), pp. 79-87.

MOMENTUM IN JAPAN : THE EXCEPTION THAT P ROVES THE RULE SUMMER 2011

JPM-ASNESS.indd 74 7/13/11 9:55:52 AM

 Banz, R. “The Relationship between Return and Market ——. “Size, Value, and Momentum in International Stock
Value of Common Stocks.” Journal of Financial Economics, Returns.” Working paper, University of Chicago, 2010.
Vol. 9, No. 1 (1981), pp. 3-18.
Griffin, J., S. Ji, and S. Martin. “Momentum Investing and
Blitz, D., and P. Van Vliet. “Global Tactical Cross-Asset Business Cycle Risk: Evidence from Pole to Pole.” Journal of
Allocation: Applying Value and Momentum across Asset Finance, 58 (2003), pp. 2515-2547.
Classes.” The Journal of Portfolio Management, Vol. 45, No. 1
(2008), pp. 23-38. Israel, R., and T. Moskowitz. “The Role of Shorting, Firm
Size, and Time-on-Market Anomalies.” Working paper,
Chui, A., J. Wei, and S. Titman. “Momentum, Legal Systems AQR Capital Management, 2011.
and Ownership Structure: An Analysis of Asian Stock Mar-
kets.” Working paper, Hong Kong Polytechnic University, Jegadeesh, N., and S. Titman. “Returns to Buying Winners
2000. and Selling Losers: Implications for Stock Market Efficiency.”
Journal of Finance, Vol. 48, No. 1 (1993), pp. 65-91.
——. “Individualism and Momentum around the World.”
Journal of Finance, Vol. 65, No. 1 (2010), pp. 361-392. Lakonishok, J., A. Shleifer, and R. Vishny. “Contrarian
Investment, Extrapolation, and Risk.” Journal of Finance,
De Groot, W., J. Pang, and L. Swinkels. “Value and Momentum Vol. 49, No. 5 (1994), pp. 1541-1578.
in Frontier Emerging Markets.” Working paper, Robeco Asset
Management, 2010. Okunev, J., and D. White. “Do Momentum-Based Strate-
gies Still Work in Foreign Currency Markets?” Working
Fama, E., and K. French. “The Cross-Section of Expected paper, 2001.
Stock Returns.” Journal of Finance, Vol. 47, No. 2 (1992), pp.
427-465. Rosenberg, B., K. Reid, and R. Lanstein. “Persuasive Evi-
dence of Market Inefficiency.” The Journal of Portfolio Manage-
——. “Common Risk Factors in the Returns on Stocks and ment, 11 (1985), pp. 9-17.
Bonds.” Journal of Financial Economics, 33 (1993), pp. 3-56.
Rouwenhorst, G. “International Momentum Strategies.”
——. “Multifactor Explanations of Asset Pricing Anomalies.” Journal of Finance, Vol. 53, No. 1 (1998), pp. 267-284.
Journal of Finance, Vol. 51, No. 1 (1996), pp. 55-84.

——. “Value versus Growth: The International Evidence.” To order reprints of this article, please contact Dewey Palmieri
Journal of Finance, Vol. 53, No. 6 (1998), pp. 1975-1999. at dpalmieri@ iijournals.com or 212-224-3675.

SUMMER 2011 THE JOURNAL OF PORTFOLIO M ANAGEMENT

JPM-ASNESS.indd 75 7/13/11 9:55:52 AM