THE JOURNAL OF FINANCE • VOL LIII, NO.

1 • FEBRUARY 1998

International Momentum Strategies

K. GEERT ROUWENHORST*

ABSTRACT
International equity markets exhibit medium-term return continuation. Between
1980 and 1995 an internationally diversified portfolio of past medium-term Win-
ners outperforms a portfolio of medium-term Losers after correcting for risk by
more than 1 percent per month. Return continuation is present in all twelve sam-
ple countries and lasts on average for about one year. Return continuation is neg-
atively related to firm size, but is not limited to small firms. The international
momentum returns are correlated with those of the United States which suggests
that exposure to a common factor may drive the profitability of momentum strategies.

MANY PAPERS HAVE DOCUMENTED that average stock returns are related to past
performance. Jegadeesh and Titman (1993) document that over medium-
term horizons performance persists: firms with high returns over the past
three months to one year continue to outperform firms with low past returns
over the same period. By contrast, DeBondt and Thaler (1985, 1987) docu-
ment return reversals over longer horizons. Firms with poor three- to five-
year past performance earn higher average returns than firms that performed
well in the past. There has been an extensive literature on whether these
return patterns reflect an improper response by markets to information, or
whether they can be explained by market microstructure biases or by prop-
erly accounting for risk.1 Fama and French (1996) show that long-term re-
versals can be consistent with a multifactor model of returns, but their model
fails to explain medium-term performance continuation. Chan, Jegadeesh,
and Lakonishok (1996) find that medium-term return continuation can be
explained in part by underreaction to earnings information, but price mo-
mentum is not subsumed by earnings momentum.
Return reversal and continuation are only two of many patterns that em-
pirical researchers have uncovered using substantially the same database of
U.S. stocks. It can therefore not be ruled out that these apparent anomalies
are simply the outcome of an elaborate data snooping process. This paper is

*The author is at the Yale School of Management. I thank Ray Ball, Jennifer Conrad, Eugene
Fama, Narasimhan Jegadeesh, Ken French, N. Prabhala, Jake Thomas, Roberto Wessels, sem-
inar participants at Erasmus Universiteit Rotterdam, Katholieke Universiteit Leuven, the Uni-
versity of North Carolina, the University of Pennsylvania, René Stulz (the editor), and an
anonymous referee for their comments, and ARCAS-Wessels Roll Ross for kindly providing the
data. I acknowledge support from a Fred Frank Grant.
1
See for example Chan (1988), Ball and Kothari (1989), Ball, Kothari, and Shanken (1995),
Conrad and Kaul (1993, 1996), Chan, Hamao, and Lakonishok (1991), Lakonishok, Schleifer,
and Vishny (1994), and DeLong et al. (1990).

267
268 The Journal of Finance

an attempt to address this concern by studying return patterns in an inter-

national context. Although Asness, Liew, and Stevens (1996) and Richards
(1996) study return patterns across markets at the country index level, this
paper primarily focuses on international return continuation within markets
and across markets at the individual stock level using a sample of 2,190
stocks from 12 European countries in the period 1978 to 1995.2 Because of
the length of the sample period, the paper concentrates only on patterns in
medium-term returns. The sample period partly overlaps with the United
States samples of Jegadeesh and Titman (1993) and Fama and French (1996),
and is thus not strictly independent because of common factors in inter-
national markets. However, return continuation in the United States does
not seem to be related to common factors or conventional measures of risk.
If return continuation is absent in international markets or, when present,
can be rationalized using conventional measures of risk, this suggests that
the U.S. experience may simply have been unusual. Return continuation
that is common to many markets and cannot be accounted for by risk points
either toward a more serious misspecification of commonly used asset pric-
ing models or a general tendency of markets to underreact to information.
The main finding of the paper is that an internationally diversified rela-
tive strength portfolio that invests in medium-term Winners and sells past
medium-term Losers earns approximately 1 percent per month. This mo-
mentum in returns is not limited to a particular market, but is present in all
12 markets in the sample. It holds across size deciles, although return con-
tinuation is stronger for small than large firms. The outperformance lasts
for about one year, and cannot be attributed to conventional measures of
risk. In fact, controlling for market risk or exposure to a size factor increases
the abnormal performance of relative strength strategies. The paper, how-
ever, presents some evidence that European and U.S. momentum strategies
have a common component, which suggests that exposure to a common fac-
tor may drive the profitability of momentum strategies.
The remainder of the paper is organized as follows. Section I describes the
sample and documents the profitability of medium-term international mo-
mentum strategies. Section II shows that momentum is not restricted to
stocks of a particular country or size category. Section III examines whether
the returns to momentum strategies can be explained by conventional asset
pricing models. Section IV provides conclusions.

I. Returns of Relative Strength Portfolios

The sample consists of monthly total returns in local currency for 2,190
firms from 12 European countries from 1978 through 1995: Austria (60 firms),
Belgium (127), Denmark (60), France (427), Germany (228), Italy (223), The
Netherlands (101), Norway (71), Spain (111), Sweden (134), Switzerland (154),
and the United Kingdom (494). The sample covers 60 to 90 percent of each

2
Foerster, Prihar, and Schmitz (1995) provide evidence on momentum strategies in the Ca-
nadian market.
 International Momentum Strategies 269

country’s market capitalization.3 All returns are converted to deutsche marks

(DM) using exchange rate information taken from the Financial Times.
The relative strength portfolios are constructed as in Jegadeesh and Tit-
man (1993). At the end of each month, all stocks with a return history of at
least 12 months are ranked into deciles based on their past J-month return
(J equals 3, 6, 9, or 12) and assigned to one of ten relative strength portfolios
(1 equals lowest past performance, or “Loser”, 10 equals highest past per-
formance, or “Winner”). These portfolios are equally weighted at formation,
and held for K subsequent months (K equals 3, 6, 9, or 12 months) during
which time they are not rebalanced.4 The holding period exceeds the interval
over which return information is available (monthly), which creates an over-
lap in the holding period returns. The paper follows Jegadeesh and Titman
(1993) who report the monthly average return of K strategies, each starting
one month apart. This is equivalent to a composite portfolio in which each
month 10K of the holdings are revised. For example, toward the end of month
t the J 5 6, K 5 3 portfolio of Winners consists of three parts: a position
carried over from an investment of one DM at the end of month t 2 3 in the
10 percent of firms with highest prior six-month performance as of t 2 3,
and two similar positions resulting from a DM invested in the top-performing
firms at the end of months t 2 2 and t 2 1. At the end of month t, the first
of these holdings will be liquidated and replaced with a unit DM investment
in the stocks with highest six-month performance as of time t.
Table I presents the average monthly returns on these composite portfolio
strategies from 1980 to 1995.5 Panel A shows that an equally weighted port-
folio formed from the stocks in the bottom decile of previous three-month
performance returns 1.16 percent per month, 0.70 percent less than the top
decile portfolio, which returns 1.87 percent. For the three-month holding
period (K 5 3), the excess return from buying Winners and selling Losers
increases with the length of the return interval used for ranking ~J !. Ir-
respective of the interval used for ranking, average returns tend to fall for
longer holding periods. For each of the ranking and holding periods, how-
ever, past Winners outperformed past Losers by about 1 percent per month.
The returns range from 0.64 to 1.35 percent per month earned by portfolios
based on 12-month ranked returns held for 12 and 3 months respectively. All
excess returns in Panel A are significant at the 5 percent level.
The portfolios in Panel A are formed at the end of the performance ranking
period. Because bid-ask bounce can attenuate the continuation effect, Panel
B reports the average returns if the portfolio formation is delayed relative to

3
Although the sample is not comprehensive, and is biased to the larger firms in each mar-
ket, there is no selection bias in the sense that the data are not backfilled.
4
An exception arises when a stock is delisted. In that case the liquidating proceeds are
invested in the value-weighted Morgan Stanley Capital International (MSCI) index of the 12
countries in the sample. The conclusions of the paper are unchanged if the proceeds are re-
invested in the remaining stocks in the same decile portfolio.
5
Return data are available from 1978, but two years are lost due to performance ranking:
the J 5 12, K 5 12 strategy consists in part of positions taken 12 months ago based on prior
12-month performance.
 270
Table I
Returns of Relative Strength Portfolios
At the end of each month all stocks are ranked in ascending order based on previous J-month performance. The stocks in the bottom decile
(lowest previous performance) are assigned to the Loser portfolio, those in the top decile to the Winner portfolio. The portfolios are initially
equally weighted and held for K months. The table gives the average monthly buy-and-hold returns on these portfolios for the period 1980 to
1995. In Panel A the portfolios are formed immediately after ranking, in Panel B the portfolio formation occurs one month after the ranking takes
place. t-stat is the average return divided by its standard error.

Panel A Panel B

The Journal of Finance

Holding Period ~K ! Holding Period ~K !
Ranking Period
~J ! Portfolio 3 6 9 12 3 6 9 12

3 Loser 0.0116 0.0104 0.0108 0.0109 0.0077 0.0087 0.0094 0.0105

Winner 0.0187 0.0192 0.0190 0.0191 0.0185 0.0191 0.0190 0.0184
Winner 2 Loser 0.0070 0.0088 0.0082 0.0082 0.0109 0.0105 0.0095 0.0079
(t-stat) (2.59) (3.86) (4.08) (4.56) (4.29) (4.74) (4.99) (4.64)
6 Loser 0.0095 0.0090 0.0092 0.0104 0.0072 0.0076 0.0088 0.0106
Winner 0.0208 0.0206 0.0204 0.0195 0.0204 0.0205 0.0200 0.0187
Winner 2 Loser 0.0113 0.0116 0.0112 0.0091 0.0131 0.0128 0.0112 0.0081
(t-stat) (3.60) (4.02) (4.35) (3.94) (4.27) (4.59) (4.50) (3.62)
9 Loser 0.0088 0.0083 0.0097 0.0111 0.0064 0.0077 0.0095 0.0114
Winner 0.0212 0.0213 0.0204 0.0193 0.0209 0.0207 0.0197 0.0184
Winner 2 Loser 0.0124 0.0129 0.0107 0.0082 0.0145 0.0130 0.0102 0.0070
(t-stat) (3.71) (4.19) (3.78) (3.19) (4.50) (4.36) (3.77) (2.83)
12 Loser 0.0084 0.0094 0.0108 0.0121 0.0077 0.0093 0.0110 0.0125
Winner 0.0219 0.0209 0.0197 0.0185 0.0208 0.0198 0.0188 0.0176
Winner 2 Loser 0.0135 0.0115 0.0089 0.0064 0.0131 0.0105 0.0078 0.0051
(t-stat) (3.97) (3.66) (3.07) (2.40) (4.03) (3.48) (2.80) (1.98)
 International Momentum Strategies 271

the ranking by one month. For the shorter ranking and holding intervals,
delaying the portfolio formation indeed increases the payoff to buying Win-
ners and selling Losers. This increase is primarily due to a lower return to
the Loser portfolio.
Bid-ask bounce can also affect the measurement of the holding period re-
turns. Blume and Stambaugh (1983) show that long-term performance mea-
sures, obtained by averaging short-term returns over time, will be biased
upward due to measurement error in the returns and bid-ask bounce. This
bias affects the apparent profitability of momentum strategies because Los-
ers are on average smaller than Winners.6 In addition to the average monthly
return on K-month strategies given in Table I, I also compute the average
K-month holding period returns on the various strategies, and find the re-
sults to be very similar.
The remainder of the paper will concentrate on portfolios formed on the
basis of six-month ranked returns, formed at the end of the ranking period
and held for six months. Table II presents the summary statistics for the 10
decile portfolios of this strategy. The Average Return column shows that the
average performance of the decile portfolios is monotonically increasing in
previous six-month return. Higher past six-month return is on average as-
sociated with stronger future six-month performance. An F-test strongly re-
jects the equality of average returns of the relative strength portfolios. The
next column of Table II shows that the standard deviation of the decile port-
folios is u-shaped. The Winner and Loser portfolios have standard devia-
tions that are 30 and 40 percent higher than the portfolios in the middle
deciles. All else equal, stocks with higher standard deviations are more likely
to show unusual performance, and past unusual performance is cross-
sectionally correlated with volatility. The standard deviation of the excess
return of Winners over Losers is about 4 percent per month, which is similar
to the volatility of a long position in the middle decile portfolios. This indi-
cates that an “unrestricted” international momentum portfolio may not be
well-diversified. The MSCI b column shows that the excess return of Win-
ners over Losers is unlikely to be explained by its covariance with the mar-
ket. The sample average excess return on the market is about 0.6 percent
per month. For market risk to explain a continuation effect of 1.2 percent
per month would require, loosely speaking and ignoring standard errors,
that the beta of Winners exceeds the beta of Losers by about two. Instead,
both betas with respect to the value-weighted Morgan Stanley Capital In-
ternational (MSCI) index are close to unity, and the beta of the excess return
of Winners over Losers is insignificantly different from zero.7 The last col-
umn of Panel A reveals two interesting characteristics of the relative strength
portfolios. First, the average size of the Losers is smaller than the average size

6
For example, Conrad and Kaul (1993) and Ball, Kothari, and Wasley (1995) show that this
bias overstates the profitability of contrarian strategies.
7
Allowing for a delayed market response due to nonsynchronous trading does not change
these conclusions.
272 The Journal of Finance

Table II
Returns of Relative Strength Decile Portfolios
At the end of each month all stocks are ranked in ascending order based on previous six-month
performance. The stocks in the bottom decile (lowest previous performance) are assigned to the
Loser portfolio, those in the top decile to the Winner portfolio. The portfolios are initially equally
weighted and held for six months. The table gives the average monthly buy-and-hold returns
and standard deviations of the 10 portfolios for the period 1980 to 1995. The Morgan Stanley
Capital International (MSCI) b is the beta of the decile portfolio computed relative to the value-
weighted MSCI index of the 12 countries in the sample. The average size is the average natural
logarithm of the market value of equity of the stocks in the portfolio, computed in deutsche
marks. The F-statistic tests for equality of average returns of the 10 relative strength portfolios.

Prior Return Average Standard MSCI Average

Decile Return Deviation b Size
Loser 0.0090 0.0564 1.00 5.55
2 0.0096 0.0459 0.89 6.01
3 0.0101 0.0420 0.85 6.19
4 0.0112 0.0402 0.83 6.29
5 0.0114 0.0398 0.84 6.36
6 0.0125 0.0403 0.86 6.40
7 0.0135 0.0414 0.89 6.44
8 0.0144 0.0431 0.91 6.43
9 0.0165 0.0450 0.93 6.41
Winner 0.0206 0.0527 1.02 6.22
Winner 2 Loser 0.0116 0.0397 0.02
(t-stat) (4.02)
F 5 3.58 ~ p-value , 0.001!

of the Winners.8 Although Section III of the paper deals with risk-adjustment
in more detail, the fact that average returns are negatively related to firm size
suggests that size as a risk factor cannot explain the continuation effect. Sec-
ond, both Winners and Losers are on average smaller than the average firm
in the sample. This suggests that implementation of the Winners 2 Losers
(W 2 L) strategy may be difficult because it predominantly requires positions
in small stocks. The next section shows, however, that this is not the case.

II. Relative Strength Strategies That Control

for Country and Size
The relative strength portfolios in the previous section combine stocks from
12 national markets, some of which are larger in size than others. More than
half of the 2,190 stocks in the sample are from the United Kingdom (494), France

8
This size differential is in part a manifestation of the continuation effect, because the
J 5 6, K 5 6 relative strength portfolios at time t contain positions taken at time t 2 6. Of two
firms that have equal size but different past performance at time t 2 6, the firm with higher
past returns will at time t on average be larger than the firm with lower past returns because
performance persists.
 International Momentum Strategies 273

(427), or Germany (228). The average market capitalization of these firms is

larger than that of firms in the smaller European markets. This raises three
questions about the source and the pervasiveness of the continuation effect.
First, the continuation effect may be confined to only a subset of the 12 mar-
kets: either the three largest markets, which contribute the majority of sample
firms, or alternatively the smaller European markets, which contain relatively
many small and thinly traded issues. Second, no restrictions have been placed
on the geographical composition of the relative strength portfolios and the coun-
try weights vary over time. The continuation effect may therefore in part be
due to country momentum. It is interesting therefore to see to what extent the
continuation effect holds in individual countries, and is present in relative
strength portfolios that are country-neutral. Finally, because both the Winner
and Loser portfolios in Table II are tilted toward small stocks, I will examine
the influence of firm size on the returns to relative strength strategies.As pointed
out before, country membership and firm size are not independent, and I also
present results for portfolios that are both size- and country-neutral.

A. Relative Strength Portfolios by Country

Return decompositions by Heston and Rouwenhorst (1994) and Griffin and
Karolyi (1996) show the presence of large country-specific factors in inter-
national stock returns. Large country-specific shocks can potentially lead to
poor international diversification of the relative strength portfolios. For exam-
ple, a strong performance of German stocks relative to other markets will sub-
sequently cause the Winner portfolio to be overweighted in Germany relative
to the European equally weighted index. Similarly, the Loser portfolio will be
tilted toward stocks from markets with poor past performance. One possible
explanation for return continuation is that country-specific market perfor-
mance persists (Asness et al. (1996), Richards (1996)). However, if return con-
tinuation is primarily due to country momentum, controlling for the geographical
composition of relative strength portfolios should significantly reduce the av-
erage payoffs to buying Winners and selling Losers. If on the other hand medium-
term persistence reflects idiosyncratic firm performance, return continuation
will remain present in country-neutral relative strength portfolios as well.
Country-neutral relative strength portfolios are formed by ranking stocks
into deciles based on past performance relative only to stocks from the same
local market. The 10 percent of stocks from each country with lowest past
six-month return are assigned to the Loser portfolio, the top 10 percent to
the Winner portfolio. Except for integer constraints, the resulting decile port-
folios are well-diversified in the sense that they have the same country al-
location, and are country-neutral relative to the equally weighted index of
the 12 countries in the sample.9 Panel A of Table III shows that controlling
for country composition only slightly reduces the average excess return of

9
This is only approximately true. The relative strength portfolios consist of K separate hold-
ings, and each of these K positions is only country-neutral at origination. Because the positions
are not rebalanced over time they lose their equal weighting in subsequent periods, due to
performance differences and as securities are added to (or removed from) the sample.
274 The Journal of Finance

Table III
Returns of Relative Strength Portfolios that Control
for Country and Size
At the end of each month all stocks are ranked in ascending order based on previous six-month
performance, relative to other stocks in its country (Panel A), size decile (Panel B), or size-
country group (Panel C). The bottom decile of stocks are assigned to the Loser (L) portfolio, the
top decile to the Winner (W) portfolio. The portfolios are initially equally weighted and held for
six months. Each panel gives the average monthly buy-and-hold return and standard deviation
of an internationally diversified relative strength portfolio and its components for the period
1980 to 1995. The W 2 L excess returns for Austria, Denmark, and Norway in Panel A are
based on Winner and Loser quintile portfolios due the small number of firms in the sample. The
size assignments in Panel C correspond to the ranking of stocks in each country on size relative
to other stocks in that country: small (bottom 30 percent), medium (middle 40 percent), and
large (top 30 percent). t(mean) is the mean divided by its standard error.

Portfolio Mean Std. Dev. t(mean)

Panel A: Country-Neutral Momentum Strategies

All stocks (country-neutral) 0.0093 0.0239 5.36

By country:
Austria 0.0080 0.0498 2.23
Belgium 0.0110 0.0444 3.42
Denmark 0.0109 0.0478 3.16
France 0.0097 0.0496 2.72
Germany 0.0072 0.0395 2.52
Italy 0.0093 0.0508 2.53
Netherlands 0.0126 0.0497 3.51
Norway 0.0099 0.0658 2.09
Spain 0.0132 0.0801 2.28
Sweden 0.0016 0.0632 0.36
Switzerland 0.0064 0.0428 2.08
United Kingdom 0.0089 0.0408 3.02

Panel B: Size-Neutral Momentum Strategies

All stocks (size-neutral) 0.0117 0.0376 4.30

By size decile:
Smallest 0.0145 0.0588 3.42
2 0.0165 0.0542 4.21
3 0.0130 0.0495 3.64
4 0.0156 0.0455 4.75
5 0.0120 0.0409 4.04
6 0.0100 0.0453 3.04
7 0.0084 0.0463 2.51
8 0.0089 0.0451 2.73
9 0.0102 0.0479 2.96
Largest 0.0073 0.0473 2.13
 International Momentum Strategies 275

Table III—Continued

Portfolio Mean Std. Dev. t(mean)

Panel C: Size-Country-Neutral Momentum Strategies
All stocks 0.0085 0.0221 5.32
(size-country-neutral)
Size-neutral country portfolios:
France 0.0099 0.0463 2.94
Germany 0.0065 0.0373 2.40
UK 0.0087 0.0363 3.31
Other 0.0087 0.0236 5.07
Country-neutral size portfolios:
Small 0.0105 0.0304 4.79
Medium 0.0092 0.0249 5.09
Large 0.0055 0.0216 3.51

Winners over Losers (W 2 L) from 1.16 to 0.93 percent per month. This
suggests that country momentum is relatively unimportant for explaining
the continuation effect.10 The better diversification of the country-neutral
relative strength portfolios lowers the standard deviations of both the Win-
ner and Loser portfolios and increases their correlation from 0.74 to 0.88. As
a result, the standard deviation of the excess return falls from 3.97 to 2.39
percent per month, and the significance of the average excess return in-
creases (t 5 5.36).
The remainder of Panel A gives the W 2 L excess returns by country.
Winners have outperformed Losers in all 12 countries. In 11 countries the
W 2 L excess return has a t-statistic exceeding two, including the largest
markets of France, Germany, and the United Kingdom. Only in Sweden is
the excess return insignificantly different from zero. The strongest continu-
ation effect occurred in Spain, followed by The Netherlands, Belgium, and
Denmark. The standard deviations of the individual country excess returns
are about two to three times larger than the standard deviation of the in-
ternationally diversified momentum strategy. This implies that a large por-
tion of the W 2 L excess return variance is country-specific and can be
diversified internationally. The conclusion from Panel A is that return con-
tinuation is not due to country momentum. It is pervasive, and not restricted
to a few individual markets.11

10
This is consistent with the relatively weak momentum in country index returns reported
in Richards (1996), Bekaert et al. (1996), and Ferson and Harvey (1996).
11
I also perform a similar analysis of sector momentum, by constructing sector-neutral port-
folios based on assignments to 7 broad industry groups obtained from the Financial Times. The
returns on sector-neutral relative strength strategies were all positive, and significantly differ-
ent from zero for Basic Industries, Capital Goods, Consumer Goods, and Finance. For the En-
ergy, Transportation, and Utilities sectors, which contain relatively few stocks and hence are
poorly diversified, the equality of Winner and Loser returns could not be rejected.
276 The Journal of Finance

B. Size-Neutral Relative Strength Strategies

The unrestricted and country-neutral relative strength strategies in
Table II and in Panel A of Table III are not size-neutral in two respects.
First, Loser firms are on average smaller than firms in the Winner decile.
Because Winners are on average larger than Losers, a size effect may at-
tenuate the Winner 2 Loser effect. Second, both Winners and Losers are
smaller than the average firm in the sample. This raises the question whether
the continuation effect is only limited to smaller stocks.
To control for size I first sort all stocks based on size (market equity), and
within each size decile on past six-month return. The Loser portfolio con-
tains the 10 percent of firms with the lowest previous performance from
each size decile; the firms with the highest past return in each size decile
end up in the Winner portfolio. Both the Winner and the Loser portfolios will
therefore contain the same number of stocks from each size decile, and are
in that sense approximately size-neutral. Panel B of Table III shows that
after controlling for size, past Winners significantly outperform past Losers
by 1.17 percent per month (t 5 4.30). Moreover, return continuation exists in
all size deciles and is not limited to small stocks. However, there is a neg-
ative relation between firm size and the excess return of the relative strength
portfolios. Winners from the smallest size decile outperform the Losers on
average by 1.45 percent per month, with a standard deviation of 5.88 per-
cent. The excess return in the largest size decile is on average 0.73 percent
per month with standard deviation of 4.73 percent. The conclusion from Panel
B is that the continuation effect is not merely a reflection of firm size. Al-
though the continuation effect is stronger for smaller firms, past Winners
outperform Losers in every size category.

C. Size-Country-Neutral Relative Strength Portfolios

Although return continuation is present in many countries and across size
deciles, country membership and size are not independent. The country-
specific relative strength portfolios take significant size bets, and the size-
sorted relative strength portfolios take significant country bets. This section
explores the effectiveness of relative strength strategies that avoid taking
significant country and size positions, in order to separate the influence of
size and country membership.
The number of sample firms is not sufficient to construct 10 relative strength
portfolios for each size decile in every country, but a coarser sort can provide
information about the influence of size independent of country. Size-country-
neutral portfolios are formed by first sorting stocks by country into three
size groups: small (bottom 30 percent), medium (middle 40 percent), and
large (top 30 percent). Within each size-country group, stocks are ranked
into deciles based on past six-month performance. The size-country-neutral
Loser (Winner) portfolio contains the stocks from the lowest (highest) past
performance decile from each of the 36 country-size groups. Panel C of
Table III shows that an internationally diversified portfolio of Winners that
controls for country and size has outperformed Losers by 0.85 percent per
 International Momentum Strategies 277

month (t 5 5.32). The performance cannot be attributed to a particular geo-

graphical market. The size-neutral W 2 L excess returns are significantly
different from zero in the three largest markets in the sample (France, Ger-
many, and the United Kingdom) and comparable to the excess return on a
size-country-neutral W 2 L portfolio constructed of stocks from the other
9 markets.
Although Winners outperform Losers in each of the three size categories,
the excess return on the country-neutral W 2 L portfolio of small stocks is
about twice as large as the excess return on the W 2 L portfolio of large
stocks.12 Interestingly, the country-neutral W 2 L strategy of stocks from
the middle 40 percent of the size distribution has on average earned 0.92
percent per month, which is not significantly different from the 0.85 percent
earned on the overall size-country-neutral strategy (t 5 0.91). Thus, the
conclusion is that although return continuation varies by country and size,
profitability of international relative strength strategies does not require
investors to take significant size or country positions.

III. Risk-Adjusted Returns

A. Adjustment for Market and Size Factors

Panel A of Table IV confirms that the excess return on the unrestricted
relative strength strategy cannot be accounted for by a simple adjustment
for beta-risk, because the betas of the Winner and Loser portfolios are very
similar. The alphas of Losers and Winners are 20.27 percent (t 5 21.05)
and 0.88 percent (t 5 4.53) per month respectively, and their difference of
1.14 percent (t 5 3.94) is highly significant. Allowing for exposure to size, as
measured by an international version of Fama and French’s (1996) SMB
factor, increases the risk-adjusted return to 1.46 percent per month (t 5
5.05). Similarly to the U.S. experience, Losers are on average smaller than
Winners and load more on the international SMB factor.13 The size-country-
neutral W 2 L portfolio, however, shows a similar negative size exposure.
Unreported results show that all 10 size-sorted W 2 L portfolios summa-
rized in Panel B of Table III have negative loadings on the international
SMB portfolio. It suggests that Losers behave more like small stocks than
Winners irrespective of size. The overall conclusion from Table IV is that a
risk adjustment for market and size makes the continuation effect appear
more at odds with the joint hypothesis of market efficiency and the two-
factor model.

12
These size results are stronger than Jegadeesh and Titman (1993) and Asness (1995) find
for the United States, where relative strength portfolios of medium-sized firms outperform both
small and large firms.
13
The SMB portfolio is constructed by sorting the sample firms by country on size in each
month. Firms smaller than the median size in a country are assigned to the internationally
diversified S portfolio, the largest 50 percent to the B portfolio. SMB is the excess return of S
minus B. The average return and standard deviation of the international SMB portfolio are
0.29 and 1.16 percent per month from 1980 through 1995.
278 The Journal of Finance

Table IV
Risk-Adjusted Excess Returns
The table gives the results from regressing the monthly returns of Loser and Winner portfolios
in excess of the deutsche mark risk free rate on the excess return on the value weighted Morgan
Stanley Capital International index of the twelve sample countries over the deutsche mark risk
free rate, Rm,t 2 r f,t , and the excess return on an internationally diversified portfolio of small
stocks over a portfolio of large stocks, SMBt :

Ri, t 2 r f, t 5 a 1 b@Rm, t 2 r f, t # 1 gSMBt 1 ei, t .

SMB is constructed by ranking all stocks in each country in ascending order on market equity.
The stocks below the median size in a country end up in the international portfolio of S, the
stocks above the median in B. The relative strength portfolios in Panel A are formed based on
past performance only, the Winner and Loser portfolios in Panel B are constrained to have a
similar size and country composition. R 2 is the coefficient of determination adjusted for degrees
of freedom and t~{! is the coefficient divided by its standard error.

Portfolio a t~a! b t~ b! g t~g! R2

Panel A: Unrestricted Relative Strength Portfolios

Loser 20.0027 21.05 1.00 17.77 0.62

Winner 0.0088 4.53 1.02 23.76 0.75
Winner–Loser 0.0114 3.94 0.02 0.33 0.00
Loser 20.0090 24.52 1.08 25.10 2.00 12.01 0.78
Winner 0.0056 3.09 1.06 26.95 1.00 6.56 0.79
Winner–Loser 0.0146 5.05 20.02 20.30 21.00 24.13 0.07

Panel B: Size-Country-Neutral Relative Strength Portfolios

Loser 20.0027 21.58 0.98 25.4 0.77
Winner 0.0062 4.81 0.92 32.12 0.84
Winner–Loser 0.0089 5.55 20.06 21.73 0.01
Loser 20.0074 25.81 1.04 37.57 1.47 13.79 0.89
Winner 0.0036 3.18 0.95 38.66 0.81 8.59 0.89
Winner–Loser 0.0110 7.00 20.09 22.57 20.65 24.98 0.12

Chan (1988) and DeBondt and Thaler (1987) find that abnormal returns
associated with long-term return reversal strategies disappear once betas
are allowed to vary with market conditions. For the continuation effect to be
consistent with market-dependent betas requires that Losers have a higher
beta in down markets than Winners, and a lower beta in up markets.
Table V shows that empirically the opposite is true. Although the betas do
vary with market conditions, Losers uniformly have a higher beta in up
markets and a lower beta in down markets than Winners, which makes the
alphas appear more anomalous. As a consequence, the beta of the W 2 L
excess returns are significantly negative in up markets and positive in down
markets. The resulting alphas are 1.41 and 1.99 percent per month respec-
tively for the size-country-neutral and the unrestricted W 2 L portfolios.
 International Momentum Strategies 279

Table V
Market Dependent Risk-Adjusted Returns
The table gives the results from regressing the monthly returns of Loser and Winner portfolios
in excess of the deutsche mark risk free rate on the excess return on the value weighted Morgan
Stanley Capital International (MSCI) index of the twelve sample countries, Rm,t :

Ri, t 2 r f, t 5 a 1 b 1Dt @Rm, t 2 r f, t # 1 b 2~1 2 Dt !@Rm, t 2 r f, t # 1 ei, t .

Dt is a dummy variable that is one if the MSCI return is positive in month t and zero otherwise.
The relative strength portfolios in Panel A are formed based on past performance only, the
Winner and Loser portfolios in Panel B are constrained to have a similar size and country
composition. R 2 is the coefficient of determination adjusted for degrees of freedom, and t~{! is
the coefficient divided by its standard error.

Portfolio a t~a! b1 t~ b 1! b2 t~ b 2! R2

Panel A: Unrestricted Relative Strength Portfolios

Loser 20.0065 21.69 1.13 10.24 0.90 9.58 0.62

Winner 0.0134 4.57 0.87 10.46 1.14 16.00 0.75
Winner 2 Loser 0.0199 4.56 20.25 22.04 0.24 2.26 0.02
Panel B: Size-Country-Neutral Relative Strength Portfolios
Loser 20.0044 21.65 1.03 13.68 0.94 14.51 0.77
Winner 0.0098 5.05 0.80 14.52 1.01 21.43 0.85
Winner 2 Loser 0.0141 5.88 20.23 23.37 0.07 1.26 0.05

B. Relative Strength Strategies in Event Time

As noted earlier, the return on the ~J, K ! relative strength portfolio at
time t is determined by the payoffs to K separate positions put into place at
times t 2 1 through t 2 K, with each position based on past J-month per-
formance rankings at those times. In this Section I look at the performance
of each of these components in event time: what is the average excess return
on buying Winners and selling Losers (J 5 6) in the kth month after the
strategy is put into place? This provides information about the duration of
the continuation effect, as well as the extent to which it is permanent.
Table VI gives the monthly average excess return (W 2 L) in the first two
years after portfolio formation, both before and after risk adjustment. The
raw excess returns are uniformly positive in the first 11 months after port-
folio formation, after which time they turn negative. The risk-adjusted ex-
cess returns are significantly positive in the first 11 months after portfolio
formation. There is some indication of time variation in the risk exposure of
the event time portfolios, but it is not sufficient to explain the excess re-
turns. In fact, all event time portfolios have negative loadings on the SMB
factor, which tends to increase the abnormal returns relative to the raw
excess returns. The sample average risk premium of SMB is 0.29 percent per
month, which is about half the sample average excess return of the market
factor of 0.62 percent per month. Because the absolute value of the loadings
280 The Journal of Finance

Table VI
Relative Strength Excess Returns in Event Time
The table reports the results of regressing the monthly excess returns of a portfolio of Win-
ners 2 Losers (W 2 L), formed by ranking stocks on six-month past performance, in the kth
month after portfolio formation on the excess return on the value weighted Morgan Stanley
Capital International index of the 12 sample countries over the deutsche mark risk free rate,
Rm,t 2 r f,t , and the excess return on an internationally diversified portfolio of small stocks over
a portfolio of large stocks, SMBt :

Wk, t 2 L k, t 5 ak 1 bk @Rm, t 2 r f, t # 1 gkSMBt 1 ek, t .

R 2 is the coefficient of determination, adjusted for degrees of freedom, and t~{! is the point
estimate divided by its standard error.

mean
k ~Wk 2 L k ! t ~mean! ak t~ak ! bk t~ bk ! gk t~gk ! R2
1 0.0072 1.94 0.0120 3.13 20.14 21.77 21.14 23.35 0.06
2 0.0136 4.07 0.0181 5.26 20.06 20.83 21.14 23.94 0.07
3 0.0153 4.60 0.0194 5.68 20.01 20.09 21.11 23.92 0.07
4 0.0125 3.84 0.0162 4.86 0.05 0.73 21.08 23.89 0.08
5 0.0106 3.28 0.0141 4.26 0.05 0.76 21.02 23.72 0.07
6 0.0127 4.10 0.0149 4.62 0.10 1.53 20.74 22.78 0.05
7 0.0143 4.82 0.0153 4.96 0.14 2.22 20.49 21.93 0.04
8 0.0102 3.52 0.0114 3.81 0.13 2.02 20.55 22.24 0.05
9 0.0092 3.37 0.0106 3.73 0.10 1.62 20.54 22.32 0.04
10 0.0062 2.45 0.0085 3.29 0.05 0.87 20.72 23.37 0.06
11 0.0035 1.35 0.0061 2.32 0.05 0.82 20.84 23.89 0.08
12 20.0006 20.25 0.0031 1.25 0.01 0.20 21.07 25.12 0.12
13 20.0052 22.03 20.0019 20.73 0.05 0.99 21.02 24.80 0.12
14 20.0046 21.68 20.0011 20.40 0.07 1.27 21.10 24.86 0.13
15 20.0059 22.17 20.0027 21.02 0.09 1.54 21.04 24.61 0.12
16 20.0071 22.58 20.0041 21.53 0.10 1.80 21.03 24.56 0.12
17 20.0059 22.24 20.0036 21.39 0.09 1.60 20.81 23.70 0.08
18 20.0018 20.73 20.0005 20.21 0.12 2.27 20.61 22.85 0.07
19 0.0010 0.41 0.0025 1.00 0.07 1.31 20.57 22.68 0.04
20 20.0009 20.38 20.0006 20.25 0.11 2.12 20.29 21.39 0.03
21 20.0044 21.95 20.0038 21.61 0.04 0.73 20.25 21.27 0.00
22 20.0035 21.60 20.0021 20.95 20.04 20.85 20.35 21.84 0.01
23 20.0034 21.50 20.0016 20.69 20.05 21.09 20.46 22.38 0.02
24 20.0043 21.80 20.0022 20.91 20.06 21.07 20.57 22.82 0.03

on SMB is more than twice as large as the market factor loadings, the SMB
factor dominates the risk correction of the raw returns. The excess returns
turn negative in the second year after portfolio formation, although the ab-
normal returns are never significant. This does suggest, however, that part
of the continuation effect may be temporary and is reversed in the second
year after portfolio formation. These results are strikingly similar to the
results of Jegadeesh and Titman (1993) for the U.S. market. They also re-
port significant raw excess returns in months 2 through 10, although the
return reversal for the U.S. market in the second year is somewhat less
pronounced than in our European sample.
 International Momentum Strategies 281

Figure 1. Cumulative payoff to momentum strategies in event time. The solid line gives
the average cumulative payoff to a buy-and-hold strategy that invest a deutsche mark (DM) in
a portfolio of Winners financed by a unit DM portfolio of Losers, in the kth month after portfolio
formation. The payoff is measured in pfennigs (equals 0.01 DM). At the time of formation, the
Winners and Losers are equally weighted portfolios constructed to be both size- and country-
neutral. They contain from each of the 36 size-country groups the top and bottom decile of
stocks ranked in ascending order based on past six-month return. The dashed lines give the
95 percent confidence interval of the average payoff, computed using autocorrelation consistent
standard errors.

Figure 1 presents the evolution of the cumulative payoff to buying Win-

ners and selling Losers in event time. Both portfolios are size-country-
neutral. The solid line is computed as the average difference between the
K-period buy-and-hold returns of the long and short positions, and is free
from the potential bias induced by summing short-term returns to obtain
long-term performance measures. The dashed lines mark a 95 percent con-
fidence interval for the average payoff, using standard errors that take into
account the autocorrelation of the payoffs. The size-country-neutral relative
strength strategy has on average a significantly positive payoff up to 24
months after portfolio formation. The payoff initially peaks 12 months after
formation at 11.54 pfennig per DM invested in the long position, after which
it stays mostly flat.
Figure 1 can also be used to assess the profitability of momentum strategy
after transactions costs. Because the sample focuses on the larger and more
liquid stocks in the European market, transactions costs for a single round-
trip are typically below 1 percent. This would imply round-trip transactions
282 The Journal of Finance

costs below 2 percent or about 2 pfennig for buying the Winner and selling
the Loser portfolios. Figure 1 shows that the payoff to the size-country-
neutral strategy significantly exceeds a 2 pfennig transaction cost for hold-
ing periods between 4 and 24 months, and transactions costs of 4 pfennig for
holding periods between 7 and 24 months.

C. Are There Common Components among European

and U.S. Momentum Strategies?
Part of the motivation of this paper is that a sample of international firms
can provide “independent” evidence about the profitability of momentum strat-
egies. However, the similarity between the European and U.S. findings does
not directly address the question of independence. Jegadeesh and Titman
(1993) conclude that the profitability of momentum strategies in the United
States cannot be attributed to contemporaneous or delayed stock price re-
sponses to common factors, but is consistent with a delayed price reaction to
firm-specific information. If momentum returns only reflect a delayed price
response to firm-specific information, the standard deviation of inter-
national momentum strategies that simultaneously buy and sell more than
200 stocks should be very small. The fact that the country-neutral European
W 2 L portfolio in this paper and the U.S. W 2 L portfolio in Jegadeesh and
Titman (1993) have standard deviations of 2.4 and 3.1 percent per month
indicates that both strategies are not perfectly diversified. It is therefore
quite conceivable that momentum (W 2 L) returns have common compo-
nents across markets.
A preliminary answer to this question can be obtained by examining the
correlation between European and U.S. momentum returns, and evaluating
the profitability of the European momentum strategy conditional on the U.S.
experience.14 The sample correlation between the country-neutral European
and U.S. momentum returns, cor(W 2 LEUR , W 2 LUS ), is 0.43 over the 1980
to 1995 period, indicating strong positive dependence across markets.15 A
regression of W 2 LEUR on W 2 LUS can be used to evaluate the profitability
of the European strategy conditional on the U.S. experience:

W 2 L EUR, t 5 0.0065 1 0.222 W 2 L US, t 1 et , R 2 5 0.19,

~4.04! ~6.62!

where t-statistics are given in parentheses. Assuming joint normality, the

intercept of this regression measures the average excess return of the com-
ponent of the European momentum portfolio which is independent of U.S.
momentum returns. Conditioning on the United States reduces the average
excess return of the European momentum portfolio from 0.93 (Table III, Panel

14
I am grateful to the referee for suggesting this point.
15
I construct the ~J 5 6, K 5 6! buy-and-hold U.S. momentum (W 2 L) portfolio using all
available NYSE and AMEX firms on CRSP in the same way as the European W 2 L portfolio.
The sample average return and standard deviation of the U.S. momentum portfolio are 1.24
and 4.65 percent per month from 1980 through 1995.
 International Momentum Strategies 283

A) to 0.65 percent per month, but the high t-statistic of the intercept implies
profitability of European momentum strategies that is independent of a com-
mon component with the United States. In this sense the European sample
provides independent evidence of profitability of momentum strategies. Al-
though these results can be consistent with the presence of a “momentum
factor” in returns, the dependence can also be due to non-zero exposures to
other common priced risk factors (such as SMB), common unpriced factors
(industry factors), or a combination of both. A more detailed analysis of this
issue is beyond the scope of the current paper, however, and is left for future
research.

IV. Conclusions
This paper documents international return continuation in a sample of 12
European countries during the period 1980 to 1995. An internationally di-
versified portfolio of past Winners outperformed a portfolio of past Losers by
about 1 percent per month. These relative strength strategies load nega-
tively on conventional risk factors such as size and the market. The payoffs
are therefore inconsistent with the joint hypotheses of market efficiency and
commonly used asset pricing models. Return continuation is present in all
countries, and holds for both large and small firms, although it is stronger
for small firms than large firms. The European evidence is remarkably sim-
ilar to findings for the United States by Jegadeesh and Titman (1993), and
makes it unlikely that the U.S. experience was simply due to chance. Re-
turns on European momentum portfolios are significantly correlated with
relative strength strategies in the United States. Whether this correlation
reflects a priced momentum factor that is common across markets remains a
topic for future research.

REFERENCES
Asness, Clifford S., 1995, The power of past stock returns to explain future stock returns,
Working paper, Goldman Sachs Asset Management.
Asness, Clifford S., John M. Liew, and Ross L. Stevens, 1996, Parallels between the cross-
sectional predictability of stock returns and country returns, Working paper, Goldman Sachs
Asset Management.
Ball, Ray, and S. P. Kothari, 1989, Nonstationary expected returns: Implications for market
efficiency and serial correlations in returns, Journal of Financial Economics 25, 51–74.
Ball, Ray, S. P. Kothari, and Jay Shanken, 1995, Problems in measuring portfolio performance:
An application to contrarian investment strategies, Journal of Financial Economics 38,
79–107.
Ball, Ray, S. P. Kothari, and Charles E. Wasley, 1995, Can we implement research on stock
trading rules: The case of short term contrarian strategies, Journal of Portfolio Manage-
ment 21, 54–63.
Bekaert, Geert, Claude B. Erb, Campbell R. Harvey, and Tadas E. Viskanta, 1996, The cross-
sectional determinants of emerging equity market returns, Working paper, Duke Univer-
sity.
Blume, Marshall, and Robert F. Stambaugh, 1983, Biases in computed returns: An application
to the size effect, Journal of Financial Economics 12, 387–404.
Chan, K. C., 1988, On the contrarian investment strategy, Journal of Business 61, 147–163.
284 The Journal of Finance

Chan, Louis K. C., Yasushi Hamao, and Josef Lakonishok, 1991, Fundamentals and stock re-
turns in Japan, Journal of Finance 46, 1739–1764.
Chan, Louis K. C., Narasimhan Jegadeesh, and Josef Lakonishok, 1996, Momentum Strategies,
Journal of Finance 51, 1681–1713.
Conrad, Jennifer, and Gautam Kaul, 1993, Long-term market overreaction or biases in com-
puted returns?, Journal of Finance 48, 39–64.
Conrad, Jennifer, and Gautam Kaul, 1996, An anatomy of trading strategies, Working paper,
University of North Carolina.
DeBondt, Werner F. M., and Richard H. Thaler, 1985, Does the stock market overreact?, Jour-
nal of Finance 40, 793–805.
DeBondt, Werner F. M., and Richard H. Thaler, 1987, Further evidence on investor overreaction
and stock market seasonality, Journal of Finance 42, 557–581.
DeLong, J. Bradford, Andrei Schleifer, Lawrence H. Summers, and Robert J. Waldman, 1990,
Positive feedback investment strategies and destabilizing rational speculation, Journal of
Finance 45, 379–395.
Fama, Eugene F., and Kenneth R. French, 1996, Multifactor explanations of asset pricing anom-
alies, Journal of Finance 51, 55–84.
Ferson, Wayne E., and Campbell R. Harvey, 1996, Fundamental determinants of national eq-
uity market returns, NBER Working paper 5860.
Foerster, Stephen, Anoop Prihar, and John Schmitz, 1995, Back to the future, Canadian In-
vestment Review 7, 9–13.
Griffin, John M., and G. Andrew Karolyi, 1996, Another look at the role of the industrial struc-
ture of markets for international diversification strategies, Working paper, Ohio State Uni-
versity.
Heston, Steven L., and K. Geert Rouwenhorst, 1994, Does industrial structure explain the
benefits of industrial diversification?, Journal of Financial Economics 36, 3–27.
Jegadeesh, Narasimhan, and Sheridan Titman, 1993, Returns to buying winners and selling
losers: Implications for stock market efficiency, Journal of Finance 48, 65–91.
Lakonishok, Josef, Andrei Schleifer, and Robert W. Vishny, 1994, Contrarian investment, ex-
trapolation and risk, Journal of Finance 49, 1541–1578.
Richards, Anthony J., 1996, Winner-loser reversals in national stock market indices: Can they
be explained?, Journal of Finance 52, 2129–2144.