Upload
naranjoa2
View
215
Download
0
Embed Size (px)
Citation preview
8/3/2019 Final JIMF Paper
1/25
Risk factor and industry effects in the cross-country
comovement of momentum returns
Andy Naranjo a,*, Burt Porter b
a University of Florida, Warrington College of Business, Department of Finance, P.O. Box 117168, Gainesville, FL 32611-7168, USAb Iowa State University, College of Business, 3345 Gerdin Business Building, Ames, IA 50011-1350, USA
JEL classification:
G12
G15
Keywords:
International return momentum
Comovement
Asset pricing
Risk factors
Industry effects
Integration
a b s t r a c t
This paper examines the sources of cross-country comovement of
momentum returns over the 19752004 period. Using data on
more than 17,000 individual firms across 100 industries from 40
countries, we document the profitability of country-neutral indi-
vidual firm, industry, and industry-adjusted return momentum.
We show that country-neutral momentum returns are signifi-cantly correlated across countries, the correlation is time-varying,
and that comovement among industries cannot explain the
comovement of country-neutral momentum returns. However, we
find that standard risk factor models do explain a significant
portion of the cross-country comovement of momentum returns,
even though they do not explain average momentum returns.
Published by Elsevier Ltd.
1. Introduction
Over the past decade, numerous researchers have provided evidence on the profitability of
momentum trading strategies in both the U.S. and many other countries around the world (e.g.,
Jegadeesh and Titman, 1993, 2001; Rouwenhorst, 1998, 1999; Chui et al., 2003; Griffin et al., 2003).
While there is strong evidence that momentum strategies earn positive profits, there is no consensus as
to the source of these profits. In this paper, we provide evidence consistent with the existence of
systematic risk exposure of momentum portfolio returns by investigating the sources of the
comovement of country-neutral momentum returns across developed and emerging markets.1
* Corresponding author. Tel.: 1 (352) 392 3781; fax: 1 (352) 392 0301.
E-mail address: [email protected] (A. Naranjo).1 Country-neutral momentum is defined as a zero-cost investment strategy long recent winners and short recent losers
where winners and losers are defined relative to all stocks within a specific country. Factor betas are likely to differ between
long and short portfolios, so country-neutral does not imply factor-neutral.
Contents lists available at ScienceDirect
Journal of International Money
and Financej o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / l o c a t e / j i m f
0261-5606/$ see front matter Published by Elsevier Ltd.
doi:10.1016/j.jimonfin.2009.06.007
Journal of International Money and Finance 29 (2010) 275299
mailto:[email protected]://www.sciencedirect.com/science/journal/02615606http://www.elsevier.com/locate/jimfhttp://www.elsevier.com/locate/jimfhttp://www.sciencedirect.com/science/journal/02615606mailto:[email protected]8/3/2019 Final JIMF Paper
2/25
The comovement of country-neutral momentum returns is a puzzle that is closely related to the prof-
itability of momentum trading strategies. Rouwenhorst (1998) finds that European and U.S. momentum
returns have a correlation of 0.43 during the 19801995 period and that by conditioning the returns of
a European momentum strategy on the returns to a U.S.-only strategy, the average return of the European
momentum portfolio is reduced from 0.93 to 0.65 percent per month, implying a common component to
the two series.2 Rouwenhorst (1998) argues that these results could be consistent with a momentum factorin returns, but the dependence could also be due to non-zero exposures to other common priced factors
(such as SMB), common unpriced factors (industry factors) or a combination of both.
In this paper, we examine the extent to which the comovement in momentum returns is driven by
systematic risk factors. To investigate the sources of cross-country comovement of momentum returns,
we assemble a large data set that consists of more than 17,000 individual firms across 100 industries
from 40 developed and emerging markets over the 19752004 period. Using momentum portfolio
sorts, we examine the profitability and cross-country comovement of unadjusted, industry-adjusted,
and risk-adjusted momentum returns across countries and over time.
We show that country-neutral momentum strategies are profitable for both developed and
emerging markets, yielding an average of 54 and 75 basis points per month respectively over the 1975
2004 sample period. This result is consistent with the international evidence presented by Rou-wenhorst (1998, 1999) and Griffin et al. (2003). We also find both an industry momentum effect and
a firm momentum effect independent of industry for individual countries and on average across
developed countries using country-neutral momentum strategies. In contrast, for the emerging market
countries, we do not find significant industry momentum effects, though we do find significant firm
effects independent of industry. Using both domestic and global versions of a single factor market
model and of a Fama-French three factor model, we confirm that simple risk factor models generally do
a poor job of explaining the level of momentum returns in both developed and emerging markets.
We find that the payoffs to country-neutral momentum returns are significantly correlated across
countries. This relationship is positive and time varying, with a small average correlation in the early
1980s and an increasing average correlation in the 1990s through early 2000s. Conditioning on
industry versus firm independent of industry momentum effects, we find that comovement amongindustries cannot explain the comovement of country-neutral momentum returns. However, we
provide evidence that common asset pricing models explain a large portion of the comovement of
country-neutral momentum returns. We find that the rational determinants of value explain the cross-
country comovement of momentum returns, even though these very same rational determinants are
unable to explain the level of momentum returns. That is, the correlation of known risk factor payoffs
across countries induces correlated time varying factor loadings in country-neutral momentum
strategies that in turn induce cross country correlation in momentum portfolio payoffs. While standard
factor models are unsuccessful in explaining average momentum returns, they are successful in
explaining a large part of the cross-country comovement of momentum returns.
The rest of the paper proceeds as follows. Section 2 provides a brief review of the momentum
literature. Section 3 describes our methodology and data. Section 4 documents the cross-countryevidence on the profitability of country-neutral momentum returns, while Section 5 examines cross-
country comovement of momentum returns. Section 6 provides evidence on risk factor and
momentum return comovement across countries. Section 7 concludes.
2. Momentum returns and comovement: the literature
Many researchers have documented the profitability of momentum trading strategies in both
developed and emerging markets. For example, using U.S. data, Jegadeesh and Titman (1993) find that
a strategy that is long recent winners and short recent losers earns statistically significant average
2 In contrast to Rouwenhorst (1998), Griffin et al. (2003) report low intraregional and interregional correlations of
momentum profits, although these correlations are not the focus of their paper. However, we find that the averaging of
country-neutral momentum profits within and across regions, across developed and emerging markets, and over time
significantly attenuates country-neutral momentum return correlations resulting in low correlations similar to those reported
by Griffin, Ji, and Martin.
A. Naranjo, B. Porter / Journal of International Money and Finance 29 (2010) 275299276
8/3/2019 Final JIMF Paper
3/25
returns of 0.95% per month over the 19651989 period. In later work,Jegadeesh and Titman (2001) find
similar results using U.S. data from 1990 to 1998, suggesting the earlier result was not a manifestation of
datamining. In terms of international evidence, Rouwenhorst (1998) finds significant momentum profits
for 11 of 12 European countries, Chui et al. (2003) find mixed results for 8 Asian countries, and Rou-
wenhorst (1999) finds mixed results for 20 emerging markets. Rouwenhorst (1998) also finds that
country-neutral momentum returns are highly correlated across markets.Much research has also focused on explaining why momentum trading strategies are so widely
profitable. Jegadeesh (1990), Lo and MacKinlay (1990), andJegadeesh and Titman (1993) show that the
source of momentum profits can be decomposed into three different components: the cross-sectional
dispersion in expected returns, momentum in systematic factors, and momentum in the idiosyncratic
component of security returns (i.e., over-reaction/under-reaction component). With this framework
as a backdrop, two lines of research have emerged in the investigation of momentum profits: one based
on rational models and the other based on behavioral models.3
Within the rational paradigm, the profitability of moment strategies arises from compensation for
risk. Conrad and Kaul (1998) argue that momentum profits reflect cross-sectional variation in expected
returns rather than predictable time-series variation in stock returns. They argue that the momentum
strategys average profits reflect the result of buying (on average) high mean return stocks and selling(on average) low mean return stocks. If the cross-sectional differences in the mean returns reflect
expected returns, then the momentum profits can be attributed to cross-sectional differences in risk
characteristics (i.e., common identified or as yet unidentified priced risk factors).4 If common risk
factors play a role in explaining momentum profits and the risk factor is correlated across countries,
then country-neutral momentum returns will be correlated in a manner consistent with risk-based
explanations. This correlation in momentum returns across markets will also likely have time-varying
characteristics to the extent that markets have become more integrated over time.
When testing risk-based explanations of momentum profits, many authors find that the estimated
coefficients from regressions of momentum returns on commonly used risk factors are indistin-
guishable from zero and often have the wrong sign. However, Grundy and Martin (2001) show that
momentum portfolios by their very nature have time varying factor coefficients, so a simple exami-nation of the factor loadings of the long and short portfolios is misleading. After controlling explicitly
for the time variation in the factor loadings, Grundy and Martin (2001) show that risk-based models
still do not explain positive average momentum returns.
Industry effects are also a potential source of common risk, and hence are a good candidate for
understanding the source of country-neutral momentum returns and their comovement.5 For instance,
comovement of industry portfolio returns due to common fundamentals may result in similar industry
concentrations in the long and short momentum portfolios across countries, inducing comovement in
country-neutral momentum portfolios. Grinblatt and Moskowitz (1999) provide evidence suggesting
3 The behavioral models suggest that momentum profits arise from inherent biases in the way that investors interpretinformation (e.g., Barberis et al., 1998; Daniel et al., 1998; Hong and Stein, 1999). Each of these models addresses particular
constraints on investor rationality that cause an under-reaction of prices to information in the short-run, thereby producing
return persistence. Brav and Heaton (2002) distinguish between behavioral models and rational structural uncertainty models
where investors are rational but have imperfect information, and they find that each class of model has similar implications for
asset pricing and therefore provides an explanation for the momentum effect. While the behavioral models attempt to explain
the level of momentum profits, they do not address implications for the comovement of country-neutral momentum strategies.
In order for the comovement evidence to be consistent with the behavioral models, the information about firm value to which
investors underreact must be correlated across markets.4 Many researchers have examined various factors as potential explanations for momentum profits. These include market
beta, firm size, book-to-market, cash-flow-to-price, business cycles, and macroeconomic risks among others (see, for instance,
Jegadeesh and Titman, 1993; Fama and French,1996; Carhart, 1997; Daniel et al., 1997; Grinblatt and Moskowitz,1999; Chordia
and Shivakumar, 2002; Griffin et al., 2003).5 Industry factors can be viewed as a catch-all risk proxy or an unpriced factor. Barberis et al. (2005), argue that industry
factors in returns are at least in part due to industry-level common factors in cash flows, resulting in fundamentals-induced
comovement. However, to the extent that investors employ common industry asset allocation strategies across countries, the
comovement can also be category-induced. Behavioral underpinnings can also play a role in determining industry sources of
momentum effects and their comovement to the extent that there is slow information diffusion within industries and across
countries (see Hou, 2007; Hong et al., 2000; Chan et al., 1996).
A. Naranjo, B. Porter / Journal of International Money and Finance 29 (2010) 275299 277
8/3/2019 Final JIMF Paper
4/25
that momentum in industry risk factors explains the profitability of momentum strategies. That is,
momentum profits in individual stocks disappear after controlling for industry effects. However,
Grundy and Martin (2001) and Asness et al. (2001) find that the profitability of momentum strategies is
not fully explained by industry risk exposure. In particular, Grundy and Martin (2001) find that neither
industry effects nor cross-sectional differences in expected returns explain the level of momentum
profits. Rather, the strategies profitability reflects momentum in the stock-specific (idiosyncratic)component of returns. Asness et al. (2001) also find both an industry effect and a firm effect inde-
pendent of industry in momentum.
In recent work on the general comovement of security returns, Barberis et al. (2005) argue that there
are three explanations for the observed comovement among different securities: fundamentals-induced,
category-induced, and habitat-induced comovement. The fundamental approach explains the comove-
ment of securities by positive correlations in the rational determinants of their values. Category-induced
comovement occurs when investors classify different securities into the same asset class and shift
resources in and out of this class in correlated ways. Habitat-induced comovement arises when a group of
investors restricts its trading to a given set of securities and moves in and out of that set in tandem.
Consistent with the earlier research, these three explanations also suggest that comovement of securities
arises from fundamental factors, behavioral/mechanically-induced factors, or a combination of both.The fundamentals-induced comovement explanation proposed by Barberis et al. (2005) suggest that
the payoffs of country-neutral momentum portfolios should be correlated if momentum portfolios are
exposed to one or more risk factors and markets are integrated. That is, we would expect to see cross-
countrycomovement in the payoffsto country-neutralmomentum strategies whenmarkets are integrated
and if momentum returns are compensation for systematic risk or if momentum returns have a behavioral
or mechanical interpretation but share a common exposure to an unpriced risk factor. When markets are
not integrated, we would expect little or no comovement in country-neutral momentum returns.
The comovement in country-neutral momentum returns could also arise from the habitat or
category-induced comovement caused by institutional traders pursuing international momentum
trading strategies across countries or within industries across countries. A pervasive behavioral bias
might also induce comovement in momentum returns with a common unpriced risk factor exposure ifinformation flows to which investors over/under react are correlated across markets. Since cross
country institutional trading data is not available over large sample periods and for most markets and
measuring the information flows that induce behavioral biases is also problematic, our paper focuses
on fundamentals-induced momentum return comovement.6
3. Methodology and data
3.1. Methodology
We construct monthly relative strength momentum portfolios by ranking stocks within each
country on their total return during the previous 12 months, omitting the month immediately prior.We refer to this variable as Past(2,12). Several papers find that skipping a month between portfolio
ranking and investment periods avoids confounding bid/ask bounce and nonsyncronous trading
effects. We define winning and losing stocks as those in the top and bottom three deciles, respectively.
The self-financing momentum portfolio return is the return on an equally-weighted portfolio of
winners less the equally-weighted portfolio of losers. For each portfolio strategy, we examine the
average monthly profit from a strategy of buying a portfolio of winners and shorting a portfolio of
losers and holding the position for 1 month.7 The portfolios are rebalanced monthly. Although our
results are reported in local currency returns, U.S. dollar returns are similar.
6 Froot et al. (2001), for instance, are able to acquire proprietary institutional international flow data for only a 4 12 year
sample period from 1994 to 1998. They find that there is a small correlation in contemporaneous cross-country flows, which
suggests that comovement in flows cannot explain comovement in momentum returns.7 As further robustness checks, we also employed FamaMacBeth regression techniques in place of momentum portfolio
sorts, defined winners and losers as the top and bottom 10%, used value-weighting, and employed alternative prior month
return ranking periods and alternative holding periods. In each case, we obtained similar results.
A. Naranjo, B. Porter / Journal of International Money and Finance 29 (2010) 275299278
8/3/2019 Final JIMF Paper
5/25
We first examine the profitability of unadjusted, industry-adjusted, and risk-adjusted momentum
returns in both developed and emerging markets. We then examine the cross-country comovement
of momentum returns associated with each of these components. To the extent that there are
industry effects in momentum and industry returns are correlated across countries, the correlation
in country neutral momentum returns could arise from across-country industry effects in
momentum returns.
3.2. Sample construction
Our primary data source is the Thomson Datastream (TDS) database. TDS has return data for more
than 30,000 firms in over 50 countries. Our initial sample includes all firms covered by TDS from
January 1975 through December 2004. Following Ince and Porter (2006) who show that the TDS must
be carefully screened, we screen our sample for issues other than common equity and for firms that are
not traded on the countrys major exchange(s).8 To eliminate double counting and to isolate country
effects, we also exclude all listings other than those on a firms home country exchange (i.e., cross-
listings and depository receipts).9
We also exclude from our portfolios any firm whose market capitalization is below the 25thpercentile market capitalization of all NYSE stocks, converted to the local currency, in any month
during the ranking period. These firms are excluded to avoid significant liquidity constraints, to
mitigate any survivorship bias problems, and as Ince and Porter (2006) demonstrate, because many of
the difficulties encountered with TDS are concentrated in the smallest deciles. Since smaller stocks
typically have higher volatility and we sort on past returns, our market capitalization screen may
induce a bias against finding significant momentum returns. However, our country-neutral
momentum results are consistent with those reported by Jegadeesh and Titman (1993, 2001), Rou-
wenhorst (1998, 1999), Chui et al. (2003), and Griffin et al. (2003). Our reported conclusions are also
robust to alternative capitalization screens.
Emerging markets often report erroneous return data. Therefore, following Rouwenhorst (1999),
we assign as missing returns below the 2.5 percentile and above the 97.5 percentile of the returndistribution for each emerging market in each month during the ranking period. Finally, to insure that
we have a series of sufficient length to examine for evidence of comovement, we eliminate countries
from the sample that have fewer than 30 valid momentum return months. Our momentum screens
eliminate nine emerging market countries from the sample, leaving 17,449 firms representing 101
industries and 40 countries over the 19752004 time period.10
A strength of TDS is that they use the same criteria for defining industries across countries. As
argued by Griffin and Stulz (2001), this consistent classification minimizes the risk of finding low cross-
country industry comovement because of firm misclassification. Griffin and Karolyi (1998) argue that
using broad industrial classifications leads to lumping together heterogeneous industries that often
include widely disparate lines of business, which can mask the importance of industry effects. In our
investigation, we use TDSs most disaggregated industry classification, level 6, which includesapproximately 100 separate industry definitions.11
8 For example, in the U.S., we include only firms that trade on the NYSE, Amex, or Nasdaq and exclude such non-common
equity issues as preferred stock, closed end funds, real estate trusts and investment companies.9 Cross-listings and depository receipts can induce a positive dependence in country-neutral momentum returns. For
instance, Karolyi and Stulz (1996) find significant return comovements between depository receipts and share returns of the
underlying security in the home market as well as aggregate market returns in the home market.10 Following the momentum screens, the nine emerging market countries dropped from the sample because they have less
than 30 valid momentum return months are: Columbia, Cyprus, Czech Republic, Egypt, Hungary, Morocco, Pakistan, Peru, and
Venezuela.11 As a robustness check, we compare industry results from the U.S. using the CRSP database with firms sorted into industries
by their 4 digit SIC classification and using TDSs level six industry definitions. We obtain very similar results using either
dataset. As a further robustness check, we also compare our results for individual countries with those from Rouwenhorst
(1998, 1999) who uses a different data source. The results are similar.
A. Naranjo, B. Porter / Journal of International Money and Finance 29 (2010) 275299 279
8/3/2019 Final JIMF Paper
6/25
3.3. Summary statistics
Table 1 reports summary statistics for our sample. There are 22 developed and 18 emerging market
countries in our sample. Looking at the second column ofTable 1, we see that most of the start dates for
the developed market countries begin in 1975, while the majority of the emerging markets begin in the
late 1980s and early 1990s. The U.S., Japan, and the U.K. have the largest number of firms, the broadestindustry representation, and the largest market capitalizations of all the countries in our sample.
Together, these three countries represent over 55% of the firms in the sample and 48% of December
2004 market capitalization. Although much smaller on average, the larger emerging markets have both
more firms and more industries than the smaller developed markets. Fig. 1 illustrates the overlap in
market size between developed and emerging markets in our sample.12
4. Cross-country evidence on the profitability of momentum returns
4.1. Profitability of unadjusted momentum returns across countries
The first three columns of Table 2 report unadjusted momentum profits for portfolios formed onwithin country Past(2,12) and held for 1 month. We report the average monthly return, t-statistic, and
number of months with valid momentum returns, where a valid month is one in which both the long
and short portfolios have at least three stocks. In the next set of columns, we report the difference in
return between the portfolio of past winners and the local market return and the difference between
the local market return and the portfolio of past losers.
The results in Table 2 show significant momentum profits in many, but not all countries. We find
that 14 of 22 developed and 5 of 18 emerging markets have a significant momentum effect in our
sample period. Of the 13 developed markets that have data for the entire time period, only Japan, the
second largest market, does not have a significant momentum effect. Chui et al. (2003) and Griffin et al.
(2003) also find an insignificant momentum effect for Japan.13 We also do not find a significant
momentum effect for several of the other Asian and Scandinavian countries. Although the proportion
of emerging markets with significant momentum returns is much lower than for developed markets,
this result is due in part to higher standard errors resulting from shorter time series and the smaller
median number of firms in emerging markets. 10 of 18 emerging markets have higher average
momentum returns than the smallest significant average return in the developed markets.
To further examine developed versus emerging market returns, we create country-neutral
momentum portfolios for all developed markets by forming an equally weighted portfolio of all stocks
classified as winners when ranked within country and shorting all stocks classified as losers. We also
form a similar portfolio for emerging markets. We find statistically significant average returns of 54
basis points per month for the developed markets and 75 basis points per month for the emerging
markets over our full sample. This higher average return for the emerging market portfolio comes from
both higher returns on the long portfolio (10 basis points) and on the short portfolio (11 basis points).
Returns from the loser portfolio exceed those of the winner portfolio for both developed and emerging
markets, although the difference is only 6 basis points for developed markets and 7 basis points for
emerging markets.14 Finally, the results reported at the bottom of each panel show that there is time
variation in the magnitude of the momentum effect in both developed and emerging market countries.
4.2. Industry and firm effects in momentum returns
Firms within an industry may have similar exposure to both priced and unpriced risk factors. The
extent to which industry effects are present in momentum returns can be a potential source of
12 We omit three countries from Fig. 1: U.S. (6,036 firms), Japan (2,199) and U.K. (1,496). Each of these markets contains more
than twice the number of firms than the largest country included in the figure.13 Though not reported, if we exclude the two largest countries (Japan and the U.S.) from our sample, both the magnitude and
significance level of the average, country-neutral momentum effect increases.14 Bris et al. (2007), however, show that short selling is difficult to implement and not feasible in many markets.
A. Naranjo, B. Porter / Journal of International Money and Finance 29 (2010) 275299280
8/3/2019 Final JIMF Paper
7/25
comovement of country-neutral momentum returns. That is, if there are industry effects in momentum
returns, similar industry concentrations in both long and short momentum portfolios across countries,
and if industry returns significantly comove across countries, then country-neutral momentum returns
would be correlated.
Table 1
Summary statistics by country: 19752004.
Starting date Number of firms Number of
industries
Market capitalization
(December 2004, $Bil)
Developed markets
Australia 1975:01 364 77 $632Austria 1975:01 107 40 $427
Belgium 1975:01 124 49 $231
Canada 1975:01 610 83 $883
Denmark 1975:01 113 37 $134
Finland 1987:01 100 45 $166
France 1975:01 658 88 $1,535
Germany 1975:01 622 80 $1,311
Hong Kong 1975:01 384 69 $710
Ireland 1975:01 46 29 $102
Italy 1975:01 300 64 $725
Japan 1975:01 2,199 89 $3,508
Luxembourg 1991:03 30 17 $162
Netherlands 1975:01 206 57 $583
New Zealand 1986:01 45 28 $29
Norway 1975:01 156 40 $126
Singapore 1975:01 156 44 $192
Spain 1986:01 167 52 $623
Sweden 1975:01 316 61 $330
Switzerland 1975:01 403 69 $131
UK 1975:01 1,496 95 $2,586
USA 1975:01 6,036 101 $14,192
All developed 14,638 $29,318
Emerging markets
Argentina 1992:01 47 22 $31
Brazil 1993:01 151 42 $163Chile 1989:07 86 32 $82
Greece 1988:01 211 56 $93
India 1988:07 324 59 $285
Indonesia 1990:04 133 39 $50
Israel 1989:11 48 25 $56
Korea 1980:10 304 65 $330
Malaysia 1975:01 318 70 $117
Mexico 1988:01 132 37 $147
Philippines 1976:11 86 22 $17
Poland 1993:11 25 16 $53
Portugal 1988:01 62 26 $65
Russia 1994:01 61 18 $146
South Africa 1985:01 117 47 $210
Taiwan 1987:09 441 60 $279Thailand 1987:01 175 53 $76
Turkey 1988:03 90 31 $66
All emerging 2,811 $2,268
All 17,449 $31,586
We list countries as Developed or Emerging based on the International Finance Corporations (IFC) categorizations. Start date is
the first month that a firm meeting all the data requirements outlined in Section 3 enters the sample. Number of Firms and
Industries are the total number of unique firms and industries represented in the sample. Market capitalization is the total value
(in billions of U.S. dollars) of firms in the sample in December 2004.
A. Naranjo, B. Porter / Journal of International Money and Finance 29 (2010) 275299 281
8/3/2019 Final JIMF Paper
8/25
We create country-neutral industry momentum portfolios and industry-adjusted momentum
portfolios by first creating equally-weighted industry portfolios within each country using level 6 TDS
industry classifications. We require each industry portfolio to have a minimum of three firms. We
create industry-adjusted returns by taking the difference between the firms monthly return and thereturn on its country-specific industry portfolio. Then, we create across-industry momentum portfolios
by ranking the total return to each industry portfolio over Past(2,12) and forming an equally-weighted
portfolio of the top 30% of industries within the country and shorting the bottom 30% and holding for
one month before rebalancing. Within-industry portfolios are formed by ranking total industry-
adjusted returns over Past(2,12) and going long the top 30% and short the bottom 30%.
The last six columns ofTable 2 show returns to both across-industry and within-industry (industry-
adjusted) momentum portfolios. For developed markets, we find significant across-industry
momentum for 7 of the 19 countries with sufficient data to calculate across-industry momentum. Of
these 7 countries, 6 also have a significant average unadjusted momentum return. We also find that the
magnitude of the average developed market across-industry momentum return of 56 basis points per
month to be similar to the unadjusted momentum effect of 54 basis points per month. Similar to theunadjusted market-wide momentum results reported earlier for the developed markets, we also find
time-variation in across-industry momentum.
There is sufficient data to calculate industry momentum for only 9 of the 18 emerging markets. We
find no evidence of significant positive across-industry momentum in emerging markets; however, the
number of industries in each market and the average number of firms in each industry portfolio is
much smaller than in the developed markets. Overall, the across-industry momentum effect is nearly
zero for the emerging markets in both the full sample period and across sub-periods.
The last three columns ofTable 2 show returns to within-industry (industry-adjusted) momentum.
For the developed markets, 6 of 20 countries have a significant positive within-industry momentum
effect. Of these 6, all also have significant unadjusted momentum and 3 also have significant positive
across-industry momentum returns. For the emerging markets, 2 of 16 countries have a significantpositive within-industry momentum effect, one of which also has significant unadjusted momentum
returns. Taken together, the across-industry and within-industry results show that industry
momentum does not subsume unadjusted momentum, consistent with the evidence in Grundy and
Martin (2001) and Asness et al. (2001).
Fig. 1. We categorize countries as Developed or Emerging based on the International Finance Corporations (IFC) categorizations.
Number of Firms is the total number of unique firms represented in the sample. The largest three developed markets: U.S., Japan,
and U.K are not shown.
A. Naranjo, B. Porter / Journal of International Money and Finance 29 (2010) 275299282
8/3/2019 Final JIMF Paper
9/25
Table 2
Momentum returns for unadjusted, across-industry, and within-industry portfolios: portfolios formed on Past(2,12) ranking
period and held for one month 19752004.
Unadjusted Returns versus local market index Across-industry Within-industry
Average
monthly
return
t-stat n Winner
portfolio
t-stat Loser
portfolio
t-stat Average
monthly
return
t-stat n Average
monthly
return
t-stat n
Developed markets
Australia 0.98 4.19 347 0.37 3.06 0.61 4.33 0.29 1.09 347 0.55 2.91 347
Austria 0.15 0.40 198 0.16 0.82 0.31 1.07 0.38 0.42 44 0.50 0.93 80
Belgium 0.94 4.08 347 0.33 2.73 0.61 3.92 0.49 1.15 176 0.42 1.30 223
Canada 0.83 3.54 347 0.33 3.13 0.49 3.28 0.47 1.59 347 0.45 2.48 347
Denmark 0.88 2.98 347 0.31 2.14 0.57 2.85 1.15 0.85 87 0.71 1.45 161
Finland 0.25 0.45 189 0.35 1.35 0.10 0.27 0.72 0.32 12 0.09 0.05 38
France 0.83 3.44 347 0.40 3.76 0.42 2.74 0.82 3.07 347 0.36 1.93 347
Germany 0.73 3.09 347 0.32 3.16 0.41 2.54 0.67 2.39 347 0.25 1.53 347
Hong Kong 0.74 2.65 347 0.43 2.69 0.31 1.75 0.05 0.12 287 0.31 1.19 347
Ireland 1.18 2.21 174 0.33 1.27 0.85 2.24
Italy 1.42 4.73 347 0.73 4.82 0.68 3.65 1.32 3.06 306 0.47 1.71 306 Japan 0.08 0.32 347 0.00 0.03 0.08 0.62 0.41 1.55 347 0.18 1.05 347
Luxembourg 0.58 0.90 103 0.27 0.80 0.31 0.70 1.51 2.92 4
Netherlands 0.90 3.38 347 0.36 2.82 0.54 3.24 1.13 2.65 281 0.57 2.09 296
New Zealand 0.71 1.30 121 0.44 1.76 0.26 0.67
Norway 0.54 1.18 236 0.22 1.01 0.32 0.96 0.61 0.71 111 0.80 1.50 144
Singapore 0.38 1.13 325 0.06 0.30 0.33 1.43 0.08 0.12 211 0.30 0.95 253
Spain 0.54 2.48 194 0.42 3.20 0.12 0.81 0.80 1.82 161 0.39 1.92 192
Sweden 0.32 0.85 263 0.21 1.26 0.11 0.44 0.90 1.96 200 0.44 1.45 212
Switzerland 0.68 3.75 347 0.20 2.34 0.48 4.13 0.21 0.76 347 0.31 2.03 347
United Kingdom 0.71 3.14 347 0.33 3.68 0.38 2.49 0.64 2.74 347 0.47 3.82 347
USA 0.66 2.96 347 0.33 3.85 0.33 2.17 0.67 3.34 347 0.48 3.70 347
Total developed: 0.54 3.34 347 0.24 3.90 0.30 2.77 0.56 3.88 347 0.25 2.53 347
19751980 0.87 3.47 59 0.40 3.16 0.47 3.51 0.89 3.96 59 0.48 3.11 59
19811985 0.32 1.38 60 0.09 0.82 0.23 1.71 0.19 0.77 60 0.14 0.84 60
19861990 0.46 1.78 60 0.16 1.28 0.30 1.91 0.42 1.57 60 0.19 1.07 60
19911995 0.35 1.41 60 0.14 1.52 0.21 1.26 0.45 2.29 60 0.09 0.42 60
19962000 0.94 1.70 60 0.46 2.24 0.48 1.27 0.94 2.00 60 0.60 1.95 60
20012004 0.23 0.33 48 0.18 0.78 0.06 0.12 0.48 0.75 48 0.05 0.13 48
Emerging markets
Argentina 0.19 0.31 75 0.15 0.48 0.05 0.09 3.49 1.35 5
Brazil 1.34 1.71 92 0.57 1.13 0.77 1.77 2.35 2.29 45
Chile 0.30 0.85 138 0.49 2.39 0.19 0.77 0.86 0.83 3 0.31 0.73 74
Greece 1.23 2.18 123 0.64 2.24 0.59 1.38 0.57 0.53 37
India 0.61 1.57 167 0.36 1.69 0.25 0.90 1.01 1.33 87 0.63 1.15 114
Indonesia 1.10 1.78 105 0.79 2.25 0.31 0.67 0.24 0.21 32 0.61 0.64 35Israel 0.75 1.54 82 0.46 1.72 0.28 0.88 0.46 0.27 3
Korea 0.53 1.18 168 0.20 0.86 0.33 1.15 1.35 2.28 106 0.03 0.08 125
Malaysia 0.31 0.99 182 0.21 1.18 0.10 0.45 0.40 1.00 140 0.14 0.45 158
Mexico 1.29 3.30 184 0.55 2.14 0.74 2.67 0.47 0.47 57 1.02 1.63 81
Philippines 2.49 0.62 80 0.69 1.76 3.18 0.80 0.47 0.39 29
Poland 1.48 1.60 33 0.50 1.20 0.98 1.27
Portugal 1.27 2.50 98 0.96 3.54 0.31 0.86 1.90 1.23 14
Russia 0.76 0.54 29 0.77 0.80 0.01 0.01 3.94 0.77 9
South Africa 0.97 2.63 203 0.47 2.45 0.50 2.05 1.95 1.56 58 0.94 1.92 95
Taiwan 0.40 1.10 181 0.31 1.50 0.10 0.33 0.05 0.10 165 0.54 1.82 168
Thailand 1.39 2.28 145 1.15 3.61 0.23 0.59 0.91 0.61 30 2.11 2.53 63
Turkey 0.45 0.44 88 0.02 0.04 0.42 0.56
Total emerging: 0.75 3.64 225 0.34 2.23 0.41 2.17 0.02 0.05 194 0.72 3.96 194
19861990 0.81 1.23 57 0.26 0.54 0.55 0.92 1.24 1.34 26 0.70 1.16 26
19911995 0.75 3.29 60 0.31 1.65 0.44 1.96 0.19 0.70 60 0.10 0.38 60
19962000 0.62 1.80 60 0.34 1.34 0.28 0.81 0.60 1.05 60 0.94 2.59 60
(continued on next page)
A. Naranjo, B. Porter / Journal of International Money and Finance 29 (2010) 275299 283
8/3/2019 Final JIMF Paper
10/25
We create country-neutral momentum portfolios for all markets in a manner similar to that used for
developed and emerging markets. The results are reported in the last panel of Table 2. The results arenearly identical to those for developed markets due to the large fraction of firms traded in developed
markets.
We also examine alternative methods for constructing industry portfolios, including value-
weighting rather than equal-weighting, using global rather than local industry portfolios and using
a mix of local and global industry portfolios for those industries where the primary finished product is
tradable. See Griffin and Karolyi (1998) for more on the role of traded vs. non-traded industries. Our
reported results are robust to each of the alternative methods.
4.3. Risk adjusted momentum returns
In this section, we verify that simple factor models cannot explain the level of average momentumreturns. We use a single-factor market model and a three-factor model similar to that of Fama and
French (1992, 1993, 1996). Our measure of the market return is the value-weighted average of all
stocks local currency return. Our size factor, SMB, is calculated using the monthly NYSE 30th and 70th
percentile breakpoints converted to local currency.15 The Small and Big portfolios are value-weighted
and must each have at least three stocks. The SMB factor is the simple difference between the two
portfolios. The HML factor is from Ken Frenchs website.16
Since Grundy and Martin (2001) show that the factor coefficients are time varying, we estimate
momentum portfolio factor coefficients from individual stock factor coefficients. For each month and
for each stock, we estimate each factor model using a minimum of 24 and a maximum of 36 past
months of stock returns and factor realizations to estimate stock specific factor loadings. The
Table 2 (continued)
Unadjusted Returns versus local market index Across-industry Within-industry
Average
monthly
return
t-stat n Winner
portfolio
t-stat Loser
portfolio
t-stat Average
monthly
return
t-stat n Average
monthly
return
t-stat n
20012004 0.84 3.15 48 0.46 2.75 0.38 1.92 0.38 0.41 48 1.23 3.60 48
Total All: 0.55 3.57 347 0.24 4.08 0.30 2.96 0.55 3.83 347 0.25 2.66 347
19751980 0.87 3.47 59 0.40 3.16 0.47 3.51 0.89 3.96 59 0.49 3.16 59
19811985 0.32 1.38 60 0.09 0.87 0.23 1.67 0.19 0.77 60 0.14 0.84 60
19861990 0.46 1.80 60 0.13 1.01 0.32 2.04 0.40 1.51 60 0.19 1.08 60
19911995 0.38 1.64 60 0.15 1.69 0.23 1.42 0.42 2.24 60 0.08 0.42 60
19962000 0.90 1.75 60 0.49 2.46 0.41 1.18 0.89 1.95 60 0.58 1.99 60
20012004 0.31 0.46 48 0.19 0.91 0.12 0.25 0.47 0.75 48 0.01 0.03 48
We report average monthly momentum profit for portfolios of winners minus losers using monthly return data from January
1975 through December 2004. An average return of 0.50 is 50 basis points per month. Winners and losers are defined based on
total return during the ranking period of month t 2 through t 12 (Past(2,12)). Unadjusted refers to the ranking of all stocks by
average monthly return within a country during the ranking period. Winners and losers are defined as the top and bottom threedeciles, respectively. Across-Industry Portfolios are formed based on total return to equal-weighted industry portfolios during
the ranking period. Winning industries are the top 30% and losers the bottom 30% of industries with valid data during the
ranking period. The average monthly profit from a strategy of buying an equally weighted portfolio of the winning industry
portfolios and shorting the losing industries and holding for 1 month is reported. Within-Industry refers to the ranking of all
stocks by difference in return from their equal-weighted industry average. Winners are defined as the top 30% of stocks during
the ranking period and losers the bottom 30%. The average monthly profit from a strategy of buying an equal-weighted portfolio
of winners and shorting an equal-weighted portfolio of losers and holding the position for 1 month is reported. To be included in
any momentum portfolio, a firm must be above the 25th percentile market capitalization of all NYSE stocks in every month
during the ranking period. All returns are in local currency. Industry portfolios have a minimum of three firms. Across-industry
strategies must have a minimum of four valid industries from which winning and losing industries are selected.
15 The sample of stocks used to calculate SMB is not subject to the 25th NYSE market capitalization decile screen used to
calculate momentum portfolio returns.16 For countries that do not have an HML factor available, we use a two factor model.
A. Naranjo, B. Porter / Journal of International Money and Finance 29 (2010) 275299284
8/3/2019 Final JIMF Paper
11/25
momentum portfolio factor loading for month t is then calculated from the individual stock factor
loadings calculated from t 1 through t 36 (i.e., it is the simple average of the factor loadings of the
individual stocks). We calculate the fitted values for time tusing the estimated portfolio factor loadings
and the factor realizations from time t.17
Table 3 provides the risk-adjusted momentum return results. Since we require that a stock must
have a minimum of 24 months of prior returns to calculate its factor loadings, the sample used to formmomentum portfolios is slightly smaller than that used in Table 2. For comparison purposes, the first
column of Table 3 reports the average unadjusted momentum returns using this smaller sample.
The results in Table 3 show that the local market model, Model 1, does a poor job of explaining the
level of momentum returns. Of the 13 developed markets with significant momentum returns, the
average fitted value is positive and significant for only 2. The fraction of unadjusted momentum return
variance that is explained by the variance of the fitted returns, %VarE, averages 28%.18 These results are
consistent with Grundy and Martin (2001) who show that a single factor model does a poor job of
explaining the level of momentum returns.
A local version of the FamaFrench three-factor model, Model 2, also does a poor job of explaining
the level of momentum returns. The fitted values are significant in only 3 of the developed countries. In
unreported results, we find that of the 13 developed markets with significant momentum returns, 10have significant average factor loadings on one or more of the factors, although their magnitudes are
too low to explain the level of momentum profits.
The bottom of Table 3 provides the country-specific risk adjustment results for the emerging
markets. We again find that the country-level single and multi-factor models do a poor job of
explaining average momentum returns. Of the 5 countries with significant momentum returns, the
fitted values are significant for only one. The fraction of unadjusted momentum return variance that is
explained by the variance of the fitted returns is also less than half that of the developed markets.
In unreported results, we use global versions of the single and multifactor models to risk adjust
returns. The global market return is the value-weighted average of local market U.S. dollar returns. The
global market return is then converted into each local currency. The global SMB factor value-weights
the country specific Small and Big portfolio U.S. dollar returns, calculates the difference, and thenconverts to each local currency. The global HML factor is from Ken Frenchs website. The global factor
models are less successful than the local versions at explaining the level of country-neutral momentum
returns. This result is consistent with Griffin (2002) who finds that domestic versions of the Fama
French model have lower pricing errors than do global versions.
5. Cross-country comovement of momentum returns
The previous section shows that a significant momentum effect exists for a broad cross-section of
countries, and it has a time varying component. The results also show that there are both industry and
firm effects independent of industry in momentum returns and standard risk factors do a poor job of
explaining average momentum returns.19
In this section, we examine the cross-country comovementof momentum returns.
5.1. Correlation of unadjusted momentum returns across countries
Table 4 provides evidence that country-neutral momentum returns significantly comove. Panel A of
Table 4 reports the average correlation of country-neutral momentum returns within developed
markets over the full sample period 19752004, while Panel B reports the average correlation within
17 Korajczyk and Sadka (2004) model the time variation of the coefficients on the winner and loser portfolios as a function of
average factor realizations in the portfolio selection period. Although easier to compute, this approach is more restrictive than
the method we use.18 %VarE (Variance of the fitted return/Variance of the total return)*100. Since the coefficients are estimated out of sample, it
is possible to have the variance of the fitted exceed the variance of the total.19 In unreported results, we also find that the factor models do a poor job of explaining the level of across-industry and
within-industry momentum.
A. Naranjo, B. Porter / Journal of International Money and Finance 29 (2010) 275299 285
8/3/2019 Final JIMF Paper
12/25
Table 3
Risk-adjusted momentum returns 19752004.
Unadjusted
momentum returns
Model 1 Model 2
Avg t-stat n Fitted t-stat Errors t-stat % VarE Fitted t-stat Errors t-stat % VarE
Developed markets
Australia 0.99 4.12 334 0.05 0.45 1.05 4.99 25.02 0.12 0.71 0.87 4.08 47.87
Austria 0.14 0.38 192 0.32 1.59 0.18 0.49 27.82 0.04 0.12 0.43 1.07 44.31
Belgium 1.02 4.24 334 0.01 0.12 1.01 4.55 9.64 0.40 2.21 0.62 2.41 38.51
Canada 0.87 3.63 334 0.14 0.95 0.74 3.56 34.92 0.17 1.08 0.70 3.52 43.57
Denmark 0.86 2.85 334 0.31 2.34 0.55 1.91 19.35 0.28 0.87 0.20 0.48 53.20
Finland 0.27 0.45 177 0.45 1.50 0.18 0.32 24.80 0.53 1.02 0.47 0.76 52.07
France 0.80 3.26 334 0.06 0.51 0.74 3.47 25.99 0.01 0.06 0.77 3.74 44.81
Germany 0.74 3.21 334 0.25 1.63 0.49 3.07 45.60 0.18 0.92 0.56 3.67 74.24
Hong Kong 0.75 2.64 334 0.30 1.58 0.46 1.82 43.66 0.30 1.31 0.38 1.34 61.32
Ireland 1.06 1.84 161 0.21 1.46 0.85 1.51 6.04 0.71 1.51 0.46 0.62 36.00
Italy 1.34 4.35 334 0.26 1.68 1.08 3.93 25.77 0.62 2.48 0.44 1.61 57.43
Japan 0.01 0.04 334 0.02 0.150.01 0.04 21.98 0.06 0.32 0.07 0.34 54.76
Luxembourg 0.15 0.21 89 0.29 1.67 0.44 0.59 5.57 0.30 0.69 0.73 0.61 12.73Netherlands 0.89 3.31 334 0.09 0.98 0.80 3.23 11.77 0.21 1.44 0.68 2.83 28.98
New Zealand 0.61 1.00 88 0.04 0.22 0.65 1.08 8.44
Norway 0.60 1.26 230 0.05 0.27 0.55 1.19 13.61 0.68 0.90 1.32 1.47 49.40
Singapore 0.36 1.03 312 0.06 0.35 0.30 0.94 23.50 0.41 1.15 0.30 0.73 46.98
Spain 0.51 2.40 187 0.26 2.26 0.25 1.23 28.64 0.26 1.72 0.25 1.21 49.36
Sweden 0.25 0.67 251 0.24 1.28 0.01 0.04 24.76 0.44 2.05 0.08 0.26 29.47
Switzerland 0.67 3.47 334 0.12 1.23 0.55 3.15 26.84 0.22 1.59 0.43 2.50 49.43
UK 0.72 3.13 334 0.05 0.65 0.66 3.18 11.97 0.08 0.49 0.63 3.38 54.29
USA 0.63 2.83 334 0.02 0.16 0.61 3.42 29.93 0.08 0.46 0.55 3.18 65.56
Total developed: 0.52 3.22 334 0.06 0.68 0.46 3.68 28.19 0.08 0.63 0.42 3.69 55.25
Emerging markets
Argentina 0.07 0.10 64 0.10 0.57 0.03 0.04 5.92 0.18 0.59 0.45 0.57 15.60
Brazil 1.67 1.88 64 0.09 0.24 1.76 2.15 19.36 0.18 0.34 1.49 1.67 34.41
Chile 0.33 0.90 136 0.07 0.62 0.41 1.17 10.51 0.34 1.25 0.06 0.13 31.54
Greece 1.33 2.28 109 0.31 1.57 1.02 1.79 11.41 0.47 1.37 1.42 2.03 21.82
India 0.34 0.80 155 0.16 1.31 0.18 0.41 8.73 0.24 1.41 0.10 0.23 15.95
Indonesia 0.46 0.68 100 0.39 1.10 0.85 1.19 26.72 0.58 0.49 0.14 0.09 149.32
Israel 0.51 0.80 56 0.25 1.53 0.26 0.40 6.71 0.51 0.60 0.72 0.44 17.59
Korea 0.42 0.89 162 0.06 0.42 0.48 1.02 8.70 0.67 2.89 0.94 1.86 20.10
Malaysia 0.27 0.85 179 0.28 1.76 0.01 0.02 24.68 0.18 0.70 0.05 0.18 47.62
Mexico 1.07 2.64 170 0.01 0.07 1.08 3.02 26.06 0.29 1.13 0.63 1.85 33.84
Philippines 2.99 0.61 65 0.77 0.83 3.76 0.76 3.59 0.85 1.18 11.60 0.89 0.32
Poland 1.48 1.43 27 1.03 2.95 0.45 0.45 11.50
Portugal 1.46 2.55 81 0.83 3.17 0.63 1.11 20.76 0.88 2.21 1.02 1.23 19.35
Russia 0.54 0.38 27 0.06 0.11 0.60 0.47 13.24 0.66 0.87 1.73 1.11 22.29South Africa 0.94 2.43 190 0.05 0.66 0.99 2.47 4.21 0.42 2.53 0.52 1.27 18.53
Taiwan 0.41 1.07 168 0.14 0.96 0.56 1.50 15.03 0.02 0.07 0.43 1.22 32.07
Thailand 1.53 2.34 137 0.19 1.00 1.34 2.12 8.52 0.46 0.83 1.91 2.77 43.90
Turkey 0.14 0.14 83 0.72 2.98 0.86 0.86 5.87
Total emerging: 0.78 2.17 261 0.27 2.23 0.50 1.51 11.80 0.39 2.35 0.06 0.28 16.59
Model 1 Country-specific market model.
Model 2 Country-specific FamaFrench three-factor model.
We risk adjust momentum returns using either a single-factor market model or a three-factor model similar to that of Fama
French. The country-specific market return is the value-weighted average of all stocks local currency returns where there are
three or more firms. For the FamaFrench three factor country-specific model, we use the local market return along with an SMB
factor calculated using NYSE 30/70 breakpoints converted to local currency. Small and Big portfolios are value weighted and
must have at least three stocks. The SMB factor is the simple difference between the two portfolios. The HML factor is from Ken
Frenchs website. Since we require that a stock must have a minimum of 24 months of prior returns to calculate its factorloadings, the sample used to form momentum portfolios is slightly smaller here than that used in Table 2. For comparative
purposes, the first column reports the average unadjusted momentum returns from using this smaller sample. %Var-
E (Variance of the fitted return/Variance of the total return) * 100.
A. Naranjo, B. Porter / Journal of International Money and Finance 29 (2010) 275299286
8/3/2019 Final JIMF Paper
13/25
Table 4
Average pairwise correlations of unadjusted momentum returns across countries: portfolios formed on Past(2,12) ranking period and he
Average
of all
correlations
n t-statistic Average of
significant
correlations
# of significant
() correlations
# of significant
() correlations
Range
() co
Panel A: Average pairwise correlations of unadjusted momentum returns for developed markets (19752004)
Developed markets
Australia 0.19 21 11.03 0.21 0 16 0.13
Austria 0.18 21 10.56 0.22 0 14 0.14
Belgium 0.18 21 12.15 0.21 0 16 0.12
Canada 0.27 21 11.22 0.28 0 20 0.11
Denmark 0.20 21 11.61 0.23 0 16 0.16
Finland 0.24 21 10.20 0.28 0 17 0.16
France 0.34 21 12.41 0.35 0 20 0.19
Germany 0.32 21 10.78 0.33 0 20 0.12
Hong Kong 0.15 21 8.66 0.20 0 12 0.12
Ireland 0.21 21 9.41 0.26 0 14 0.15
Italy 0.23 21 12.72 0.24 0 19 0.15
Japan 0.15 21 9.39 0.20 0 11 Luxembourg 0.27 21 11.85 0.33 0 15 0.21
Netherlands 0.30 21 11.78 0.31 0 20 0.11
New Zealand 0.12 21 7.25 0.25 0 2 0.23
Norway 0.19 21 10.13 0.23 0 17 0.13
Singapore 0.13 21 8.07 0.18 0 11 0.11
Spain 0.18 21 10.31 0.22 0 15 0.15
Sweden 0.28 21 9.97 0.34 0 16 0.17
Switzerland 0.31 21 14.97 0.31 0 21 0.180
United Kingdom 0.33 21 11.68 0.34 0 20 0.180
USA 0.32 21 10.86 0.33 0 20 0.160
Total developed: 0.23 231 29.19 0.27 0 176 0.110
19751980 0.11 91 8.30 0.33 0 11 0.26019811985 0.02 105 1.43 0.23 2 7 0.260
19861990 0.08 136 5.29 0.34 0 21 0.260
19911995 0.16 190 14.40 0.35 0 54 0.260
19962000 0.27 231 21.62 0.42 0 123 0.260
20012004 0.41 210 28.05 0.51 0 149 0.290
8/3/2019 Final JIMF Paper
14/25
Table 4 (continued)
Average
of all
correlations
n t-statistic Average of
significant
correlations
# of significant
() correlations
# of significant
() correlations
Range
() co
Panel B: Average pairwise correlations of unadjusted momentum returns for emerging markets (19902004)
Emerging markets
Argentina 0.02 15 0.73 0.29 0 1 0.29
Brazil 0.08 16 2.56 0.21 0 1 0.21
Chile 0.02 16 0.76 0.22 1 0
Greece 0.03 16 0.69 0.02 1 1 0.23
India 0.05 16 2.25 0.21 0 1 0.21
Indonesia 0.00 15 0.13 0.27 0 1 0.27
Israel 0.01 14 0.68
Korea 0.01 16 0.45 0.23 0 1 0.23
Malaysia 0.01 16 0.34 0.04 1 1 0.29
Mexico 0.05 16 2.26
Philippines 0.06 14 1.52 0.27 0 1 0.27
Poland 0.07 11 1.21
Portugal 0.00 15 0.15 0.22 1 0
South Africa 0.01 16 0.58 0.21 1 0
Taiwan 0.00 16 0.12 0.21 2 0
Thailand 0.03 16 1.26 0.21 1 0 Turkey 0.01 16 0.26
Total emerging: 0.01 130 1.41 0.02 4 4 0.21
19911995 0.02 44 0.99 0.32 1 0
19962000 0.01 78 0.61 0.17 1 2 0.38
20012004 0.05 86 2.96 0.32 0 2 0.31
Panel C: Average pairwise correlations of unadjusted momentum returns across developed and emerging markets
All
Emerging w/developed 0.08 373 12.58 0.20 5 62 0.14
19901995
Emerging w/developed 0.04 200 3.27 0.23 3 18 0.26
19962000Emerging w/developed 0.04 308 4.22 0.24 5 27 0.26
20012004
Emerging w/developed 0.12 294 11.64 0.34 2 54 0.29
We report correlations of unadjusted momentum returns for portfolios of winners minus losers from 1975 to 2004 for developed countri
markets in Panel B. In Panel C, we report average correlations of unadjusted momentum returns between developed and emerging m
average monthly return during the ranking period. Unadjusted refers to the ranking of all stocks by total return within a country during th
as the top and bottom 30% of firms. To be included in a portfolio, a firm must be above the 25th percentile market capitalization of all NYS
returns are in local currency.
8/3/2019 Final JIMF Paper
15/25
emerging markets over the 19902004 sample period. The begin date for the emerging markets
correlation results in Panel B corresponds to the starting date available for many of the emerging
markets. We also report the average correlation between equally-weighted country market indices for
comparison.
For each developed market in Panel A, we find that the average pairwise correlations of country-
neutral momentum returns with the other developed markets are significant. The average of thepairwise correlations range from 0.30 and above for France, the U.K., the U.S., Germany, Switzerland,
and the Netherlands to 0.12 for New Zealand. With the exception of New Zealand, we find that for each
country well over half of the pairwise correlations are individually significant.20 Seven countries have
20 or more of a possible 21 positive and significant pairwise correlations, while only one developed
country, New Zealand, has less than 10 positive and significant pairwise correlations. No pair of
developed countries has a significant negative correlation in their momentum returns.
The range of statistically significant pairwise correlations for the developed markets in the eighth
column shows that the averages often mask the magnitude of the relationship between country-
neutral momentum returns. For example, the momentum returns for the U.K. and the U.S. have
a pairwise correlation of 0.55 over the full sample period. Many of the other developed markets also
have substantial pairwise momentum return correlations. At the bottom of Panel A, we see that thecomovement in momentum has grown over time. If we broaden our correlation analysis to the
intraregional and interregional levels and mix the developed and emerging markets, we obtain much
smaller correlations similar to those reported in Griffin et al. (2003).
For comparison, we provide the correlation of country return indices. However, it is important to
note that the correlation of country return indices does not imply that country-neutral momentum
returns should necessarily be correlated. The contemporaneous market return average correlations are
all positive, significant, and larger than the average momentum return correlations.
Panel B ofTable 4 provides the average pairwise correlations for the emerging markets. The magnitude
and significance of the momentum return correlations are much smaller than with the developed
markets. The average of all correlations among the emerging markets is only 0.01, compared with 0.23
among the developed markets. Of the 17 emerging markets shown in Table 4, the average correlation ofeach emerging market with the others ranges from0.06 to 0.08. The average market index correlations
are all significantly positive, though lower in magnitude than within developed markets.21
Panel C of Table 4 shows average correlations of developed with emerging markets. The average
correlation is positive and significant, though smaller in magnitude than the average within only
developed countries.
Although emerging markets display higher average momentum returns than developed markets,
country-neutral momentum returns within emerging markets do not generally comove. Comovement in
country-neutral momentum returns is concentrated primarilyamong developed countries. Thispattern is
consistent with integrated equity markets across developed countries but not across emerging countries
and is also consistent with the broader literature on the integration of global equity markets (e.g., Puchkov
et al., 2005; Bekaert et al., 2002; Bekaert and Harvey, 1995 among others.) The concentration ofcomovement among integrated markets observed here is also similar to that documented for other asset
classes, for example: syndicated loans (Carey and Nini, 2007; Itzkowitz et al., 2008), bonds (Barr and
Priestley, 2004), and real estate (Hastings and Nordby, 2007; Ling and Naranjo, 1999).
5.2. Correlation of country neutral across-industry and within-industry momentum returns
If country-specific industry portfolio returns are correlated across countries, then industry effects
might explain the comovement of country-neutral momentum returns. For example, if the auto
20 The correlations for the long portfolios across countries and short portfolios across countries are also similar, suggesting
that contagion is not the source of the dependence in momentum returns across markets. Cross country contagion effects can
potentially be a source of the comovement in momentum returns to the extent that momentum profits arise primarily from the
relation in short positions across countries. Contagion strategies entail selling assets from one country when crisis hits another
(see, for instance, Bekaert et al., 2005; Kaminsky et al., 2004).21 This is consistent with the findings of others. See, for instance, Goetzmann et al. (2005) and Longin and Solnik (1995, 2001).
A. Naranjo, B. Porter / Journal of International Money and Finance 29 (2010) 275299 289
8/3/2019 Final JIMF Paper
16/25
industry has a good year in both the U.S. and Japan, then momentum portfolios in both countries would
be long auto industry stocks. To the extent that cross-country industry comovement explains the
comovement in unadjusted country-neutral momentum, we would expect country-neutral across-
industry momentum to comove and within-industry momentum portfolios not to comove.
In Table 5 we examine the extent to which across-industry and within-industry momentum returns
comove. Panel A of Table 5 reports the average pairwise correlation between developed markets forcountries where we can calculate across-industry momentum returns in at least 30 months of the
sample period. Although from Table 2 we see that the level of across-industry momentum of 56 basis
points per month is slightly higher than unadjusted momentum return of 54 basis points per month,
the average developed market across-industry pairwise correlation of 0.08 is much lower than the
unadjusted momentum correlation of 0.23. The country-neutral within-industry momentum portfo-
lios are more highly correlated than the across-industry momentum portfolios. It is clear from Table 5
that industry comovement does not by itself explain the comovement in momentum returns.
We repeat the analysis for emerging markets in Panel B. Similar to the earlier results for the
comovement of unadjusted momentum returns within emerging markets reported in Table 4, we find
little evidence of comovement of across and within-industry momentum returns in emerging markets.
6. Risk factor and momentum return comovement
Industry comovement does not explain the comovement in country-neutral momentum returns.
Although unable to explain the level of momentum returns, it is possible that common asset pricing
models may be able to explain the comovement of momentum returns. As Grundy and Martin (2001)
point out, buying recent winners and selling recent losers guarantees time-varying factor exposures in
accordance with the performance of common risk factors during the ranking period. If risk factors covary
across countries, then the time varying factor exposures of the long and short portfolios must also covary.
6.1. Correlation of risk adjusted momentum returns across countries
In Table 6 we report risk-adjusted correlations corresponding to the country-specific versions of the
factor models. In Panel A, we use the domestic version of the market model, while in Panel B we use the
domestic version of the FamaFrench model. We only include countries that have data available to
estimate both models so that we can more accurately compare across models without deviations
caused by different countries or sample periods.22
When using a one factor model to risk adjust momentum returns, we find that the correlations of the
fitted returns are large and significant. The average correlation across countries of the fitted values is 0.37,
with a t-statistic of 9.03. Although the correlations of the errors from the one-factor model are also large
and significant, the average correlation of 0.16 is less than half as big as the average fitted correlations.
However, we must be careful in comparing the correlations of the fitted values and of the residuals as the
variance of the residuals is typically much higher, potentially masking differences in covariance.
We decompose the covariance of country-neutral momentum returns as:
CovRA; RB Cov
RFitA RErrA ; R
FitB R
ErrB
Cov
RFitA ; RFitB
Cov
RErrA ; R
ErrB
h
Cov
RFitA ; RErrB
Cov
RErrA ; R
FitB
i
1 Percent Fit Percent Error Percent Other; (1)
where RA is the country-neutral momentum of countryA, RFit
A is the return explained by the model, and
RErrA is the return not explained by the model. RFit
A and RErr
A are not orthogonal since the factor loadings
are estimated out of sample.
22 If we allow the sample to vary across models, we obtain qualitatively similar results for the developed countries. Due to
insufficient data, many of the models cannot be estimated for the emerging markets, and the fit is generally poorer when the
estimates can be obtained.
A. Naranjo, B. Porter / Journal of International Money and Finance 29 (2010) 275299290
8/3/2019 Final JIMF Paper
17/25
The covariance of the fitted values explains 30% of the variance of the country-neutral momentum
returns and the errors 45%. For 16 of 19 developed markets, the covariance of the errors explains
a greater percentage of the covariance of the momentum returns than does the covariance of the fitted
values. The sub-sample results show that there is considerable time variation in the fitted and error
correlations and confirms that the fitted values explain a smaller portion of the covariance relative to
the errors with the exception of the 20012004 sub-period.In Panel B of Table 6, we report the results from using the domestic version of the FamaFrench
model to risk adjust momentum returns. The fitted values explain an average of 40% of the variance of
the country-neutral momentum returns compared to 30% for the one factor model. The portion of the
variance explained by the errors falls from 45% to 33%. For 14 of 19 developed markets, the fitted values
explain a greater percentage of the covariance in momentum returns than the errors. The sub-sample
results at the bottom of the panel also show that there is considerable time variation in the fitted and
error correlations. The proportion of the covariance explained by the fitted values and errors also varies
over time, with the fitted values explaining an increasing portion of the covariance through time.
In unreported results, we use global versions of the one and three factors models and get quali-
tatively similar results. However, the percentage of the covariance explained by the errors is larger with
the global versions of the factor models, suggesting a general deterioration in the fit of these modelsrelative to the domestic versions. These results are consistent with the relatively poorer fit of global
factor models relative to the domestic versions as reported by Griffin (2002).
6.2. Does the comovement of risk factors explain the comovement of country-neutral momentum returns
across countries?
The risk-adjusted momentum results in Table 6 suggest a relationship between the correlation of
momentum returns and risk factors. To explore this relationship in greater detail, we first examine if
country pairs with higher factor covariance also have higher country-neutral momentum return
covariance. We then examine if the covariance of factor returns during the measurement period
predict the covariance of momentum returns in the holding period.
6.2.1. Risk factor and momentum return covariation
For the first set of tests, we estimate the sample pairwise covariance of country-neutral momentum
returns and risk factors. We then regress the covariance of momentum returns on the covariance of risk
factors for all pairs of developed countries with sufficient data to calculate local factors, have a valid
momentum return on or within 12 months of the factor realization, and where the covariances are
estimated with a minimum of 30 data points. We report the regression results in Panel A ofTable 7. The
first three columns are results from univariate regressions corresponding to each risk factor, while the
fourth column contains the multivariate regression results with all three risk factors included. The fifth
and sixth columns contain the multivariate results over two equal length 15-year sub-periods:
19751989 and 19902004.Each risk factor in the univariate regressions has a positive and significant coefficient. In other
words, countries whose risk factors have high covariance also tend to have high covariance of country-
neutral momentum. In the last two columns, we see that the factor covariances do a better job at
explaining the covariance of momentum returns in the second sub-sample, which corresponds to the
period when momentum return covariances are the largest over the entire sample.
6.2.2. Does factor return covariation predict the covariation of momentum returns?
In the second set of tests, we examine whether or not the covariance in factor returns during the
measurement period predicts the covariance of momentum returns in the holding period. For each
country-neutral momentum strategy and for each local factor realization, we calculate the following
deviations from country specific time-series means:
3i;t Ri;t RigRMRF
i;t RMRFi;t RMRFi; (2)
A. Naranjo, B. Porter / Journal of International Money and Finance 29 (2010) 275299 291
8/3/2019 Final JIMF Paper
18/25
Table 5
Average pairwise correlations of across-industry and within-industry momentum returns across countries: portfolios formed on Past(2,1
Pairwise correlations of industry and industry-adjusted momentum returns
Across-industry momentum Within-industry-adjusted
All correlations Individually significant correlations All correlations
Average n t-stat. Average N () N () Range Average n t-stat.
Panel A: Developed markets (19752004)
Developed markets
Australia 0.07 17 2.61 0.22 0 4 0.130.31 0.10 18 7.43
Austria 0.00 16 0.07 0.00 1 1 0.310.31 0.10 17 2.59
Belgium 0.03 17 1.75 0.16 18 4.44
Canada 0.13 17 4.32 0.23 0 8 0.160.35 0.13 18 6.59
Denmark 0.01 16 0.42 0.05 18 2.48
Finland 0.17 17 5.92
France 0.12 17 3.41 0.15 1 9 0.130.34 0.18 18 7.05
Germany 0.12 17 4.40 0.23 0 8 0.160.35 0.23 18 7.07
Hong Kong 0.10 17 3.83 0.20 0 8 0.160.29 0.06 18 3.34
Ireland
Italy 0.10 17 3.14 0.20 0 8 0.120.32 0.16 18 6.35
Japan 0.09 17 5.56 0.17 0 4 0.150.21 0.14 18 9
Luxembourg
Netherlands 0.14 17 4.23 0.23 0 10 0.130.38 0.17 18 5.55
New Zealand
Norway 0.02 17 0.83 0.12 18 5.96
Singapore 0.06 17 3.73 0.19 0 1 0.190.19 0.15 18 7.15
Spain 0.05 17 2.74 0.19 0 1 0.190.19 0.09 18 4.36
Sweden 0.10 17 4.66 0.19 0 5 0.160.24 0.13 18 6.92
Switzerland 0.11 17 6.02 0.18 0 8 0.120.22 0.16 18 7.20
UK 0.17 17 4.51 0.26 0 12 0.150.38 0.21 18 7.19
USA 0.03 17 0.97 0.16 0 5 0.120.22 0.22 18 9.29
Total developed: 0.08 152 8.17 0.20 1 46 0.120.38 0.14 170 16.50
19751980 0.06 45 2.47 0.24 1 4 0.310.47 0.09 45 4.81
19811985 0.05 36 1.65 0.16 2 3 0.440.52 0.06 55 2.70
19861990 0.04 91 2.23 0.16 2 6 0.260.37 0.06 91 4.26
19911995 0.07 136 5.15 0.23 2 13 0.260.42 0.12 153 10.19
19962000 0.17 136 8.62 0.45 0 42 0.290.70 0.15 153 11.58
20012004 0.21 91 10.34 0.41 1 29 0.300.66 0.34 105 16.44
8/3/2019 Final JIMF Paper
19/25
Panel B: Emerging Markets (19902004)
Emerging markets
Argentina
Brazil 0.06 4 0.69
Chile 0.16 7 2.77
Greece 0.08 2 0.33
India 0.03 7 0.34 0.38 0 1 0.38 0.38 0.03 9 0.45
Indonesia 0.06 3 0.36 0.38 0 1 0.38 0.38 0.00 5 0.03
IsraelKorea 0.03 6 1.39 0.04 7 0.80
Malaysia 0.04 7 0.91 0.06 9 1.44
Mexico 0.09 4 1.35 0.08 7 1.42
Philippines
Poland
Portugal
Russia
South Africa 0.00 4 0.04 0.01 7 0.15
Taiwan 0.01 7 0.21 0.02 10 0.42
Thailand 0.05 4 0.43 0.06 7 0.85
Turkey
Total Emerging: 0.02 21 0.52 0.38 0 1 0.380.38 0.02 37 0.75
19911995 0.02 6 0.57 0.00 10 0.0919962000 0.03 3 0.35 0.09 5 0.81
20012004 0.07 1 0.10 11 1.54
Panel C: Across developed and emerging markets
All
Emerging w/Developed 0.03 134 2.81 0.06 4 9 0.160.41 0.03 192 2.95
19901995
Emerging w/Developed 0.04 66 2.20 0.21 1 6 0.280.34 0.03 90 1.84
19962000
Emerging w/Developed 0.06 63 2.92 0.02 1 1 0.330.33 0.02 86 1.20
20012004
Emerging w/Developed 0.09 28 3.11 0.01 1 1 0.320.32 0.07 90 4.50
We report average correlations of across-industry and within-industry momentum portfolio returns from 1975 to 2004 for developed countr
markets in Panel B. In Panel C, we report average correlations of across-industry and within-industry momentum portfolio returns betwee
losers are defined based on total return during the ranking period. Across-Industry portfolios are formed based on total return to equal-weig
Winning industries are the top 30% and losers the bottom 30% of industries with valid data during the ranking period. The average monthly
portfolio of winning industry portfolios and shorting the losing industries and holdingfor 1 month is reported. Within-Industry portfolios refe
from their equal-weighted industry average. Winners are defined as the top 30% of stocks during the ranking period and losers the bottom 3
buying an equal-weighted portfolio of winnersand shorting an equally-weighted portfolio of losers and holding the positionfor 1 month is re
above the 25th percentile of market capitalizationof all NYSE stocks, converted to the local currency, in every month of the portfolio formatio
portfolios have a minimum of three firms. Across-industry strategies must have a minimum of four valid industries from which winning a
8/3/2019 Final JIMF Paper
20/25
8/3/2019 Final JIMF Paper
21/25
8/3/2019 Final JIMF Paper
22/25
8/3/2019 Final JIMF Paper
23/25
second and third columns, we see that for all three factors, covariances of the risk factor returns during
the measurement period are significant predictors of the covariance of country-neutral momentum
returns during the holding period. In the last two columns, we see that the covariance of the risk factor
returns during the measurement period play an increasingly significant role in predicting covariances
of country-neutral momentum returns during the holding period, with increasing coefficient esti-
mates, significance levels and explanatory power over time.We also perform several robustness checks. First, we include the intersection of developed with
emerging markets in the regression analysis and again find that the covariance of market factor returns
during the measurement period are significant in explaining subsequent covariances of country
momentum returns, though the explanatory power is lower due to the sparseness of emerging market
observations.23 Second, we include several measures of industry overlap to examine the possibility that
industry concentration in momentum portfolios drives the cross sectional results. It does not. We also
examine five-year sub-periods to see if there is a stationarity problem with the covariances in the
regression analysis. Though the early five-year sample periods are sparse, the time varying significance
of the results over each of the five-year periods suggests that there is not a stationarity problem in the
regression analysis. Finally, to assess if there are errors-in-variables problems in the regression anal-
ysis, we use instrumental variables procedures and again obtain similar results to those we report.
7. Conclusion
We examine the profitability and cross-country comovement of country-neutral momentum
returns for both developed and emerging markets. Using relative strength portfolio sorts, we first
document the profitability of momentum returns across countries and then decompose those returns
by industry effects, firm effects independent of their industry and with single and multifactor models.
In addition, we examine the relationship in momentum returns across countries and over time.
We show that country-neutral momentum strategies are profitable on average for both developed
and emerging markets, yielding 54 and 75 basis points per month respectively over our sample period
extending from 1975 to 2004. This result is consistent with the international evidence presented byRouwenhorst (1998, 1999) and Griffin et al. (2003). We also find both an industry momentum effect
and a firm momentum effect independent of its industry for individual countries and on average across
countries using country-neutral momentum strategies. Using a market model and the FamaFrench
three factor model, we find that the models do a poor job at explaining average momentum returns.
We find that country-neutral momentum returns are significantly correlated across countries. This
comovement is positive and time varying. Conditioning on industry, we find that industry comovement
does not explain the comovement in country-neutral momentum returns. We also find that while
standard factor models do not explain average momentum returns, they do explain a significant
portion of the comovement of country-neutral momentum returns. This finding is of particular
importance to the asset allocation, diversification, and risk management strategies of investors who
employ international momentum strategies since countries whose capital markets are more highlyintegrated are likely to have correlated pay-offs to country-specific momentum portfolios. Taken
together, our results suggest that risk factors play an important role in explaining the cross-country
comovement of momentum returns.
Acknowledgements
We thank Mark Flannery, Kent Hargis, Campbell Harvey, Jason Karceski, Andrew Karolyi, Michael
Melvin (the Editor), M. Nimalendran, an anonymous referee, and seminar participants at the University
of Miami, University of New Orleans, and CIBER at the University of Florida for helpful comments and
suggestions.
23 The intersection of emerging markets with emerging markets is very sparse across the multivariate specifications and
largely insignificant. The HML factor, for instance, is largely unavailable for the emerging markets.
A. Naranjo, B. Porter / Journal of International Money and Finance 29 (2010) 275299 297
8/3/2019 Final JIMF Paper
24/25
References
Asness, C.S., Porter, R.B., Stevens, R.L., 2001. Predicting stock returns using industry-relative firm characteristics. UnpublishedWorking Paper, Iowa State University, Ames, IA.
Barberis, N., Shleifer, A., Vishny, R., 1998. A model of investor sentiment. Journal of Financial Economics 49, 307343.Barberis, N., Shleifer, A., Wurgler, J., 2005. Comovement. Journal of Financial Economics 75, 283317.
Barr, D., Priestley, R., 2004. Expected returns, risk, and the integration of international bond markets. Journal of InternationalMoney and Finance 23, 7197.
Bekaert, G., Harvey, C., 1995. Time-varying world market integration. Journal of Finance 50, 403444.Bekaert, G., Harvey, C., Lumsdaine, R., 2002. Dating the integration of world equity markets. Journal of Financial Economics 65,
203247.Bekaert, G., Harvey, C., Ng, A., 2005. Market integration and contagion. Journal of Business 78, 3969.Brav, A., Heaton, J.B., 2002. Competing theories of financial anomalies. Review of Financial Studies 15, 575606.Bris, A., Goetzman, W.M., Zhu, N., 2004. Efficiency and the bear: short sales and markets around the world. Journal of Finance
62, 10291079.Carey, M., Nini, G., 2007. Is the corporate loan market globally integrated? A pricing puzzle. Journal of Finance 62, 29693007.Carhart, M., 1997. On persistence in mutual fund performance. Journal of Finance 52 (1), 5782.Chan, L.K.C., Jegadeesh, N., Lakonishok, J., 1996. Momentum strategies. Journal of Finance 51, 16811713.Chui, A.C.W., Titman, S., Wei, K.C.J., 2003. Momentum, legal systems and ownership structure: an analysis of Asian stock market.
Unpublished Working Paper, University of Texas, Austin, TX.Chordia, T., Shivakumar, L., 2002. Momentum, business cycle, and time-varying expected returns. Journal of Finance 57,
9851019.Conrad, J., Kaul, G., 1998. An anatomy of trading strategies. Review of Financial Studies 11, 489519.Daniel, K., Hirshleifer, D., Subrahmanyam, A., 1998. Investor psychology and security market under- and overreactions. Journal
of Finance 53, 18391886.Daniel, K., Grinblatt, M., Titman, S., Wermers, R., 1997. Measuring mutual fund performance with characteristic-based bench-
marks. Journal of Finance 52, 12171218.Fama, E.F., French, K.R., 1992. The cross-section of expected stock returns. Journal of Finance 47, 427465.Fama, E.F., French, K.R., 1993. Common risk factors in the returns on stocks and bonds. Journal of Financial Economics 33, 356.Fama, E.F., French, K.R., 1996. Multifactor explanations of asset pricing anomalies. Journal of Financial Economics 51, 5584.Froot, K., OConnell, P.G.J., Seasholes, M., 2001. The portfolio flows of international investors. Journal of Financial Economics 59,
151193.Goetzmann, W., Li, L., Rouwenhorst, K.G., 2005. Long-term global market correlations. Journal of Business 78, 138.Griffin, J.M., Karolyi, A.G., 1998. Another look at the role of industrial structure of markets for international diversification
strategies. Journal of Financial Economics 50, 351373.
Griffin, J.M., Stulz, R., 2001. International competition and exchange rate shocks: a cross-country industry analysis of stockreturns. Review of Financial Studies 14, 215241.
Griffin, J.M., 2002. Are the Fama and French factors global or country specific? Review of Financial Studies 15, 783803.Griffin, J.M., Ji, X., Martin, J.S., 2003. Momentum investing and business cycle risk: evidence from pole to pole. Journal of Finance
58, 25152547.Grinblatt, M., Moskowitz, T., 1999. Do industries explain momentum? Journal of Finance 54, 12491290.Grundy, B.D., Martin, J.S., 2001. Understanding the nature of the risks and the source of the rewards to momentum investing.
Review of Financial Studies 14, 2978.Hastings, A., Nordby, H., 2007. Benefits of global diversification on a real estate portfolio. The Journal of Portfolio Management,
5362 (Special issue).Hong, H., Lim, T., Stein, J.C., 2000. Bad news travels slowly: size, analyst coverage, and the profitability of momentum strategies.
Journal of Finance 55, 265295.Hong, H., Stein, J.C., 1999. A unified theory of underreaction, momentum trading and overreaction in asset markets. Journal of
Finance 54, 21432184.Hou, K., 2007. Industry information diffusion and the lead-lag effect in stock returns. Review of Financial Studies 20, 11131138.
Ince, O., Porter, R.B., 2006. Individual equity return data from Thomson datastream: handle with care!. Journal of FinancialResearch 29, 463479.
Itzkowitz, J., Houston, Naranjo, A., 2008. Borrowing beyond borders: the geography and pricing of Syndicated Bank loans.Unpublished Working Paper, University of Florida, Gainesville, FL.
Jegadeesh, N., 1990. Evidence of predictable behavior of security returns. Journal of Finance 45, 881898.Jegadeesh, N., Titman, S., 1993. Returns to buying winners and selling losers: implications for stock market efficiency. Journal of
Finance 48, 6591.Jegadeesh, N., Titm