Final JIMF Paper

Journal of International Money and Finance 29 (2010) 275299
Contents lists available at ScienceDirect
Journal of International Money and Finance

journal homepage: www.elsevier.com/locate/jimf
Risk factor and industry effects in the cross-country comovement of momentum returns
Andy Naranjo a, *, Burt Porter b
a b
University of Florida, Warrington College of Business, Department of Finance, P.O. Box 117168, Gainesville, FL 32611-7168, USA Iowa State University, College of Business, 3345 Gerdin Business Building, Ames, IA 50011-1350, USA
a b s t r a c t
JEL classication: G12 G15 Keywords: International return momentum Comovement Asset pricing Risk factors Industry effects Integration
This paper examines the sources of cross-country comovement of momentum returns over the 19752004 period. Using data on more than 17,000 individual rms across 100 industries from 40 countries, we document the protability of country-neutral individual rm, industry, and industry-adjusted return momentum. We show that country-neutral momentum returns are signicantly correlated across countries, the correlation is time-varying, and that comovement among industries cannot explain the comovement of country-neutral momentum returns. However, we nd that standard risk factor models do explain a signicant 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 protability 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; Grifn et al., 2003). While there is strong evidence that momentum strategies earn positive prots, there is no consensus as to the source of these prots. 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: andy.naranjo@cba.u.edu (A. Naranjo). 1 Country-neutral momentum is dened as a zero-cost investment strategy long recent winners and short recent losers where winners and losers are dened relative to all stocks within a specic country. Factor betas are likely to differ between long and short portfolios, so country-neutral does not imply factor-neutral. 0261-5606/$ see front matter Published by Elsevier Ltd. doi:10.1016/j.jimonn.2009.06.007
276
A. Naranjo, B. Porter / Journal of International Money and Finance 29 (2010) 275299
The comovement of country-neutral momentum returns is a puzzle that is closely related to the profitability of momentum trading strategies. Rouwenhorst (1998) nds 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 factor in 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 rms across 100 industries from 40 developed and emerging markets over the 19752004 period. Using momentum portfolio sorts, we examine the protability 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 protable 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 Rouwenhorst (1998, 1999) and Grifn et al. (2003). We also nd both an industry momentum effect and a rm 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 nd signicant industry momentum effects, though we do nd signicant rm 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 conrm that simple risk factor models generally do a poor job of explaining the level of momentum returns in both developed and emerging markets. We nd that the payoffs to country-neutral momentum returns are signicantly 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 rm independent of industry momentum effects, we nd that comovement among industries 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 nd that the rational determinants of value explain the crosscountry 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-country evidence on the protability of country-neutral momentum returns, while Section 5 examines crosscountry 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 protability of momentum trading strategies in both developed and emerging markets. For example, using U.S. data, Jegadeesh and Titman (1993) nd that a strategy that is long recent winners and short recent losers earns statistically signicant average
2 In contrast to Rouwenhorst (1998), Grifn et al. (2003) report low intraregional and interregional correlations of momentum prots, although these correlations are not the focus of their paper. However, we nd that the averaging of country-neutral momentum prots within and across regions, across developed and emerging markets, and over time signicantly attenuates country-neutral momentum return correlations resulting in low correlations similar to those reported by Grifn, Ji, and Martin.
277
returns of 0.95% per month over the 19651989 period. In later work, Jegadeesh and Titman (2001) nd similar results using U.S. data from 1990 to 1998, suggesting the earlier result was not a manifestation of data mining. In terms of international evidence, Rouwenhorst (1998) nds signicant momentum prots for 11 of 12 European countries, Chui et al. (2003) nd mixed results for 8 Asian countries, and Rouwenhorst (1999) nds mixed results for 20 emerging markets. Rouwenhorst (1998) also nds that country-neutral momentum returns are highly correlated across markets. Much research has also focused on explaining why momentum trading strategies are so widely protable. Jegadeesh (1990), Lo and MacKinlay (1990), and Jegadeesh and Titman (1993) show that the source of momentum prots 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 prots: one based on rational models and the other based on behavioral models.3 Within the rational paradigm, the protability of moment strategies arises from compensation for risk. Conrad and Kaul (1998) argue that momentum prots reect cross-sectional variation in expected returns rather than predictable time-series variation in stock returns. They argue that the momentum strategys average prots reect 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 reect expected returns, then the momentum prots can be attributed to cross-sectional differences in risk characteristics (i.e., common identied or as yet unidentied priced risk factors).4 If common risk factors play a role in explaining momentum prots 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 prots, many authors nd that the estimated coefcients from regressions of momentum returns on commonly used risk factors are indistinguishable 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 coefcients, so a simple examination 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 prots arise from inherent biases in the way that investors interpret information (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 nd 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 prots, 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 rm value to which investors underreact must be correlated across markets. 4 Many researchers have examined various factors as potential explanations for momentum prots. These include market beta, rm size, book-to-market, cash-ow-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; Grifn 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 ows, 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).
278
that momentum in industry risk factors explains the protability of momentum strategies. That is, momentum prots in individual stocks disappear after controlling for industry effects. However, Grundy and Martin (2001) and Asness et al. (2001) nd that the protability of momentum strategies is not fully explained by industry risk exposure. In particular, Grundy and Martin (2001) nd that neither industry effects nor cross-sectional differences in expected returns explain the level of momentum prots. Rather, the strategies protability reects momentum in the stock-specic (idiosyncratic) component of returns. Asness et al. (2001) also nd both an industry effect and a rm effect independent 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 comovement 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 crosscountry comovement in the payoffs to country-neutral momentum strategies when markets 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 if information ows 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 ows 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 nd that skipping a month between portfolio ranking and investment periods avoids confounding bid/ask bounce and nonsyncronous trading effects. We dene winning and losing stocks as those in the top and bottom three deciles, respectively. The self-nancing 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 prot 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 ow data for only a 4 12 year sample period from 1994 to 1998. They nd that there is a small correlation in contemporaneous cross-country ows, which suggests that comovement in ows cannot explain comovement in momentum returns. 7 As further robustness checks, we also employed FamaMacBeth regression techniques in place of momentum portfolio sorts, dened 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.
279
We rst examine the protability 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 rms in over 50 countries. Our initial sample includes all rms 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 rms 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 rms home country exchange (i.e., crosslistings and depository receipts).9 We also exclude from our portfolios any rm whose market capitalization is below the 25th percentile market capitalization of all NYSE stocks, converted to the local currency, in any month during the ranking period. These rms are excluded to avoid signicant liquidity constraints, to mitigate any survivorship bias problems, and as Ince and Porter (2006) demonstrate, because many of the difculties 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 nding signicant momentum returns. However, our country-neutral momentum results are consistent with those reported by Jegadeesh and Titman (1993, 2001), Rouwenhorst (1998, 1999), Chui et al. (2003), and Grifn 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 return distribution for each emerging market in each month during the ranking period. Finally, to insure that we have a series of sufcient 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 rms representing 101 industries and 40 countries over the 19752004 time period.10 A strength of TDS is that they use the same criteria for dening industries across countries. As argued by Grifn and Stulz (2001), this consistent classication minimizes the risk of nding low crosscountry industry comovement because of rm misclassication. Grifn and Karolyi (1998) argue that using broad industrial classications 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 classication, level 6, which includes approximately 100 separate industry denitions.11
8 For example, in the U.S., we include only rms 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) nd signicant 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 rms sorted into industries by their 4 digit SIC classication and using TDSs level six industry denitions. 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.
280
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 of Table 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 rms, the broadest industry representation, and the largest market capitalizations of all the countries in our sample. Together, these three countries represent over 55% of the rms in the sample and 48% of December 2004 market capitalization. Although much smaller on average, the larger emerging markets have both more rms 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 protability of momentum returns 4.1. Protability of unadjusted momentum returns across countries The rst three columns of Table 2 report unadjusted momentum prots for portfolios formed on within 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 signicant momentum prots in many, but not all countries. We nd that 14 of 22 developed and 5 of 18 emerging markets have a signicant 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 signicant momentum effect. Chui et al. (2003) and Grifn et al. (2003) also nd an insignicant momentum effect for Japan.13 We also do not nd a signicant momentum effect for several of the other Asian and Scandinavian countries. Although the proportion of emerging markets with signicant 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 rms in emerging markets. 10 of 18 emerging markets have higher average momentum returns than the smallest signicant 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 classied as winners when ranked within country and shorting all stocks classied as losers. We also form a similar portfolio for emerging markets. We nd statistically signicant 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 rm 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 rms), Japan (2,199) and U.K. (1,496). Each of these markets contains more than twice the number of rms than the largest country included in the gure. 13 Though not reported, if we exclude the two largest countries (Japan and the U.S.) from our sample, both the magnitude and signicance level of the average, country-neutral momentum effect increases. 14 Bris et al. (2007), however, show that short selling is difcult to implement and not feasible in many markets.
A. Naranjo, B. Porter / Journal of International Money and Finance 29 (2010) 275299 Table 1 Summary statistics by country: 19752004. Starting date Developed markets Australia Austria Belgium Canada Denmark Finland France Germany Hong Kong Ireland Italy Japan Luxembourg Netherlands New Zealand Norway Singapore Spain Sweden Switzerland UK USA All developed Emerging markets Argentina Brazil Chile Greece India Indonesia Israel Korea Malaysia Mexico Philippines Poland Portugal Russia South Africa Taiwan Thailand Turkey All emerging All Number of rms Number of industries 77 40 49 83 37 45 88 80 69 29 64 89 17 57 28 40 44 52 61 69 95 101
281
Market capitalization (December 2004, $Bil) $632 $427 $231 $883 $134 $166 $1,535 $1,311 $710 $102 $725 $3,508 $162 $583 $29 $126 $192 $623 $330 $131 $2,586 $14,192 $29,318
1975:01 1975:01 1975:01 1975:01 1975:01 1987:01 1975:01 1975:01 1975:01 1975:01 1975:01 1975:01 1991:03 1975:01 1986:01 1975:01 1975:01 1986:01 1975:01 1975:01 1975:01 1975:01
364 107 124 610 113 100 658 622 384 46 300 2,199 30 206 45 156 156 167 316 403 1,496 6,036 14,638
1992:01 1993:01 1989:07 1988:01 1988:07 1990:04 1989:11 1980:10 1975:01 1988:01 1976:11 1993:11 1988:01 1994:01 1985:01 1987:09 1987:01 1988:03
47 151 86 211 324 133 48 304 318 132 86 25 62 61 117 441 175 90 2,811 17,449
22 42 32 56 59 39 25 65 70 37 22 16 26 18 47 60 53 31
$31 $163 $82 $93 $285 $50 $56 $330 $117 $147 $17 $53 $65 $146 $210 $279 $76 $66 $2,268 $31,586
We list countries as Developed or Emerging based on the International Finance Corporations (IFC) categorizations. Start date is the rst month that a rm meeting all the data requirements outlined in Section 3 enters the sample. Number of Firms and Industries are the total number of unique rms and industries represented in the sample. Market capitalization is the total value (in billions of U.S. dollars) of rms in the sample in December 2004.
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 signicantly comove across countries, then country-neutral momentum returns would be correlated.
282
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 rms represented in the sample. The largest three developed markets: U.S., Japan, and U.K are not shown.
We create country-neutral industry momentum portfolios and industry-adjusted momentum portfolios by rst creating equally-weighted industry portfolios within each country using level 6 TDS industry classications. We require each industry portfolio to have a minimum of three rms. We create industry-adjusted returns by taking the difference between the rms monthly return and the return on its country-specic 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 industryadjusted returns over Past(2,12) and going long the top 30% and short the bottom 30%. The last six columns of Table 2 show returns to both across-industry and within-industry (industryadjusted) momentum portfolios. For developed markets, we nd signicant across-industry momentum for 7 of the 19 countries with sufcient data to calculate across-industry momentum. Of these 7 countries, 6 also have a signicant average unadjusted momentum return. We also nd 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 the unadjusted market-wide momentum results reported earlier for the developed markets, we also nd time-variation in across-industry momentum. There is sufcient data to calculate industry momentum for only 9 of the 18 emerging markets. We nd no evidence of signicant positive across-industry momentum in emerging markets; however, the number of industries in each market and the average number of rms 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 of Table 2 show returns to within-industry (industry-adjusted) momentum. For the developed markets, 6 of 20 countries have a signicant positive within-industry momentum effect. Of these 6, all also have signicant unadjusted momentum and 3 also have signicant positive across-industry momentum returns. For the emerging markets, 2 of 16 countries have a signicant positive within-industry momentum effect, one of which also has signicant 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).
283
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 Average t-stat monthly return Developed markets Australia Austria Belgium Canada Denmark Finland France Germany Hong Kong Ireland Italy Japan Luxembourg Netherlands New Zealand Norway Singapore Spain Sweden Switzerland United Kingdom USA Total developed: 19751980 19811985 19861990 19911995 19962000 20012004 0.98 0.15 0.94 0.83 0.88 0.25 0.83 0.73 0.74 1.18 1.42 0.08 0.58 0.90 0.71 0.54 0.38 0.54 0.32 0.68 0.71 0.66 0.54 0.87 0.32 0.46 0.35 0.94 0.23 4.19 0.40 4.08 3.54 2.98 0.45 3.44 3.09 2.65 2.21 4.73 0.32 0.90 3.38 1.30 1.18 1.13 2.48 0.85 3.75 3.14 2.96 n Returns versus local market index Across-industry Winner t-stat portfolio Loser t-stat portfolio Average t-stat monthly return 1.09 0.42 1.15 1.59 0.85 0.32 3.07 2.39 0.12 n Within-industry Average t-stat monthly return 2.91 0.93 1.30 2.48 1.45 0.05 1.93 1.53 1.19 n
347 0.37 198 0.16 347 0.33 347 0.33 347 0.31 189 0.35 347 0.40 347 0.32 347 0.43 174 0.33 347 0.73 347 0.00 103 0.27 347 0.36 121 0.44 236 0.22 325 0.06 194 0.42 263 0.21 347 0.20 347 0.33 347 0.33 0.24 0.40 0.09 0.16 0.14 0.46 0.18
3.06 0.82 2.73 3.13 2.14 1.35 3.76 3.16 2.69 1.27 4.82 0.03 0.80 2.82 1.76 1.01 0.30 3.20 1.26 2.34 3.68 3.85
0.61 0.31 0.61 0.49 0.57 0.10 0.42 0.41 0.31 0.85 0.68 0.08 0.31 0.54 0.26 0.32 0.33 0.12 0.11 0.48 0.38 0.33
4.33 0.29 1.07 0.38 3.92 0.49 3.28 0.47 2.85 1.15 0.27 0.72 2.74 0.82 2.54 0.67 1.75 0.05 2.24 3.65 1.32 0.62 0.41 0.70 3.24 1.13 0.67 0.96 0.61 1.43 0.08 0.81 0.80 0.44 0.90 4.13 0.21 2.49 0.64 2.17 0.67 2.77 3.51 1.71 1.91 1.26 1.27 0.12 0.56 0.89 0.19 0.42 0.45 0.94 0.48
347 0.55 44 0.50 176 0.42 347 0.45 87 0.71 12 0.09 347 0.36 347 0.25 287 0.31
347 80 223 347 161 38 347 347 347
3.06 306 0.47 1.55 347 0.18 1.51 2.65 281 0.57 0.71 0.12 1.82 1.96 0.76 2.74 3.34 111 0.80 211 0.30 161 0.39 200 0.44 347 0.31 347 0.47 347 0.48 0.25
1.71 306 1.05 347 2.92 4 2.09 296 1.50 0.95 1.92 1.45 2.03 3.82 3.70 144 253 192 212 347 347 347
3.34 347 3.47 1.38 1.78 1.41 1.70 0.33 59 60 60 60 60 48
3.90 0.30 3.16 0.82 1.28 1.52 2.24 0.78 0.47 0.23 0.30 0.21 0.48 0.06
3.88 347 3.96 0.77 1.57 2.29 2.00 0.75
2.53 347 3.11 0.84 1.07 0.42 1.95 0.13 59 60 60 60 60 48
59 0.48 60 0.14 60 0.19 60 0.09 60 0.60 48 0.05
Emerging markets Argentina 0.19 Brazil 1.34 Chile 0.30 Greece 1.23 India 0.61 Indonesia 1.10 Israel 0.75 Korea 0.53 Malaysia 0.31 Mexico 1.29 Philippines 2.49 Poland 1.48 Portugal 1.27 Russia 0.76 South Africa 0.97 Taiwan 0.40 Thailand 1.39 Turkey 0.45 Total emerging: 19861990 19911995 19962000 0.75 0.81 0.75 0.62
0.31 1.71 0.85 2.18 1.57 1.78 1.54 1.18 0.99 3.30 0.62 1.60 2.50 0.54 2.63 1.10 2.28 0.44
75 92 138 123 167 105 82 168 182 184 80 33 98 29 203 181 145 88
0.15 0.57 0.49 0.64 0.36 0.79 0.46 0.20 0.21 0.55 0.69 0.50 0.96 0.77 0.47 0.31 1.15 0.02 0.34 0.26 0.31 0.34
0.48 1.13 2.39 2.24 1.69 2.25 1.72 0.86 1.18 2.14 1.76 1.20 3.54 0.80 2.45 1.50 3.61 0.04
0.05 0.77 0.19 0.59 0.25 0.31 0.28 0.33 0.10 0.74 3.18 0.98 0.31 0.01 0.50 0.10 0.23 0.42
0.09 1.77 0.77 0.86 1.38 0.90 1.01 0.67 0.24 0.88 1.15 1.35 0.45 0.40 2.67 0.47 0.80 1.27 0.86 0.01 2.05 1.95 0.33 0.05 0.91 0.59 0.56 2.17 0.02 0.92 1.24 1.96 0.19 0.81 0.60
0.83 1.33 0.21
3 87 32
2.28 106 1.00 140 0.47 57
3.49 2.35 0.31 0.57 0.63 0.61 0.46 0.03 0.14 1.02 0.47
1.35 5 2.29 45 0.73 74 0.53 37 1.15 114 0.64 35 0.27 3 0.08 125 0.45 158 1.63 81 0.39 29 1.23 14 0.77 9 1.92 95 1.82 168 2.53 63 3.96 194 1.16 0.38 2.59 26 60 60
1.90 3.94 1.56 58 0.94 0.10 165 0.54 0.61 30 2.11 0.05 194 1.34 0.70 1.05 26 60 60 0.72 0.70 0.10 0.94
3.64 225 1.23 3.29 1.80 57 60 60
2.23 0.41 0.54 0.55 1.65 0.44 1.34 0.28
(continued on next page)
284 Table 2 (continued)
Unadjusted Average t-stat monthly return 20012004 Total All: 19751980 19811985 19861990 19911995 19962000 20012004 0.84 0.55 0.87 0.32 0.46 0.38 0.90 0.31 3.15 n
Returns versus local market index Across-industry Winner t-stat portfolio 48 0.46 0.24 0.40 0.09 0.13 0.15 0.49 0.19 Loser t-stat portfolio Average t-stat monthly return 0.41 n
Within-industry Average t-stat monthly return 1.23 0.25 0.49 0.14 0.19 0.08 0.58 0.01 3.60 n
2.75 0.38 4.08 0.30 3.16 0.87 1.01 1.69 2.46 0.91 0.47 0.23 0.32 0.23 0.41 0.12
1.92 0.38 2.96 3.51 1.67 2.04 1.42 1.18 0.25 0.55 0.89 0.19 0.40 0.42 0.89 0.47
48
48
3.57 347 3.47 1.38 1.80 1.64 1.75 0.46 59 60 60 60 60 48
3.83 347 3.96 0.77 1.51 2.24 1.95 0.75 59 60 60 60 60 48
2.66 347 3.16 0.84 1.08 0.42 1.99 0.03 59 60 60 60 60 48
We report average monthly momentum prot 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 dened 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 dened as the top and bottom three deciles, 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 prot 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 dened as the top 30% of stocks during the ranking period and losers the bottom 30%. The average monthly prot 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 rm 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 rms. Across-industry strategies must have a minimum of four valid industries from which winning and losing industries are selected.
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 are nearly identical to those for developed markets due to the large fraction of rms traded in developed markets. We also examine alternative methods for constructing industry portfolios, including valueweighting 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 nished product is tradable. See Grifn 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 momentum returns. 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 coefcients are time varying, we estimate momentum portfolio factor coefcients from individual stock factor coefcients. 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 specic factor loadings. The
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.
285
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 tted values for time t using 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 form momentum portfolios is slightly smaller than that used in Table 2. For comparison purposes, the rst 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 signicant momentum returns, the average tted value is positive and signicant for only 2. The fraction of unadjusted momentum return variance that is explained by the variance of the tted 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 tted values are signicant in only 3 of the developed countries. In unreported results, we nd that of the 13 developed markets with signicant momentum returns, 10 have signicant average factor loadings on one or more of the factors, although their magnitudes are too low to explain the level of momentum prots. The bottom of Table 3 provides the country-specic risk adjustment results for the emerging markets. We again nd that the country-level single and multi-factor models do a poor job of explaining average momentum returns. Of the 5 countries with signicant momentum returns, the tted values are signicant for only one. The fraction of unadjusted momentum return variance that is explained by the variance of the tted 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 specic Small and Big portfolio U.S. dollar returns, calculates the difference, and then converts 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 Grifn (2002) who nds 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 signicant 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 rm 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 comovement of momentum returns. 5.1. Correlation of unadjusted momentum returns across countries Table 4 provides evidence that country-neutral momentum returns signicantly 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 coefcients 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 tted return/Variance of the total return)*100. Since the coefcients are estimated out of sample, it is possible to have the variance of the tted exceed the variance of the total. 19 In unreported results, we also nd that the factor models do a poor job of explaining the level of across-industry and within-industry momentum.
286
Table 3 Risk-adjusted momentum returns 19752004. Unadjusted momentum returns Avg Developed markets Australia Austria Belgium Canada Denmark Finland France Germany Hong Kong Ireland Italy Japan Luxembourg Netherlands New Zealand Norway Singapore Spain Sweden Switzerland UK USA Total developed: 0.99 0.14 1.02 0.87 0.86 0.27 0.80 0.74 0.75 1.06 1.34 0.01 0.15 0.89 0.61 0.60 0.36 0.51 0.25 0.67 0.72 0.63 0.52 t-stat 4.12 0.38 4.24 3.63 2.85 0.45 3.26 3.21 2.64 1.84 4.35 0.04 0.21 3.31 1.00 1.26 1.03 2.40 0.67 3.47 3.13 2.83 3.22 n 334 192 334 334 334 177 334 334 334 161 334 334 89 334 88 230 312 187 251 334 334 334 334 Model 1 Fitted 0.05 0.32 0.01 0.14 0.31 0.45 0.06 0.25 0.30 0.21 0.26 0.02 0.29 0.09 0.04 0.05 0.06 0.26 0.24 0.12 0.05 0.02 0.06 t-stat 0.45 1.59 0.12 0.95 2.34 1.50 0.51 1.63 1.58 1.46 1.68 0.15 1.67 0.98 0.22 0.27 0.35 2.26 1.28 1.23 0.65 0.16 0.68 Errors 1.05 0.18 1.01 0.74 0.55 0.18 0.74 0.49 0.46 0.85 1.08 0.01 0.44 0.80 0.65 0.55 0.30 0.25 0.01 0.55 0.66 0.61 0.46 t-stat 4.99 0.49 4.55 3.56 1.91 0.32 3.47 3.07 1.82 1.51 3.93 0.04 0.59 3.23 1.08 1.19 0.94 1.23 0.04 3.15 3.18 3.42 3.68 Model 2 % VarE Fitted 25.02 27.82 9.64 34.92 19.35 24.80 25.99 45.60 43.66 6.04 25.77 21.98 5.57 11.77 8.44 13.61 23.50 28.64 24.76 26.84 11.97 29.93 28.19 t-stat Errors 0.87 0.43 0.62 0.70 0.20 0.47 0.77 0.56 0.38 0.46 0.44 0.07 0.73 0.68 1.32 0.30 0.25 0.08 0.43 0.63 0.55 0.42 t-stat 4.08 1.07 2.41 3.52 0.48 0.76 3.74 3.67 1.34 0.62 1.61 0.34 0.61 2.83 1.47 0.73 1.21 0.26 2.50 3.38 3.18 3.69 % VarE 47.87 44.31 38.51 43.57 53.20 52.07 44.81 74.24 61.32 36.00 57.43 54.76 12.73 28.98 49.40 46.98 49.36 29.47 49.43 54.29 65.56 55.25
0.12 0.71 0.04 0.12 0.40 2.21 0.17 1.08 0.28 0.87 0.53 1.02 0.01 0.06 0.18 0.92 0.30 1.31 0.71 1.51 0.62 2.48 0.06 0.32 0.30 0.69 0.21 1.44 0.68 0.90 0.41 1.15 0.26 1.72 0.44 2.05 0.22 1.59 0.08 0.49 0.08 0.46 0.08 0.63
Emerging markets Argentina 0.07 0.10 64 0.10 0.57 0.03 1.76 Brazil 1.67 1.88 64 0.09 0.24 0.90 136 0.07 0.62 0.41 Chile 0.33 Greece 1.33 2.28 109 0.31 1.57 1.02 India 0.34 0.80 155 0.16 1.31 0.18 Indonesia 0.46 0.68 100 0.39 1.10 0.85 Israel 0.51 0.80 56 0.25 1.53 0.26 Korea 0.42 0.89 162 0.06 0.42 0.48 Malaysia 0.27 0.85 179 0.28 1.76 0.01 Mexico 1.07 2.64 170 0.01 0.07 1.08 Philippines 2.99 0.61 65 0.77 0.83 3.76 Poland 1.48 1.43 27 1.03 2.95 0.45 Portugal 1.46 2.55 81 0.83 3.17 0.63 Russia 0.54 0.38 27 0.06 0.11 0.60 South Africa 0.94 2.43 190 0.05 0.66 0.99 Taiwan 0.41 1.07 168 0.14 0.96 0.56 Thailand 1.53 2.34 137 0.19 1.00 1.34 Turkey 0.14 0.14 83 0.72 2.98 0.86 Total emerging: 0.78 2.17 261 0.27 2.23 0.50
0.04 2.15 1.17 1.79 0.41 1.19 0.40 1.02 0.02 3.02 0.76 0.45 1.11 0.47 2.47 1.50 2.12 0.86 1.51
5.92 19.36 10.51 11.41 8.73 26.72 6.71 8.70 24.68 26.06 3.59 11.50 20.76 13.24 4.21 15.03 8.52 5.87 11.80
0.18 0.59 0.18 0.34 0.34 1.25 0.47 1.37 0.24 1.41 0.58 0.49 0.51 0.60 0.67 2.89 0.18 0.70 0.29 1.13 0.85 1.18 0.88 2.21 0.66 0.87 0.42 2.53 0.02 0.07 0.46 0.83 0.39 2.35
0.45 1.49 0.06 1.42 0.10 0.14 0.72 0.94 0.05 0.63 11.60 1.02 1.73 0.52 0.43 1.91
0.57 1.67 0.13 2.03 0.23 0.09 0.44 1.86 0.18 1.85 0.89 1.23 1.11 1.27 1.22 2.77
15.60 34.41 31.54 21.82 15.95 149.32 17.59 20.10 47.62 33.84 0.32 19.35 22.29 18.53 32.07 43.90 16.59
0.06
0.28
Model 1 Country-specic market model. Model 2 Country-specic 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-specic market return is the value-weighted average of all stocks local currency returns where there are three or more rms. For the FamaFrench three factor country-specic 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 factor loadings, the sample used to form momentum portfolios is slightly smaller here than that used in Table 2. For comparative purposes, the rst column reports the average unadjusted momentum returns from using this smaller sample. %VarE (Variance of the tted return/Variance of the total return) * 100.
Table 4 Average pairwise correlations of unadjusted momentum returns across countries: portfolios formed on Past(2,12) ranking period and held for one month. Average of all correlations n t-statistic Average of signicant correlations # of signicant () correlations # of signicant () correlations Range of signicant () correlations Average local market index correlations t-statistic
Panel A: Average pairwise correlations of unadjusted momentum returns for developed markets (19752004) Developed markets Australia Austria Belgium Canada Denmark Finland France Germany Hong Kong Ireland Italy Japan Luxembourg Netherlands New Zealand Norway Singapore Spain Sweden Switzerland United Kingdom USA Total developed: 19751980 19811985 19861990 19911995 19962000 20012004 0.19 0.18 0.18 0.27 0.20 0.24 0.34 0.32 0.15 0.21 0.23 0.15 0.27 0.30 0.12 0.19 0.13 0.18 0.28 0.31 0.33 0.32 0.23 0.11 0.02 0.08 0.16 0.27 0.41 21 21 21 21 21 21 21 21 21 21 21 21 21 21 21 21 21 21 21 21 21 21 231 91 105 136 190 231 210 11.03 10.56 12.15 11.22 11.61 10.20 12.41 10.78 8.66 9.41 12.72 9.39 11.85 11.78 7.25 10.13 8.07 10.31 9.97 14.97 11.68 10.86 29.19 8.30 1.43 5.29 14.40 21.62 28.05 0.21 0.22 0.21 0.28 0.23 0.28 0.35 0.33 0.20 0.26 0.24 0.20 0.33 0.31 0.25 0.23 0.18 0.22 0.34 0.31 0.34 0.33 0.27 0.33 0.23 0.34 0.35 0.42 0.51 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 0 0 0 0 16 14 16 20 16 17 20 20 12 14 19 11 15 20 2 17 11 15 16 21 20 20 176 11 7 21 54 123 149 0.130.39 0.140.33 0.120.27 0.110.53 0.160.31 0.160.43 0.190.55 0.120.55 0.120.32 0.150.34 0.150.41 0.130.27 0.210.45 0.110.51 0.230.27 0.130.30 0.110.29 0.150.26 0.170.48 0.180.50 0.180.55 0.160.55 0.110.55 0.260.51 0.260.59 0.260.53 0.260.54 0.260.74 0.290.88 0.36 0.32 0.43 0.41 0.35 0.46 0.12 0.47 0.32 0.12 0.38 0.27 0.45 0.49 0.34 0.44 0.33 0.44 0.43 0.53 0.48 0.45 0.38 0.27 0.30 0.50 0.45 0.41 0.56 11.01 12.02 13.40 12.83 12.89 14.11 12.14 13.65 10.78 7.24 12.87 17.50 12.20 13.90 12.71 15.99 12.26 11.76 13.26 15.16 16.56 14.80 34.21 17.07 24.69 48.84 41.24 26.50 41.42
(continued on next page) 287
Table 4 (continued) Average of all correlations n t-statistic Average of signicant correlations # of signicant () correlations # of signicant () correlations Range of signicant () correlations Average local market index correlations t-statistic
288
Panel B: Average pairwise correlations of unadjusted momentum returns for emerging markets (19902004) Emerging markets Argentina Brazil Chile Greece India Indonesia Israel Korea Malaysia Mexico Philippines Poland Portugal South Africa Taiwan Thailand Turkey Total emerging: 19911995 19962000 20012004 0.02 0.08 0.02 0.03 0.05 0.00 0.01 0.01 0.01 0.05 0.06 0.07 0.00 0.01 0.00 0.03 0.01 0.01 0.02 0.01 0.05 15 16 16 16 16 15 14 16 16 16 14 11 15 16 16 16 16 130 44 78 86 0.73 2.56 0.76 0.69 2.25 0.13 0.68 0.45 0.34 2.26 1.52 1.21 0.15 0.58 0.12 1.26 0.26 1.41 0.99 0.61 2.96 0.29 0.21 0.22 0.02 0.21 0.27 0.23 0.04 0.27 0.22 0.21 0.21 0.21 0.02 0.32 0.17 0.32 0 0 1 1 0 0 0 1 0 1 1 2 1 4 1 1 0 1 1 0 1 1 1 1 1 1 0 0 0 0 4 0 2 2 0.210.29 0.380.50 0.310.32 0.290.29 0.210.21 0.230.23 0.210.21 0.270.27 0.230.23 0.290.29 0.270.27 0.21 0.13 0.16 0.10 0.14 0.25 0.12 0.14 0.22 0.24 0.03 0.26 0.19 0.14 0.17 0.18 0.12 0.17 0.13 0.21 0.26 7.43 3.74 6.83 4.46 5.28 7.91 2.45 8.99 6.68 7.32 2.24 8.54 4.28 7.10 6.65 5.56 5.32 15.38 8.41 12.28 20.74
Panel C: Average pairwise correlations of unadjusted momentum returns across developed and emerging markets All Emerging w/developed 19901995 Emerging w/developed 19962000 Emerging w/developed 20012004 Emerging w/developed 0.08 0.04 0.04 0.12 373 200 308 294 12.58 3.27 4.22 11.64 0.20 0.23 0.24 0.34 5 3 5 2 62 18 27 54 0.140.50 0.260.43 0.260.52 0.290.50 0.22 0.17 0.27 0.36 31.65 17.05 27.88 43.19
We report correlations of unadjusted momentum returns for portfolios of winners minus losers from 1975 to 2004 for developed countries in Panel A and from 1990 to 2004 for emerging markets in Panel B. In Panel C, we report average correlations of unadjusted momentum returns between developed and emerging markets. Winners and losers are dened based on average monthly return during the ranking period. Unadjusted refers to the ranking of all stocks by total return within a country during the ranking period. Winners and losers are dened as the top and bottom 30% of rms. To be included in a portfolio, a rm must be above the 25th percentile market capitalization of all NYSE stocks in every month of the ranking period. All returns are in local currency.
289
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 nd that the average pairwise correlations of countryneutral momentum returns with the other developed markets are signicant. The average of the pairwise 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 nd that for each country well over half of the pairwise correlations are individually signicant.20 Seven countries have 20 or more of a possible 21 positive and signicant pairwise correlations, while only one developed country, New Zealand, has less than 10 positive and signicant pairwise correlations. No pair of developed countries has a signicant negative correlation in their momentum returns. The range of statistically signicant pairwise correlations for the developed markets in the eighth column shows that the averages often mask the magnitude of the relationship between countryneutral 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 the comovement 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 Grifn 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, signicant, and larger than the average momentum return correlations. Panel B of Table 4 provides the average pairwise correlations for the emerging markets. The magnitude and signicance 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 of each emerging market with the others ranges from 0.06 to 0.08. The average market index correlations are all signicantly 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 signicant, 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 primarily among developed countries. This pattern 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 of comovement 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-specic 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 prots 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 ndings of others. See, for instance, Goetzmann et al. (2005) and Longin and Solnik (1995, 2001).
290
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 acrossindustry 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 for countries 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 portfolios 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 nd 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-specic 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 nd that the correlations of the tted returns are large and signicant. The average correlation across countries of the tted 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 signicant, the average correlation of 0.16 is less than half as big as the average tted correlations. However, we must be careful in comparing the correlations of the tted 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 RFit RErr ; RFit RErr B B A A h i Cov RFit ; RFit Cov RErr ; RErr Cov RFit ; RErr Cov RErr ; RFit B B B B A A A A 1 Percent Fit Percent Error Percent Other; (1)
where RA is the country-neutral momentum of country A, RFit is the return explained by the model, and A RErr is the return not explained by the model. RFit and RErr are not orthogonal since the factor loadings A A A 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 insufcient data, many of the models cannot be estimated for the emerging markets, and the t is generally poorer when the estimates can be obtained.
291
The covariance of the tted 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 tted values. The sub-sample results show that there is considerable time variation in the tted and error correlations and conrms that the tted 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 tted 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 tted 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 tted and error correlations. The proportion of the covariance explained by the tted values and errors also varies over time, with the tted 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 qualitatively 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 t of these models relative to the domestic versions. These results are consistent with the relatively poorer t of global factor models relative to the domestic versions as reported by Grifn (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 rst 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 rst 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 sufcient 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 of Table 7. The rst 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 fth 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 signicant coefcient. In other words, countries whose risk factors have high covariance also tend to have high covariance of countryneutral 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 specic time-series means:
3i;t Ri;t Ri ; gRMRF RMRF i;t RMRF i i;t
(2)
292
Table 5 Average pairwise correlations of across-industry and within-industry momentum returns across countries: portfolios formed on Past(2,12) ranking period and held for one month. Pairwise correlations of industry and industry-adjusted momentum returns Across-industry momentum All correlations Average n t-stat. Individually signicant correlations Average N () N () Range Within-industry-adjusted momentum All correlations Average n t-stat. Individually signicant correlations A. Naranjo, B. Porter / Journal of International Money and Finance 29 (2010) 275299 Average N () N () Range
Panel A: Developed markets (19752004) Developed markets Australia Austria Belgium Canada Denmark Finland France Germany Hong Kong Ireland Italy Japan Luxembourg Netherlands New Zealand Norway Singapore Spain Sweden Switzerland UK USA Total developed: 19751980 19811985 19861990 19911995 19962000 20012004 0.07 0.00 0.03 0.13 0.01 0.12 0.12 0.10 0.10 0.09 0.14 0.02 0.06 0.05 0.10 0.11 0.17 0.03 0.08 0.06 0.05 0.04 0.07 0.17 0.21 17 16 17 17 16 17 17 17 17 17 17 17 17 17 17 17 17 17 152 45 36 91 136 136 91 2.61 0.07 1.75 4.32 0.42 3.41 4.40 3.83 3.14 5.56 4.23 0.83 3.73 2.74 4.66 6.02 4.51 0.97 8.17 2.47 1.65 2.23 5.15 8.62 10.34 0.22 0.00 0.23 0 1 0 4 1 8 0.130.31 0.310.31 0.160.35 0.10 0.10 0.16 0.13 0.05 0.17 0.18 0.23 0.06 0.16 0.14 0.17 0.12 0.15 0.09 0.13 0.16 0.21 0.22 0.14 0.09 0.06 0.06 0.12 0.15 0.34 18 17 18 18 18 17 18 18 18 18 18 18 18 18 18 18 18 18 18 170 45 55 91 153 153 105 7.43 2.59 4.44 6.59 2.48 5.92 7.05 7.07 3.34 6.35 9.22 5.55 5.96 7.15 4.36 6.92 7.20 7.19 9.29 16.50 4.81 2.70 4.26 10.19 11.58 16.44 0.14 0.33 0.28 0.18 0.23 0.38 0.25 0.29 0.17 0.22 0.17 0.25 0.21 0.21 0.18 0.20 0.20 0.25 0.26 0.23 0.10 0.19 0.24 0.31 0.35 0.50 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 0 0 8 3 9 10 1 2 11 13 4 11 9 11 4 9 3 7 12 14 13 77 2 4 6 29 37 59 0.120.18 0.280.37 0.190.39 0.130.32 0.230.23 0.330.42 0.140.43 0.120.43 0.120.28 0.140.32 0.140.28 0.130.42 0.180.23 0.130.34 0.170.20 0.140.24 0.120.33 0.130.39 0.180.39 0.120.43 0.300.32 0.270.40 0.260.44 0.250.51 0.260.54 0.290.77
0.15 0.23 0.20 0.20 0.17 0.23
1 0 0 0 0 0
9 8 8 8 4 10
0.130.34 0.160.35 0.160.29 0.120.32 0.150.21 0.130.38
0.19 0.19 0.19 0.18 0.26 0.16 0.20 0.24 0.16 0.16 0.23 0.45 0.41
0 0 0 0 0 0 1 1 2 2 2 0 1
1 1 5 8 12 5 46 4 3 6 13 42 29
0.190.19 0.190.19 0.160.24 0.120.22 0.150.38 0.120.22 0.120.38 0.310.47 0.440.52 0.260.37 0.260.42 0.290.70 0.300.66
Panel B: Emerging Markets (19902004) Emerging markets Argentina Brazil Chile Greece India Indonesia Israel Korea Malaysia Mexico Philippines Poland Portugal Russia South Africa Taiwan Thailand Turkey Total Emerging: 19911995 19962000 20012004
0.03 0.06 0.03 0.04 0.09
7 3 6 7 4
0.34 0.36 1.39 0.91 1.35
0.38 0.38
0 0
1 1
0.38 0.38 0.38 0.38
0.06 0.16 0.08 0.03 0.00 0.04 0.06 0.08
4 7 2 9 5 7 9 7
0.69 2.77 0.33 0.45 0.03 0.80 1.44 1.42
0.34 0.34 0.35
0 0 0
2 1 1
0.340.35 0.340.34 0.350.35 A. Naranjo, B. Porter / Journal of International Money and Finance 29 (2010) 275299
0.00 0.01 0.05 0.02 0.02 0.03 0.07
4 7 4 21 6 3 1
0.04 0.21 0.43 0.52 0.57 0.35 0.38 0 1 0.380.38
0.01 0.02 0.06 0.02 0.00 0.09 0.10
7 10 7 37 10 5 11
0.15 0.42 0.85 0.75 0.09 0.81 1.54
0.30 0.30
0 0
1 1
0.300.30 0.300.30
0.33 0.39 0.39
0 0 0
3 1 2
0.300.35 0.390.39 0.300.48
Panel C: Across developed and emerging markets All Emerging w/Developed 19901995 Emerging w/Developed 19962000 Emerging w/Developed 20012004 Emerging w/Developed 0.03 0.04 0.06 0.09 134 66 63 28 2.81 2.20 2.92 3.11 0.06 0.21 0.02 0.01 4 1 1 1 9 6 1 1 0.160.41 0.280.34 0.330.33 0.320.32 0.03 0.03 0.02 0.07 192 90 86 90 2.95 1.84 1.20 4.50 0.13 0.19 0.40 0.37 4 2 0 0 17 8 6 5 0.170.32 0.270.39 0.300.49 0.330.42
We report average correlations of across-industry and within-industry momentum portfolio returns from 1975 to 2004 for developed countries in Panel A and from 1990 to 2004 for emerging markets in Panel B. In Panel C, we report average correlations of across-industry and within-industry momentum portfolio returns between developed and emerging markets. Winners and losers are dened based on total return during the ranking period. 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 prot from a strategy of buying an equal-weighted portfolio of winning industry portfolios and shorting the losing industries and holding for 1 month is reported. Within-Industry portfolios refer to the ranking of all stocks by difference in return from their equal-weighted industry average. Winners are dened as the top 30% of stocks during the ranking period and losers the bottom 30%. The average monthly prot from a strategy of buying an equal-weighted portfolio of winners and shorting an equally-weighted portfolio of losers and holding the position for 1 month is reported. To be included in a portfolio a rm must be above the 25th percentile of market capitalization of all NYSE stocks, converted to the local currency, in every month of the portfolio formation period. All returns are in local currency. Industry portfolios have a minimum of three rms. Across-industry strategies must have a minimum of four valid industries from which winning and losing industries are selected.
293
294
Table 6 Average pairwise correlations of risk-adjusted momentum returns across countries: portfolios formed on Past(2,12) ranking period and held for one month (19752004). Correlations of unadjusted momentum returns All Avg t-stat Signicant pairs Avg t-stat n Range Correlation of tted values All Avg t-stat Signicant pairs Avg t-stat n Range Correlation of model residuals All Avg t-stat Signicant pairs Avg t-stat n Range Cov. decomposition All % Fit % Err % Other
Panel A: Correlations of risk-adjusted momentum returns using country-specic market model Australia Austria Belgium Canada Denmark Finland France Germany Hong Kong Ireland Italy Japan Luxembourg Netherlands New Zealand Norway Singapore Spain Sweden Switzerland UK USA Total 19751980 19811985 19861990 19911995 19962000 20012004 0.19 0.20 0.22 0.26 0.33 0.38 0.37 0.17 0.26 0.27 0.18 0.35 0.37 0.15 0.22 0.33 0.36 0.37 0.35 0.28 0.13 0.04 0.08 0.19 0.28 0.44 9.74 8.53 10.17 9.03 10.51 12.44 11.65 8.03 9.77 8.48 12.78 12.19 10.08 5.80 9.91 8.90 15.58 14.13 11.81 10.50 2.96 0.71 1.78 5.80 7.42 11.71 0.22 0.30 0.26 0.33 0.37 0.38 0.37 0.20 0.34 0.31 0.21 0.36 0.45 0.24 0.26 0.39 0.36 0.37 0.36 0.32 0.33 0.41 0.33 0.34 0.40 0.50 13.55 11.56 11.66 16.42 16.32 12.44 11.65 7.72 26.03 11.11 14.69 13.90 14.22 7.44 16.97 12.50 15.58 14.13 12.49 13.71 27.68 17.04 16.84 40.49 12.70 17.97 13 7 13 13 16 18 18 11 11 15 11 17 12 8 14 14 18 18 17 0.130.33 0.210.39 0.180.39 0.190.42 0.210.51 0.170.57 0.130.56 0.110.44 0.270.41 0.140.59 0.110.28 0.180.51 0.310.60 0.150.44 0.170.33 0.180.60 0.200.50 0.190.56 0.150.56 0.17 0.31 0.28 0.44 0.36 0.50 0.48 0.22 0.31 0.35 0.17 0.46 0.60 0.18 0.37 0.44 0.43 0.46 0.42 0.37 0.01 0.05 0.10 0.27 0.26 0.56 3.91 6.09 9.86 10.77 8.85 10.80 8.71 6.86 7.00 7.20 6.31 11.01 17.62 5.21 12.24 10.39 8.34 10.05 10.31 9.03 0.09 0.93 1.28 6.50 5.18 14.62 0.25 0.45 0.30 0.46 0.42 0.50 0.53 0.26 0.42 0.39 0.24 0.46 0.60 0.27 0.39 0.46 0.46 0.46 0.43 4.55 9.94 11.54 12.21 13.95 10.80 11.85 7.70 22.74 7.65 11.46 11.01 17.62 6.85 13.77 11.56 9.63 10.05 10.85 11 10 17 17 15 18 16 13 13 15 12 18 17 10 17 17 17 18 17 0.160.55 0.210.61 0.140.51 0.210.66 0.230.61 0.120.76 0.120.74 0.120.60 0.290.53 0.160.73 0.160.40 0.160.73 0.280.79 0.170.57 0.190.68 0.180.67 0.130.79 0.160.71 0.140.68 0.15 0.06 0.13 0.15 0.18 0.23 0.22 0.06 0.17 0.17 0.11 0.23 0.18 0.09 0.12 0.20 0.19 0.24 0.23 0.16 0.10 0.04 0.07 0.13 0.22 0.23 9.51 3.16 7.35 6.93 6.53 7.98 10.45 4.86 7.33 6.36 7.25 10.83 5.87 4.97 5.78 6.79 10.95 10.15 7.86 7.42 1.81 0.85 1.80 3.84 6.06 6.07 11.37 10.31 16.06 7.40 13.38 13.26 10.75 16.44 12.19 9.54 11.84 41.55 42.53 12.40 14.92 12.74 11.24 9 7 9 6 12 14 3 8 10 5 14 4 1 8 13 13 14 14 0.120.26 0.140.28 0.190.32 0.210.44 0.210.42 0.130.36 0.110.15 0.220.36 0.150.36 0.150.27 0.110.37 0.360.39 0.280.28 0.190.22 0.170.44 0.140.34 0.150.45 0.140.45 19% 29% 18% 40% 37% 34% 39% 31% 19% 37% 18% 24% 37% 19% 42% 33% 31% 21% 33% 30% 2% 17% 8% 14% 14% 38% 57% 30% 47% 50% 44% 42% 36% 31% 53% 44% 52% 50% 32% 44% 52% 47% 39% 52% 47% 45% 65% 60% 76% 67% 65% 33% 24% 41% 35% 10% 19% 24% 25% 38% 29% 19% 30% 26% 32% 37% 6% 20% 30% 27% 20% 26% 33% 23% 15% 19% 21% 29%
Panel B: Correlations of risk-adjusted momentum returns using country-specic FamaFrench three factor model Australia Austria Belgium Canada Denmark Finland France Germany Hong Kong Ireland Italy Japan Luxembourg Netherlands New Zealand Norway Singapore Spain Sweden Switzerland UK USA Total 19751980 19811985 19861990 19911995 19962000 20012004 0.19 0.20 0.22 0.26 0.33 0.38 0.37 0.17 0.26 0.27 0.18 0.35 0.37 0.15 0.22 0.33 0.36 0.37 0.35 0.28 0.13 0.04 0.08 0.19 0.28 0.44 9.74 8.53 10.17 9.03 10.51 12.44 11.65 8.03 9.77 8.48 12.78 12.19 10.08 5.80 9.91 8.90 15.58 14.13 11.81 10.50 2.96 0.71 1.78 5.80 7.42 11.71 0.22 0.30 0.26 0.33 0.37 0.38 0.37 0.20 0.34 0.31 0.21 0.36 0.45 0.24 0.26 0.39 0.36 0.37 0.36 0.32 0.33 0.41 0.33 0.34 0.40 0.50 13.55 11.56 11.66 16.42 16.32 12.44 11.65 7.72 26.03 11.11 14.69 13.90 14.22 7.44 16.97 12.50 15.58 14.13 12.49 13.71 27.68 17.04 16.84 40.49 12.70 17.97 13 7 13 13 16 18 18 11 11 15 11 17 12 8 14 14 18 18 17 0.130.33 0.210.39 0.180.39 0.190.42 0.210.51 0.170.57 0.130.56 0.110.44 0.270.41 0.140.59 0.110.28 0.180.51 0.310.60 0.150.44 0.170.33 0.180.60 0.200.50 0.190.56 0.150.56 0.17 0.20 0.14 0.31 0.28 0.35 0.35 0.16 0.16 0.23 0.13 0.28 0.39 0.12 0.26 0.28 0.30 0.31 0.26 0.25 0.08 0.04 0.06 0.17 0.22 0.43 7.14 7.06 6.12 10.51 9.64 9.46 9.10 8.36 7.21 6.53 8.85 8.03 9.32 4.47 9.08 9.58 9.14 9.35 7.46 8.23 1.71 0.77 1.18 4.40 5.60 11.84 0.22 0.32 0.22 0.34 0.34 0.37 0.38 0.20 0.28 0.29 0.17 0.32 0.48 0.23 0.31 0.33 0.32 0.33 0.32 10.66 8.25 12.93 11.26 10.33 10.15 11.23 9.34 16.43 8.15 12.59 9.04 15.13 5.64 12.08 13.43 9.43 10.61 10.73 11 6 8 15 11 16 16 11 4 13 8 15 12 5 13 14 16 16 14 0.120.34 0.240.47 0.140.29 0.170.51 0.220.58 0.130.60 0.150.60 0.110.38 0.250.32 0.140.57 0.110.22 0.110.57 0.290.62 0.150.38 0.170.52 0.160.49 0.120.62 0.140.57 0.160.57 0.09 0.00 0.05 0.13 0.15 0.18 0.13 0.06 0.10 0.13 0.08 0.18 0.20 0.09 0.07 0.19 0.17 0.20 0.15 0.12 0.07 0.05 0.07 0.11 0.11 0.22 7.29 0.04 2.84 4.69 4.92 7.70 9.54 4.23 3.99 4.98 4.45 8.30 5.61 4.88 3.40 7.69 10.83 11.89 5.48 5.93 1.41 1.05 1.63 3.31 3.19 6.30 12.85 7.72 10.57 14.52 10.55 12.69 9.33 9.71 7.35 6.87 11.23 14.44 14.93 8.00 9.42 13.00 11.64 11.10 5 2 7 6 11 10 3 3 8 4 13 7 4 3 12 11 12 11 0.120.19 0.150.20 0.160.32 0.220.34 0.120.39 0.120.27 0.120.17 0.220.31 0.150.37 0.130.23 0.130.32 0.290.46 0.160.21 0.160.24 0.160.46 0.130.27 0.140.32 0.130.31 39% 37% 25% 52% 41% 45% 55% 46% 25% 42% 34% 29% 51% 37% 54% 33% 40% 43% 42% 40% 20% 26% 24% 31% 35% 39% 37% 11% 24% 46% 30% 30% 21% 40% 34% 30% 36% 37% 38% 39% 33% 38% 35% 37% 29% 33% 62% 39% 53% 53% 24% 33% 24% 52% 51% 2% 28% 26% 24% 15% 42% 28% 30% 34% 10% 24% 13% 30% 25% 20% 29% 27% 18% 35% 23% 16% 41% 28%
A. Naranjo, B. Porter / Journal of International Money and Finance 29 (2010) 275299 295
We report average correlations of risk-adjusted momentum returns from 1975 to 2004 for developed countries. In Panel A, we use a country-specic market model, while in Panel B we use a country-specic three factor model similar to that of FamaFrench. For the country-specic market model, we use the equal weighted average of all stocks local currency return within the country where there are three or more rms. For the FamaFrench three factor country-specic model, we use the market factor 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. The last three columns decompose the pairwise covariance of country-neutral momentum as described in Equation (1).
296
Table 7 Predicting momentum portfolio comovement with factor comovement (19752004). Variables Full sample period 19752004 (1) (2) (3) (4) Sub-periods 19751989 (5) 19902004 (6)
Panel A: Regressions of country-neutral momentum return covariance on factor covariance for developed countries Constant RMKT Rf SMB HML Country pairs: Adj R2 Variables 231 0.08 210 0.15 3.15 (4.55) 0.19 (4.68) 5.12 (15.95) 5.06 (17.31) 1.78 (3.73) 0.16 (3.74) 0.21 (3.57) 0.19 (3.64) 171 0.33 0.33 (0.57) 0.10 (2.31) 0.08 (0.76) 0.06 (0.49) 66 0.06 Sub-periods 19751989 (2) (3) (4) 19902004 (5) 3.67 (3.54) 0.13 (2.41) 0.22 (2.67) 0.22 (3.48) 171 0.23
0.33 (6.16) 0.26 (5.08) 190 0.12
Full sample period 19752004 (1)
Panel B: Panel regressions of country-neutral momentum return covariances during the holding period on covariances in factor returns during the measurement period for developed countries Constant RMKT Rf SMB HML 2.14 (0.70) 0.17 (14.51) 0.66 (0.17) 0.19 (12.08) 0.19 (9.13) 0.56 (0.14) 0.19 (11.59) 0.18 (8.07) 0.05 (3.59) 171 0.24 0.87 (0.32) 0.02 (0.90) 0.06 (1.45) 0.12 (3.67) 65 0.08 0.70 (0.16) 0.24 (11.02) 0.19 (7.89) 0.04 (2.70) 171 0.24
Country Pairs: Adj R2
231 0.21
210 0.24
For Panel A, we calculate the sample covariance of country-neutral momentum returns and country specic risk factors. We then regress the covariance of momentum returns on the covariance of risk factors for all pairs of developed countries that meet the sample selection criteria, have local factors available, and have a valid momentum return on or within 12 months of the factor realization. For Panel B, we examine if the covariance in factor returns during the measurement period predict the covariance of momentum returns in the holding period. For each country-neutral momentum strategy and for each local factor realization, we calculate the deviations from the country time series mean using Equation (2). We do similar calculations for SMB and HML. We then calculate the cross product for country i and j using Equation (3). For the regression analysis, we use panel regressions of y on x and include time and country-pair dummies (t-statistics in parenthesis).
where Ri,t is the country-neutral momentum return in country i in month t, and RMRFi,t is the return to the excess local market factor in country i in month t. We do similar calculations for SMB and HML. We then calculate the cross product for country i and j:
yi;j;t 3i;t 3j;t P xi;j;t 1=11 12 2 gRMRF ,gRMRF k i;tk j;tk
(3)
For the regression analysis, we use panel regressions of y on x and include time and country-pair dummies. We report the results from these regressions in Panel B of Table 7. The rst three columns of Panel B in Table 7 contain panel regression results for univariate and multivariate specications using only developed countries, while the last two columns contain results for the three-factor specication over two equal 15 year sub-periods. Looking at the rst column, we see that covariance of the market factor returns during the measurement period are signicant predictors of the covariance of country-neutral momentum returns during the holding period. In the
297
second and third columns, we see that for all three factors, covariances of the risk factor returns during the measurement period are signicant 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 signicant role in predicting covariances of country-neutral momentum returns during the holding period, with increasing coefcient estimates, signicance 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 nd that the covariance of market factor returns during the measurement period are signicant 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 ve-year sub-periods to see if there is a stationarity problem with the covariances in the regression analysis. Though the early ve-year sample periods are sparse, the time varying signicance of the results over each of the ve-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 analysis, we use instrumental variables procedures and again obtain similar results to those we report. 7. Conclusion We examine the protability and cross-country comovement of country-neutral momentum returns for both developed and emerging markets. Using relative strength portfolio sorts, we rst document the protability of momentum returns across countries and then decompose those returns by industry effects, rm 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 protable 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 by Rouwenhorst (1998, 1999) and Grifn et al. (2003). We also nd both an industry momentum effect and a rm 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 nd that the models do a poor job at explaining average momentum returns. We nd that country-neutral momentum returns are signicantly correlated across countries. This comovement is positive and time varying. Conditioning on industry, we nd that industry comovement does not explain the comovement in country-neutral momentum returns. We also nd that while standard factor models do not explain average momentum returns, they do explain a signicant portion of the comovement of country-neutral momentum returns. This nding is of particular importance to the asset allocation, diversication, and risk management strategies of investors who employ international momentum strategies since countries whose capital markets are more highly integrated are likely to have correlated pay-offs to country-specic 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 specications and largely insignicant. The HML factor, for instance, is largely unavailable for the emerging markets.
298
References
Asness, C.S., Porter, R.B., Stevens, R.L., 2001. Predicting stock returns using industry-relative rm characteristics. Unpublished Working 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 International Money 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 nancial anomalies. Review of Financial Studies 15, 575606. Bris, A., Goetzman, W.M., Zhu, N., 2004. Efciency 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 benchmarks. 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 ows 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. Grifn, J.M., Karolyi, A.G., 1998. Another look at the role of industrial structure of markets for international diversication strategies. Journal of Financial Economics 50, 351373. Grifn, J.M., Stulz, R., 2001. International competition and exchange rate shocks: a cross-country industry analysis of stock returns. Review of Financial Studies 14, 215241. Grifn, J.M., 2002. Are the Fama and French factors global or country specic? Review of Financial Studies 15, 783803. Grifn, 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. Benets of global diversication 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 protability of momentum strategies. Journal of Finance 55, 265295. Hong, H., Stein, J.C., 1999. A unied 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 Financial Research 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 efciency. Journal of Finance 48, 6591. Jegadeesh, N., Titman, S., 2001. Protability of momentum strategies: an evaluation of alternative explanations. Journal of Finance 56, 699720. Kaminsky, G., Lyons, R.K., Schmukler, S., 2004. Managers, investors, and crises: mutual fund strategies in emerging markets. Journal of International Economics 64, 113134. Karolyi, G.A., Stulz, R.M., 1996. Why do markets move together? An investigation of U.S. Japan stock return comovements. Journal of Finance 51, 951986. Korajczyk, R., Sadka, R., 2004. Are momentum prots robust to transactions costs? Journal of Finance 59, 10391082. Ling, D., Naranjo, A., 1999. The integration of commercial real estate markets and stock markets. Journal of Real Estate Economics 27, 483515. Lo, A.W., MacKinlay, A.C., 1990. When are contrarian prots due to stock market overreaction? Review of Financial Studies 3, 175208. Longin, F., Solnik, B., 1995. Is correlation in international equity returns constant: 19601990? Journal of International Money and Finance 14, 326.
299
Longin, F., Solnik, B., 2001. Extreme correlation of international equity markets. Journal of Finance 56, 649676. Puchkov, A., Stefek, D., Davis, M., 2005. Sources of return in global investing. The Journal of Portfolio Management 31, 1221. Rouwenhorst, K.G., 1998. International momentum strategies. Journal of Finance 53, 267284. Rouwenhorst, K.G., 1999. Local return factors and turnover in emerging stock markets. Journal of Finance 54, 14391464.

Final JIMF Paper

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Final JIMF Paper

Uploaded by

Copyright:

Available Formats

Journal of International Money and Finance 29 (2010) 275299

Contents lists available at ScienceDirect

Journal of International Money and Finance

A. Naranjo, B. Porter / Journal of International Money and Finance 29 (2010) 275299

A. Naranjo, B. Porter / Journal of International Money and Finance 29 (2010) 275299

A. Naranjo, B. Porter / Journal of International Money and Finance 29 (2010) 275299

A. Naranjo, B. Porter / Journal of International Money and Finance 29 (2010) 275299

A. Naranjo, B. Porter / Journal of International Money and Finance 29 (2010) 275299

A. Naranjo, B. Porter / Journal of International Money and Finance 29 (2010) 275299

A. Naranjo, B. Porter / Journal of International Money and Finance 29 (2010) 275299

347 80 223 347 161 38 347 347 347

3.34 347 3.47 1.38 1.78 1.41 1.70 0.33 59 60 60 60 60 48

3.88 347 3.96 0.77 1.57 2.29 2.00 0.75

2.53 347 3.11 0.84 1.07 0.42 1.95 0.13 59 60 60 60 60 48

59 0.48 60 0.14 60 0.19 60 0.09 60 0.60 48 0.05

0.83 1.33 0.21

2.28 106 1.00 140 0.47 57

3.64 225 1.23 3.29 1.80 57 60 60

2.23 0.41 0.54 0.55 1.65 0.44 1.34 0.28

(continued on next page)

284 Table 2 (continued)

A. Naranjo, B. Porter / Journal of International Money and Finance 29 (2010) 275299

3.57 347 3.47 1.38 1.80 1.64 1.75 0.46 59 60 60 60 60 48

3.83 347 3.96 0.77 1.51 2.24 1.95 0.75 59 60 60 60 60 48

2.66 347 3.16 0.84 1.08 0.42 1.99 0.03 59 60 60 60 60 48

A. Naranjo, B. Porter / Journal of International Money and Finance 29 (2010) 275299

A. Naranjo, B. Porter / Journal of International Money and Finance 29 (2010) 275299

A. Naranjo, B. Porter / Journal of International Money and Finance 29 (2010) 275299

(continued on next page) 287

A. Naranjo, B. Porter / Journal of International Money and Finance 29 (2010) 275299

A. Naranjo, B. Porter / Journal of International Money and Finance 29 (2010) 275299

A. Naranjo, B. Porter / Journal of International Money and Finance 29 (2010) 275299

A. Naranjo, B. Porter / Journal of International Money and Finance 29 (2010) 275299

3i;t Ri;t Ri ; gRMRF RMRF i;t RMRF i i;t

0.15 0.23 0.20 0.20 0.17 0.23

0.130.34 0.160.35 0.160.29 0.120.32 0.150.21 0.130.38

0.03 0.06 0.03 0.04 0.09

0.34 0.36 1.39 0.91 1.35

0.38 0.38 0.38 0.38

0.06 0.16 0.08 0.03 0.00 0.04 0.06 0.08

0.69 2.77 0.33 0.45 0.03 0.80 1.44 1.42

0.34 0.34 0.35

0.00 0.01 0.05 0.02 0.02 0.03 0.07

0.04 0.21 0.43 0.52 0.57 0.35 0.38 0 1 0.380.38

0.01 0.02 0.06 0.02 0.00 0.09 0.10

0.15 0.42 0.85 0.75 0.09 0.81 1.54

0.33 0.39 0.39

0.300.35 0.390.39 0.300.48

A. Naranjo, B. Porter / Journal of International Money and Finance 29 (2010) 275299

A. Naranjo, B. Porter / Journal of International Money and Finance 29 (2010) 275299

0.33 (6.16) 0.26 (5.08) 190 0.12

Full sample period 19752004 (1)

Country Pairs: Adj R2

yi;j;t 3i;t  3j;t   P xi;j;t 1=11  12 2 gRMRF ,gRMRF k i;tk j;tk

A. Naranjo, B. Porter / Journal of International Money and Finance 29 (2010) 275299

A. Naranjo, B. Porter / Journal of International Money and Finance 29 (2010) 275299

A. Naranjo, B. Porter / Journal of International Money and Finance 29 (2010) 275299

You might also like

yi;j;t 3i;t 3j;t P xi;j;t 1=11 12 2 gRMRF ,gRMRF k i;tk j;tk