Industry nd Country ffects in Interntionl Stock Returns Implictions for sset lloction. Steven L. Heston nd K. Geert Rouwenhorst STVN L. HSTON is ssistnt professor of finnce t the John M. Olin School of Business of Wshington University in St. Louis (MO 61-48). K. GRT ROUWNHORST is ssocite professor of finnce t the School of Mngement of Yle University in New Hven (CT 65). I n the prctice of interntionl equity mngement, mngers often use two-stge pproch to portfolio selection. In the first stge the mnger lloctes portions of the portfolio to severl industries; in the second stge, the mnger uses industry nlysts to select the most ttrctive stocks from those sectors. This strtegy is used by mngers who believe tht interntionl returns re predominntly driven by industry fctors. Mngers who believe tht domestic mrket fctors re more importnt for returns thn industry fctors decide on country lloction first, then in the second stge select the most promising stocks from ech country. This rticle presents simple model to mesure country nd industry effects in interntionl stock returns, nd provides quntittive frmework for nlyzing these two pproches to portfolio selection.' We show tht there re three resons for portfolio mngers to py more ttention to the geogrphicl thn to the industril composition of n interntionl portfolio. ch of these resons is bsed on the finding tht country effects in interntionl stock returns re lrger thn industry effects. First, tilting n interntionl portfolio geogrphiclly leds, on verge, to lrger nd more vrible trcking errors thn tilting the industril composition of the portfolio. Second, stocks from the sme domestic mrket but in different industries re closer substitutes thn stocks from the sme industry but in different countries. Finlly, the benefits of interntionl SPRING 15 TH JOURNAL OF PORTFOLIO MANAGMNT 5
XHIBIT 1 Number of Firms by Country nd Industry Industry - Consumer Finnce, Bsic Cpitl Goods Insurnce, Trnsporttion Country Industries Goods nd Services nergy nd Rel stte nd Storge Utilities Totl Austri Belgium Denmrk Frnce Germny Itly Netherlnds Norwy Spin Sweden Switzerlnd United Kingdom urope 16 --- 16 7 1 7 6 14 6 17 17 4 14 46 1 1 1 1 1 5 6 4 186 1 1 7 1 1 4 1 8 8 1 5 1 5 1 5 18 1 6 5 85 7 --- 5 --- 4 1 5 5 --- 55 1 --- 7 8 4 18 18 18 47 8, diversifiction stem lrgely from geogrphicl diversifiction nd not from industril diversifiction. DATA The smple consists of monthly totl returns for ll 8 firms tht were included in the Morgn Stnley Cpitl Interntionl (MSCI) country indexes of twelve uropen countries from 178 through 1. Country ssignments for ech firm re bsed on the country clssifictions of MSCI. ch firm is ssigned to one of seven brod industry ctegories, s defined by the Finncil Times Acturies. The distribution of firms over countries nd industries is given in xhib:it 1. If firm is dropped from the MSCI[ indexes, but remins listed on n exchnge, it is kept in the smple. Al returns re converted into deutschemrks using exchnge rtes tken from the Finncil Times. MODL The model ssumes tht we cn write the return on stock i tht belongs to industry j nd country k s follows: qution (1) sttes tht ll returns shre common fctor. Pj is the industry effect for industry j, yk is the country effect for country k, nd ei is firm-specific component of the return in period t. For ech month in the smple, we estimte, p, nd y by running cross-sectionl regression of the returns of ll 8 firms in our smple on set of industry nd cou-ntry dummies: where I.. = 1 if firm i belongs to industry j (zero oth- 'J erwise), nd Ci, = 1 if firm i belongs to country k (zero otherwise). By running cross-sectionl regression for ech month we obtin time series of estimted in.dustry nd country effects. The only compliction in estimting qution () is tht there is perfect multicolhnerity mong the regressors. The problem is tht every firm belongs to both n industry nd country, nd therefore we cn mesure only differences between countries nd dxerences between industries. To get round this protllem, we choose to mesure industry nd country effect!; reltive to the common fctor, which is the uropen eqully weighted (W) index. ch of the estimted industry effects, Pj, cn be interpreted s the excess return over the uropen W index on portfolio tht invests in industry j, nd hs no net position in other industries. This industry port- 54 INDUSTRY AND COUNTRY FFCTS IN INTRNATIONAL STOCK RTURNS SPRING 15
XHIBIT Industry nd Country ffects 178-1 (monthly returns expressed in % per yer) Men Stndrd Devition Industry ffects Bsic Industries -.6 4. Cpitl Goods -. 4.74 Consumer Goods 1.5.1 nergy. 14.56 Finnce -.64.86 Trnsporttion.4 8.41 Utilities 1. 1.51 Averge Absolute Vlue 1.4 7. Country ffects Austri Belgium Denmrk Frnce Germny Itly Netherlnds Norwy Spin Sweden Switzerlnd United Kingdom -5.5.7-1.6.8 -.41. -.6-5.7 -.15 4.6-5.85.. 1.7 16.6 15.7..5 1.67.55.17..4 1.1 Averge Absolute Vlue. 16.54 folio is geogrphiclly diversified in the sense tht it hs the sme country composition s the uropen W index, nd is therefore pure industry bet. Similrly, yk is the excess return of n industrilly &versified investment in country k, nd represents pure country ply. Becuse the portfolio weights of these industry nd country bets depend only on the industry nd country ssignments of the stocks, which re known t the beginning of ech period, the ps nd ys re excess returns on fesible investment strtegies. RLATIV SIZ OF COUNTRY AND INDUSTRY FFCTS xhibit gives the men nd stndrd devition, expressed in percent per yer, of the industry nd country effects. The first row of the tble indctes tht the verge portfolio of stocks in bsic industries with the sme country composition s the uropen W index underperformed the uropen index by.6% per yer with stndrd devition of 4.% per yer. The strikmg feture of xhibit is tht the bsolute vlue of the country effects is on verge twice s lrge s the bsolute vlue of the industry effects. The verge stndrd devition of the country effects is more thn twice the verge stndrd devition of the industry effects. xcept for the energy sector, the stndrd devitions of the industry effects re ll smller thn the stndrd devitions of the country effects. The reltive size of country nd industry effects is importnt for portfolio mngers becuse these effects hve the interprettion of trchng errors reltive to the uropen index. Suppose mnger uses the uropen index s the benchmrk, nd considers either 1% country tilt towrd the U.K. or 1% industry tilt towrd the finncil sector. Becuse the finncil sector nd the U.K. ech contribute 18 securities to the smple, these tilts re in tht sense eqully diversified. The country effect for the U.K. indictes tht replcing 1% of the stocks in the uropen index portfolio with U.K. stocks, while mintining the industry composition of the portfolio, would hve led to outperformnce of the benchmrk by 1% x. =.% per yer, with stndrd devition of the trcking error of 1% x 1.1 = 1.1% per yer. A 1% tilt towrd the finncil sector would hve led to underperformnce of the uropen index by.6% per yer with stndrd devition of only.%. Becuse country effects re lrger thn industry effects, devition from the country composition of the benchmrk portfolio while mintining the industry composition will induce much lrger trcking errors thn tilting portfolio towrd Qfferent industry, holding the country composition fixed.* Firms in different industries but locted in the sme country shre the common fctor nd the country effect, whde firms in the sme industry but in different countries shre the common fctor nd the industry effect. Becuse the vrince of the country effect is lrger thn the vrince of the industry effect, the verge correltion between securities within country is higher thn the verge correltion between firms tht re in the sme industry. In other words, two firms tht re locted in the sme country re closer substitutes thn two firms tht belong to the sme industry. XPLAINING COUNTRY PRFORMANC The model is prticulrly useful in explining SPRING 15 TH JOURNAL OF PORTFOLIO MANAGMNT 55
XHIBIT DCOMPOSITION OF COUNTRY INDX RTURNS INTO INDUSTRY AND CUNTR.Y FFCTS PANL A. QUALLY WIGHTD COUNTRY INDX RTURNS IN DM 178-1 18 c," 16- v 14- l z 1-1 - - ur Ger Swi Bel Neth Der? Fr UH Aus Swe It Nor 8 8-1.5 15. 17.5..5 5.C 7.5. Stndrd devition (Z PA.) SP PANL B. COMMON FACTOR -t COUNTRY FFCT c 14 t Swe It 18 "C Bel Aeth t I - - - - 1.5 15. 17.5..!5 5. 7.5 5. Stndrd devition (Z p..) m PANL C. COMMON FACTOR + INDUSTRY FFCTS PANL D. COMMON FACTOR UROPAN QUALLY WIGHTD MARKT ' 18 e N 16 v 14 m F 1 urope 1.5 15. 17.5..5 5. 7.5. Stndrd devition (7. p) differences in country performnce. Summing the individul returns by country, we cn write the return on the eqully weighted mrket of country k s the sum of three components: the common fctor, the verge of the industry effects of the stocks in the index, nd country effect: Common Industry Country Fctor ffects ffect where nk is the number of stocks in country k. Becuse ll countries shre the common fctor 1.5 15.Q 17.5..5 5. 7.5. Stondrd devition (1. p.) (), there re two resons for differences in country performnce. The first fctor is tht countries spec:ihze in different industries, nd re therefore subject to ;I different set of industry effects. This is mesured by the second term on the right-hnd side of qution (). For exmple, xhibit shows tht the utilities sector hs positive nd cpitl goods negtive industry effect. Both hve positive effect on the reltive performnce of Spin, becuse Spin hs proportionlly more utilities thn other countries, nd fewer firms in the cpitl goods sector (see xhibit 1). The second fctor driving country performnce is the country effect, given by the lst term in qution (). It mesures the prt of the performnce of Spnish firms reltive to firms in the sme industry but locted 56 INDUSTRY AND COUNTRY FFCTS IN INTRNATIONAL STOCK RTURNS SPRING 15
outside Spin. It controls for the fct tht Spin hs proportionlly more firms in sectors tht performed well (utdities), nd fewer firms in industries tht performed poorly (cpitl goods). The four pnels ofxhbit give grphicl view of this decomposition in terms of the mens nd stndrd devitions of the country returns. The smple mens nd stndrd devitions of returns of the twelve uropen countries during 178-1 re given in Pnel A. Pnel D plots the men nd stndrd devition of the common fctor. The horizontl difference between two pnels illustrtes the role of industry effects; the verticl difference represents the country effects. Pnel A cn be obtined from Pnel D by dding to the common fctor the country effects (D B) nd the industry effects ( A), or by dding the industry effects first (D C) nd the country effects next (C A). ither wy, the decomposition shows tht crosscountry differences in verge return nd return voltility re primrily due to country fctors. Industry fctors cnnot ccount for the differences in the men nd voltility of county index returns. These results re surprising in the sense tht we find lrge country effects in smple where one would expect them to be reltively unimportnt, becuse the uropen countries re economiclly nd finncilly integrted, yet industrilly diverse. We conjecture tht country effects will ply n even lrger role in smple tht lso includes the U.S. nd Jpn, or number of emerging mrkets. TH BNFITS OF INTRNATIONAL DIVRSIFICATION By selecting interntionl stocks, portfolio mnger chieves risk reduction through the benefits of both geogrphicl nd industril diversifiction. When building portfolios, however, it is more importnt to be geogrphiclly &versified thn to be industrilly diversified. This is shown in xhibit 4, which compres the vrince reduction tht cn be chieved by three different ~trtegies.~ The top line plots the verge vrince of strtegy tht diversifies cross industries within country, s function of the number of stocks in the portfolio. This strtegy is well-diversified industrilly but not geogrphiclly. As the number of stocks in the portfolio becomes lrge, the portfolio vrince becomes 8% of the verge vrince of the securities in the portfolio. XHIBIT 4 BRAKDOWN OF TH BNFITS OF INTRNATIONAL DIVRSIFICATION 8 s (d.e 1 8 P Y v), l * (d ru s 6 4, u.d P Ir 8 1. 1.. \. -.-. -.-.-. country diversifiction limit. 8 z industry diversific otion z diversiticotion ocross urope 1 4 number of stocks in portfolio The strtegy tht corresponds to the second line chooses stocks from single industry, but is well-diversified cross countries. For lrge portfolios, the verge portfolio vrince is pproximtely % of the vrince of the typicl stock in the portfolio. A lrge portfolio tht simultneously &versifies over industries s well s countries hs vrince of 18% of the typicl security vrince. Becuse we merely llocte the sme securities into portfolios ccording to different criteri, the verge return of ll three strtegies is the sme. Hence, the verticl difference between the lines is n estimte of the verge hfference in the diversifible risk of these portfolio strtegies. The figure therefore shows tht geogrphicl diversifiction is much more effective tool for risk reduction thn industril diversifiction. SUMMARY AND CONCLUSIONS We hve presented simple method to mesure the importnce of industry nd country effects in interntionl stock returns. We show tht, for twelve SPRING 15 TH JOURNAL OF PORTFOLIO MANAGMNT 57
I terms of economic policies, country effkcts dominte industry effects. ConsequendY the pelformnce of interntionl portfolios is lrgely country-driven, nd interntionl portfolio mngers should py more ttention to the geogrphicl composition thn to the industril composition of their portfolios. NDNOTS The uthors thnk Richrd Lindsey nd ]<oberto Wesels for comments, nd ARCAS-Wessels Roll Ross for mking the dt vilble. The nlysis is in the spirit of Grinold, Rudd, nd Stefek [18]. The frmework we develop is eqully useful I:O mngers who do not explicitly use either of these two-stge pproches, becuse interntionl portfolio. These conclusions re qulittively similr if we choose the vlue-weighted uropen index s benchmrk. A complete disc)xsion of these results is found in Heston nd Rowenhorst [14]. This figure ws first suggested by Solnik [174]. RFRNCS Grinold, Richrd, Andrew Rudd, nd Dn Stefek. Globl Fictors: Fct or Fiction? Jouml of Porlfolio Mngement, 16 (18). pp. 7-88. Heston, Steven L., nd K. Geert Rouwenhorst. Does Industril Structure xplin the Benefits of Interntiond Diversifiction?,hrml qf Finncil conomics, 6 (14). pp. -7. Solnik, Bruno. Why Not Diversify Interntionlly Rther Thn Domesticlly? Finncil Anlystsjocrml, (174), pp. 48-54. 58 INDUSTRY AND COUNTRY FFCTS IN INTRNATIONAL STOCK RTURNS SPRING 15