JHD-2014-02

Discussion Paper Number: JHD-2014-02 Discussion Paper Regional and Product Diversification and the Performance of Retail Multinationals March 2014 Chang Hoon Oh Beedie School of Business, Simon Fraser University Timo Sohl IESE Business School and University of St. Gallen Alan M Rugman Henley Business School, University of Reading ii Oh, Sol and Rugman, March 2014Theaimofthisdiscussionpaperseriesisto disseminatenewresearchofacademic distinction.Papersarepreliminarydrafts, circulatedtostimulatediscussionandcritical comment.HenleyBusinessSchoolistriple accreditedandhometoover100academic faculty,whoundertakeresearchinawiderange offieldsfromethicsandfinancetointernational business and marketing. [email protected] www.henley.reading.ac.uk/dunning Oh, Sol and Rugman, March 2014 Henley Discussion Paper Series Oh, Sohl and Rugman, March 20141 RegionalandProductDiversificationandthePerformance of Retail Multinationals Abstract Despitetheimportanceofgeographicexpansionintheservicessector,fewstudieshave analyzedtherelationshipsbetweeninternationaldiversification,productdiversificationand performanceforservicesfirms.Hereweinvestigatewhetherandhowfirmsintheretail-trade sector may benefit by spreading their boundaries within and across regional boundaries. Using a panel data set of 68 large European retailers from 19 countries for the period between 1997 and 2010,wefindthatintra-regionaldiversificationhasahorizontalS-curverelationshipandinter-regional diversification has an S-curve relationship with firm performance. Moreover, the results show that unrelated product diversification has a negative moderating effect on the relationship betweeninter-regionaldiversificationandfirmperformance.Overall,theseresultsaddsupport intheservicessectorforthethree-stageparadigmofinternationaldiversificationand performance. Keywords retailfirms,geographicdiversification,productdiversification,regionaldiversification, performance, S curve Contact ChangHoonOh:[email protected],BeedieSchoolofBusiness,SimonFraserUniversity,500Granville Street, Vancouver, B.C., V6C 1W6, Canada. Tel: 1 905 401 3137. Fax: 1 905 378 5716. Timo Sohl: [email protected], IESE Business School, Av. Pearson 21, 08034 Barcelona, Spain. Tel: 34 936026422.Fax:34932534343.Also:[email protected],UniversityofSt.Gallen,Dufoustr. 40a, 9000 St. Gallen, Switerland. Tel: 41 71 224 2856. Fax: 41 71 224 7194. Alan M Rugman: [email protected], Henley Business School, University of Reading, Henley-on-Thames,Oxon,RG93AU,UnitedKingdom.Tel:44(0)1183787340.Fax:44(0)118378 6229. John H Dunning Centre for International Business 2Oh, Sohl and Rugman, March 2014 1Introduction OfthefirmsonthemostrecentFortuneGlobal500list,242firmsaremanufacturingfirms,111 firms are financial firms, 100 firms are services firms and 47 firms are utility firms, based on their listed main industry sectors in 2012. The combined revenue of the 100 services firms was $5,431 billion. Thus, the services sector makes up a substantial part of the global economy, and its rapid growth deserves more scholarly attention (Kundu and Merchant, 2008). Amongthese100servicesfirms,45firmsarefromtheretail-tradesectorwithcombined revenueof2,627billionU.S.dollars.Thesenumbersincreasedfrom40firmswhosecombined revenuewas1,399U.S.dollarsin2003.Mostoftheseretailfirmsaremultinationalenterprises (MNEs),buttheirinternationalfootprintisrelativelyweakcomparedtoMNEsinotherindustry sectors(RugmanandOh,2010;CorstjensandLal,2012).AccordingtoRugmanandGirod (2003),retailMNEsmainlyfocusontheirdomesticmarketsorhome-regionmarkets.Possible explanationsforthishome-regionfocusarethehighcapitalintensity(RugmanandVerbeke, 2008), asset specificity and government regulations (Campbell and Verbeke, 1994) in the retail-trade sector. Tobothacademicscholarsandmanagers,akeyquestioninregardtoaretailfirm'srelatively weakinternationalactivitiesiswhetherandhowitmaybenefitbyspreadingitsboundaries within and across regional boundaries. In other words, what are the performance implications of intra-andinter-regionalgeographicdiversificationofretailMNEs?Previousresearchonthe relationshipbetweeninternationaldiversification(ormultinationality)andperformance providesimportantguidanceinansweringthiscriticalquestion.Inparticular,thethree-stage paradigmofinternationalization(Contractor,Kundu,andHsu,2003)generalizesdifferent argumentsandfindings,andimpliesthattheperformanceofMNEsmaydependupontheir international trajectory. Recentstudieshavesharpenedthetheoreticalandempiricalapproachofthethree-stage paradigm by accommodating regional characteristics of MNEs (Li, 2005; Rugman and Oh, 2010; OhandContractor,2012,2014).ThesestudiesfocusedonU.S.manufacturingfirmsandfound thathome-regionexpansionimprovedtheperformanceofU.S.firms,whileforeigntriadregion expansionreducedtheirperformance,aspredictedbythethree-stageparadigm(Contractoret al.,2003;Contractor,2007)andregionalMNEperspective(RugmanandVerbeke,2004).Triad regions arewell-defined geographicblocksofNorth America,Europe andAsia-Pacific (Rugman, 2005).Thesestudiesalsoshowedtheimportanceofproductdiversificationasamoderating effect on the relationship between international expansion and performance. Henley Discussion Paper Series Oh, Sohl and Rugman, March 20143 Inordertoanswertheresearchquestion,weuseinformationon68Europeanretailersforthe periodbetween1997and2010,whichtotal385observations.WefocusonEuropeanretailers becausetheyarethemostinternationalizedamongtheworldwideleadingretailfirms,with comparativelyhighoperationsoutsidetheslow-growingEuropeanhomemarket(Deloitte, 2014).Still,despitetheirincreasinginternationalactivities,manyEuropeanretailersare struggling to break even and survive in global markets (Gielens and Dekimpe, 2007). Whileourresultssupportedathree-stageparadigmforinternationalexpansion(Contractoret al.,2004),theyalsoprovidedadditionalinsightsintotheinternationaldiversification-performancerelationship.Ourfindingsshowedthatintra-regionaldiversificationoflarge EuropeanretailMNEshadahorizontalS-curverelationshipwithperformance,whileinter-regionaldiversificationhadanS-curverelationship.Moreover,theresultsindicatedthatthis relationshipwasmoderatedbyproductdiversificationinthatunrelatedproductdiversification had a negative moderating effectonthe relationship between inter-regional diversification and performance in the case of European retail MNEs. Interestingly, related product diversification of retailMNEsdidnothaveasignificantmoderatingeffectandintra-regionaldiversificationwas not affected by either related or unrelated product diversification. Thus, our study contributes to the literature in three ways. First, we extend the existing literature that focuseson U.S. manufacturing firms toEuropean service (retail) firms, providing additional insight into the three-stage paradigm and regional diversification. Second, we divide geographic diversificationintointra-regionalandinter-regionaldiversificationandproductdiversification intorelatedandunrelateddiversification.Third,weanalyzetheinteractionbetweenthetwo types of geographic and product diversification. Theremainingportionsofthispaperareorganizedasfollows.Thenextsectionprovidesa literaturereviewanddevelopsourhypotheses.InSection3,weexplainourdataandmethod. The empirical results appear in Section 4. In Section 5, we provide a discussion of our results, the limitations of our study, future research avenues and a conclusion. 2Theory and hypotheses 2.1Intra- and inter-regional diversification Afirmsdiversificationstrategyanditsperformanceimplicationshavebeenatthecenterof researchinstrategicmanagement,internationalbusinessandfinancialeconomics.Thegaining John H Dunning Centre for International Business 4Oh, Sohl and Rugman, March 2014 importanceofthisresearchtopicisreflectedinthegrowingphenomenonoffirmsproducing multiple lines of products and serving various geographical markets (Kim, 1999). Althoughtheincreasingvolumeandvalueofinternationaltradeandforeigninvestmentsover thepastseveraldecadesareundeniablefacts,recentevidenceshowsthatthedestinationof suchtradeandinvestmentshavebeenmostlygeographicallyboundedintohome-region countriesandneighboringcountries(RugmanandVerbeke,2004;Ghemawat,2007).In addition,pastresearchwasunabletoprovidestrongevidenceastowhethergeographic boundaries of MNEs are expanding or contracting (Rugman and Oh, 2009). Thus, the majority of MNEactivitieshaveremainedwithintheirhome-regionboundariesinthepasttenplusyears (Akhter and Beno, 2011; Rugman, Oh, and Lim, 2012). In regard to services or retail firms, home-region orientation is substantially stronger than for manufacturing firms (Rugman and Verbeke, 2008). Despiteitsimportanceforservicesfirmsintodayseconomy,fewstudieshaveanalyzedthe relationshipbetweeninternationaldiversificationandperformanceforfirmsfromtheservices sector.Forexample,CaparandKotabe(2003)foundaU-shapedrelationshipforGerman servicesfirms,whileContractoretal.(2003)foundanS-shapedrelationshipforknowledge-based services firms and a horizontal S-shaped relationship for capital intensive services firms by usingasampleoflargeservicesfirmsfrom12countries.UsingaregionalMNEperspective, SukpanichandRugman(2007)andMohretal.(2014)providedevidenceoftheperformance superiorityofhome-regionorientationforretailfirms.Li(2005)foundthatahome-region strategy provides better performance outcomes for U.S. MNEs in various service industry sectors, but his study excluded retail and food service sectors. Forservicesfirms,diversificationisassociatedwithvaluecreationorriskminimizationatthe customerend,whileformanufacturingfirmsitisassociatedwithknowledgeadoptionorcost minimizationattheproductionend(RugmanandVerbeke,2008).Thus,customer characteristicsareimportantdeterminantsingeographicdiversificationdecisionsofservices firms,supportingthenatureofhome-regionorientationintheservicessector.Inorderto considerthesimilaritiesanddifferencesofsuchdemand-basedcharacteristics,wedivided geographicdiversificationintotwoaspects:intra-regionalandinter-regionaldiversification. Intra-regionaldiversificationreflectsafirm'sdiversificationtoothercountrieswithinitshome triadregion,whereasinter-regionaldiversificationisitsdiversificationtoconsumermarkets outsideofitshometriadregion.Recentstudies(Qianetal.,2010;Qian,Li,andRugman,2013) haveanalyzedtheeffectofintra-andinter-regionaldiversificationonasampleofU.S.-and Canada-basedfirms.Forexample,usingasampleofU.S.-basedmanufacturingfirms,Qianetal. Henley Discussion Paper Series Oh, Sohl and Rugman, March 20145 (2010)foundalinearpositiveeffectofintra-regionaldiversificationonperformanceandan invertedU-shapedrelationshipbetweeninter-regionaldiversificationandperformance. However,thesestudiesneitherfocusedonservicesMNEsnoronEuropeanfirms.Moreover, thesestudiesdidnotexaminehowproductdiversificationmaymoderateperformance implications of intra- and inter-regional diversification. Inthisstudy,wefocusonretail-tradewithintheservicesectorandarguethatimportant differencesexistbetweentheinternationalizationofmanufacturingandretailfirms.For example,whilemanufacturingfirmsareabletointernationalizesalesbyexportingproductsto foreign markets, retail firms need to significantly invest into the development of a store network to achieve sales in foreign markets. Building a network of retailstores involves activitiessuch as selectingandmanaginglocalrealestate,suppliers,humanresourcesandlogisticschannels. Moreover,retailfirmsneedtodevelopadeepunderstandingofshoppingbehaviorsinforeign markets,beyondpurelyproduct-relatedconsumerpreferences.Asthisprocessinvolves considerable investments and learning processes, retail firms typically need several years before theycanreachbreak-evenvolumesinforeignmarkets(CorstjensandLal,2012).Previous researchhassuggestedthatfewretailerssucceedinachievingsimilarprofitmarginsin internationalmarketswhencomparedtotheirhomemarkets(GielensandDekimpe,2007).By distinguishingbetweenintra-andinter-regionaldiversification,wedescribethepotential performance implications of retail internationalization in more detail. Whenaretailfirmdiversifiesitsoperationsacrosscountrieswithinitshome-regionmarket,it maybeabletoquicklylearnaboutnewconsumerneedsandshoppingbehaviorsduetotheir similaritiestodomesticconsumerbehaviors.Accordingtotheinternationalizationprocess theory(JohansonandVahlne,1977),aservicesfirmfirstentersintofewkeymarketsthatare culturally,geographically,economically,andpoliticallyclosetoitshomemarket.Inaddition, productstandards,consumertastesandsuppliercharacteristicsinsuchcountriesdonotmuch differfromthoseinthefirmshomecountry.Thus,theretailfirmdoesnotneedtoextensively changeitsbusinessmodelandmayevenusethesamesuppliersasinitsdomesticmarket. Indeed,theretailerwillachievecriticalsuccessfactors,suchaseconomiesofscalein procurement and low transportation and coordination costs. Further, since the firm's entry into home-regioncountriesistypicallyearlierthanitscompetitorsfromforeignregions(Ohand Rugman,2007),itlikelyenjoysafirstmoveradvantage.Thesebenefitsleadtoapositive performance effect at the beginning stage of a retailer's intra-regional diversification. Whenaretailfirmfurtherdiversifiesintoadditionalcountriesinitshome-region,itmust coordinate its operations in these countries, requiring extra administrative costs. If the retail firm John H Dunning Centre for International Business 6Oh, Sohl and Rugman, March 2014 decidestoexpanditsoperationsintoafewkeyforeigncountriesinwhichitoperatedinitially, then it needs to start building its own supply chain in each foreign market due to the growth of itsoperationsize.Thus,itseconomiesofscaleshouldbeconsideredseparatelyineachcountry (e.g.,Hennart,2007),althoughitstillshareseconomiesofscopeinthehome-region.These factorsmayleadtonegativeperformanceeffectsinthesecondstageofintra-regional diversification. In the last stage of intra-regional diversification, a retail firm may reach a critical size for its store networkoperationsineachcountry.Eachlocalmarketstartsgainingitsownscalebenefitsina retailer'ssourcing,logisticsandmarketingactivities.Theadditionaladministrativeand marketing costs may be offset by economies of scale. At this stage, the retail firm finds the most efficientandreliablesuppliers,integratesthemandachievesaregionalsupplychainnetwork (Doner,1991;Rugman,Li,andOh,2009).Retailersmaybeabletobuild,integrate,and reconfigure various resources located in proximate countries and transfer and leverage them to countriesinthehome-region(Qianetal.,2008;Qianetal.,2010).Thesebenefitsturnintoa superior performance. Thus, we hypothesize as follows. Hypothesis1:Intra-regionaldiversificationofretailfirmhasahorizontalS-curvefit relationship with the firms performance. Theperformanceimplicationsofinter-regionaldiversificationaredifferentfromthatofintra-regionaldiversification.Atthebeginningstageofinter-regionaldiversification,aretailfirmwill servenewcustomerswhosecharacteristicsaresubstantiallydifferentfromthoseoftheir existingcustomersinthehomecountryorregion.Theretailfirmwillfaceaconsiderably differentsetofcompetitorsandmaysufferfromlatecomerdisadvantages.Inordertoserve these different markets, the companymayneed to utilize verydifferent kinds of storenetworks thanthoseusedinthehome-regionduetointer-regionalvariationsinshoppingbehaviorsand infrastructures(e.g.,Gereffi,1999;Qian,Li,andRugman,2013).Developingsuchanewstore networkrequiresadditionalcosts.Yet,theeconomiesofscalemaynotoffsetsuchcosts. Although the retail firm will likely choose few, large, key countries in foreign regions, it needs to recombineitsexistingpartnersandnewpartnersinitssupplychain.Therecombinationwill increasetransportationandadministrativecosts.Thus,theretailerwillsufferfromthese negative aspects of foreignness and newness. Inthelaterstageofinter-regionaldiversification,asthesizeoftheoperationineachcountry grows,theretailfirmmaystartbenefitingfromeconomiesofscaleinprocurementandsales. Henley Discussion Paper Series Oh, Sohl and Rugman, March 20147 Sourcingfromlocalsupplierswillenabletheretailertoreducetransportationandfactorcosts, improveresponsetime,andlowerseveraltradebarriersinforeignoperations(Kogutand Kulatilaka,1994;Arntzenetal.,1995).Inaddition,theretailfirmwillreceivebenefitsfrom productandindustrylifecycles,whichenableittoefficientlyuseitsinventories.Atthisstage, therefore,thefirm'sperformancemayimproveascomparedtothebeginningstageofinter-regional diversification. In the last stage of inter-regional diversification, the retail firm has to control and coordinate its extensively dispersed worldwide operations. Global competitiveness is hard to achieve (Rugman et al., 2012). Although a retailer may be able to find the most efficient suppliers and build a new supplychainnetwork,itwillnotlikelyimplementglobalsupplychainintegrationduetohigh transportationcostsbetweentriadregionsanddifferentregulationsandbusinesspractices across regions (Li, 2005; Oh and Rugman, 2011). Thus, the supply chain network and distribution systemswillbeorganizedandcontrolledthrougheachregionalheadquarters,requiring additionaladministrativecosts.Inaddition,duetodecreasingmarginalreturnstoscale,the scalebenefitsinthisstagearenotsignificantlyhigherthanthoseinthepreviousstage.Thus, overallperformanceinthelaststagewilldecline.Combiningtheperformanceimplicationsof each stage of inter-regional diversification, we hypothesize as follows. Hypothesis 2: Inter-regional diversification of retail firm has an S-curve fit relationship with the firms performance. 2.2Moderating effects of product diversification TallmanandLi(1996)developedatheoreticalframeworkthatsuggeststhatthesimilaritiesof geographicandproductdiversificationsproposepotentialinteractioneffectsonafirm's performance.Accordingly,severalstudieshaveinvestigatedtheinteractionsbetween geographicandproductdiversifications(e.g.,TallmanandLi,1996;Hitt,Hoskisson,andKim, 1997;Geringer,Tallman,andOlsen,2000;SukpanichandRugman,2007;OhandContractor, 2012, 2014). While the results in the existing studies are inconsistent, a study by Sukpanich and Rugman(2007),whichiscloselyrelatedtothispaper,foundthathigherlevelsofhome-region salesenhancethepositiveimpactofproductdiversityontheperformanceoflarge merchandizers.Previous research referred to a retail format based on its specific product assortment and service offering, store design, and location, aiming to match specific consumer segments and shopping situations(e.g.,Gonzales-Benito,Munoz-Gallego,andKopalle,2005).Accordingly,Keep, John H Dunning Centre for International Business 8Oh, Sohl and Rugman, March 2014 HollanderandCalantone(1995)arguedthatintheretail-tradesector,productdiversification indicatesthataretailfirmhasdiversifiedacrossretailformats.1Overthepastdecades,most worldwide leading retailers have started to diversify their retail formats, indicating that retailers market an increasingly diversified portfolio of product lines (Deloitte, 2014). Wearguethatproductdiversificationofaretailfirmdoesnotrequiresubstantialcoordination andcontrolcostswhenthefirmdiversifiesitsretailformatswithinitshome-region.Thisis becausegeographic,cultural,economic,andinstitutionalproximitywithinthehome-region enablesaretailertoefficientlytransferitsknowledgeandcapabilitiesfromoneretailformatto theothersandexploitmarketinefficiencyandconsumertrendsthroughretailformat diversification(e.g.,KediaandBhagat,1988).Thus,experientialknowledgefromaretailformat ismorevaluablewhenitappliestootherretailformatswhenmarketcharacteristicsaresimilar withthoseofexistingones(Barkema,Bell,andPennings,1996;KostovaandRoth,2002).In addition,geographicadjacencyprovidesmoreopportunitiestocreateintensiveinteractions, which are required when a firm diversifies its product lines (Jones and Hill, 1988). Economies of scaleandscopeinprocurementandlogisticscanimproveafirm'sbenefitsinproduct diversificationthrougharegionallyintegratedsupplychaininitshome-region.Indeed,product diversificationmayincreasethemonopolypoweroflargeretailersintheirhome-region. Therefore,productdiversificationstrengthensthepositivebenefitsofintra-regional diversification. Specifically, Hypothesis3:Productdiversificationhasapositivemoderatingeffectonthe relationship between intra-regional diversification and performance, in a way that it hardens the positive slopes and softens the negative slope in the horizontal S-curve fit.Whenaretailermarketsitsretailformatportfolioindistantforeignregions,theretailermay sufferfromincreasedcomplexityintheformofaddedcoordinationandmarketingcosts.In addition, increasing differences between existing and new markets may make the firm unable to use learning benefits from experiences in existing markets because accumulated knowledge is a location-boundcapability(Li,2005).Thus,eachnewretailformatmayrequiresubstantial

1Intheretail-tradesector,productmarketswithintwo-digitStandardIndustrialClassification(SIC) industry groups are classified by retail formats. For example, the two-digit industry "food stores" (SIC 54) encompassesretailformatssuchconveniencestores,supermarkets,hypermarketsandsuperstores. Moreover,inthetwo-digitnon-foodretailindustries,productmarketsareclassifiedbyretailformats suchasdepartmentstores(withinSIC53"generalmerchandisestores"),clothingstores(withinSIC56 "apparelandaccessorystores")andconsumerelectronicsstores(withinSIC57"homefurniture, furnishings and equipment stores"). Henley Discussion Paper Series Oh, Sohl and Rugman, March 20149 market research in each foreign region. Further, retail format diversification increases the level of competitionandnumberofcompetitorsinforeignregions(GreenwaldandKahn,2005).In general,becauseincumbentdomesticorhome-regionretailershavealreadyachievedhigher levels of economies of scale and scope as first movers (Bhattacharya and Michael, 2008), it is not easyforlatemoverstoachievesuchscaleandscopeeconomiesinforeignregions.Therefore, product diversification aggravates the negative effects of inter-regional diversification. Hence, Hypothesis4:Productdiversificationhasanegativemoderatingeffectonthe relationship between inter-regional diversification and performance, in a way that it hardens the negative slopes and softens the positive slope in the S-curve fit. 3Methodology 3.1Sample and data Inordertotestthesehypotheses,weobtainedpaneldataon68leadingEuropeanretailfirms over a time period of 14 years (1997 2010) from the database of Planet Retail, a leading private retail research company.2Specifically, based on Planet Retail's ranking lists of 1997 and 2010, we extractedpaneldataonthelargestEuropeanretailfirmsforwhichprofitabilitydatawere available.ThePlanetRetaildatabasehasalsobeenusedbypreviousresearchtoinvestigate topics of international diversification in the retail-trade sector (e.g., Gielens and Dekimpe, 2007; Gielens and Dekimpe, 2001; Mohr et al., 2014).Oursampledfirmsoriginatefrom19Europeancountriesindifferentretailindustries.For example,oursampleincludedleadingEuropeanretailfirms,suchasFrance-basedCarrefour, Netherlands-basedAhold,andU.K.-basedTescoforsupermarketsandhypermarkets;Spain-basedInditex,Sweden-basedH&M,andIreland-basedPrimarkforclothingstores;andSweden-based IKEA, U.K.-based Kingfisher, and Germany-based Hornbach for home and garden stores. Of thesampledfirms,26%originatedfromtheU.K.,9%fromSweden,7%fromGermanyandthe remainder from countries such as Croatia, France or Italy. Moreover, 60% of these firms had their core business in food retail (e.g., convenience stores, supermarkets and hypermarkets), while the remainderhadtheircorebusinessinnon-foodretail(e.g.,consumerelectronics,homeand garden stores, and department stores).

2 http://www1.planetretail.net/ John H Dunning Centre for International Business 10Oh, Sohl and Rugman, March 2014 3.1.1Dependent variable We used return on sales (ROS) as a performance measure. In accordance with previous studies of firmsintheservicessector(e.g.,CaparandKotabe,2003;Contractoretal.,2003;Mohretal., 2014),wechoseROSmainlyduetodataavailability.Asthefirmsinoursamplehavedifferent capitalstructuresandtaxrequirementsthatvaryacrosstheirhomecountries,wecomputed ROSbytakingafirm'searningsbeforeinterestsandtaxes(EBIT)dividedbyitsnetsales.We lagged the ROS data by one year (t + 1) to facilitate causal inference (Lu and Beamish, 2004). 3.1.2Independent variables We operationalized geographic diversification in several ways. First, following previous research, wecalculatedafirm'sforeigntototalsalestooperationalizeitsdegreeofinternational diversification (e.g., Geringer, Beamish, and daCosta, 1989). Second, we calculated a firm's home-regiontototalsalestooperationalizeitsdegreeofdiversificationwithintheEuropeanhome-region(e.g.,OhandRugman,2010).Third,wecomputedinternationaldiversificationusingthe entropyindex(e.g.,Vachani,1991;Hittetal.,2006).Theentropyindexof(total)geographic diversification is expressed by the following equation: Geographic diversification = 1ln(1/ )Ni iiP P

wherePi=proportionofsalesincountryiforafirmwithoperationsinNdifferentcountries. Accordingly, the entropy measure captures a firms degree of geographic diversification in terms ofbreath(i.e.,numberofforeigncountries)anddepth(i.e.,relativeimportanceofeachforeign country).Fourth,wemeasuredintra-regionaldiversification(home-regionaldiversification)using the entropy index. Intra-regional diversification captures a firms degree of diversification across countrieswithintheEuropeanhome-region.Finally,weoperationalizedafirm'sdegreeof inter-regionaldiversification.Inordertoclassifygeographicalregions,wefollowedRugmans(2005) triad region classification. We then calculated inter-regional diversification by using the entropy index on a firm's diversification across these regions. Figure 1 shows changes in intra- and inter-regionaldiversificationforoursampledfirms,asmeasuredbytheentropyindex.AsFigure1 illustrates, European retail firms tended to spread their store operations throughout their home-region,Europe,between1997and2010,whiletheirextentofinter-regionaldiversification remained relatively unchanged. Followingpreviousresearch,weoperationalizedproductdiversificationusingtheabove describedentropyindex(JacqueminandBerry,1979;Palepu,1985).Specifically,we operationalizedretailfirms'productdiversificationinthreeways.First,sincetheretailformats Henley Discussion Paper Series Oh, Sohl and Rugman, March 201411 reported in the Planet Retail database correspond to product markets in the Standard Industrial Classification(SIC)system,wemeasuredafirm's(total)formatdiversificationbycapturingits diversificationacrossretailformats,suchassupermarkets,electronicssuperstoresandclothing stores. Second, since Planet Retail also provides data on a retailer's percentage of food and non-foodproductsales,wemeasuredafirm'sdegreeofassortmentdiversificationbycapturingits diversification across food and non-food products. Third, we distinguished between related and unrelatedformatdiversification.Followingpreviousresearch,wecalculatedaretailfirm's unrelatedformatdiversificationacrosstwo-digitSICgroups,suchasgeneralmerchandisestores (SIC53),foodstores(SIC54)andapparelandaccessorystores(SIC56).Relatedformat diversificationwasthencomputedasthedifferencebetweenaretailfirm'stotalformat diversificationandunrelatedformatdiversification(Palepu,1985).Finally,wemean-centered theintra-andinter-regionaldiversificationas wellasourproductdiversificationvariableswhen we calculated the interaction terms (Aiken and West, 1991). 3.1.3Control variables Consistent with previous studies (e.g., Mayer and Whittington, 2003; Wan and Hoskisson, 2003; Qian et al., 2010), we used several firm- and home country-specific control variables in order to addressalternativeexplanationsforourresults.First,wecontrolledforfirmsizeusingthe logarithmofafirm'ssales.Second,wecontrolledforfirmgrowthbycomputingafirmsannual sales growth.3 Third, in order to account for differences in investments between small and large chainstoreretailers,wefollowedGielensandDekimpes(2001)exampleandcomputedfirm efficiency by taking a retail firm's sales over investments (i.e., sales per unit of selling area). Fourth, wecontrolledforretailindustrysizepercapitaandretailindustrygrowthinaretailershome country.Weobtainedthishomecountry-specificdatafromtheEuromonitordatabase.Fifth,we includeddummyvariablesforeachyearinordertocontrolforunobservedseasonaland macroeconomictrendsintheeconomy.Finally,wewerealsointerestedinaccountingforfirm-specific resources that are difficult to approximate with control variables. Therefore, we decided touseafirmfixed-effectsmodelforcoefficientestimation.Thefixed-effectsmodelconsiders unobserved heterogeneity across sampled firms.

3 For our robustness checks, we alternatively used the selling area in square meters as firm size and annual growthofthesellingareaasfirmgrowth.Theresultsshowedthatourkeyfindingswererobusttothis alternative measures. John H Dunning Centre for International Business 12Oh, Sohl and Rugman, March 2014 3.2Model Theequationexplainingtherelationshipbetweeninternationaldiversificationandfirm performance is expressed as: Yit = + Xit + Zit + i + it where Yit is the vector of performance observations for firm i (i = 1,,68) and time t (t = 1,,14); Xit is the vector of the international and product diversification measures and their interactions; andZitisthevectorofthecontrolvariables.Furthermore,i+itistheresidualwithi asthe fixed-effectthatdiffersbetweenfirms,butisconstantovertimeforeachsinglefirm.Thus,the fixed-effects model relates changes in profitability (ROS) for a given retail firm to changes in the independentandcontrolvariablesforthesamefirm.Wealsoestimatedarandom-effects model,buttheHausmanspecificationtestwassignificant,suggestingthatthefixed-effects model should be preferred to obtain consistent coefficient estimates. 4Results Table 1 illustrates the descriptive statistics and Pearson correlations for the variables used in our study. The descriptive statistics show that, on average, the sampled firms achieve less than 9% of theirsalesoutsidetheirhome-region,indicatingthatcross-regionalmovesarestillrareamong Europeanretailfirms.Allofthecorrelationcoefficientsbetweenexplanatoryvariablesusedin ourregressionswerebelowtherecommendedthresholdof0.8(MasonandPerreault,1991). Moreover,weperformedpooledestimationsofourmodelsinordertocalculatethevariance inflationfactor(VIF)andfoundthatthemaineffectsandcontrolvariableshadaVIFlowerthan thesuggestedcriticallimitof10.OurresultsrevealedthatthemodelVIFsassociatedwithour independentvariablesdidnotexceed2.8andtheVIFsofthe individualvariabledidnotexceed 5.4.4Thus,weconcludedthatmulticollinearitywasnotasignificantconcernbetweenour variables. Table2showsourresultsforHypotheses1and2.Inordertochecktheconsistencyofour results,wefirsttestedourmodelwithforeigntototalsalesasameasureofgeographic diversification. Column 1 of Table 1 shows an S-curve fit between geographic diversification and performance(ROS),whichsupportstheseminalfindingofthethree-stageparadigmof internationalization(Contractoretal.,2003).InColumn2ofTable1,weusedhome-regionto totalsalesandfoundahorizontalS-curvefit,whichsupportstheregionalnatureofgeographic

4TheVIFsweretestedforallofthemodelsinTables2and3,includingtheinteractionmodels.When calculating the VIFs, we did not take into account the quadratic and cubic terms. Henley Discussion Paper Series Oh, Sohl and Rugman, March 201413 diversification and performance relationship (Rugman and Oh, 2010). We note here that home-regiontototalsalesincludesdomesticsalesandthatthefindingofhome-regiontototalsales (horizontalS-curve)shouldbeoppositefromthefindingofforeigntototalsales(S-curve).In Column3,weusedtheentropymeasureofglobaldiversification(Vachani,1991)andallofthe diversification variables were insignificant. In Column 4 of Table 2, we included the intra-regional and inter-regional diversification variables and their quadratic and cubic terms to test our Hypotheses 1 and 2. The results showed that the lineartermoftheintra-regionaldiversificationvariablewaspositiveandsignificant(=0.061, p