JournalofEconomicLiterature,Vol.XXXIX December2001 ...jrauch/pdfs/JEL_Dec_2001.pdfallied families....

27
Journal of Economic Literature Vol. XXXIX (December 2001) pp. 1177–1203 Rauch: Business and Social Networks in International Trade Journal of Economic Literature, Vol. XXXIX (December 2001) Business and Social Networks in International Trade JAMES E. RAUCH 1 Today . . . new transportation and com- munications technologies allow even the smallest firms to build partnerships with for- eign producers to tap overseas expertise, cost-savings, and markets . . . The scarce re- source in this new environment is the ability to locate foreign partners quickly and to man- age complex business relationships across cul- tural and linguistic boundaries ... [T]he Chinese and Indian entrepreneurs of Silicon Valley . . . are creating social structures that enable even the smallest producers to locate and maintain mutually beneficial collabo- rations across long distances. [AnnaLee Saxenian 1999, pp. 54–55] 1. Introduction Nations appear to trade too much with themselves and too little with each other (John McCallum 1995; Daniel Trefler 1995; John Helliwell 1998). Jonathan Eaton and Samuel Kortum (2000, p. 27) calculate that “zero grav- ity” (no geographic barriers to trade) would imply a more than fivefold in- crease in world trade. Attempts to ex- plain this “mystery of the missing trade” have increasingly focused on informal trade barriers, especially weak enforce- ment of international contracts (James Anderson and Douglas Marcouiller forthcoming) and inadequate informa- tion about international trading oppor- tunities (Richard Portes and Hélène Rey 1999). Business and social networks that operate across national borders can help to overcome these kinds of infor- mal trade barriers. In the work quoted above, for example, Saxenian (1999) shows that a transnational community of Indian engineers has facilitated out- sourcing of software development from Silicon Valley to regions like Bangalore and Hyderabad. Research can provide us with insight into how transnational networks overcome informal trade bar- riers, and at the same time can serve to document and even quantify the exis- tence of these barriers. Determining the relative importance of contract enforce- ment versus informational barriers is especially important since they point to quite different areas of concern for policy makers. Sections 3 and 4 of this survey are devoted to research moti- vated by the roles of transnational networks in alleviating problems of con- tract enforcement and providing in- formation about trading opportunities, respectively. Whereas transnational networks have primarily been studied as means of overcoming informal trade barriers, 1177 1 Department of Economics, University of Cali- fornia, San Diego. I am grateful to Alessandra Casella, Robert Feenstra, Joel Watson, and three referees for helpful comments and discussions. Financial support was provided by NSF grant #SBR–9709237.

Transcript of JournalofEconomicLiterature,Vol.XXXIX December2001 ...jrauch/pdfs/JEL_Dec_2001.pdfallied families....

Page 1: JournalofEconomicLiterature,Vol.XXXIX December2001 ...jrauch/pdfs/JEL_Dec_2001.pdfallied families. Typical mechanisms serving to integrate the firms include ... and case study. An

Journal of Economic LiteratureVol. XXXIX (December 2001) pp. 1177–1203

Rauch: Business and Social Networks in International TradeJournal of Economic Literature, Vol. XXXIX (December 2001)

Business and Social Networksin International Trade

JAMES E. RAUCH1

Today . . . new transportation and com-munications technologies allow even thesmallest firms to build partnerships with for-eign producers to tap overseas expertise,cost-savings, and markets . . . The scarce re-source in this new environment is the abilityto locate foreign partners quickly and to man-age complex business relationships across cul-tural and linguistic boundaries . . . [T]heChinese and Indian entrepreneurs of SiliconValley . . . are creating social structures thatenable even the smallest producers to locateand maintain mutually beneficial collabo-rations across long distances. [AnnaLeeSaxenian 1999, pp. 54–55]

1. Introduction

Nations appear to trade too muchwith themselves and too little with eachother (John McCallum 1995; DanielTrefler 1995; John Helliwell 1998).Jonathan Eaton and Samuel Kortum(2000, p. 27) calculate that “zero grav-ity” (no geographic barriers to trade)would imply a more than fivefold in-crease in world trade. Attempts to ex-plain this “mystery of the missing trade”have increasingly focused on informaltrade barriers, especially weak enforce-ment of international contracts (James

Anderson and Douglas Marcouillerforthcoming) and inadequate informa-tion about international trading oppor-tunities (Richard Portes and HélèneRey 1999).

Business and social networks thatoperate across national borders canhelp to overcome these kinds of infor-mal trade barriers. In the work quotedabove, for example, Saxenian (1999)shows that a transnational community ofIndian engineers has facilitated out-sourcing of software development fromSilicon Valley to regions like Bangaloreand Hyderabad. Research can provideus with insight into how transnationalnetworks overcome informal trade bar-riers, and at the same time can serve todocument and even quantify the exis-tence of these barriers. Determining therelative importance of contract enforce-ment versus informational barriers isespecially important since they pointto quite different areas of concern forpolicy makers. Sections 3 and 4 of thissurvey are devoted to research moti-vated by the roles of transnationalnetworks in alleviating problems of con-tract enforcement and providing in-formation about trading opportunities,respectively.

Whereas transnational networks haveprimarily been studied as means ofovercoming informal trade barriers,

1177

1 Department of Economics, University of Cali-fornia, San Diego. I am grateful to AlessandraCasella, Robert Feenstra, Joel Watson, and threereferees for helpful comments and discussions.Financial support was provided by NSF grant#SBR–9709237.

Page 2: JournalofEconomicLiterature,Vol.XXXIX December2001 ...jrauch/pdfs/JEL_Dec_2001.pdfallied families. Typical mechanisms serving to integrate the firms include ... and case study. An

much of the research on the impact ofdomestic networks on internationaltrade has been motivated by the per-ception that they constitute informaltrade barriers in themselves, with net-work members colluding to increasetheir market power by restricting for-eign competition. There is also a newline of work that investigates the effectof domestic networks on the composi-tion of international trade. Section 5 ofthis survey covers both these strands ofresearch on domestic networks.

These three main sections dealingprimarily with the role of networks inovercoming or creating trade barrierswill be further unified by two over-arching questions. First, what conse-quences for economic efficiency areassociated with the impact of networkson international trade? We will see thatin many instances the answers to thisquestion turn on the extent to whichnetworks are open to new members.Second, will networks grow or shrink inimportance for international trade? Thefirst reaction of most economists ispresumably the latter: institutions forinternational contract enforcement areprobably improving, and technology forinformation dissemination certainly is,hence the need for networks to over-come informal trade barriers is declin-ing. To forestall judgment, let us simplynote for the time being that this onlyaddresses the demand for transnationalnetworks and leaves out entirely thequestion of changes in their supply.

Two other important areas of re-search on business and social networksin international trade are: the role of in-termediaries who can connect foreignagents to domestic networks, and theability of transnational production net-works to facilitate technology trans-fer. These two directions have beenpursued more by noneconomists andeconomists based outside of academia

(in institutions such as the World Bank)than by academic economists. Theoriz-ing in these areas has been rather infor-mal, and empirical work has consistedmainly of case studies. The purpose ofthe two brief sections (6 and 7) devotedto this research is therefore less to criti-cally review the literature than to tryto establish agendas for academic econo-mists interested in working in theseareas.

This survey is framed by the exam-ples and definitions of economic net-works given in the next section. Thethree main sections follow. After thetwo brief sections on network interme-diaries and learning in transnationalproduction networks, a final sectionreviews the answers to the two over-arching questions above and highlightsadditional issues for future research.

2. Examples and Definitions

Empirical research into the impact ofnetworks on international trade hastended to lead theorizing, as will beclear below. Most of the empirical worksurveyed concerns coethnic networksand business groups with publicly re-corded membership such as the Japa-nese keiretsu. Coethnic networks arecommunities of individuals or busi-nesses that share a demographic attri-bute such as ethnicity or religion. Busi-ness groups are “sets of firms that areintegrated neither completely nor barelyat all” (Mark Granovetter 1995, pp. 96–97), and where the lineages of themembers can often be traced back to afounding family or small number ofallied families. Typical mechanismsserving to integrate the firms includemutual stockholdings and frequentmeetings of top executives. This focuson coethnic networks and businessgroups is explained by observabilityrather than primacy of importance:

1178 Journal of Economic Literature, Vol. XXXIX (December 2001)

Page 3: JournalofEconomicLiterature,Vol.XXXIX December2001 ...jrauch/pdfs/JEL_Dec_2001.pdfallied families. Typical mechanisms serving to integrate the firms include ... and case study. An

census takers will not record the char-acteristic “former employee of IBM,”yet the fact that many of the key deci-sion makers in the hard disk drive indus-try shared this characteristic contrib-uted to the rapid spread of popularity ofSingapore as a site for FDI, accordingto industry observers (David McKen-drick 1998). However ubiquitous2 net-works may be in the conduct of interna-tional trade, little can be learned abouttheir impact unless they are observable.

In some contexts the key feature ofthe networks studied below is that theirmembers are engaged in repeated ex-change that helps sustain cooperation/collusion.3 In other contexts the keyfeature is that network members havethorough knowledge of each others’characteristics, which helps them matchwith each other or refer each other tooutside business opportunities. Thesekey features roughly correspond to twodefinitions of economic networks usedin the sociological literature. The first,based on Joel Podolny and Karen Page(1998, p. 59), defines an economic net-work as a group of agents that pursuerepeated, enduring exchange relationswith one another. The second, weakerdefinition is based on the work ofGranovetter (1973, 1995 [1974]): a setof actors who know each others’ rele-vant characteristics or can learn themthrough referral.4

For some purposes these definitionsare clearly too broad to be useful. Orga-nized international spot markets do notexist for most traded goods (especiallymanufactures), so it is entirely possiblethat systematic micro surveys would re-veal that even by the repeated exchangedefinition most international trade couldbe viewed as taking place through net-works.5 Egan and Mody (1992, p. 325),for example, state of trade in bicyclesand footwear, “Most U.S. buyers inter-viewed for this study preferred long-term, stable and direct relationshipswith both developed and developingcountry suppliers.” This broad defini-tion of the role of networks in interna-tional trade may provide insight in somecontexts, one being the literature sur-veyed in section 7, which suggests thatthis network view of trade could pro-vide a new way of thinking about theconnection between trade and technol-ogy transfer. It clearly will not do, how-ever, when evaluating the claim thatnetworks overcome informal barriers totrade, because it admits networks thatwere created by trade. Hence in consid-ering the impact of networks in over-coming trade barriers in sections 3 and4, I restrict the discussion to networksthat were formed domestically and be-came transnational through migrationor foreign direct investment. This inter-nationalization may have taken place

2 Discussing imports of manufactures from lessdeveloped countries, Mary Egan and AshokaMody (1992, p. 329) report, “When evaluating po-tential suppliers, virtually all buyers first seek in-formation within their own network. This networkis a tight system of product-specific buyers andsuppliers of both finished goods and components.The first source of information is the personaljudgment of other buyers.”

3 The trust produced within these networks mayfacilitate flows of financial capital as well as goodsbetween network members.

4 If one is engaged in repeated, enduring ex-change one presumably learns the relevant charac-teristics of one’s partner. Hence the first defini-tion is clearly stronger than the second and

corresponds to “strong ties” among network mem-bers as opposed to the “weak ties” emphasized byGranovetter.

5 The same might be said of domestic trade. It isthe contention of the “Uppsala School” of interna-tional business analysis that “markets are more orless stable networks of business relationships”(Mats Forsgren and Jan Johanson 1992, p. 5). Theinsight that the study of business and social net-works in international trade can provide into thefunctioning of market economies in general wouldhave to be the subject of a separate paper. Su-bramanian Rangan and Robert Lawrence (1999,ch. 4) have shown how a network perspective can helpto understand multinational enterprise responsesto changes in real exchange rates.

Rauch: Business and Social Networks in International Trade 1179

Page 4: JournalofEconomicLiterature,Vol.XXXIX December2001 ...jrauch/pdfs/JEL_Dec_2001.pdfallied families. Typical mechanisms serving to integrate the firms include ... and case study. An

generations ago, but in this case wewill see that it is more accurate to saythat these networks are sustained bythe trade they are creating than to saythat these networks were created bytrade.

The concentration of empirical re-search on business groups and coethnicnetworks has two unfortunate potentialside effects. First, conclusions might bebiased by the difficulty of entry intothese networks relative to others. Sec-ond, the impression might be conveyedthat networks are a vestigial or at leastculture-bound phenomenon. This im-pression would come from the beliefsthat coethnic or coreligious ties are ofdeclining importance in commerce andthat business groups are confined toless developed countries (the gruposeconómicos of Latin America or “busi-ness houses” of India) or East Asia (thechaebol of Korea or keiretsu of Japan).Though trends for coethnic networksand business groups need not apply tonetworks as a whole, we should never-theless note that the former belief isat least debatable (see Joel Kotkin1992) and the latter belief is incor-rect: business groups are ubiquitous inthe developed countries of continentalEurope (see David Encaoua and AlexisJacquemin 1982 and the referencestherein).6

That said, nowhere in this survey isthe impression that the impact of net-works on international trade is a waningphenomenon more strongly conveyedthan in the next section. Subsequentsections give much more ambivalentanswers to the question of whethernetworks are growing or shrinking inimportance for international trade.

3. Networks and Opportunism

Enforcement of contracts in interna-tional trade presents a difficult prob-lem. This section is concerned with hownetworks facilitate trade across politiesby building, or substituting for, trustwhen contract enforcement is weak tononexistent. Regarding the overseasChinese network, Maurray Weiden-baum and Samuel Hughes (1996, p. 51)report, “If a business owner violates anagreement, he is blacklisted. This is farworse than being sued, because the en-tire Chinese network will refrain fromdoing business with the guilty party.”This kind of description suggests thattransnational networks deter opportun-ism in a modern setting, yet the bulk ofthe more analytical literature has fo-cused on the distant past or extremelyunderdeveloped environments. This fo-cus is related to the exclusive relianceof this literature on the tools of theoryand case study. An important futuretask should be collection of data forcontemporary networks that facilitatestatistical testing of hypotheses.

The literature surveyed invariablyuses the repeated exchange definitionof networks. Its subject is always tradediasporas, which are ethnic or religiousgroups with settlements at endpointsand transshipment points of a traderoute. According to the survey by PhilipCurtin (1984), trade diasporas domi-nated cross-cultural trade in most partsof the world until the nineteenth

6 It is the virtual absence of business groups inthe United Kingdom and the United States thatgives rise to this common misperception amongeconomists, plus their low public profile inEurope. Encaoua and Jacquemin (1982, p. 26)note that business groups in France “have no legalexistence and are not identified in official cen-suses. Each subsidiary maintains its legal auton-omy and keeps separate accounts.” A related mis-perception is that the typical large corporation inwealthy countries is widely held, whereas RafaelLa Porta, Florencio Lopez-de-Silanes, and AndreiSchleifer (1999) find that these firms are typicallycontrolled by families or the state. Again, the mis-perception arises from the fact that the typicallarge corporation is indeed widely held in theUnited Kingdom and the United States.

1180 Journal of Economic Literature, Vol. XXXIX (December 2001)

Page 5: JournalofEconomicLiterature,Vol.XXXIX December2001 ...jrauch/pdfs/JEL_Dec_2001.pdfallied families. Typical mechanisms serving to integrate the firms include ... and case study. An

century. Curtin uses the term “cross-cultural” because the term “interna-tional” is an anachronism for the periodpreceding the rise of the nation-state.To some extent trade diasporas consti-tuted polities in themselves, but theirlegal enforcement powers were veryweak (see Abner Cohen 1969, 1971;Avner Greif 1989, 1993). Sellers oftenavoided this problem by traveling withtheir goods, but agency relations werealso prevalent in many trade diasporas.The use of agency is best documentedby Curtin for Armenians and severalWest African trade diasporas. Curtinasserts (p. 197), “Within the [17th–18thcentury] Armenian community, con-tracts could be made on a handshake.”How were trade diasporas able to de-ter opportunistic behavior and therebysustain agency?

One leading answer is given by Cohen(1969): a trade diaspora created trust byestablishing a “moral community.” Anadvantage of Cohen’s work is that hestudies a trade diaspora, the Hausa inWest Africa, that is contemporary to thetime in which he is writing, which al-lows him to observe the operation ofthe diaspora directly rather than infer itfrom historical documents. The key ac-tors in Cohen’s account are landlord-brokers resident in Ibadan, Nigeria, siteof the main Hausa trading settlement.These landlord-brokers employed com-mission agents (“clients”) to sell cattleor purchase kola nuts on behalf of deal-ers located elsewhere in the diaspora.Opportunistic behavior by the landlord-brokers vis-à-vis the dealers was limitedby the fact that they fell under theauthority of the chief of the Hausaquarter in Ibadan and by their traditionof accumulating wealth in the form ofhousing assets: “A landlord cannot sellhis houses overnight and leave the com-munity after embezzling the money oftraders. On the other hand, when it is

necessary, the Chief can put a greatdeal of pressure on a landlord in diffi-culties to sell some of his housing assetsin order to meet his financial obliga-tions to traders” (Cohen 1971, p. 274).No such mechanisms prevented cheat-ing by the clients of the landlord-broker, however, so it is here wheredevelopment of “moral community” ismost crucial.

Cohen (1969, p. 89) notes that therelationships between a landlord-brokerand his clients constitute a networkrather than a hierarchy within thelarger network-polity of the diaspora,not only because there is no bindingmeans of dispute resolution, but alsobecause “the line of demarcation [be-tween landlords and clients] is not rigidbut is continually crossed by men mov-ing to the one category or to the other.”He argues that theft was deterred by ty-ing the repeated economic exchangebetween a landlord-broker and his cli-ents to building of pseudo-kin relations:“The landlord will try to marry the cli-ent off to . . . the foster daughters ofhis wives, or the daughters of his ownrelatives or of his wives’ relatives” (p.88); “The landlord can thus be seen asthe centre of a network of kinship rela-tions with his clients” (p. 91). Of coursethis group endogamy was facilitated bythe separate identity the diaspora mi-nority maintained within the host soci-ety. Cohen concludes (p. 91) that “therelationship which thus develops be-tween landlord and client cannot be mea-sured in material or contractual termsalone. It is a relation which cannot becompletely reduced to economic or po-litical relations. There is an inescapablemoral bond as well.”

The other leading answer to thequestion of how trade diasporas wereable to sustain agency relations, antici-pated by the above quotation regardingoverseas Chinese, is given by Greif

Rauch: Business and Social Networks in International Trade 1181

Page 6: JournalofEconomicLiterature,Vol.XXXIX December2001 ...jrauch/pdfs/JEL_Dec_2001.pdfallied families. Typical mechanisms serving to integrate the firms include ... and case study. An

(1989, 1993): the threat of collectivepunishment of deviant agents by allmerchants (principals) in the diasporasubstituted for trust. Formally, a re-peated game equilibrium in which thestrategy of each merchant in a coalitionis to refuse forever to deal with anagent who cheats any merchant in thecoalition sustains a lower efficiencywage premium than a repeated gameequilibrium in which each merchant’sstrategy is to punish only an agent whocheats him (Greif 1993). The higher ef-ficiency wage required to prevent theagent from cheating under a bilateralpunishment strategy could be so high asto make agency relations unprofitable.

Greif studies the Maghribi tradersof the eleventh-century Mediterraneanthrough their documents.7 Like Cohenin his discussion of landlord-brokersand clients, Greif (1989, p. 874) empha-sizes the “horizontal” or non-hierarchicalnature of the Maghribi network: “Onedoes not observe the existence of twoseparate ‘classes’ . . . an agents classand a merchants class . . . the Magh-ribi traders group was a homogeneousgroup of middle-class traders and eachof them operated as a merchant and asan agent at the same time.” He presentsevidence (1989, pp. 868–69) for boththe transmission of information regard-ing past conduct of agents throughoutthe network required by the coalitionstrategy and for the operation of collec-tive punishment. For example, whenthe creditors of Samhun ben Da’ud, aprominent trader from Tunisia, werenot paid, he complained that “. . . theirletters filled with condemnation hadreached everyone.” Regarding collec-tive punishment, Greif reports the fol-lowing example: “an agent who lived inJerusalem, Abun ben Zedaka, was ac-

cused (although not charged in court)of embezzling the money of a Maghribitrader. When word of this accusationreached other Maghribi traders, mer-chants as far away as Sicily canceledtheir agency relations with him.”

Since establishment of a moral com-munity and collective punishment ofcheaters are not mutually exclusivemechanisms for discouraging opportu-nistic behavior among agents, it seemslikely that both were used among theHausa, Maghribis, and other trade dias-poras. Indeed, Cohen (1969, p. 173)mentions that the chief of the quartercould resort to “public scandalizing” ofa landlord-broker whom he could notpunish effectively in other ways. Thiswas accomplished through a specialmeeting, held in front of the house ofthe accused man, to which many Hausadealers who happened to be in thequarter would be invited, ensuringthe spread of the scandal throughoutthe diaspora. Similarly, kinship ties arenot entirely absent from the analysis byGreif, who states (1989, p. 875) that theMaghribi coalition was able to sur-mount the “endgame” problem because“an old agent would not cheat becausehe feared that he would be punishedthrough the punishment imposed on hisrelatives. Only moral responsibility ofrelatives was required to enable tradersto base their relations upon a reputa-tion mechanism despite the fact thateach of them lived for a finite numberof years.” To some extent it may be thatdifferent theoretical perspectives of Co-hen and Greif, rather than differencesbetween the Hausa and Maghribi tradediasporas, caused the former to empha-size moral community and the latter toemphasize collective punishment. Onone very important point, however, thetwo authors are in complete agreement:both the Hausa and the Maghribismaintained their separate identities for

7 The Maghribis were a group that maintained aseparate identity within the Jewish diaspora of theIslamic Mediterranean.

1182 Journal of Economic Literature, Vol. XXXIX (December 2001)

Page 7: JournalofEconomicLiterature,Vol.XXXIX December2001 ...jrauch/pdfs/JEL_Dec_2001.pdfallied families. Typical mechanisms serving to integrate the firms include ... and case study. An

economic reasons, i.e., to facilitate theirparticipation in profitable trade.

The efficiency consequences of usingnetworks to solve the problem of op-portunism in cross-cultural (or interna-tional) trade would seem to be straight-forward. It seems clear that there is nettrade creation, which should yield theusual gains from improved allocation ofresources, and that the network mem-bers themselves should capture a sur-plus from cooperation. In addition tothe theoretical argument of Greif thatprincipals within a repeated exchangenetwork can pay a lower wage to theiragents, Cohen shows that the Hausaout-competed both the main rival WestAfrican ethnic group (the Yoruba) andEuropeans in controlling the cattle andkola trades.

Greif (1994), however, raises the in-triguing possibility that the use of net-works to facilitate cross-cultural trade,while statically efficient, is actually dy-namically inefficient. In particular, be-cause an agent from outside the groupmust be paid a higher wage, the growthof trade handled by a network is hin-dered by the reluctance of merchantsfollowing a “collectivist strategy” to ini-tiate intereconomy relations. This isnot true for merchants following an“individualist strategy” of bilateralpunishment.

In my view this argument is not deci-sive for two reasons. First, the savingsto the traders following a collectiviststrategy from paying lower wages mayfinance greater investment and growthin the presence of imperfect capitalmarkets. Second, the network may sim-ply incorporate the trader from theother economy. The exclusiveness ofeven coethnic networks can be exagger-ated: Cohen (1969, p. 49) notes that “itis possible for some non-Hausa . . . tobecome, in effect, Sabo Hausa [Hausaof the Quarter]”; and regarding the

overseas Chinese, Constance Lever-Tracy et al. (1991, p. 79) report that“those studied in Australia were proudof trusting relations they had builtup with Anglophone Australians.” Ofcourse the process of incorporation re-quires time and resources, but it maybe cheaper in present discounted valueterms than paying a higher wage for-ever. Indeed, part of what the greatersavings of traders following a collectiv-ist strategy might finance is preciselyinvestment in cultivation of new net-work members. Whether or not net-works create inefficiencies in expandingtrade to new areas or cultures is thus anempirical question.8

Greif (1994) also argues that the useof networks to suppress opportunisticbehavior in international trade was dy-namically inefficient in the sense that ithindered institutional innovation de-signed to accomplish the same purpose.He shows that during the twelfth cen-tury the Genoese, whom he identifies asfollowing individualist strategies, “de-veloped an extensive legal system forregistration and enforcement of con-tracts” (p. 937) as well as bookkeepinginnovations that made it easier to detecttheft of goods being sent overseas. Itcertainly seems clear that the use ofnetworks to substitute for formal-legalmeans of enforcing contracts will hin-der the improvement of such means.The more general question of the rela-tionship between the use of networks ininternational trade and technologicalprogress is considered in section 7.

Improvements in international con-tract enforcement and other formal-legal

8 Hubert Schmitz (1999) finds that trust withindomestic production networks of shoe exportingfirms in Brazil and surgical instrument exportingfirms in Pakistan was able to evolve from a basis in“socio-cultural ties” to a basis in “conscious invest-ments in interfirm relationships” (p. 145) in re-sponse to the need to penetrate markets for higherquality goods.

Rauch: Business and Social Networks in International Trade 1183

Page 8: JournalofEconomicLiterature,Vol.XXXIX December2001 ...jrauch/pdfs/JEL_Dec_2001.pdfallied families. Typical mechanisms serving to integrate the firms include ... and case study. An

means of substituting for trust in inter-national trade have continued to thepresent day. Two important examplesare international commercial arbi-tration and letters of credit. Interna-tional commercial arbitration offers aprivate means of dispute resolution;Laurence Craig, William Park, and JanPaulsson (1985) is the standard refer-ence for International Chamber of Com-merce arbitration. Letters of credit al-low the trading parties to shift some oftheir commercial credit risk to the issu-ing bank and allow the buyer to deferpayment until the shipment passes qual-ity inspection; for more information ontheir benefits one can see Charles delBusto (1994). Such innovations havesurely reduced the demand for net-works as a means of deterring oppor-tunistic behavior in international tradeand have contributed, along with tech-nological innovations in communicationand transport, to “the twilight of thetrade diasporas” (Curtin 1984, ch. 11).(I postpone consideration of changes inthe supply of networks to the next sec-tion.) An important future direction forthe literature on transnational networksand opportunism should be to move be-yond trade diasporas and to integratecontemporary international trade lawand institutions into the analysis asoutside options that may influence howthe networks operate.

4. Networks and Opportunities

A more recent literature has empha-sized that, in addition to being used totransmit information about past oppor-tunistic business conduct, networks canbe used to transmit information aboutcurrent opportunities for profitable in-ternational trade (or investment). Theliterature surveyed almost always usesthe “characteristics knowledge” defini-tion of networks, since the key is know-

ing the agents’ characteristics so as tobe able to match them to opportunities.Transnational networks can facilitatethis matching through provision of mar-ket information, letting suppliers knowthat consumers in a particular countrywill be receptive to their products, orenlightening suppliers on how to adapttheir products to consumer preferencesin a given country. Korean wig exportsto the United States are an especiallywell-studied example (Ku-Sup Chin,In-Jin Yoon, and David Smith 1996,p. 498):

Korean wig importers’ contribution to theKorean wig import business was far greaterthan their numbers. From these immigrantwig importers, South Korea wig manufactur-ers could obtain information on new stylesand market trends. Since they were not ableto develop new styles of their own (promi-nent U.S. hair designers continuously devel-oped innovative styles), South Korean wigmanufacturers had to depend entirely on Ko-rean immigrant wig importers for informationon trends in U.S. wig fashion.

Within a given foreign market, transna-tional networks can also help producersof consumer goods to find appropriatedistributors, assemblers to find the rightcomponent suppliers, and investors tofind joint-venture partners. Weiden-baum and Hughes (1996, p. 55) write ofthe overseas Chinese:

[T]he members of the bamboo network oper-ate in the interstices of the trading world.They make components, manufacture for oth-ers, and perform subassembly work. They arealso heavily involved in wholesaling, financ-ing, sourcing, and transporting. . . . Theleading businessmen know each other person-ally and do deals together, with informationspreading through an informal network ratherthan through more conventional channels.

The empirical analyses reviewed be-low have provided evidence for thetrade-creating effects of immigrants andof business groups operating acrossnational borders. Immigrants know the

1184 Journal of Economic Literature, Vol. XXXIX (December 2001)

Page 9: JournalofEconomicLiterature,Vol.XXXIX December2001 ...jrauch/pdfs/JEL_Dec_2001.pdfallied families. Typical mechanisms serving to integrate the firms include ... and case study. An

characteristics of many domestic buyersand sellers and carry this knowledgeabroad. Foreign direct investment byone or more members of a domesticbusiness group has the same effect.These empirical papers have not beenguided by formal theory. The measuresof network strength used are ad hoc,and the models predicting trade towhich they are added are taken “off theshelf.” As a result it is hard to identifythe extent to which the estimated coef-ficients tell us that transnational net-works work through provision of marketinformation and matching and referralservices as opposed to other means.Readers should therefore keep in mindthat all the results reported are open toalternative interpretations.

We first consider studies of the im-pact on bilateral trade of immigrants.An immediate concern is that, ratherthan a network effect, any positive im-pact might simply reflect immigranttaste for goods from their countries oforigin or the correlation of immigrationwith country characteristics that pro-mote trade, such as proximity. DavidGould (1994) allays both concerns byestimating separate equations for ex-ports and imports and including countrydummies in his study of the immigrantimpact on U.S. bilateral trade with 47trading partners during the period1970–86. His basic estimating frame-work is Jeffrey Bergstrand’s (1985) for-mulation of the gravity equation, towhich Gould adds a lagged dependentvariable, the stock of immigrants in theUnited States from each partner, andmeasures of immigrant skill composi-tion and length of stay in the UnitedStates.9 The coefficients on the immi-

grant stock are positive and highly sig-nificant in both the export and importequations, but the coefficients on theimmigrant skill composition and lengthof stay variables are insignificant. Theimplied long-run elasticities indicatethat a 10-percent increase in immi-grants to the United States will increaseU.S. exports to the country of origin by4.7 percent and U.S. imports from thecountry of origin by 8.3 percent.10 Areasonable interpretation of the largerpoint estimate for the import elastic-ity is that it combines a taste effectand a network effect, while the exportelasticity only reflects a network effect.

Keith Head and John Ries (1998) es-sentially repeat Gould’s exercise forCanada, investigating the impact of im-migrants on Canadian bilateral importsand exports for 136 trading partnersduring the period 1980–92. Their con-stant elasticity specification using alagged dependent variable (table 2)is closest to that of Gould, the maindifferences being omission of countrydummies and use of a Tobit specifica-tion to allow for the many observations ofzero on Canadian bilateral exports andimports. For two different measures ofthe immigrant stock the estimated coef-ficients are positive and significant inboth the export and import equations.The implied long-run elasticities for the

9 Gould’s preferred specification transforms theimmigrant stock to allow for the possibility of di-minishing returns in the impact of immigration ontrade. Because the parameter that determines thedegree of diminishing returns is very imprecisely

estimated, for the purpose of quantifying the im-pact of immigrants on trade I prefer to use theresults from Gould’s alternative constant elasticityspecification (i.e., using the log of the immigrantstock). Nevertheless, it is worth noting that thepoint estimates of the diminishing returns parame-ter have the intriguing implication that most of theimmigrant effect on exports exhausts itself for amuch smaller number of immigrants than doesmost of the effect on imports.

10 These elasticities are computed using the co-efficients on the lagged dependent variables fromthe equations using the diminishing returns speci-fication because Gould does not report these co-efficients for the equations using the constantelasticity specification.

Rauch: Business and Social Networks in International Trade 1185

Page 10: JournalofEconomicLiterature,Vol.XXXIX December2001 ...jrauch/pdfs/JEL_Dec_2001.pdfallied families. Typical mechanisms serving to integrate the firms include ... and case study. An

preferred immigrant stock measure in-dicate that a 10-percent increase in im-migrants to Canada will increase Cana-dian exports to the country of origin by1.3 percent and Canadian imports fromthe country of origin by 3.3 percent.The fact that these elasticities are muchlower than those estimated by Gouldcould be due to the nature of Canadiancompared to U.S. trade: Head and Ries(p. 48) point out that “Canada’s mainexport categories, natural resources andUnited States-bound automotive goods,do not seem likely candidates for trans-actions cost reductions by immigrants.”

We turn next to studies of the trade-creating effects of business groups op-erating across national borders. Bothstudies surveyed are concerned withwhether foreign direct investment inmanufacturing by assemblers in “verti-cal” keiretsu stimulates exports by otherkeiretsu members.11 Having an assem-bler abroad whose characteristics theyknow could be helpful to supplierslooking for export opportunities. RenéBelderbos and Leo Sleuwaegen (1998)investigate the 1988 intensity of exportsto the European Community (EC) of 86Japanese firms classified in the elec-tronics or precision machinery indus-tries. They regress the log of the ratioof firm exports to the EC to total salesminus exports to the EC on various firmcharacteristics plus a dummy for whetherthe firm is a member of a verticalkeiretsu for which the assembler oper-ated one or more manufacturing plantsin the EC. The keiretsu dummy is posi-tive and significant in all specifications.Head and Ries (2001) examine total ex-ports of 96 keiretsu suppliers in theautomobile and electronics industriesfor the period 1966–90. They regressthe log of firm exports on firm charac-

teristics, counts of the firm’s foreigndistribution and manufacturing invest-ments, and the count of foreign manu-facturing investments by the assemblerof the vertical keiretsu to which thesupplier belongs. The keiretsu assem-bler investment count is positive andhighly significant.

Rauch and Vitor Trindade (forthcom-ing) attempt a partial synthesis of theliteratures surveyed in this section andthe previous one by examining thetrade-creating effects of what is, in allprobability, the largest transnationalnetwork (or set of interlinked nationalnetworks) in the world: the overseasChinese. On the one hand, the over-seas Chinese can be seen as a latter-day trade diaspora that deters op-portunistic behavior in internationaltransactions.12 On the other hand, theoverseas Chinese can promote tradeby providing market information andmatching and referral services becausethey use coethnic business societies tokeep knowledge of network members’characteristics fresh:

This networked organizational system alsodistinguishes ethnic Chinese business pat-terns from other “personalistic” methods ofconducting business. These types of businesspatterns developed along with various institu-tional supports, like clan halls, regional placeassociations, and “umbrella” organizationssuch as the Chinese Benevolent Associationsin both local and overseas communities.These institutions were primarily involvedin maintaining the social “glue” necessaryfor normative relationships and practices to

11 A vertical keiretsu consists of an assemblerand many component suppliers.

12 Although it was not formed for the purpose ofpromoting trade, some writers have argued thatthe overseas Chinese now maintain their separateidentity for that purpose, just as Cohen (1969) andGreif (1989, 1993) argued that the Hausa andMaghribi, respectively, maintained their separateidentities for economic reasons. For example, intheir fieldwork in Thailand, Chan Kwok Bun andChee Kiong Tong (1993, p. 152) found that acommon reason given for maintaining use of aChinese language was “Chinese is the businesslanguage—if you don’t speak Chinese, how canyou do business?”

1186 Journal of Economic Literature, Vol. XXXIX (December 2001)

Page 11: JournalofEconomicLiterature,Vol.XXXIX December2001 ...jrauch/pdfs/JEL_Dec_2001.pdfallied families. Typical mechanisms serving to integrate the firms include ... and case study. An

continue over time and space. The contempo-rary evolution of these types of institutionshas led to new forms of organization, such asthe Greater Chinese Entrepreneurs’ Confer-ences, the Chinese Business Fairs, and anumber of other international ethnic Chineseconferences devoted primarily to businessconnections and transactions. (KatharyneMitchell and Brian Hammer 1997, p. 78)

Rauch and Trindade try to distin-guish the trust and business opportunityimpacts of the overseas Chinese net-work on bilateral trade by estimatingseparate gravity equations for commodi-ties that have “reference prices” andcommodities that do not. A referenceprice is defined as a price that is quotedwithout mentioning a brand name orother producer identification. Commodi-ties that possess reference prices aretaken to be sufficiently homogeneousthat if traders see the price differentialbetween two countries’ markets is largeenough to cover customs and transportcosts, they know it is profitable to shipthe product. Commodities that do notpossess reference prices are taken to besufficiently differentiated that pricescannot convey enough of the informa-tion relevant for international trade:buyers and sellers must be matched incharacteristics space, and hence thethicker information that can be pro-vided by the overseas Chinese networkis much more important than for inter-national trade in homogeneous com-modities.13 In contrast, moral commu-nity or the threat of collective sanctionsshould deter equally shipments of rot-ting fruit or stockings with runs. Thesame lack of distinction between com-modities with and commodities withoutreference prices should hold for other

forms of opportunistic behavior such asfailure to pay for a shipment one hasreceived. Rauch and Trindade thus ar-gue that an economically and statisti-cally greater impact of the overseasChinese network on bilateral trade incommodities without reference pricesthan in commodities with them estab-lishes a presumption that this networkhas a quantitatively important effect bymatching traders with business opportu-nities, in addition to its effect throughbuilding or substituting for trust.

Rauch and Trindade adapt JeffreyFrankel’s (1997) formulation of thegravity equation and begin with hissample of 63 countries. They have dataon ethnic Chinese population shares for57 and 59 of these countries circa 1980and 1990, respectively, and estimateequations for the log of bilateral trade(sum of exports and imports) for eachyear separately. The strength of theoverseas Chinese network for any twotrading partners is measured by theproduct of their ethnic Chinese popula-tion shares, which gives the probabilitythat, if we select an individual at ran-dom from each country, both will beethnic Chinese. Table 1 uses the gravityequation estimates to compute the per-centage increases in bilateral trade at-tributable to the overseas Chinese net-work, evaluating all variables at theirmean values. As we would expect, thenetwork impact is much larger for tradebetween countries with ethnic Chinesepopulation shares at the levels prevail-ing in Southeast Asia than for all othercountry pairs.14 The percentage in-creases in bilateral trade attributable todirect and indirect colonial ties are alsoprovided as a standard for comparison,13 Rauch and Trindade further divide homoge-

neous commodities into commodities whose refer-ence prices are quoted on organized exchangesand those whose reference prices are quotedsolely in trade publications. Here I only report un-published results for homogeneous commoditiesas an aggregate.

14 Separate coefficients were estimated for thetwo subsets of countries to allow for diminishingreturns. Like Head and Ries (1998), Rauch andTrindade use a Tobit gravity model specificationto allow for observations of zero on bilateral trade.

Rauch: Business and Social Networks in International Trade 1187

Page 12: JournalofEconomicLiterature,Vol.XXXIX December2001 ...jrauch/pdfs/JEL_Dec_2001.pdfallied families. Typical mechanisms serving to integrate the firms include ... and case study. An

where an indirect colonial tie exists be-tween two countries that had the samecolonial power. The integration of com-mercial interests that prevailed duringcolonial periods should have establisheda common business language or linguafranca and a set of business contacts, fa-cilitating the search by producers forthe right distributors, by assemblers forthe right suppliers, and so on.15 The dif-ferences across commodity groups re-ported in table 1 are all statistically sig-nificant, and confirm that the increase

in bilateral trade attributable to theoverseas Chinese network is larger fordifferentiated than for homogeneousproducts.

The results reported in table 1 leadnaturally to a discussion of whetherover time networks will grow or shrinkin their importance for internationaltrade. We see that the impact on bilat-eral trade of both the overseas Chinesenetwork and direct and indirect colonialties is lower in 1990 than in 1980 forboth the homogeneous and the differ-entiated commodity groups. This couldreflect strengthening of internationalcontract enforcement mechanisms andimprovements in communications tech-nology, or it could reflect weakening ofethnic bonds and direct colonial tiesand the spread of English as a commonbusiness language. Countering these(possible) trends are two tendenciesthat should increase rather than de-crease the importance of networks forinternational trade. First, the informa-tion intensity of trade is increasing:Rauch (1999) finds that differentiatedproducts have increased their share of

TABLE 1PERCENTAGE INCREASES IN BILATERAL TRADE ATTRIBUTABLE TO OVERSEAS CHINESE NETWORK

AND TO COLONIAL TIES

HomogeneousProducts

DifferentiatedProducts

Overseas Chinese network, both countries with ethnicChinese population shares > 1% in 1980

95.8 177.8

1980 Overseas Chinese network, all other country pairs 4.1 6.2

Direct and indirect colonial ties 8.6 18.6

Overseas Chinese network, both countries with ethnicChinese population shares > 1% in 1990

32.0 59.2

1990 Overseas Chinese network, all other country pairs 1.7 5.5

Direct and indirect colonial ties 5.2 13.8

Source: Computations based on unpublished estimates from Rauch and Trindade (forthcoming) for the “conserva-tive aggregation.”

15 In addition to the dummy variable for thepresence of a direct or indirect colonial tie, a vari-able was included in the gravity equations thatmeasured the probability that, if we select an indi-vidual at random from each country, they will havea common birth language. The point estimates ofthe coefficients on this variable were alwayssmaller for the differentiated than the homogene-ous commodity group and not statistically signifi-cant for the former group. If common mothertongue and common Chinese ancestry haveroughly equal associations with taste similarity(most emigration from China occurred beforeWorld War I), these results argue against an inter-pretation of the greater impact of the ethnic Chi-nese variable on bilateral trade in differentiatedthan homogeneous products as due to its acting asa proxy for taste similarity.

1188 Journal of Economic Literature, Vol. XXXIX (December 2001)

Page 13: JournalofEconomicLiterature,Vol.XXXIX December2001 ...jrauch/pdfs/JEL_Dec_2001.pdfallied families. Typical mechanisms serving to integrate the firms include ... and case study. An

world trade from 56.5 percent in 1970to 67.1 percent in 1990, and it seemslikely that product differentiation itselfis becoming ever finer. Second, supplyof the other transnational networks cov-ered by studies in this section is in-creasing: a rising flow of migrants, espe-cially from poor to rich countries, hasresulted in the foreign-born accountingfor a growing share of the population ofmost rich countries (Peter Stalker 1994),and business groups are increasinglyspread across national borders.

The implications for economic effi-ciency of transnational networks thatprovide information about profitabletrading opportunities can be evaluatedusing the model of Rauch and Alessan-dra Casella (1998). In their model pro-duction takes place as follows: produc-ers match pairwise, and if the match isacceptable an internationally immobileresource (“labor”) is employed to real-ize the productive opportunity. Domes-tic matching is characterized by com-plete information—every producer knowsthe type of every other—while interna-tional matching is hampered by incom-plete information: because producers’types are not observable to foreigners,matching is effectively random. How-ever, each producer has access only tothe labor in his own country, so domes-tic matches must employ domestic la-bor, while international matches canemploy labor in whichever country it ischeaper. International matches can thusserve to transfer labor demand (pro-ducer services) from the country wherelabor is scarce to where it is abundant,yielding the standard gains from trade.

Rauch and Casella model a transna-tional network as extending to the inter-national market the complete informa-tion that prevails for nonmembers onlyin the domestic market: each networkmember knows the types of all othernetwork members in both countries.

This facilitates the transfer of labor de-mand and increases the extent to whichthe countries’ factor endowment ratioscan differ without ruling out achieve-ment of full efficiency. In this respectthe network acts like an improvementin the matching technology betweencountries. The network differs fromsuch an improvement, however, in thatit only reduces informational barriersfor a subset of producers, so that dis-tributive effects between network mem-ber and nonmember producers canarise. Rauch and Casella find that inequilibria where full efficiency is notachieved the network may cause aggre-gate profits of producers who are notmembers to fall even though worldwages and world profits as a whole in-crease.16 Moreover, although the net-work must (weakly) increase the valueof world output in a two-country model,Rauch and Casella show that this neednot be true if there is more than oneinternational market so that the net-work does not necessarily link the coun-tries with the largest difference in fac-tor endowment ratios: they demonstratethat in a three-country model a networkcan have an effect analogous to harmful“trade diversion.”17

A crucial next step for the literaturesurveyed in this section is better integra-tion of theory and empirical work. Ide-ally, a structural model of international

16 The negative distributional consequences ofthe network for nonmember producers result fromthe decision of less attractive member producersto enter the anonymous market rather than usethe network. It is assumed that nonmember pro-ducers cannot distinguish members from non-members. This assumption may hold even for em-pirically distinguishable groups like the overseasChinese—a Thai national may recognize that an-other Thai businessman is of ethnic Chineseorigin, but an Indonesian national (not of ethnicChinese origin) may not be able to.

17 The same caveats regarding the benefits oftransnational networks might arise if these net-works promote trade by deterring opportunisticbehavior.

Rauch: Business and Social Networks in International Trade 1189

Page 14: JournalofEconomicLiterature,Vol.XXXIX December2001 ...jrauch/pdfs/JEL_Dec_2001.pdfallied families. Typical mechanisms serving to integrate the firms include ... and case study. An

trade that incorporates incomplete in-formation could be estimated that tellsus the overall reduction in trade thatcan be attributed to informational barri-ers, allowing us to ascertain whethertransnational networks only scratch thesurface or make a serious dent in thisproblem.

5. Networks as Market Structure

Whereas transnational networks canhelp to overcome informal barriers tointernational trade, domestic networkscan create informal barriers by facilitat-ing collusion to restrict the marketaccess of foreign firms. The dominantstrand of the literature surveyed in thissection investigates whether keiretsuact this way. The repeated exchangedefinition of networks is always used inthis work. A new line of research cov-ered at the end of this section movesaway from investigation of the relation-ship between networks and informaltrade barriers to examine how the mar-ket structure created by domestic busi-ness groups influences the compositionof trade.

Studies of whether keiretsu act asbarriers to imports (especially fromthe United States) are surveyed byLawrence (1993) and Gary Saxonhouse(1993). Two influential papers are K. C.Fung (1991) and Lawrence (1991). Fungexamines U.S. net exports to Japan for22 industries in 1980. In various specifi-cations he consistently finds a negative,highly statistically significant, but quan-titatively small effect of either the per-centage of industry sales or the percent-age of industry employment accountedfor by keiretsu-affiliated companies. Col-lusion by keiretsu members to restrictimports is certainly not the only possi-ble explanation for this result, and Law-rence (1991) is especially concernedabout the alternative explanation that

efficiency gains realized by keiretsureduce the competitiveness of imports.He reasons that the latter explanationshould imply that keiretsu increaseindustry exports as well as reduce im-ports. He examines the ratio of importsto Japanese domestic demand and theratio of Japanese exports to total worldexports for 37 industries in 1985. Thepercentage of industry sales accountedfor by keiretsu-affiliated firms is nega-tively and significantly related to importpenetration, consistent with the find-ings of Fung, and is insignificantly re-lated to the Japanese world exportshare. The keiretsu impact on importsfound by Lawrence is quantitativelylarge, unlike that found by Fung; set-ting the keiretsu sales percentage tozero would roughly double the averageimport share. When Lawrence splits thekeiretsu variable into the percentage ofindustry sales accounted for by firms af-filiated with “horizontal”18 versus verti-cal keiretsu, both are still negativelyassociated with import share but thevertical keiretsu variable has a positiveand weakly significant effect on exportshare.

If vertical keiretsu do indeed reduceimports and increase exports because ofefficiency gains, it is most likely be-cause the long-term supply relationshipsthat exist within these repeated ex-change networks encourage relationship-specific investments. Barbara Spencerand Larry Qiu (2001) model relationship-specific investments by keiretsu suppli-ers in a paper designed to shed light onthe United States-Japan auto parts dis-pute. In their model keiretsu auto partssuppliers make nonverifiable invest-ments that create rents for the keiretsuassembler, perhaps by reducing assem-bly costs by improving the “fit” with

18 Horizontal keiretsu stretch across many un-related industries; a main bank rather than anassembler forms the “center” of the network.

1190 Journal of Economic Literature, Vol. XXXIX (December 2001)

Page 15: JournalofEconomicLiterature,Vol.XXXIX December2001 ...jrauch/pdfs/JEL_Dec_2001.pdfallied families. Typical mechanisms serving to integrate the firms include ... and case study. An

other keiretsu parts. Payments to thesuppliers cannot be based on their in-vestments and instead the assemblerengages in simultaneous Nash bargain-ing over price with each supplier,where if bargaining breaks down thesupplier does not produce and the as-sembler purchases a “generic” part,which could be produced by a U.S. sup-plier. Note that the equilibrium priceexclusive of the rent to the assemblercould exceed the price of the genericpart.

Spencer and Qiu use their model toshow how a number of equilibriumkeiretsu behaviors could yield a falseimpression of collusion to restrict U.S.market access. First, under the reason-able assumption that the rents to theassembler generated by relationship-specific investments are unobservableoutside the keiretsu, it could appearthat the assembler is sourcing partsfrom within the keiretsu even thoughprices are higher than those chargedby outside (U.S.) suppliers. Second, ifrelationship-specific investments aremore valuable to the assembler forparts with higher cost shares, importscould be restricted to lower value parts.Third, a reduction in trade barriers orprices charged by U.S. auto parts sup-pliers could lead to an expansion ofJapanese auto production without anycorresponding rise in imports by im-proving the bargaining power of the as-sembler with keiretsu suppliers. More-over, as shown by Qiu and Spencer(forthcoming), a policy that directly re-quires an increase in the U.S. contentof Japanese autos could paradoxicallyreduce the value of U.S. auto partsexports to Japan by reducing Japaneseauto output due to lost benefits fromrelationship-specific investment.

Suppose for the sake of argumentthat the model of Spencer and Qiu cap-tures the true state of affairs regard-

ing vertical keiretsu. The question ofwhether vertical keiretsu or similarly or-ganized business groups in other coun-tries act as trade barriers then comesdown to whether more efficient foreignsuppliers can become members of thenetwork.19 This is another version ofthe unanswered empirical question thatarose in connection with evaluation ofGreif’s (1994) claim of dynamic ineffi-ciency in the ability of transnationalnetworks to overcome trade barriers.

The work surveyed so far, in its focuson whether business groups (specifi-cally, keiretsu) are “efficient or exclu-sionist,” has nothing to say about thenumbers or sizes of business groups,which could be important features ofdomestic market structure that influ-ence foreign trade. In a series of papersand a book manuscript, Robert Feen-stra, Gary Hamilton, and coauthorshave explored the implications for in-ternational trade of the differences be-tween the size distributions and average“internalization” of Korean versus Tai-wanese business groups. Figures 1 and2 show the size and internalization ofthe 44 largest business groups (knownas chaebol) in Korea in 1989 and theeighty largest business groups in Tai-wan in 1994 for which data on internaltransactions are available. The verticalaxis measures internalization by the ra-tio of sales to other firms in each groupto total group sales. The points labeled“without retail” exclude the purchasesfrom other firms within the group of

19 While it could be that the behavior of the net-work as a whole somehow produces an exclusion-ary outcome, exclusion could also be a rational de-cision made unilaterally by the assemblers: if theyare uncertain about how the foreign auto parts,say, would “fit in” (literally and figuratively), itmay be optimal for them to “stick with what theyknow” despite the greater foreign efficiency. Inother words, the assembler may also have to makesome (irreversible) relationship-specific invest-ments, which in combination with uncertaintyproduce hysteresis (Avinash Dixit 1992).

Rauch: Business and Social Networks in International Trade 1191

Page 16: JournalofEconomicLiterature,Vol.XXXIX December2001 ...jrauch/pdfs/JEL_Dec_2001.pdfallied families. Typical mechanisms serving to integrate the firms include ... and case study. An

group trading companies and all othergroup firms within the wholesale andretail sector. We see that in Korea thereexist five very large business groups,the smallest of which is larger than thelargest business group in Taiwan. Theweighted average internal sales ratios,with and without retail, respectively,are 22.1 and 12.2 percent for the Ko-rean sample versus 9.5 and 4.5 percentfor the Taiwanese sample. The groupsshown in figures 1 and 2 account forroughly 15 percent of Korean GDP in1989 and 7 percent of Taiwanese GDPin 1994.

With their large size and high inter-nalization the Korean groups are well-suited to realization of economies ofscale through long production runs ofstandardized products, whereas Taiwan-ese groups are better suited to servingniche markets by changing product de-sign to meet the demands of overseascustomers. Feenstra, Tzu-Han Yang, andHamilton (1999) thus predict that U.S.imports from Taiwan will display greaterwithin-industry product variety thanU.S. imports from Korea. This predic-

tion is generally confirmed for the peri-ods 1978–82 and 1983–88 in that theU.S. expenditure share within a five-digit SIC industry on country varietiesthat are not supplied by both countriestends to be greater for Taiwan than Ko-rea, except in a few intermediate goodsindustries where some of the largestand most internalized Taiwanese busi-ness groups are located. Feenstra, Yang,and Hamilton also find for these twoperiods that within intermediate goodsindustries the United States tended toimport higher-priced varieties fromTaiwan than from Korea, whereas theopposite was true within final goodsindustries. Taking price as a proxy forquality, they interpret these results assupporting the hypothesis of DaniRodrik (1994) that, if reputations forquality tend to spill over to all exportersfrom a given country, large businessgroups will be better able to internalizethis externality and hence have a greaterincentive to produce high-quality goods.

Feenstra, Deng-Shing Huang, andHamilton (1997) and Feenstra, Hamilton,and Huang (2001) develop a theoretical

40

35

30

25

20

15

10

5

00 5 10 15

Sales ($ billion)

Inte

rnal

Sal

es (p

erce

nt)

Figure 1. Korean Business Groups, 1989

20 25

with retail

w/out retail

30

Source: Data from Feenstra, Hamilton, and Huang (2001).

1192 Journal of Economic Literature, Vol. XXXIX (December 2001)

Page 17: JournalofEconomicLiterature,Vol.XXXIX December2001 ...jrauch/pdfs/JEL_Dec_2001.pdfallied families. Typical mechanisms serving to integrate the firms include ... and case study. An

model of multiple business groups.They offset the complexity of dealingwith many rather than one businessgroup by dispensing with the view ofbusiness groups as networks and insteadmodeling them as hierarchies: eachbusiness group is a centrally organizedset of firms that maximizes joint profits.The two papers begin with a model inwhich there is monopolistic competitionin supply of both intermediate and finalgoods. This generates an incentive forbusiness groups (vertically integratedfirms) to form and eliminate the mark-ups on intermediate goods in internalgroup transactions, each group tradingoff increased collective profits against afixed “governance cost” for the group asa whole. For most parameter values thepapers find two stable equilibria: a low-concentration equilibrium with a largenumber of low-internalization groupsthat sell to outsiders at low markups,and a high-concentration equilibrium witha small number of high-internalizationgroups that sell to outsiders at highmarkups. The intuition for the existence

of multiple stable equilibria is as fol-lows. Start in the low-concentrationequilibrium and artificially raise mark-ups charged to outsiders for intermedi-ate goods. This creates an incentive forexisting groups to expand and incorpo-rate more product varieties, but themarket can only support a smaller num-ber of these larger groups, the resultbeing greater concentration that vali-dates the higher markups. Feenstra andHamilton (forthcoming) argue thatcultural and historical factors pushedTaiwan towards the low-concentration andKorea towards the high-concentrationequilibrium. Due to the existence of lowmarkups and greater product variety,consumer welfare is greater in thelow-concentration (Taiwan) equilibrium.

Whereas the literature surveyed inthis section has a lot to say about theefficiency consequences of businessgroups, it is virtually silent on the ques-tion of whether business groups willbecome more or less important for in-ternational trade. Many writers on busi-ness groups in a context other than

Source: Data from Feenstra, Hamilton, and Huang (2001).

40

35

30

25

20

15

10

5

00 5 10 15

Sales ($ billion)

Inte

rnal

Sal

es (p

erce

nt)

Figure 2. Taiwanese Business Groups, 1994

20 25

with retail

w/out retail

30

Rauch: Business and Social Networks in International Trade 1193

Page 18: JournalofEconomicLiterature,Vol.XXXIX December2001 ...jrauch/pdfs/JEL_Dec_2001.pdfallied families. Typical mechanisms serving to integrate the firms include ... and case study. An

international trade take the functional-ist position that they are responses tovarious “market imperfections,” andthat these market imperfections will ul-timately vanish with economic develop-ment, taking business groups along withthem. As we noted in section 2, thereis no evidence to support this predic-tion. The fate of business groups is, ofcourse, bound up with the fate of thenetwork form of organization in general,which at one time was widely believedto be a transitional stage on the wayto the centrally administered multidivi-sional firm (Alfred Chandler 1977) butnow seems to be gaining ground againstthat hierarchical form (Michael Pioreand Charles Sabel 1984). Entering thisdebate would take us well beyond thescope of this survey.

A next step for the literature on net-works as market structure could be toconnect with the literatures surveyed insections 3 and 4. For example, past in-vestments by keiretsu suppliers to adapttheir products to the specifications ofassemblers in the home country couldgive them an advantage even overequally knowledgeable competing sup-pliers when the assembler establishesfacilities abroad, providing an alterna-tive explanation for the trade-creatingeffect of such foreign direct investment.Similarly, the East Asian businessgroups might be solving problems ofweak contract enforcement or lack ofinformation regarding business oppor-tunities in a purely domestic context.Taken as a whole, the internationaltrade literature on business groups alsoneeds to outgrow its exclusive focus onEast Asia. Not only does this convey themisleading impression that businessgroups are not important for interna-tional trade elsewhere, but it may leadto biased inferences about the typicalimpact of business groups on trade.

Another important step for this litera-

ture is micro-level studies of the open-ness of business groups to membershipby foreign firms or of how businessgroups engage in importing and export-ing. With firm-level empirical studiesproliferating in the international tradefield (e.g., Mark Roberts and James Ty-bout 1997; Andrew Bernard and Brad-ford Jensen 1999), there is less and lessexcuse for treating business groups asempirical (though not theoretical) blackboxes. In particular, a great deal of in-sight into the questions of whetherbusiness groups exclude imports or howthey provide variety and quality in ex-ports could be had by studying the trad-ing companies that are at the center ofmany groups.20 The literature surveyedin the next section makes a start in thisdirection.

6. Network Intermediaries

Suppose that a firm looking for for-eign buyers or sellers does not belongto an appropriate network. Conven-tional search is an option, but often apoor one if what the firm is searchingfor is not sufficiently homogeneous tohave an informative price. An alterna-tive is to engage a network intermedi-ary, i.e., an agent who sells access toand use of his network, typically for acommission on the value of the transac-tions realized. These intermediaries goby the names agent, broker, trader, etc.,but not all actors so labeled should beconsidered network intermediaries. Net-work intermediaries are distinguished

20 Feenstra, Hamilton, and Huang (2001) findthat roughly two-thirds of Korean and half of Tai-wanese business groups in their samples havetrading companies. Trading companies appear tobe more central to the functioning of Korean thanTaiwanese business groups in that excluding theirpurchases from other group firms from computa-tion of the internal sales ratio reduces theweighted average internalization from 22.1 to 13.8percent for the Korean sample compared to 9.5 to8.5 percent for the Taiwanese sample.

1194 Journal of Economic Literature, Vol. XXXIX (December 2001)

Page 19: JournalofEconomicLiterature,Vol.XXXIX December2001 ...jrauch/pdfs/JEL_Dec_2001.pdfallied families. Typical mechanisms serving to integrate the firms include ... and case study. An

by what Yung Rhee and ChristineSoulier (1989, p. 25) call their “deepknowledge” of the capabilities and pref-erences of the sellers or buyers in theirnetworks:

As highlighted in our Hong Kong survey, themost important resource that ETCs (exporttrading companies) have is their deep knowl-edge about external markets/buyers and localproduction capabilities/producers. Without suchinformation, ETCs can hardly be effective inmatching potential overseas buyers to localproducers . . . the effectiveness of Japanese,Korean, and Hong Kong GTCs (general trad-ing companies) has been based on the depthof their product-market knowledge and of thesupplier-buyer network.

The networks to which these inter-mediaries sell access may consist ofdensely interconnected firms, such asbusiness groups, or of firms that do vir-tually no business with each other, sothat if one removed the intermediarythey would not constitute a network atall. Descriptions of the activities of theformer type of intermediaries suggestthat they may be crucial to both theability of small, low-internalizationbusiness groups to provide productvariety and the ability of large, high-internalization business groups to real-ize economies of scale in the produc-tion of standardized goods. Of tradingcompanies for groups of fashion shoeproducers in Taiwan, You-tien Hsing(1999, p. 106) writes, “A typical shoetrading company usually had 12–15partner manufacturers. To offset lim-ited demand for each order, tradingcompanies coordinated and allocatedorders in accordance with the specialtyof individual factories.” Describing thelarge, general trading companies at theheart of many keiretsu, the sogo shoshaof Japan,21 M. Y. Yoshino and ThomasLifson (1986, p. 38) state, “the sogo

shosha’s uniqueness lies in its capacityto provide essential links betweenstages in a product system for a clientfirm.”

Studies of network intermediariesthat sell access to unconnected firmshave mainly concerned buyers who pro-vide less developed country manufac-turers with access to sets of developedcountry retailers. The principal interestof these studies (e.g., Schmitz 1995;Khalid Nadvi 1999; and papers sur-veyed by Schmitz and Nadvi 1999) is inhow such network intermediaries trans-late the preferences of retailers intoproduct designs and models and pro-vide other technical assistance to the(low-tech) manufacturers. This subjectwill be taken up in more detail in thenext section of this survey.

Whatever their virtues in supportingprovision of product variety, realizationof economies of scale, and transfer oftechnical know-how, the possibility thatnetwork intermediaries are not gettingthe basic job of connecting buyers orsellers to foreign opportunities done issuggested by the trade-creating impactsof immigrants, foreign direct invest-ment by business groups, and the over-seas Chinese reviewed in section 4. Theapparent importance of belonging tothese transnational business and socialnetworks does not prove that networkintermediation is undersupplied, how-ever, because the costs of establishingeach of these transnational networkswere sunk for purposes other than cre-ating trade: it could be that the costs ofsetting up new network intermediarieswould outweigh the benefits of theadditional trade they would create.

A crucial input to any evaluation ofthe case for market failure in provisionof network intermediation is knowledgeof the determinants of supply: how doactors become network intermediaries?The literature is virtually silent on this

21 The general trading companies of the largestKorean chaebol were explicitly modeled after thesogo shosha.

Rauch: Business and Social Networks in International Trade 1195

Page 20: JournalofEconomicLiterature,Vol.XXXIX December2001 ...jrauch/pdfs/JEL_Dec_2001.pdfallied families. Typical mechanisms serving to integrate the firms include ... and case study. An

question, but there exist many sugges-tive anecdotes: a former mergers andacquisitions officer for Chase Manhat-tan in Hong Kong who now matchesleisure-related, California-based busi-nesses with Asian partners (MatthewMiller 1997); an industrial engineeringconsultant who had designed factorylayouts throughout Asia who nowmatches U.S. toy designers with Chi-nese toy manufacturers (Bruce Bigelow1997); a leader of Daewoo’s team ofmachine installation specialists and pro-duction line experts assigned to Bangla-desh who subsequently opened a firm“engaged exclusively with the importand export trade of Bangladesh’s newgarment factories” (Rhee 1990, p. 342).Such anecdotes, and intuition, suggestthat actors become network intermedi-aries by accumulating “deep knowledge”of buyers and sellers through workingwith them in a non-intermediary capac-ity, then selling access to those withwhom they formerly worked when thisbecomes more profitable.

At the very least, it seems unlikelythat this process will supply intermedi-aries whose networks span many indus-tries. Realization of business opportuni-ties often requires coordination acrossseveral industries, however. An entre-preneur seeking many components toassemble into a final product for whichhe has a buyer, for example, could em-ploy many specialized intermediaries,but this is likely to be cumbersome.More importantly, the components needto be matched not only to the clientbut to each other, so only a diversifiedintermediary could put together anattractive package.22

If diversified network intermediariesare not generated as a by-product of

non-intermediary activity, could an ade-quate supply result from more plannedaccumulation of contacts by actors whohave already dedicated themselves tointermediation? There is good reason tobelieve that this kind of investment willfall short of the socially optimal level,because the intermediary may not beable to garner a large enough share ofthe surplus he creates for his clients torecover his investment. The fact thatthe intermediary needs deep knowledgeof the members of his network in orderto know which is the best match for hisclient means that the quality of servicehe provides his client is inherently non-contractible. In the absence of an en-forceable contract based on paymentfor surplus created the intermediarymust rely on his bargaining power, butthis is limited because the specificity ofeach match leaves the intermediarywith poor alternative transactions ifbargaining breaks down.23 In additionto this contracting problem, there is adifficulty if one needs to be actually in-volved in the production process in or-der to acquire deep knowledge. It isone thing to leave a career in produc-tion to become an intermediary whenthe opportunity to earn a better livingpresents itself, and quite another toparticipate in production with the planof augmenting the network to whichone will sell access as an intermediary,given the risk that by the time one haslearned enough, markets or technologywill have changed so as to destroy thesynergy with one’s existing network.

22 In his discussion of the sogo shosha,Terutomo Ozawa (1987, p. 4) affirms that “tradingin multiple products creates synergistic tradingopportunities.”

23 In practice many intermediaries charge acommission on sales rather than taking title to thegoods transacted and risking being stuck withthem if a deal breaks down. The commission isoften fixed at a customary rate such as five or tenpercent. The point here is that, if the expectationof these commissions does not provide adequateincentive to make socially profitable investmentsin contacts, the intermediary may have little lever-age to raise his return through side payments orother means.

1196 Journal of Economic Literature, Vol. XXXIX (December 2001)

Page 21: JournalofEconomicLiterature,Vol.XXXIX December2001 ...jrauch/pdfs/JEL_Dec_2001.pdfallied families. Typical mechanisms serving to integrate the firms include ... and case study. An

The case for undersupply of networkintermediation thus appears to bestrongest for large diversified interme-diaries. Some governments have evi-dently come to the same conclusion:Rauch (1996) describes how the afore-mentioned sogo shosha and their imita-tors, the general trading companies(GTCs) of Korea and the foreign tradecompanies (FTCs) of Turkey, all bene-fitted from implicit or explicit govern-ment subsidies during their start-upyears, after which they became viablewithout subsidies. This evidence thatgovernment policy can be used, ineffect, to promote the formation oftransnational networks is intriguing, butfailed attempts to establish generaltrading companies in Taiwan and theUnited States make it clear that muchfurther theoretical and empirical re-search is required before reliable policyrecommendations can be formulated.24

7. Learning in Networks

Economists have often modeled in-ternational technology transfer as anarm’s-length phenomenon. Firms are nottaught the new technology. Rather theyengage in purposive imitative activity

on their own (e.g., Gene Grossman andElhanan Helpman 1991), employ ma-chinery and equipment that embodiesforeign knowledge (e.g., David Coe,Helpman, and Alexander Hoffmaister1997), license the new technology, andso on. However, it is difficult to learnnew technology through these mecha-nisms (Howard Pack and Larry West-phal 1986). There is a growing body ofevidence that for less developed coun-try (LDC) firms in particular a majorand perhaps predominant source oftechnology transfer (and transfer ofmanagerial know-how) is instruction bydeveloped country buyers: producersseeking cheaper suppliers of inputs anddistributors seeking cheaper suppliersof final goods. Pack and John Page(1994, pp. 220–21) state:

The motivation of the purchasers is to obtainstill lower-cost, better quality products frommajor suppliers whose products account for asignificant percentage of profits. To achievethis they are willing to transmit tacit and oc-casionally proprietary knowledge from theirother OECD suppliers. Such transfers ofknowledge are more likely to characterizesimpler production sectors such as clothingand footwear or more generally those oldertechnologies that are not hedged by restric-tions adopted to increase appropriability,such as patents and trade secrets.

As mentioned in section 2, these buyer–seller relationships are long-term andthus fit our repeated exchange definitionof networks.

One example of such evidence is astudy by Egan and Mody (1992), who sur-veyed U.S. buyers operating in LDCs,including “manufacturers, retailers, im-porters, buyers’ agents, and joint venturepartners” (p. 322). They found (p. 328):

Buyers also render long-term benefits to sup-pliers in the form of information on pro-duction technology. This occurs principallythrough various forms of in-plant training.The buyer may send international expertsto train local workers and supervisors . . .

24 In 1982 the United States passed and signedinto law the Export Trading Company Act with theclear intent of generating sogo shosha imitators.This Act eased antitrust constraints for registeredexport trading companies and allowed banks toparticipate indirectly in exporting, but no subsi-dies accompanied these regulatory changes. Thefew subsequent attempts to establish large-scale,diversified U.S. trading companies all failed (MikePeng 1998, pp. 37–41). Karl Fields (1995, p. 214)attributes the failure of Taiwan’s Large TradingCompany program to “the feeble nature of incen-tives,” though it may also have been the case thatthe overseas Chinese network made general trad-ing companies redundant for Taiwan. In contrast,Feenstra, Hamilton, and Huang (2001) argue thatthe difference between Korea and Taiwan showsthat GTCs must build upon existing large do-mestic business groups. Both views could be rightif existing large domestic business groups areneeded to push through subsidy policies for GTCs.

Rauch: Business and Social Networks in International Trade 1197

Page 22: JournalofEconomicLiterature,Vol.XXXIX December2001 ...jrauch/pdfs/JEL_Dec_2001.pdfallied families. Typical mechanisms serving to integrate the firms include ... and case study. An

Buyers may also arrange short-term workertraining in a developed country plant.

Rhee, Bruce Ross-Larson, and GarryPursell (1984) surveyed Korean exportersof manufactures. Their findings (p. 61)were similar to those of Egan and Mody:

The relations between Korean firms and theforeign buyers went far beyond the negotia-tion and fulfillment of contracts. Almost halfthe firms said they had directly benefitedfrom the technical information foreign buyersprovided: through visits to their plants by en-gineers or other technical staff of the foreignbuyers, through visits by their engineeringstaff to the foreign buyers . . .

The Rhee, Ross-Larson, and Pursellsurvey was conducted in 1975. More re-cently Korea and the other advancedEast Asian countries have played therole for LDCs that foreign buyers usedto play for them. The role of Korea indeveloping garment exports from Ban-gladesh is an especially interesting casethat is studied in Rhee (1990). This caseis part of the broader phenomenon of“triangle manufacturing” (Gary Gereffi1999) in East Asia: countries such asKorea and Taiwan continue to accept andfulfill the orders of developed countrybuyers for labor-intensive goods, buthave “outsourced” the actual produc-tion to countries with lower wages.

This process of learning foreign tech-nology can be thought of as takingplace within international productionnetworks or “global commodity chains”(Gereffi 1994, 1999). This theoreticalframework predicts that once LDC firmsare incorporated into the “bottoms” ofthe chains, their learning will continueby movement up the chains.25 There

are two types of chains: “producer-driven” and “buyer-driven,” which areillustrated in figure 3. In the former,large manufacturers play the central rolesin coordinating the production net-works. Producer-driven chains are typi-cal in capital- and technology-intensiveindustries such as automobiles, aircraft,semiconductors, and heavy machinery.In the latter, large retailers, brandedmarketers,26 and branded manufactur-ers play the coordinating roles. Buyer-driven commodity chains are typical inlabor-intensive, consumer goods indus-tries such as garments, footwear, toys,housewares, and consumer electronics.Profitability is highest at the tops of thechains where barriers to entry are great-est: scale and technology in producer-driven chains, design and marketingexpertise in buyer-driven chains.

In buyer-driven commodity chains,one mode through which learning ispredicted to continue is organizationalsuccession: from assembler to originalequipment manufacturer (OEM) to origi-nal brand-name manufacturer (OBM),which is from more subordinate, com-petitive, and low-profit positions to morecontrolling, oligopolistic, high-profit posi-tions. In the apparel industry, Gereffi(1999) finds that LDC firms that haveparts provided to them for assemblylearn how to find on their own the partsneeded to make the product accordingto the design specified by the buyer(and may then subcontract the assem-bly); firms that have reached this levellearn how to design and sell their ownmerchandise, becoming branded manu-facturers (and may then subcontract theproduction, becoming branded market-ers). Additional study is needed todetermine whether this pattern of

25 The global commodity chain framework alsohas implications for changes in the pattern ofinternational trade. It predicts that as a countryaccumulates physical and human capital it willnot upgrade to a random set of capital- or skill-intensive industries but rather to products thatare organizationally related through the globalcommodity chains in which it is participating.

26 The growth of this type of firm led to the intro-duction of a new category, “Own brand importer-marketers,” in the 1997 Census of WholesaleTrade (Robert Steiner 1997).

1198 Journal of Economic Literature, Vol. XXXIX (December 2001)

Page 23: JournalofEconomicLiterature,Vol.XXXIX December2001 ...jrauch/pdfs/JEL_Dec_2001.pdfallied families. Typical mechanisms serving to integrate the firms include ... and case study. An

learning is common in other consumergoods industries,27 and what kind oflearning modes might be present in

producer-driven commodity chains. Atthe same time, work is needed to recon-cile the kind of findings discussed inthis section with econometric analysessuch as Sofronis Clerides, Saul Lach,and James R. Tybout (1998) that

Figure 3. The Organization of Producer-driven and Buyer-driven Global Commodity Chains

Retailersand

DealersManufacturers

Domestic and Foreign Subsidiariesand Subcontractors

Producer-driven Commodity Chains

Buyer-driven Commodity Chains

Distributors

Factories Retailers

BrandedMarketers

U.S. MARKET

BrandedManufacturers

Traders

OverseasBuyers

OVERSEAS

Source: Gary Gereffi, based on Gereffi (1999).Notes: Solid arrows are primary relationships; dashed arrows are secondary relationships. Retailers, branded marketers, and traders require full-package supply from overseas factories. Branded manufacturers ship parts for overseas assembly and then export to the manufacturer’s home market.

27 Schmitz and Peter Knorringa (1999) apply thebuyer-driven commodity chain framework to theshoe industry.

Rauch: Business and Social Networks in International Trade 1199

Page 24: JournalofEconomicLiterature,Vol.XXXIX December2001 ...jrauch/pdfs/JEL_Dec_2001.pdfallied families. Typical mechanisms serving to integrate the firms include ... and case study. An

conclude that more productive LDCfirms export, but exporting does notmake LDC firms more productive.28

8. Review of Main Themes andIssues for Future Research

Numerous statistical and case studiesprovide evidence that transnationalbusiness and social networks promoteinternational trade by alleviating prob-lems of contract enforcement and pro-viding information about trading oppor-tunities. Some studies established alarge quantitative importance for thesenetwork effects. Developments in thedemand for networks will probably tendto reduce this quantitative importancein the future, while trends in the supplyof networks will probably tend to in-crease it. On the demand side, institu-tional developments such as internationalcommercial arbitration and letters ofcredit and technological improvementsin communication reduce the need fornetworks, although this effect is offsetby the increased share of differentiatedproducts in international trade. On thesupply side, increased migration and in-creased foreign direct investment areextending ever more domestic networksacross international borders. Perhapsthe biggest unknown is the impact ofe-commerce on the importance of net-works for international trade. Certainlythe use of search engines is likely to re-duce the demand for networks, but thiseffect could be limited if, as it is said,the internet disseminates informationbut not understanding. New technology

could interact with networks in sur-prising ways:29 for example, e-mail prob-ably retards the decay of establishednetworks.

The efficiency implications of trans-national networks would appear to bestraightforward: they improve the allo-cation of resources by creating trade,and they generate surplus from coop-eration for their members. Closer theo-retical examination, however, raises threecaveats. First, less desirable networkmembers may choose to enter theanonymous international market wheretheir characteristics are not known,harming nonmembers even though theexistence of a transnational network stillincreases world output in the aggregate.Second, a transnational network canhave an effect analogous to harmfultrade diversion if it links the “wrong”countries. Third, organization of inter-national trade through networks mayhinder its growth if transnational net-works tend to be closed to new mem-bers. This last caveat is an especiallyimportant topic for empirical researchbecause the openness of networks isalso the key to whether otherwiseefficiency-enhancing domestic networksimpede international trade.

As little as we know about the abilityof existing transnational (or domestic)networks to incorporate new members,we know even less about how entirely newtransnational networks are formed. Thissurvey has discussed how domestic net-works, when spread across international

28 One possibility is that organizational succes-sion is too lengthy a process to be captured by thedata sets used in the econometric analyses per-formed to date, and that the productivity benefitsof the other learning that takes place accrue to thebuyers because of the superior bargaining powerthey possess by virtue of being higher up thecommodity chain.

29 Though not directly relevant, I find the re-sults of Jess Gaspar and Edward Glaeser (1998)suggestive in this regard. They show that the ratioof business trips in the United States to real U.S.GDP rose sharply in the early 1990s even control-ling for the fall in airline costs, which they inter-pret as evidence that new telecommunicationstechnology is a complement to rather than a sub-stitute for face-to-face contact. It appears that in-formation is still an “experience good” and thatface-to-face contact still helps to build the trustneeded to close deals.

1200 Journal of Economic Literature, Vol. XXXIX (December 2001)

Page 25: JournalofEconomicLiterature,Vol.XXXIX December2001 ...jrauch/pdfs/JEL_Dec_2001.pdfallied families. Typical mechanisms serving to integrate the firms include ... and case study. An

boundaries by migration or foreigndirect investment, promote trade, andalso how trade can help to maintaintransnational networks once they areestablished, but nothing has been saidabout if and how trade creates transna-tional networks. An important questionfor future research is whether forma-tion of new business and social net-works can help us understand apparenthysteresis effects in bilateral trade, suchas those found by Barry Eichengreenand Douglas Irwin (1998).

REFERENCES

Anderson, James E. and Douglas Marcouiller.Forthcoming. “Insecurity, and the Pattern ofTrade: An Empirical Investigation,” Rev. Econ.Statist.

Belderbos, René and Leo Sleuwaegen. 1998. “Tar-iff Jumping DFI and Export Substitution: Japa-nese Electronics Firms in Europe,” Int. J. Ind.Org. 16, pp. 601–38.

Bergstrand, Jeffrey H. 1985. “The Gravity Equa-tion in International Trade: Some Micro-economic Foundations and Empirical Evi-dence,” Rev. Econ. Stat. 67, pp. 474–81.

Bernard, Andrew B. and Bradford J. Jensen. 1999.“Exceptional Exporter Performance: Cause, Ef-fect, or Both?” J. Int. Econ. 47:1, pp. 1–25.

Bigelow, Bruce V. 1997. “The Toy Maker: Chapter8,” San Diego Union-Tribune, Oct. 26, p. I–1.

Bun, Chan Kwok and Chee Kiong Tong. 1993.“Rethinking Assimilation and Ethnicity: TheChinese in Thailand,” Int. Mig. Rev. 27:1, pp.140–68.

Chandler, Alfred D. Jr. 1977. The Visible Hand:The Managerial Revolution in American Busi-ness. Cambridge: Harvard U. Press.

Chin, Ku-Sup; In-Jin Yoon and David A Smith.1996. “Immigrant Small Business and Interna-tional Economic Linkage: A Case of the KoreanWig Business in Los Angeles, 1968–77,” Int.Mig. Rev. 30, pp. 485–510.

Clerides, Sofronis K.; Saul Lach and James R. Ty-bout. 1998. “Is Learning by Exporting Impor-tant? Micro-Dynamic Evidence from Colombia,Mexico, and Morocco,” Quart. J. Econ. 113:3,pp. 903–47.

Coe, David T.; Elhanan Helpman and AlexanderW. Hoffmaister. 1997. “North-South R&DSpillovers,” Econ. J. 107:440, pp. 134–49.

Cohen, Abner. 1969. Custom and Politics in Ur-ban Africa: A Study of Hausa Migrants in Yo-ruba Towns. Berkeley: U. California Press.

———. 1971. “Cultural Strategies in the Organi-zation of Trading Diasporas,” in The Develop-ment of Indigenous Trade and Markets in West

Africa. Claude Meillassoux, ed. London: OxfordU. Press, pp. 266–78.

Craig, Laurence W.; William W. Park and JanPaulsson. 1985. International Chamber of Com-merce Arbitration. NY: Oceana Publications.

Curtin, Philip D. 1984. Cross-Cultural Trade inWorld History. Cambridge: Cambridge U.Press.

del Busto, Charles. 1994. ICC Guide to Documen-tary Credit Operations for the UCP 500. Paris:ICC.

Dixit, Avinash. 1992. “Investment and Hysteresis.”J. Econ. Perspect. 6, pp. 107–32.

Eaton, Jonathan and Samuel Kortum. 2000.“Technology, Geography, and Trade,” workingpaper, econ. dept. Boston U.

Egan, Mary L. and Ashoka Mody. 1992. “Buyer-Seller Links in Export Development,” WorldDevel. 20:3, pp. 321–34.

Eichengreen, Barry and Douglas A. Irwin. 1998.“The Role of History in Bilateral Trade Flows,”in The Regionalization of the World Economy.Jeffrey A. Frankel, ed. Chicago: U. ChicagoPress, pp. 33–62.

Encaoua, David and Alexis Jacquemin. 1982. “Or-ganizational Efficiency and Monopoly Power:The Case of French Industrial Groups,” Eur.Econ. Rev. 19, pp. 25–51.

Feenstra, Robert C. and Gary G. Hamilton. Forth-coming. Emergent Economies, Divergent Paths:Business Groups and Economic Organization inSouth Korea and Taiwan. Cambridge: Cam-bridge U. Press.

Feenstra, Robert C.; Gary G. Hamilton and Deng-Shing Huang. 2001. “The Organization of theTaiwanese and South Korean Economies: AComparative Equilibrium Analysis,” in Net-works and Markets. James E. Rauch and Ales-sandra Casella, eds. NY: Russell Sage Founda-tion, pp. 86–142.

Feenstra, Robert C.; Deng-Shing Huang and GaryG. Hamilton. 1997. “Business Groups andTrade in East Asia, Part I: Networked Equilib-ria,” working paper 5886, NBER, Cambridge,MA.

Feenstra, Robert C.; Tzu-Han Yang and Gary G.Hamilton. 1999. “Business Groups and ProductVariety in Trade: Evidence from South Korea,Taiwan and Japan,” J. Int. Econ. 48:1, pp. 71–100.

Fields, Karl J. 1995. Enterprise and the State inKorea and Taiwan. Ithaca: Cornell U.

Forsgren, Mats and Jan Johanson. 1992. “Man-aging Internationalization in Business Net-works,” in Managing Networks in InternationalBusiness. Mats Forsgren and Jan Johanson,eds. Philadephia: Gordon and Breach, pp. 1–16.

Frankel, Jeffrey A. 1997. Regional Trading Blocsin the World Economic System. Washington,DC: Inst. Int. Econ.

Fung, K. C. 1991. “Characteristics of Japanese In-dustrial Groups and Their Potential Impact onU.S.–Japanese Trade,” in Empirical Studies of

Rauch: Business and Social Networks in International Trade 1201

Page 26: JournalofEconomicLiterature,Vol.XXXIX December2001 ...jrauch/pdfs/JEL_Dec_2001.pdfallied families. Typical mechanisms serving to integrate the firms include ... and case study. An

Commercial Policy. Robert Baldwin, ed. Chi-cago: U. Chicago Press, pp. 137–64.

Gaspar, Jess and Edward L. Glaeser. 1998. “Infor-mation Technology and the Future of Cities,” J.Urban Econ. 43, pp. 136–56.

Gereffi, Gary. 1994. “The Organization of Buyer-Driven Global Commodity Chains: How U.S.Retailers Shape Overseas Production Net-works,” in Commodity Chains and Global Capi-talism. Gary Gereffi and Miguel Korzeniewicz,eds. Westport, CT: Praeger, pp. 95–122.

Gereffi, Gary. 1999. “International Trade and In-dustrial Upgrading in the Apparel CommodityChain,” J. Int. Econ. 48:1, pp. 37–70.

Gould, David M. 1994. “Immigrant Links to theHome Country: Empirical Implications for U.S.Bilateral Trade Flows,” Rev. Econ. Stat. 76, pp.302–16.

Granovetter, Mark. 1973. “The Strength of WeakTies,” Amer. J. Soc. 78, pp. 1360–80.

———. 1995. “Coase Revisited: Business Groupsin the Modern Economy,” Ind. Corp. Change4:1, pp. 93–129.

———. 1995. Getting a Job: A Study of Contactsand Careers. 2nd ed. Chicago: U. ChicagoPress.

Greif, Avner. 1989. “Reputation and Coalitions inMedieval Trade: Evidence on the MaghribiTraders,” J. Econ. Hist. 49:4, pp. 857–82.

———. 1993. “Contract Enforceability and Eco-nomic Institutions in Early Trade: The Magh-ribi Traders’ Coalition,” Amer. Econ. Rev. 83:3,pp. 525–48.

———. 1994. “Cultural Beliefs and the Organi-zation of Society: A Historical and TheoreticalReflection on Collectivist and Individualist So-cieties,” J. Polit. Econ. 102:5, pp. 912–50.

Grossman, Gene and Elhanan Helpman. 1991. “En-dogenous Product Cycles,” Econ. J. 101:408,pp. 1214–29.

Head, Keith and John Ries. 1998. “Immigrationand Trade Creation: Econometric Evidencefrom Canada,” Can. J. Econ. 31:1, pp. 47–62.

———. 2001. “Overseas Investment and FirmExports,” Rev. Int. Econ. 9:1, pp. 108–22.

Helliwell, John F. 1998. How Much Do NationalBorders Matter? Washington, DC: BrookingsInstitution.

Hsing, You-tien. 1999. “Trading Companies inTaiwan’s Fashion Shoe Networks,” J. Int. Econ.48:1: 101–20.

Kotkin, Joel. 1992. Tribes: How Race, Religion,and Identity Determine Success in the NewGlobal Economy. NY: Random House.

La Porta, Rafael; Florencio Lopez-de-Silanes andAndrei Shleifer. 1999. “Corporate OwnershipAround the World,” J. Finance 54:2, pp. 471–517.

Lawrence, Robert Z. 1991. “Efficient or Exclu-sionist? The Import Behavior of Japanese Cor-porate Groups,” Brookings Pap. Econ. Act. 1.

———. 1993. “Japan’s Different Trade Regime:An Analysis with Particular Reference toKeiretsu,” J. Econ. Perspect. 7:3, pp. 3–19.

Lever-Tracy, Constance; David Ip, Jim Kitay,Irene Phillips and Noel Tracy. 1991. Asian En-trepreneurs in Australia: Ethnic Small Businessin the Chinese and Indian Communities of Bris-bane and Sydney. Canberra: OMA/AGPS.

McCallum, John. 1995. “National Borders Matter:Canada-U.S. Regional Trade Patterns,” Amer.Econ. Rev. 85, pp. 615–23.

McKendrick, David. 1998. Personal communica-tion. UCSD Data Storage Project.

Miller, Matthew. 1997. “The Asian Connection,”San Diego Union-Tribune, June 17, pp. C1, C4.

Mitchell, Katharyne and Brian Hammer. 1997.“Ethnic Chinese Networks: A New Model?” inThe People Link: Human Resource Linkagesacross the Pacific. A. E. Safarian and WendyDobson, eds. Toronto: U. Toronto Press, pp.73–97.

Nadvi, Khalid. 1999. “The Cutting Edge: Collec-tive Efficiency and International Competitive-ness in Pakistan.” Oxford Devel. Stud. 27:1, pp.143–77.

Ozawa, Terutomo. 1987. “The Nature and Roleof General Trading Firms as a Foreign-TradeInfrastructure: Conceptualization,” in Role ofGeneral Trading Firms in Trade and Develop-ment: Some Experiences. Terutomo Ozawa, ed.Tokyo: Asian Productivity Org.

Pack, Howard and John M. Page Jr. 1994. “Accu-mulation, Exports, and Growth in the High-performing Asian Economies,” Carnegie Roch-ester Conf. Ser. Public Pol. 40, pp. 199–257.

Pack, Howard and Larry E. Westphal. 1986. “In-dustrial Strategy and Technological Change:Theory versus Reality,” J. Devel. Econ. 22:1, pp.87–128.

Peng, Mike W. 1998. Behind the Success and Fail-ure of U.S. Export Intermediaries. Westport,CT: Quorum Books.

Piore, Michael J. and Charles F. Sabel. 1984. TheSecond Industrial Divide: Possibilities for Pros-perity. NY: Basic Books.

Podolny, Joel M. and Karen L. Page. 1998. “Net-work Forms of Organization,” Ann. Rev. Sociol.24, pp. 57–76.

Portes, Richard and Hélène Rey. 1999. “The De-terminants of Cross-Border Equity Flows,”working paper 7336, NBER, Cambridge,MA.

Qiu, Larry D. and Barbara J. Spencer. Forthcom-ing. “Keiretsu and Relationship-Specific Invest-ment: Implications for Market-Opening TradePolicy,” J. Int. Econ.

Rangan, Subramanian and Robert Z. Lawrence.1999. A Prism on Globalization: Corporate Re-sponses to the Dollar. Washington, DC: Brook-ings Institution.

Rauch, James E. 1996. “Trade and Search: SocialCapital, Sogo Shosha, and Spillovers,” workingpaper 5618, Cambridge, MA: NBER.

———. 1999. “Networks Versus Markets in Inter-national Trade,” J. Int. Econ. 48:1, pp. 7–35.

Rauch, James E. and Alessandra Casella. 1998.“Overcoming Informational Barriers to Interna-

1202 Journal of Economic Literature, Vol. XXXIX (December 2001)

Page 27: JournalofEconomicLiterature,Vol.XXXIX December2001 ...jrauch/pdfs/JEL_Dec_2001.pdfallied families. Typical mechanisms serving to integrate the firms include ... and case study. An

tional Resource Allocation: Prices and GroupTies,” working paper 6628, Cambridge, MA:NBER.

Rauch, James E. and Vitor Trindade. “Ethnic Chi-nese Networks in International Trade,” Rev.Econ. Stat. forthcoming.

Rhee, Yung-Whee. 1990. “The Catalyst Model ofDevelopment: Lessons from Bangladesh’s Suc-cess with Garment Exports.” World Devel. 18:2,pp. 333–46.

Rhee, Yung Whee; Bruce Ross-Larson and GarryPursell. 1984. Korea’s Competitive Edge: Man-aging the Entry into World Markets. Baltimore:Johns Hopkins.

Rhee, Yung Whee and Christine Soulier. 1989.“Small Trading Companies and a Successful Ex-port Response: Lessons from Hong Kong.” In-dustry Series paper 16, World Bank Industryand Energy Dept.

Roberts, Mark J. and James R. Tybout. 1997. “TheDecision to Export in Colombia: An EmpiricalModel of Entry with Sunk Costs.” Amer. Econ.Rev. 87:4, pp. 545–64.

Rodrik, Dani. 1994. “Industrial Organization andProduct Quality: Evidence from South Koreanand Taiwanese Exports,” in Empirical Studiesof Strategic Trade Policy. Paul Krugman andAlisdair Smith, eds. Chicago: U. Chicago Press,pp. 195–210.

Saxenian, AnnaLee. 1999. Silicon Valley’s NewImmigrant Entrepreneurs. San Francisco: Pub-lic Policy Institute of California.

Saxonhouse, Gary R. 1993. “What Does Japanese

Trade Structure Tell Us about Japanese TradePolicy?” J. Econ. Perspect. 7:3, pp. 21–43.

Schmitz, Hubert. 1995. “Small Shoemakers andFordist Giants: Tale of a Supercluster,” WorldDevel. 23:1, pp. 9–28.

———. 1999. “From Ascribed to Earned Trust inExporting Clusters,” J. Int. Econ. 48:1, pp. 139–50.

Schmitz, Hubert and Peter Knorringa. 1999.“Learning from Global Buyers,” working paper100, Sussex: Institute for Development Studies.

Schmitz, Hubert and Khalid Nadvi. 1999. “Clus-tering and Industrialization: Introduction,”World Dev. 27:9, pp. 1503–14.

Spencer, Barbara J. and Larry D. Qiu. 2001.“Keiretsu and Relationship-Specific Invest-ment: A Barrier to Trade?” Int. Econ. Rev. 42,pp. 871–901.

Stalker, Peter. 1994. The Work of Strangers: ASurvey of International Labour Migration.Geneva: ILO.

Steiner, Robert L. 1997. “The ‘Own-Brand Mar-keter’ Debuts in the 1997 Census,” Rev. Ind.Org. 12, pp. 809–16.

Trefler, Daniel. 1995. “The Case of the MissingTrade and Other Mysteries,” Amer. Econ. Rev.85:5, pp. 1029–46.

Weidenbaum, Murray and Samuel Hughes. 1996.The Bamboo Network. NY: Free Press.

Yoshino, M. Y. and Thomas B. Lifson. 1986. TheInvisible Link: Japan’s Sogo Shosha and theOrganization of Trade. Cambridge, MA: MITPress.

Rauch: Business and Social Networks in International Trade 1203