EMBEDDEDNESS IN THE MAKING OF FINANCIAL CAPITAL: HOW ... · EMBEDDEDNESS IN THE MAKING OF FINANCIAL CAPITAL: HOW SOCIAL RELATIONS AND NETWORKS BENEFIT FIRMS SEEKING FINANCING* Brian

American Sociological Review, 1999, Vol. 64 (August:481–505) 481

EMBEDDEDNESS IN THE MAKING OF FINANCIAL CAPITAL:HOW SOCIAL RELATIONS AND NETWORKS BENEFIT

FIRMS SEEKING FINANCING *

Brian UzziNorthwestern University

I investigate how social embeddedness affects an organization’s acquisitionand cost of financial capital in middle-market banking—a lucrative but un-derstudied financial sector. Using existing theory and original fieldwork, Idevelop a framework to explain how embeddedness can influence whichfirms get capital and at what cost. I then statistically examine my claimsusing national data on small-business lending. At the level of dyadic ties, Ifind that firms that embed their commercial transactions with their lender insocial attachments receive lower interest rates on loans. At the network level,firms are more likely to get loans and to receive lower interest rates on loansif their network of bank ties has a mix of embedded ties and arm’s-lengthties. These network effects arise because embedded ties motivate networkpartners to share private resources, while arm’s-length ties facilitate accessto public information on market prices and loan opportunities so that thebenefits of different types of ties are optimized within one network. I con-clude with a discussion of how the value produced by a network is at a pre-mium when it creates a bridge that links the public information of marketswith the private resources of relationships.

cess and costs in ways that are inadequatelyincorporated into financial theory (Baker1990; Mintz and Schwartz 1985; Mizruchiand Stearns 1994b; Podolny 1993; Uzzi andGillespie 1999). Bankers and entrepreneursecho this observation and complain that fi-nancial models often do not appreciate thevalue of bank-client relationships. This sug-gests that there is a growing demand for so-ciological theory on finance (Abolafia 1997;Arrow 1998; Haunschild 1994; Mizruchi andStearns 1994b).

Most sociological research on lending hasnot focused on the availability and pricing ofcapital. Rather, classical writings take aphilosophical approach to analyzing moneyeconomies, and most contemporary workconcentrates on how bank-firm ties, coarselydefined as the incidence of a director inter-lock, affect the politics and level of a firm’sborrowing (Baker 1990; Mintz and Schwartz1985; Mizruchi and Stearns 1994a). Theseapproaches, while productive for certainproblems, leave unexplored how the embed-dedness of commercial transactions in social

hich firms get capital and at whatcost? The answer can determine theW

* Direct all correspondence to Brian Uzzi,Kellogg Graduate School of Management, North-western University, Evanston, IL ([email protected]). I thank John Wolken at the Federal ReserveBank, James Baron, Wayne Baker, Willie Ocasio,Ryon Lancaster, Nan Lin, Mitch Petersen, HollyRaider, Mike Sacks, James G. Gillespie, the bankCEOs and Relationship Managers that contrib-uted their insights to this research, and North-western University’s Institute for Policy Researchand the Kellogg Graduate School of Manage-ment’s Banking Resource Center for their re-search support.

life chances of firms, the growth of econo-mies, and how markets stratify firms and per-sons through the rationing and pricing ofcredit. In financial theory, any firm with apositive economic net present value shouldobtain credit at a competitive price (Petersenand Rajan 1994). Sociological theory doesnot necessarily reject this axiom, yet arguesthat banking transactions are embedded insocial relations that uniquely shape credit ac-

482 AMERICAN SOCIOLOGICAL REVIEW

attachments and networks affects personaland corporate financial dealings (DiMaggioand Louch 1998; Uzzi 1997). Finance re-search similarly concludes that while bank-firm ties are more critical to lending marketsthan classical theory suggests, inconsisten-cies in financial theory also signify a needfor more research on how social relation-ships and networks affect who gets capitaland at what cost (Arrow 1998; Petersen andRajan 1994).

I examine how bank-borrower relation-ships and networks affect a firm’s acquisitionand cost of capital using a social embedded-ness approach (Granovetter 1985). The so-cial embeddedness approach aims to explainwhy economic transactions become embed-ded in social relations that differentially af-fect the allocation and valuation of re-sources. Social embeddedness is defined asthe degree to which commercial transactionstake place through social relations and net-works of relations that use exchange proto-cols associated with social, noncommercialattachments to govern business dealings(Marsden 1981; Uzzi 1997). I argue that em-bedding commercial transactions in socialattachments benefits firms that are seekingfinancing by promoting distinctive gover-nance mechanisms and the transfer of privateinformation—factors that motivate banksand firms to find integrative solutions to fi-nancing problems beyond those possiblethrough market relations, which possess dif-ferent benefits.

In developing my arguments, I analyzehow both social relationships and networksaffect lending. At the level of relationships, Idraw on research that examines how proper-ties of embedded ties and arm’s-length tiespromote different kinds of access and gover-nance benefits in market exchanges. At thelevel of the network, I elaborate on the find-ing that networks incorporating a mix of em-bedded ties and market ties provide premiumbenefits because they enable a firm to syn-thesize the advantages of partnering via em-bedded ties with the advantages of brokerageoffered by arm’s-length ties. I argue thatfirms are more likely to secure loans and re-ceive lower interest rates if they are tied totheir lenders through embedded ties and iftheir networks of bank ties have a mix of em-bedded ties and arm’s-length ties.

My study aims to enhance sociologicaltheory on finance in several ways. First, I ex-amine the setting of interest rates, the touch-stone market mechanism by which value iscreated, thereby extending sociological re-search on markets to price formation(Fligstein 1996; Podolny 1993). Second,since financial capital is a substitutable com-modity and banks can write nearly completecontracts by holding collateral, this study re-veals how social embeddedness operates inthe presence of market “efficiency” condi-tions thought to supplant it (Carruthers1996). Third, I focus on what bankers dub“the midmarket”—a sector of the economythat has been neglected in research yet de-serves closer analysis (Fama 1985). Themidmarket is composed of firms with fewerthan 500 employees or less than $500 mil-lion in annual sales and is bountiful in its so-cial and economic effects. The midmarketaccounts for more than one-half of the U.S.gross domestic product (GDP), has twice theinnovations per employee as large firms, andsince 1970 has created two-thirds of the jobsin the United States. In the 1980s, it emergedas a seedbed for entrepreneurship and asource of 16 million of the 20 million newjobs created in that decade.

Before proceeding, it is worth noting theunique qualitative and quantitative materialsused in this analysis. Because the effect ofembeddedness in financial markets is con-tested and remains “in need of greater theo-retical specification” (Smelser and Swedberg1994:18), I strengthen my analysis using atriangulation of theory, fieldwork, and statis-tical analysis (King, Keohane, and Verba1994). I use original field data on bank-bor-rower ties to help explicate and illustrate themechanisms by which embeddedness pro-duces outcomes and actors construct mar-kets. I then use a national random sample of2,400 companies to test the validity andgeneralizability of my theory.

THEORY

The social embeddedness framework is oneof several sociological accounts for how so-cial structure affects financial markets(Granovetter 1985; Portes and Sensenbren-ner 1993; Romo and Schwartz 1995, Uzzi1996, 1997). Research has focused on the

EMBEDDEDNESS IN THE MAKING OF FINANCIAL CAPITAL 483

types of social relations and social networksthat exist and their economic effects. Follow-ing this literature, I treat social embed-dedness as a variable and focus on how thequality of relationships and the configurationof ties in a network influence a firm’s abilityto obtain loans and to lower the cost of bor-rowing.

Relationships vary between arm’s-lengthand embedded (Baker 1990; Lie 1997;Powell 1990; Uzzi 1996, 1997). Arm’s-length ties are characterized by lean andsporadic transactions and “function withoutany prolonged human or social contact be-tween parties . . . [who] need not enter intorecurrent or continuing relations as a resultof which they would get to know each otherwell” (Hirschman 1982:1473). Weber(1946) characterized them as lacking socialdistinction and having an expressive naturethat “knows nothing of honor” (p. 192). Themain proposition related to arm’s-length tiesis that they determine the degree to whichan actor can access heterogeneous informa-tion in a market, even if that information ispublicly available through advertising orpublicity, because actors use network ties tosearch for opportunities and investments.For example, Granovetter (1973) found thatjob-seekers tend to search for and learnabout new job openings through acquaintan-ces, even when the jobs were publicly ad-vertised. Davis (1991) found that firmsadopted “poison pills” chiefly through inter-lock ties, despite the takeover defense’spublic notoriety and marketing by many le-gal firms. Burt (1992) developed thisformulation’s most trenchant propositions.He argued that the strategic expansion ofnetworks through the use of arm’s-lengthties offers the highest possible returns tofirms and persons by linking them to di-verse pools of market information, whichthey broker among less informed actors whoreside in cloistered networks of relationsthat hinder their autonomy.

In contrast, there is less theory and empiri-cal justification for expecting that sociallyembedded ties generate exchange benefits inmarkets. Recent research on interfirm net-works suggests that embedding economic ex-changes in social attachments can both cre-ate unique value and motivate exchange part-ners to share the value for their mutual ben-

efit. Embedded ties promote these outcomesthrough the transfer of private resources andself-enforcing governance (Portes andSensenbrenner 1993; Uzzi 1997). Private re-sources and private information are distinc-tive in that they identify where an actor’s ex-pertise and dependencies reside. They mightinclude, for example, unpublished capabili-ties in products, the need to source a particu-lar material, the strategic blueprints for anexecutive succession, investment plans,failed solutions, the rollout date of a newproduct, or critical resource dependencies. Inessence, private information differs frompublic market information, such as financialstatements or job listings, in that it is not “in-formation for the asking,” but informationthat must be voluntarily transferred in an ex-change. In fact, because private knowledgecan be misappropriated, it is commonly in-accessible through arm’s-length ties and isshared only within a set of trustworthy ex-change partners.

The transfer of private knowledge pro-motes value creation in exchanges by reveal-ing to exchange partners the unique possi-bilities they possess for matching their com-petencies and resources. In contrast, publicinformation such as ask-and-bid prices canbe a source of value creation, but is less soin competitive markets because it is less re-stricted and unique than private knowledgeand resources. Hence, the solutions promptedby the transfer of private knowledge arevaluable not only because they are distinc-tive, but also because they are hard for com-petitors without private knowledge to imi-tate. Consistent with this argument, Ecclesand Crane (1988) found that investmentbankers were able to customize deals andcreate innovative risk-reducing financial in-struments for their clients when they pos-sessed information and resources beyondwhat firms made publicly available. MarkTwain Bancshares, a lucrative midmarketbank, gained recognition by using private in-formation to tailor their bank products andloan structures to the distinctive and oftenconfidential capabilities of their customers(Baker 1994).

The embedding of commercial transac-tions in social attachments promotes thebenefits discussed above by enacting expec-tations of trust and reciprocal obligation that


actors espouse as the right and proper proto-cols for governing exchange with personsthey come to know well (Blau 1964;DiMaggio and Louch 1998; Portes andSensenbrenner 1993).1 These expectationsreduce fears of misappropriation becausetransactors anticipate that others will notvoluntarily engage in opportunistic behav-ior. Instead, exchange partners share the be-lief that these motives, coupled with accessto private information, can enlarge the poolof potentially beneficial transactions thatare not available through market means.Moreover, because the protocols of embed-ded ties are borrowed from the protocols ofsocial attachments, which are learned frompre-existing structures, they are serviceablein business dealings not just as “good faithconformity” norms, but as clear expecta-tions for a “meeting of the minds” (Macneilforthcoming). Potentially this can save onthe costs of organizing other governance ar-rangements, freeing resources for futureproductive prospects (Fukuyama 1995). Inthis way, embedding transactions in socialties does not foreordain cooperative out-comes. Rather, it provides an essential prim-ing mechanism that promotes initial offersof trust and reciprocity that, if accepted andreturned, solidify through reciprocal invest-ments and self-enforcement. In contrast, theexpectation of avaricious actions that is an-ticipated in arm’s-length ties is likely toprompt distrust, even if action is credible,except for discrete cases in which economic

incentives are aligned or third parties en-force fairness (Kollock 1994).2

Extending the above arguments to lendingrelationships suggests that the availabilityand costs of a firm’s capital should varywith the degree to which its commercialtransactions with a bank are embedded insocial attachments. Embedded ties furnishgovernance and access to private informa-tion benefits that can channel resources andmotivate attempts at integrative solutions tolending problems that are not availablethrough market ties. Building on these argu-ments for how embeddedness affects dyadicexchange also suggests that an actor’s net-work of ties is pertinent to the financingprocess. Previous research has argued that alarge network of arm’s-length ties to banksexpands the firm’s pool of potential loan of-fers and the firm’s ability to play banks offagainst one another (Baker 1990; Eccles andCrane 1988). While I agree in part with thislogic, my analysis of lending suggests that“shopping the market” for potential offers isan incomplete picture of the lending pro-cess. Firms secure loans and lower theirborrowing costs by shopping the market forwhat’s available and through collaborativeproblem-solving over terms with specificlenders. This suggests that firms embeddedin networks that enable them to gather in-formation about the range of loan dealsavailable in a market and to access the pri-

1 The literature on attachments distinguishesbetween ties between persons and ties betweenorganizations (Baker, Faulkner and Fisher 1998;Blau 1964; Levinthal and Fichman 1988;Seabright, Levinthal, and Fichman 1992) andviews social attachments as personal ties (whichconstitute ties between the individuals’ firms,even if the reverse does not hold). A social at-tachment is an affiliation of shared interests andfidelity that develops when behavior that is cul-turally associated with familiar and noncommer-cial transactions is enacted as part of the commer-cial exchange (Iacobucci and Ostrom 1996). Inthis study, such social behaviors include weddinginvitations, parties, dining, sports competitions,shows, or other social events that both friends andbusinesspersons can and do commonly enactthrough time and that are valued in that personsshare these behaviors in proprietary ways withselect others.

2 While my objective is to propose and investi-gate these processes rather than to assert their va-lidity, much social psychological research on de-cision-making supports these processes. Mont-gomery (1998) showed how transactors who as-sume the identity of “friend” are likely to cooper-ate, while transactors who assume the identity of“businessperson” are unlikely to cooperate (evenif commitments are credible) because there is nopriming mechanism for trust. The logic of appro-priateness, first identified by March (1994), sug-gests that decision-makers choose actions by ask-ing, “Who am I and what is the appropriate ac-tion for my role?” rather than basing these actionson situation-free personal preferences. Cognitivedissonance research also finds that attitudes andinterests are aligned with role behavior (Kunda1990:484). This suggests that motives to createand share value are supported by psychologicalprocesses that are set in motion by embeddingprocesses.


vate resources of particular lenders gainpremium benefits from social structure be-cause their ability to “broker and partner”on loan deals is enhanced.

ETHNOGRAPHIC FIELDWORK

I conducted field research to help formulatemy embeddedness framework. Given thescarcity of research on midmarket banking,field research furnished an empirical basisfor describing the pertinent actors, resources,and relationships. It also enabled a more re-fined analysis of bank-borrower ties thanwould have been possible using coarsermethodological tools, although the smallsample size moderated generalizability. Fieldresearch also permitted a triangulation oftheory, ethnography, and statistical analysison lending.3

I conducted field research at 11 midmarketbanks in the Chicago area, a highly competi-tive banking market. Table 1 describes thedemographic and organizational back-grounds of my 26 interviewees and their 11banks. My sample of interviewees typifiedthe racial, gender, and educational profiles ofbankers, who are largely white, male, andcollege-educated. I principally interviewed“Relationship Managers” (hereafter RMs),the bank personnel who make lending deci-sions and interface with clients. I also inter-viewed two bank CEOs and two bad-debtcollectors (who deal with fraudulent clients)to understand and cross-examine the view-points of other types of lending officers. Ifocused on RMs because they make the judg-ments about a client’s loan eligibility andconsequently can reveal how social and net-work ties affect their lending decisions. Ialso reviewed Federal Reserve Bank Opin-ion Surveys on lending to corroborate RMs’accounts. Appendix A describes thesesources and the fieldwork methodology ingreater detail.

The Social Embeddedness of LendingRelationships

The fieldwork revealed that bankers segmentthe market into three strata: new corporate,midmarket, and entry-level firms. Table 2summarizes the segments, lending practices,and bank-firm relationships in these markets.While important distinctions exist betweenthe new corporate segment and the other twosegments, RMs rarely distinguish betweenmidmarket and entry-market clients andtypically maintain ties to both types of firmsin their portfolios. Thus, I treat midmarketand entry-market firms similarly in my field-work and control for possible differencesdue to organizational size in my statisticalanalysis.

Consistent with previous research, I foundthat in the new corporate segment, public andcertified financial statements provide bankswith ready access to pertinent informationabout a firm’s creditworthiness (Mizruchiand Stearns 1994b). Similarly, firms use theirlarge treasury departments to identify thelowest cost loans or to gain bargaining posi-tion vis-à-vis banks by borrowing directlyfrom money markets.4 Thus, lending ties be-tween big firms and banks are transactional,with banks chasing customers who treatloans and banks as commodities (Davis andMizruchi 1999).

The social structure of the midmarket dif-fers from the new corporate segment in waysthat have important theoretical and substan-tive implications. Firms experience ambigu-ity in evaluating banks because they lack so-phisticated financial expertise and are toosmall to borrow from money markets. Thus,they depend on banks for financial adviceand credit, yet they lack the clout and finan-cial wherewithal to ensure a bank’s probity,increasing their reluctance to share privateinformation with the bank’s RMs. For ex-ample, one RM observed,

If it’s company money it might be in the rightpocket, if it’s personal money it might be intheir left pocket, but it’s all in the same pair of

3 Lending decisions are made in two stages. Instage one, a bank decides whether to offer a loanto an applicant. Loan denial reflects cases inwhich the bank will not raise the interest rate tomake up for a bad credit risk. In stage two, thebank decides what cost of credit to charge appli-cants who were deemed creditworthy in stageone.

4 Money markets sell capital directly to firms atthe same rate as banks. London InterBank OfferRate is the current money market standard forlending rates and reflects the interest rate paid ondeposits among banks in the Eurodollar market.


Table 1. Characteristics of Interviewees in the Field Research: Relationship Managers (RMs) at Chi-cago Banks, 1988

RM’s Profile

Bank Number of Number of InterviewDepositsa RM’s Years in Firms in Time

Bank (in $1,000s) Sex /Race Industry Portfolio (in Minutes)

Entry-LevelFirst Bank of 104,181 Male/white 17 21 60 Evanston Female/white 2 9 30

1st National Bank 125,475 Male/white 40+ 50 120 of La Grange Male/white 8 17 120

Male/white 3 6 120

1st Midwest Bank 178,825 Male/white 35 —. 60Male/white 4 17 45

MidmarketBank One-Chicago 1,156,874 Male/white 15 50 45

Male/black 3 12 50

Female/white 6 26 30

Cole Taylor 1,327,893 Male/white 20 54 45

Male/white 5 13 50

BankAmerica 3,887,571 Female/white 19 60

Male/white 7 15 45


Male/white 19 50 75

American National 4,357,509 Male/white 9 25 50 Bank

Northern Trust 6,301,607 Male/white 25 27 120

Male/white 7 —. 60

Male/white 5 8 30

Male/white 15 19 70

Midmarket and New CorporateHarris Bank 8,653,638 Male/white 7 21 60


Male/white 12 14 45

LaSalle National 9,761,356 Male/white 12 25 50 Bank

1st National Bank 17,961,480 Male/white 25 50 35 of Chicago

Note: A total of 26 RMs were interviewed at 11 different Chicago banks. Interviews were conducted onsite between February 1 and May 1, 1988. Interview time totalled 26 hours.

a Bank deposits came from the Bank and Thrift Branch Office Data Book (FDIC 1998).

pants. It’s all their money. It’s very personal tothem. . . . A lot of them will feel like, “We’rejust a small guy, they’re from a big bank.”

Although banks control the flow of capi-tal, they also experience ambiguity in evalu-ating midmarket firms, which are typically

not debt rated or certified. In particular, thebundling of the business and private lives ofthe firm’s managers is viewed as a keysource of performance ambiguity. Becausethe firm’s capital and the entrepreneur’scapital are often intertwined, banks’ RMs


need to assess how a client’s private life af-fects the firm’s economic performance.These social preconditions importantly affecteconomic exchange in this market and, in thecourse of everyday business dealings, alsoencourage discussions of private matters nor-mally had with social attachments, deepen-ing the embeddedness of commercial trans-actions between firms and banks in social at-tachments and networks. An RM explained,

It’s something you wouldn’t think . . . has todo with major business, but . . . [e]very socialissue is played out in economic form. They[CEOs] have children of unequal talents; theCEO is less talented than the children. Some-body doesn’t want to give up stock. Somebodydoes. . . . Can’t see that on a balance sheet orP&L [profit and loss statement]. You need tounderstand what’s going on around the indi-vidual, . . . and that plays out in “situations.”That’s the dynamic.

So information is not efficient, and with thatcomes the need for the bank to interpret. . . .[I]mperfect information and [the firm’s] imper-fect awareness of alternatives means that most

conversations are negotiations because thereneeds to be a meeting of the minds. . . . Youalso will develop, as a byproduct of that atten-tion, a relationship.

In a fashion analogous to how financialmarkets govern exchanges and certify the va-lidity of publicly available information, theembedding of commercial transactions in so-cial attachments and networks creates amechanism by which to govern exchanges ofprivate resources (Abolafia 1997). RMs re-fer to this process as “market making,” aphrase that denotes their view of how socialties furnish governance arrangements andpromote transfers of private resources, whichin turn make deals that would not arise intheir absence. A lead RM summarized theseessential conditions:

If anybody tells you a story, it reflects theirview of the world, which doesn’t mean thatthey’re lying but it’s impossible to separate outthe storyteller from their objectives. . . . [Y]oucouldn’t just say, “Oh the truth is in the finan-cial statements.” Take a company, and based

Table 2. Characteristics of Banks by Market Segment

Market Segment

Characteristic New Corporate Midmarket Entry Level

Sales segmentation •500 million or more •10 to 500 million •Up to 10 million and market range • Multi-market • Single market •Single market

• Multi-product • Single or multi-product •Single product

Financial market •Firms debt rated •Firms rarely debt rated •Firms rarely debt rated intermediation •Certified financial •Unreliable or no certified •Unreliable or no certified

statements financial statements financial statements

Firm’s financial •Treasury department •Limited or no financial •No financial staff decision structure •Separation of CEO staff •CEO is owner-manager

and CFO positions •CEO is owner-manager •No CFO•No CFO

Firm’s capital •Firms have multiple •Banks are major source •Banks are major source dependence on sources of external of external financing of external financing banks financing for firms for firms

•Firms have multiple •Limited or no internal •Limited or no internalsources of internal financing for firms financing for firmsfinancing

Role of relationship •Solicits requests for •Assesses financial and •Assesses financial and manager in financing from corporate managerial credit- managerial making deals treasury departments worthiness creditworthiness

•Makes bids on corporate •Makes loan decision •Makes loan decisionofferings •Seeks loan approval •Seeks loan approval

from bank from bank

Sources: Ettin (1994); Berger, Kashyap, and Scalise (1995); Gorton and Rosen (1995); Federal ReserveBulletins on bank changes (November 1996); Federal Reserve Board senior loan officer opinion surveys;and original field research.


on different accounting treatments you havedifferent looking balance sheets. If all you didwas look at the numbers, you would make dif-ferent decisions on the same company! I’ll askquestions about financial statements or projec-tions or the business but getting answers is notenough. So, we need this interactive process,. . . which is this digging in and recreating ofsomething so that you understand the compo-nents. . . . That’s a relationship, . . . a marketbeing made.

Also consistent with embeddedness theory,I found that the degree to which intervieweesembed their transactions in social structurevaries. The more that commercial transac-tions are embedded in social attachments, themore expectations of trust govern exchanges.One RM expressed it this way:

[A] relationship [means] that you know a per-son like his family and you feel on a level withhim—not pure friends—but that he trusts whatyou say. That you’re taking care of him. . . .[So] the more I know a person, the more he un-derstands why I’m asking these questions. Hedoesn’t feel so defensive. Otherwise, with mar-ket ties it’s a battle.

Another RM said,

A relationship on a social basis tends to breaka lot of ice and develop a multidimensional re-lationship that’s more than cold facts, interestrates, and products. It’s an emotion-based bond. . . that’s so important to have . . . [because]the customer will let us know about problemsearly, so we can correct them.

Other RMs noted that reciprocity charac-terizes embedded ties, an outcome that isbolstered by expectations of trust. As anotherRM explained,

On the golf course, at a ball game, or the the-ater, they’ll let their guard down more often.We exchange information—not like a mar-riage—more like dating. I share informationabout me as a person. I let them see me andshare with them our company’s struggles. As Ishare that information, I get information back.It’s kind of a quid pro quo.

A distinctive aspect of the embedding ofcommercial transactions in this market is thatthey often have properties of extended clo-sure that promote the creation of a commonset of inferences among the members of anetwork. These properties of closure arisewhen RMs and entrepreneurs form direct re-lations with third persons, such as spouses

and children, which entrepreneurs and RMsconfidentially rely on for perceptions of char-acter and trustworthiness in their businessdealings with others. By enclosing privateand public information flows in a matrix ofsocial ties, shared expectations and regular-ized behavior arises not through the enforce-ment by a third-party, such as the courts, butthrough unanimity of inference. This unanim-ity of inference is important for undertakingrisky actions, such as estimating credit eligi-bility or long-term returns on loans, becauseit reduces the perceived uncertainty associ-ated with estimating an actor’s most likelybehavior. In its application and conse-quences, this property combines the conse-quences of network closure typically foundwithin common social groupings (Coleman1988) with the purposeful multivocal strate-gies used by the Medici to interlock the eco-nomic and personal fates of Florentine mer-chants and bankers (Padgett and Ansell1993). A key difference is that the Mediciconstructed this closure through formal mar-riages and contracts, whereas modern-daybankers achieve similar governance out-comes by organizing informal social eventsthat promote relationship-building amongRMs, entrepreneurs, and their significant oth-ers. Tom, a lead RM at a large midmarketbank, described the process of enclosing so-cial ties so that they crisscrossed the personaland business networks of entrepreneurs, in-creasing the embeddedness of the tie betweenthe banker and entrepreneur and the unifor-mity of inference about the banker’s credibil-ity. Here Tom refers to his client, Jim, Jim’sspouse, Ellen, and the consequences for gov-ernance of this type of embedding:

For Ellen to tell Jim, “You know, that Tom, Ireally like him and I trust him a lot,” has moreimpact on his view of me than if his controllertold him that. It’s sort of the old Nancy Reagan“pillow talk” thing with Ron. They’re integralto their spouses’ decisions. Getting to knowthem and having them get to know you, bridgesthose personal things that you talk about andknow about them. And the web gets wovendeeper in terms of the personal side.

While these results illustrate the preva-lence and material consequences of embed-ded ties, my fieldwork revealed that embed-ded relations retain “passions” that are un-characteristic of the antipathy of arm’s-


length transacting (Hirschman 1977). Even ifembedding is initiated for governance andaccess benefits, the resulting intimacy im-bues the relationship with expressive valuethat is separate from purely material stipula-tions, even if it impinges upon them. For ex-ample, one RM declared,

[Y]ou have to maintain that professional dis-tance because you never know when you’regoing to have to make that tough call [i.e., risklosing the attachment over a business differ-ence]. But having said that, . . . I have clientsthat I’m very close with, and in most circum-stances it helps. I know their kids’ names andwhen their kids have the flu. I go out sociallywith my wife and with them and their spouses.

Another RM explained how embeddingspontaneously infuses a business tie with so-cial values and attitudes that would be irrel-evant in the market model:

After he [the entrepreneur] becomes a friend,you want to see your friend succeed and thatgoes along many lines. If I can be a part ofhelping them do that, it’s a real good feelingand I’m providing a service not only to thembut their employees. . . . So there’s a lot ofthings that you kind of from a moral standpointtake into effect. . . . That is kind of a side ef-fect of your relationship.

By contrast, arm’s-length ties lack the ex-pectations of trust and reciprocity needed forknowledge transfer and collaboration, result-ing in different economic consequences. Forexample, one RM described how arm’s-length ties might be effective at governingprice data across different banks but are poorgovernance arrangements for imbuing infor-mation with credibility or transferring pri-vate information. He said,

Firms got to get comparative information, . . .[but] oftentimes entrepreneurs will negotiatewith you and they’ll tell you they’ve got a dealfrom somebody else and they don’t. That’s partof where that honesty and integrity and beingable to trust the people that you’re dealing withbecomes very important.

Another RM stated that arm’s-length ties put

. . . a relationship out for bidding. Every op-portunity a customer has to get credit they’llshop your deal. [They’ll say], “I’ve talked to acouple other banks and they’re willing to giveme this.” . . . It’s price oriented. . . . [If] I askquestions about performance, the client is ag-gressive and that’s not fun.

RELATIONAL EMBEDDEDNESS ANDLOAN ACQUISITION AND COSTS

How do these properties of embeddednessaffect an organization’s cost and acquisitionof capital? I argue that embedded ties bothcreate value in the dyad and motivate ex-change partners to share that value. Focus-ing on the mechanisms by which value iscreated and shared from the firm’s perspec-tive, the two most relevant processes have todo with the creation of contingent contractsand the leveraging of social capital from oneRM to another RM on behalf of the firm.5

The high level of trust in the relationship en-ables firms and banks to negotiate contingentloan agreements. A contingent loan agree-ment converts a loan with a homogeneousprice structure into a loan that has separatecomponents and a tiered price structure. Agradient of prices and terms in a contingentagreement enables a firm to start with andmaintain low capital costs and few loan re-strictions so long as it sustains a preset levelof performance, thereby enhancing its crediteligibility and reducing its borrowing costsrelative to one-cost, one-structure loans. Bycontrast, loans negotiated through arm’s-length ties often have homogeneous struc-tures because there is no basis of trustwor-thiness on which to estimate how credit riskscan be assessed relative to the firm’s prom-ises to repay. A case often recounted by RMsconcerned attempts to structure loans inways that gave firms the benefit of the doubtcontingent on the interpretation of ambigu-ous performance data that would otherwiseresult in loan denial, an unfavorable rate, ortight restrictions. Using a contingent agree-ment, RMs might offer, for instance, a lowinterest rate the first year that would rise insubsequent years only if the firm failed tomaintain its projected performance level. Re-markably, this specification of the contingentordering of the risk inherent in the loan dealwas often predicated on the level of trust and

5 Embeddedness creates value for banks by en-hancing their ability to reduce the costs of writ-ing loan contracts, to retain clients, and to de-commodify financial capital. These processes andtheir effects on the bank’s profit spread and con-tractual restrictions on a loan are explored in Uzzi(1999).


reciprocity in the relationship rather than oncomparative information that appeared inpublic financial statements. In particular, thetrust and reciprocity of embedded ties en-hanced the transfer and credibility of the pri-vate information that was essential to the cre-ation of a unique contingent loan agreementfor the firm. A lead RM explained how hisembedded tie to an entrepreneur motivatedan integrative solution to the firm’s credit re-quest through the design of a specializedcontingent contract, which resulted in betteraccess and lower costs than arm’s-length re-lations could have provided:

[B]ecause we knew this guy [I said], . . . “Tellyou what we’ll do: We’ll give you a price of Xtoday. We’ll base our pricing as if those ex-penses were not in your financial state-ments. . . . But after 12 months, . . . if it’s allflushed through you will continue on in thisprice level. If you don’t, boom, your pricingwill go up.” So, because of the relationship,because we knew the guy and we really be-lieved in him and trusted him, we gave him thebenefit of the doubt on the pricing for the firstyear. He has to continue to perform or it goesup. So, that’s a way we would sort of marrythe two, the objective and the subjective, if youwill.

Embedded ties also benefit firms by moti-vating bankers to leverage their personal so-cial capital at the bank on the firm’s behalf.Unlike the advantages described above, theseoutcomes are not necessarily attempts to af-fect the loan’s tangible features. Rather, RMsuse aspects of their social capital within thebank, such as their reputation or social tiesto other RMs, to influence the expectationsof other relevant bank decision-makers re-garding a firm’s creditworthiness. I observeda similar phenomenon (Uzzi 1996) when Ifound that the expectations of trust and reci-procity between two economic actors couldbe “rolled over” to a new third party, therebyestablishing trust and reciprocal obligationsbetween two parties that lacked a prior his-tory of exchange. In an analogous process,RMs seeking to act on a firm’s behalf primedfirst-time introductions between other RMsand the firm’s managers with expectations oftrust—extending the web of shared beliefsabout the firm’s creditworthiness to other rel-evant bank decision-makers. One RM de-scribed how these factors can play out for a

firm in a deal in which its creditworthinessis indefinite. In particular, she noted how herreciprocal obligee to the entrepreneur and theentrepreneur’s personal expression of needinduced her to pledge her social capital at thebank on the firm’s behalf, despite the perfor-mance ambiguity reported in the firm’s fi-nancial statements:

[T]he deal on paper is a tough deal. And he [theCEO] said, “I’m fucking scared.” I said,“Okay, as long as I know where you stand.” . . .Well, obviously that’s a long way from I’mfucking scared to there’s a deal here. [So], I goto my president and we go through the creditrisks. I said, “All the credit risks are blatantlyobvious. . . . He said, “Well, how do you over-come it?” [I said], “We’ve got to go see thebusiness and meet the people.” And he agreedand said, “Then I want to see the business andmeet the people.” Now, I can’t control what his“gut” is going to be. But I know the principalsof the firm, a regional credit officer who’schairing up a loan committee, my President andsenior lender. [So], it’s got to be a real badcredit for them to say no, especially when Ihave a 40-percent growth markup.

These arguments and findings illustratepatterns of relational embeddedness and howit operates even in well-developed financialmarkets by positively affecting a firm’s abil-ity to acquire capital and lower its cost ofcapital. Embedded ties generate surplusvalue for the firm by promoting private in-formation and resource transfers that createvalue and motivate banks to share the valuecreated in the relationship with the firm.Therefore, based on my framework andfieldwork, I expect statistical analysis to sup-port the following hypotheses:

Hypothesis 1: The more a firm’s commercialexchanges with a bank are embedded insocial attachments, the more likely thefirm is to acquire financing at that bank.

Hypothesis 2: The more a firm’s commercialexchanges with a bank are embedded insocial attachments, the lower the firm’scost of financing at that bank.

STRUCTURAL EMBEDDEDNESS ANDLOAN ACQUISITION AND COSTS

My argument has focused on the propertiesand consequences of dyadic bank-firm socialattachments, yet dyadic exchanges reside


within a larger network of ties that amplifiesor diminishes their benefits. Some financeand organization theories hold that firms withexpansive networks of arm’s-length ties tobanks optimize their bargaining power andprovide access to a large pool of price andloan possibilities, thereby increasing theirchances at getting corporate financing (Mintzand Schwartz 1985; Petersen and Rajan1994). My argument about network structurepartly agrees with these theories, but focuseson the organization rather than on networksize as the operative mechanism. Although anexpansive network of arm’s-length ties canenable a firm to effectively scan the marketfor deals and help broker information amongbanks, it lacks the embedded ties that facili-tate partnering. Conversely, while embeddedties promote collaboration, a network com-posed only of embedded ties could induceoverattentiveness to local resources and his-torical conventions, limiting a firm’s accessto market information and new ideas. Thissuggests that a network composed of bothembedded ties and arm’s-length ties can mod-erate the shortcomings of each type of tiewhile preserving their strengths, optimizingthe firm’s range of available action.

Building on prior research (Baker 1990;Uzzi 1996, 1997), I refer to a network’sability to synthesize the benefits of differenttypes of ties as network complementarity.Networks high in complementarity producepremium outcomes because the features ofdifferent ties reinforce one another’s advan-tages while mitigating their disadvantages.Thus, while I argue that embedded ties pro-vide special informational and governancebenefits with a specific lender, I acknowl-edge that a firm that maintains a networkcomposed only of embedded ties risks sub-optimal network-level outcomes by notcapitalizing on the properties of networkcomplementarity.6

Heterogeneity in the market for loans sug-gests that networks high in complementarityshould enhance a firm’s ability to get fi-nancing and lower its financing costs. Ac-cess to capital grows with a firm’s ability to(a) shop the market for a loan structure thatis compatible with its credit profile and (b)partner with a bank on the customization ofa loan structure that fits its credit profile.Thus, high network complementarity shouldenhance a firm’s access to capital by pro-moting both brokerage and partnering ben-efits. In my fieldwork, the dual benefits ofnetworks high in complementarity weremanifested in several ways. Bankers notedthat firms with networks high in comple-mentarity used their arm’s-length ties toscan the market for differences in loanprices and structures. That information wasthen transferred to their close lenderthrough an embedded tie, which imbuedmarket data and unfamiliar loan stipulationswith credibility and motivated the lender touse novel market data to the mutual benefitof the firm and bank. In this way, firms withnetworks of embedded ties and arm’s-lengthties combined the partnering benefits of em-bedded ties with the brokering benefits ofarm’s-length ties.

In an example of this process, an RM re-counted the dynamics of a recent deal inwhich his bank was one of the arm’s-lengthties in the network of a firm seeking corpo-rate financing. He noted how the entrepreneurused an arm’s-length tie to his bank to accessinformation about his bank’s loan prices andstructures. The entrepreneur then disclosedthat information to his close lender, whichcustomized a deal for the firm using thebank’s distinctive capabilities and resourcesand the novel loan ideas of other banks thatthe entrepreneur had accumulated through itsarm’s-length ties. What’s more, the RM ob-served that the embedded tie between the en-trepreneur and his bank was the main sourceof both the trust and reciprocal obligationsneeded to customize deals that benefit from a

6 My concept of network complementaritybuilds on portfolio theory, which argues that as-sets in a portfolio have a contingent value thatdepends on the other assets in the portfolio, notjust the properties of the individual asset. In asimilar manner, a tie’s value is greatest whenother types of ties complement its strengths,while the entire portfolio’s value rises if the ben-efits of the different types of ties that composethe portfolio do not coincide. In this context, a

network refers to the firm’s ego network of directties to banks, not the aggregation of all bank-firmnetworks in an arbitrary region or industry bound-ary. Previous research has also referred meta-phorically to ego networks as portfolios (Powell,Kopub, and Smith-Doerr 1996:120).


synthesis of market information and exclu-sive private resources. The RM reported:

Three banks were pitching on the same deal,and the company said to me “give us a creativeidea on how you would structure this financ-ing.” [W]e provided a very creative idea withterm loans and revolving credit [factors affect-ing price and structure]. They said, “We reallylike this structure, but X has been our bank for50 years and we don’t want to pull the agencyfrom them.” When the term sheet came backfrom X bank, X bank had basically our termsheet with their name on it. Later, the CFO saidto me, . . . “Look, you guys came up with theidea. So, we’d like to give you the first shot atour trust business or the private banking of theowners” [a conciliation prize for providingvaluable ideas]. So, we gave the banking in-sight on the marketplace to the firm [but lostthe deal].

The above argument and data suggest thatnetworks high in complementary producepremium benefits by preserving the strengthsand diminishing the weaknesses of differenttypes of ties. In the context of corporate fi-nancing, complementarity increases a firm’sability to broker market information and in-stigate partnering around custom deals bydrawing on both the novel information thatis dispersed among players in a heteroge-neous market and the private resources ob-tained through relationships. Thus, based onmy framework and the fieldwork data, I ex-pect statistical analysis to support the follow-ing hypotheses:

Hypothesis 3: A firm’s likelihood of acquir-ing financing increases when it has anetwork with an integrated mix of em-bedded ties and arm’s-length ties anddecreases when it has a network thattends toward either solely embedded tiesor solely arm’s-length ties.

Hypothesis 4: A firm’s cost of financing de-creases when it has a network with anintegrated mix of embedded and arm’s-length ties and increases when it has anetwork that tends toward either solelyembedded ties or solely arm’s-lengthties.

DATA AND METHODS

I tested my hypotheses using data from theNational Survey of Small Business Finances,

a survey administered by the Federal ReserveBank to investigate how market and organi-zational characteristics affect capital costsand availability. This in-person survey col-lected data on firms’ lending ties, sources offinancing, loans, and organizational and fi-nancial characteristics. The random sampleconsisted of 1,875 corporations and 1,529partnerships/sole-proprietorships with up to500 employees and $154 million in assets,operating in 1989 in the U.S. nonagriculturalsector. Depending on the item, the responserate was 70 to 80 percent, reducing thesample size to about 2300 cases. Nearly 90percent of the businesses were owner-man-aged; 12 percent were owned by women and7 percent were minority-owned.

Dependent Variables

The two dependent variables I examined cor-respond to the first and second stages of thecorporate financing process. The first stageconcerns whether a firm acquires capital (i.e.,a loan). The standard assumption in lendingresearch is that small- to medium-sized firmsconstantly need credit. Consequently, the lackof a loan suggests that a firm was deniedcredit or offered unattractive loan agree-ments, which in effect informally counselapplicants to withdraw their requests, mak-ing self-restricted consumption tantamount todenial of a loan (Lummer and McConnell1989; Munnell et al. 1992). Given these com-plexities, most research cannot fully deter-mine whether a lack of a loan is due to creditrationing by the bank or to the firm’s self-restricted consumption. Hence, research gen-erally adopts the convention that if a firm’sneed for credit is controlled for, firms with-out loans were probably denied credit (Coleand Wolken 1995; Hawley and Fujii 1991).Given that my data coincide with previousresearch in this area, I followed the aboveconvention, defining a firm as credit accessed(coded 1) if it obtained a loan between 1987and 1989 (coded 0 if not). This approach al-lowed me to extend prior research, even if ajudicious interpretation of this stage of themodel’s results were called for. Stage two es-timates the cost of capital, which I defined asthe interest rate on the firm’s loan —the typi-cal measure used in research and practice onlending (Petersen and Rajan 1994).


Independent Variables

A methodological issue concerns how to cre-ate valid quantitative measures that capturethe dimensions of the ethnographic findingsand yet are parsimonious and amenable tostatistical analysis. Using Miles andHuberman’s (1994) and Bollen and Paxton’s(1998) methods, I applied techniques thatlook for convergence among theory, face va-lidity, and discriminant validity. In thesemethods, validity increases if independentsources of theory and evidence converge ona consistent pattern and discriminate amongother concepts.7

Prior research holds that embedding in-creases with the duration of the relationshipand the multiplexity of the relationship(Blau 1964; Gulati 1994; Lazerson 1995;Larson 1992; Lin, Ensel, and Vaughn 1981;Marsden and Campbell 1984; Seabright etal. 1992). Time in a relationship permitsnetwork partners to learn about and shareprivate information, incur debits and creditsin the relationship, form bonds of trust, andexploit opportunities for reciprocity.Seabright et al. (1992) operationalize a so-cial attachment between an auditor and aCEO as “the length of time the individualengages in activities associated with the re-lationship . . . [the strength of the social at-tachment] is likely to increase with theyears of tenure that have elapsed since theformation of the interorganizational rela-tionships” (pp. 133–34). They also opera-tionalize attachment strength between anauditor and a CEO as the multiplexity ofoverlapping roles they interfaced around(CEO, CFO, or CAO).

I looked for convergence between mytheory, prior operationalizations, and the ac-counts of RMs (i.e., face validity) by askingRMs about how embedded ties could beoperationalized and distinguished from otherquantitative measures of the bank-firm rela-tionship (i.e., discriminant validity). For in-stance, I probed RMs with inquiries such as,“If you want to determine if your colleaguehas a close tie with a client like the one wehave been discussing, what quantitative in-formation would you use or look for?” Con-sistent with research, RMs independentlystated that sound measures of embedded tiesincluded (a) the duration of the relationshipand (b) the multiplexity of the relationshipbetween the lender and the firm. Thus, I de-fined the duration of the relationship in yearsand the multiplexity of the relationship as thenumber of business services (e.g., cash boxservices, wire transfers) and personal bankservices (e.g., personal bank accounts, wills,estate planning) used by the entrepreneur.Business and personal services included bro-kerage, capital leases, cash management ser-vices, credit card receipt processing, letter ofcredit, night depository, pension fund, per-sonal estate planning, trusts, retirement plan-ning, revolving credit arrangements, money/coins for operations, and wire transfers. Fi-nally, I performed discriminant validitychecks by inquiring if these indicators mightalso measure lower loan production costs.RMs said these factors lowered transactioncosts and increased retention, but they didnot directly lower the costs of administeringa loan. One RM stated, “It doesn’t make itless expensive to manage a tie the longer it’saround because some long-term clients wantto see the banker every month or utilize thebank’s services where that gets expensive.”

I used the same method to construct mymeasure of network complementarity, whichoperationalizes the degree to which a firmuses arm’s-length ties, embedded ties, or amix of ties to transact with the banks in itsnetwork. (Here, the term network refers tothe egocentric network of direct ties betweena firm and all its banks, not just the lendingbank.) Baker (1990) shows that a Herfindahlindex, a relative of the Gibbs-Martin indexof social heterogeneity (Blau and Schwartz1984), parsimoniously measures the mix ofdifferent types of ties in a firm’s network.

7 Triangulation determines the validity of con-structs by looking for convergence in the out-comes of different methods. As with some otherpsychometric methods (e.g., factor analysis), noformal statistical tests are involved. Triangulationworks by demonstrating that a measure accuratelyrepresents the construct even if some nuances areomitted in the same way that econometric modelsdo not explain all the variance. Moreover, theweaknesses of this method are not unique, but arepart of a class of statistical problems that intro-duce measurement error into network variables.Thus, since measurement error normally attenu-ates estimates, tests of hypotheses are conserva-tive (McPherson, Popielarz, and Drobnic 1992).


The index varies between greater than 0 andless than or equal to 1. As the index nears 0,a firm disperses its transactions among manybanks through arm’s-length ties; as it nears1, a firm consolidates its transactions withone or few embedded ties. At intermediatevalues, a firm has an integrated mix of em-bedded ties and arm’s-length ties. In theanalysis, I used a linear and quadratic termto capture these curvilinear effects. The mea-sure is defined as Σ( )Pj

2 , where j variesfrom 1 to N banks, and Pj is the proportionof the firm’s banking business that is dedi-cated to bank j. I defined Pj as the sum ofthree fundamental accounts—savings, check-ing, and line-of-credit accounts—that RMsused to indicate the intensity of exchangebetween a firm and a bank. For example, if afirm used three banks and apportions 70 per-cent of its business to bank one, 20 percentto bank two, and the remaining 10 percent tobank three, then its network structure scoreis equal to (.70)2 + (.20)2 + (.10)2 = .54. Onepotential shortcoming of this measure is thatit is difficult to compare across cases if thesizes of the firms’ banking networks varywidely. Because midmarket firms vary butnot excessively in the size of their bankingnetworks, I directly controlled for networksize in the regressions.

Consistent with the assumptions of mymeasure of network complementarity, inter-viewees stated that firms that consolidatedtheir banking with one bank tended to em-bed their commercial transactions in socialattachments. In contrast, interviewees re-vealed that firms that dispersed their bank-ing transactions tended to have arm’s-lengthties and that at least occasional arm’s-lengthbusiness at the bank was a prerequisite forthem to respond to clients’ requests for in-formation on loan pricing or structures.(RMs rarely supplied similar information tocustomers without at least an arm’s-lengthtie, such as cold-callers, that RMs consideredto be nonrelationships.) Finally, to examinethe discriminant validity of this measure withmeasures of resource dependence, I askedbankers whether their willingness to collabo-rate was motivated by reciprocal obligationsand trust or dependence on a firm’s business.They declared that banks rarely feel depen-dent on any one firm’s business, particularlyin this banking market segment where banks

are larger than firms. Thus, this measure hasthe advantage of summarizing the level ofthe embeddedness of a firm’s ego-network inone measure that has high face validity andprecedent in studies of banking.

Control Variables

I controlled for the organizational, market,and loan characteristics known to affect cor-porate financing using the standard measuresapplied in prior research (Cole and Wolken1995; Mizruchi and Stearns 1994a). Organi-zational controls included number of employ-ees, organization age, log of sales change(current year minus previous year), corpo-rate status (1 = yes), and cash on hand,which controlled for the firm’s need forcredit. I controlled for the firm’s ability toconvert assets into cash and to carry debt us-ing the firm’s acid ratio ([current assets – in-ventory]/current liabilities) and debt ratio(total liabilities/total assets), respectively. Tocontrol for gender and racial discrepancies incredit decisions (Arrow 1998), I defined afirm as women-owned (1 = yes) or minority-owned (1 = yes) if it had 51 percent or moreownership by women or minorities (the sur-vey did not collect the exact percent). Tocontrol for the size of a firm’s ego-networkfor banking, I created network size, a countof the number of financial institutions a firmuses for financial services. This variable ishighly correlated with more complex indicesof network structure, thus furnishing a rea-sonable proxy for size and degree measuresof network structure that my data did not al-low me to construct (Borgatti 1997).

To control for the loan characteristics thataffect lending, I measured the prime rate (theinterest rate to which loans are pegged) at themonth and year the loan was granted. Termstructure spread was defined as the yield ona government bond of the same maturity ofthe loan less the Treasury bill yield and ac-counts for interest rate differences that varywith the loan’s maturity (Petersen and Rajan1994: 13). Two indicator variables, collateral(1 = yes) and fixed-rate loan (1 = yes) con-trolled for systematic differences in interestrates for loans that require collateral or thathave fixed rather than variable rates. Bankcompetition in the firm’s locale was mea-sured using the Federal Reserve’s bank con-


Ta

ble

3.

Co

rre

latio

n C

oe

ffic

ien

ts b

etw

ee

n V

ari

ab

les

Use

d in

th

e A

na

lysi

s: U

.S.

No

na

gri

cultu

ral F

irm

s, 1

98

9

Va

ria

ble

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

(10

)(1

1)

(12

)(1

3)

(14

)(1

5)

(16

)(1

7)

(18

)(1

9)

(20

)

(1)

Inte

rest

ra

te1

.00

.—.—

.—.—

.—.—

.—.—

.—.—

.—.—

.—.—

.—.—

.—.—

.—

(2)

Du

ratio

n o

f b

an

k-fir

m–

.12

1.0

0.—

.—.—

.—.—

.—.—

.—.—

.—.—

.—.—

.—.—

.—.—

.—re

latio

nsh

ip

(3)

Mu

ltip

lexi

ty o

f b

an

k-–

.14

.09

1.0

0.—

.—.—

.—.—

.—.—

.—.—

.—.—

.—.—

.—.—

.—.—

firm

re

latio

nsh

ip

(4)

Ne

two

rk c

om

ple

-–

.08

–.0

0–

.02

1.0

0.—

.—.—

.—.—

.—.—

.—.—

.—.—

.—.—

.—.—

.—m

en

tari

ty

(5)

Ne

two

rk c

om

ple

-–

.03

–.0

3–

.05

.97

1.0

0.—

.—.—

.—.—

.—.—

.—.—

.—.—

.—.—

.—.—

me

nta

rity

sq

ua

red

(6)

Ne

two

rk s

ize

–.0

6.0

5.2

1–

.12

–.2

01

.00

.—.—

.—.—

.—.—

.—.—

.—.—

.—.—

.—.—

(7)

Wo

me

n-m

an

ag

ed

fir

m.0

2–

.08

–.0

3–

.04

–.0

2–

.01

1.0

0.—.—

.—.—

.—.—

.—.—

.—.—

.—.—

.—

(8)

Min

ori

ty-m

an

ag

ed

fir

m.0

3–

.04

–.0

5.0

0.0

0–

.04

.03

1.0

0.—

.—.—

.—.—

.—.—

.—.—

.—.—

.—

(9)

Nu

mb

er

of

em

plo

yee

s–

.04

.03

.11

.00

–.0

1.1

5.0

0–

.00

1.0

0.—.—

.—.—

.—.—

.—.—

.—.—

.—

(10

)A

ge

of

firm

–.1

2.4

4.0

9.0

5.0

2.1

2–

.10

–.0

9.1

01

.00.

—.—

.—.—

.—.—

.—.—

.—.—

(11

)C

orp

ora

tion

–.1

7.0

8.1

1.0

7.0

3.1

7–

.05

–.0

6.0

8.1

21

.00.—

.—.—

.—.—

.—.—

.—.—

(12

)C

ash

in r

eta

ine

d–

.25

.10

.17

.46

.34

.21

–.0

7.0

0.1

2.1

2.3

11

.00.—

.—.—

.—.—

.—.—

.—e

arn

ing

s

(13

)S

ale

s ch

an

ge

(lo

g)

–.1

1–

.00

.05

.06

.04

.10

–.0

0–

.00

.10

.00

.08

.17

1.0

0.—

.—.—

.—.—

.—.—

(14

)A

cid

ra

tio.0

4.0

2–

.06

.04

.04

–.1

2–

.00

.03

–.0

3.0

5–

.11

.03

–.0

01

.00.—

.—.—

.—.—

.—

(15

)D

eb

t ra

tio.0

2–

.09

.03

–.0

9–

.08

.13

.04

–.0

3.0

0–

.14

.12

–.1

5–

.05

–.3

41

.00

.—.—

.—.—

.—

(16

)P

rim

e r

ate

.04

–.0

1.0

2.0

0.0

0.0

2–

.02

.00

.02

–.0

7.0

8.0

6.0

4–

.02

.03

1.0

0.—

.—.—

.—

(17

)T

erm

str

uct

ure

sp

rea

d.1

4–

.06

–.1

4–

.05

–.0

3.0

4.0

4.0

1–

.04

–.0

7–

.01

–.0

7–

.00

.01

.00

.00

1.0

0.—

.—.—

(18

)C

olla

tera

l on

loa

n–

.03

–.0

6–

.00

.01

.03

.04

.00

–.0

2–

.01

–.0

6.0

8–

.00

.05

–.0

6.0

4.0

0.2

51

.00

.—.—

(19

)F

ixe

d-r

ate

loa

n.2

0–

.06

–.1

8–

.06

–.0

4–

.00

.05

.01

–.0

8–

.05

–.0

9–

.14

–.0

2.0

8–

.06

.00

.75

.15

1.0

0.—

(20

)B

an

k co

mp

etit

ion

.08

.04

–.0

3–

.06

–.0

4–

.03

.01

–.0

7.0

4–

.02

–.0

7–

.16

.00

–.0

6.0

2–

.06

–.0

0–

.00

.00

1.0

0

Me

an

10

.97

12

.98

2.6

8.8

0.7

72

.23

.12

.07

25

.61

4.0

8.5

58

.55

4.0

66

.79

.52

9.1

4.8

9.7

9.5

32

.51

Sta

nd

ard

de

via

tion

2.6

01

2.7

73

.56

.35

.38

1.5

4.—.—

54

.91

2.5

2.—

3.5

48

.84

40

.64

.75

1.9

71

.07

.—.—

.65

No

te: M

ea

ns

are

exp

ress

ed

as

pe

rce

nta

ge

s; N

= 2

,22

6.


centration index (1 = high concentration, 2 =moderate concentration, and 3 = low concen-tration). In the analysis, I reversed the scaleof this variable so that higher values corre-spond to greater bank competition. This mea-sure captures both the market pressures onbanks to offer favorable interest rates and therange of interest rates in a market (Petersenand Rajan 1995). Differences in interest ratesin the Northeast, North Central, South, andWest were controlled for with four regionalindicators; industry differences were con-trolled for with seven industry indicators us-ing two-digit SIC codes (lowest level of dis-aggregation in the data). Table 3 reports thecorrelations and descriptive statistics for thevariables used in the analysis.

Statistical Model

I modeled the effect of social ties and net-works on the acquisition and cost of capitalwith a Heckman two-stage selection model(Heckman 1976). This model is used whenone dependent variable (e.g., the cost ofcapital) depends on another dependent vari-able (e.g., having a loan). The first stage is aprobit regression that models whether thefirm accessed credit. The second stage is anOLS regression that uses the estimated Millsratio from the first stage to account for se-lection bias in estimating the interest rate onthe outstanding loan. If regressions were runon the firms with loans, it could produce bi-ased results if access to capital was not ran-dom. Fitting variables that correlate with ac-cess to credit but not with the interest ratehelps solve the problem. Thus, I chose vari-ables for each stage based on previous re-search and on my fieldwork (Mizruchi andStearns 1994b; Petersen and Rajan 1994).8

RESULTS

Table 4 presents the results of the selectionmodel, which examines the correlates of theacquisition (Model 1) and cost of corporatefinancing (Model 2). The control variable ef-fects are noteworthy because they establisha baseline for gauging the effects of socialstructure and the appropriateness of themodel’s specification. Consistent with previ-ous research, Model 1 shows that organiza-tions are more likely to access credit if theyare younger, have liquidity, have a noncor-porate form, and are located in regions withlow bank competition or low credit costs;Model 2 shows that large firms and firmswith increasing sales, variable rate loans, orloans with collateral have lower costs ofcapital (Petersen and Rajan 1994).

In Model 2 however, it is somewhat sur-prising that prime rate and term structurespread had null effects. In a post hoc analy-sis, I dropped the fixed-rate loan variablefrom the regression equation to see if it wascapturing the effects of these variables. Asexpected, prime rate had a positive effect andterm structure spread had a negative effecton the cost of capital when this fixed effectwas removed from the equation. Also of sub-stantive importance is the finding thatwomen-owned, and to a lessor degree minor-ity-owned firms (significant at the p < .05level in a one-tailed test), are less likely toaccess credit than firms managed by whitemales. The discrepancy among these socialcategories is significant because access tocapital may be more critical than its cost,particularly to liquidity-sensitive smallfirms. Taken together, these results suggestthat the models are adequately specified andthat the embeddedness measures capture neteffects not explained by current financialtheory.

Hypotheses 1 and 2 predicted that relation-ship duration and relationship multiplexityincrease a firm’s probability of access to

8 I checked my model specifications by esti-mating the interest rate and an interest rate devia-tion score (firm’s interest rate minus the primerate) with nested OLS regressions. While OLSdoes not handle selection error, it allows subsetsof variables to be entered separately—a nestedprocedure that tends to bias the Heckman model,which is sensitive to specification error. I also ex-amined possible distributional artifacts. I trun-cated the duration variable at its 95th percentileto test for sensitivity to distributional extremes. Iran separate models with multiplexity dichoto-mized (no services versus one service versus mul-

tiple services). I ran models on the subset of datawhere network complementarity was not equal to1 to investigate whether the effects were sensi-tive to observations where network comple-mentarity is equal to network size because thefirm uses just one bank. The results were substan-tively identical to those reported in Models 1 and2, supporting my specifications.


Table 4. Coefficients from the Heckman Selection Regression of Access to Credit and Interest Rateon Loan on Selected Independent Variables: U.S. Nonagricultural Firms, 1989

Model 1 Model 2(Credit Accessed) (Cost of Capital)

Independent Variable Coefficient S.E. Coefficient S.E.

Constant 1.164** (.184) 11.233** (1.155)

Embeddedness

Duration of bank/firm relationship –.001 (.002) –.013* (.005)

Multiplexity of bank/firm relationship .005 (.008) –.042* (.018)

Complementarity of firm’s bank network 1.772* (.547) –6.275** (1.134)

(Complementarity of firm’s bank network)2 –1.119* (.461) 5.030** (.960)

Size of firm’s bank network .133** (.020) .039 (.053)

Organizational CharacteristicsWomen-managed firm –.189* (.086) .020 (.221)

Minority-managed firm –.180 (.107) .371 (.288)

Number of employees .001 (.001) –.002* (.001)

Age of firm –.007** (.002) .— .—

Corporation –.132* (.063) .— .—

Cash in retained earnings –.023 (.013) .— .—

Sales change (log) .005 (.003) –.022** (.007)

Acid ratio –.025** (.004) .028* (.015)

Debt ratio .004 (.038) .070 (.150)

Loan Characteristics

Prime rate .— .— .155 (.096)

Term structure spread .— .— –.020 (.106)

Collateral on loan .— .— –.393* (.175)

Fixed-rate loan .— .— .709** (.207)

Region and Industry IndicatorsBank competition –.140** (.043) .193 (.111)

Northeast .264** (.082) .— .—

North Central .211** (.085) .— .—

South .138* (.081) .— .—

Mining .257 (.316) .— .—

Construction .104 (.093) .— .—

Manufacturing –.063 (.097) .— .—

Transportation, communication, public utilities .124 (.164) .— .—

Wholesale trade –.004 (.070) .— .—

Retail trade .178 (.141) .— .—

Wald χ2 (d.f. = 16) 131.31**

Log-likelihood –3363.2

Rho –.352

Note: N = 2,226.*p < .05 ** p < .01 (two-tailed tests)


credit and lower their cost of capital. Con-sistent with Hypothesis 2, Model 2 showsthat the duration of the relationship andmultiplexity of the relationship have largeand significant effects on reducing the costof capital net of standard market, firm, andloan control variables. Furthermore, the con-sistent effects of both independent variablessuggest that they capture a similar underly-ing construct net of each other’s effects. Inreal terms, the coefficients in Model 2 indi-cate that an additional year in a relationshiplowers a firm’s interest rate by 1.3 basispoints (.013 percentage points), while an ad-ditional dimension (service) of multiplexitylowers a firm’s interest rate by 4.2 basispoints (.042 percentage points). In this mar-ket, such interest rate reductions are signifi-cant and underscore the importance of rela-tional embeddedness relative to conventionalfinancial and organizational factors that af-fect interest rates.

Inconsistent with Hypothesis 1, neither du-ration of the relationship nor multiplexity ofthe relationship affect the probability of ac-cessing credit. These null effects indicatethat while the quality of a relationship caninfluence the competitiveness of a rate, it isunrelated to whether or not a firm “passes thebar” for credit eligibility. This suggests thatrelationships influence market allocationonce a firm has been deemed creditworthybut does not independently influencewhether a firm is categorized as credit eli-gible. This inference fits with my interviewdata, which indicated that there is a level ofrisk at which banks deny loans, even if theyare close to the client, because no level ofconfidence in the client’s competency canoffset the credit-carrying shortfalls that ema-nate from other business factors. Thus, it ap-pears that long-term and multiplex ties influ-ence the “pricing” decision only if creditwor-thiness has been established by other factors,some of which are financial and others ofwhich are sociological.

Consistent with Hypothesis 3 and 4, net-work complementarity improves a firm’s ac-cess to credit and its cost of capital. In Model1, the linear coefficient is positive and thesquared coefficient is negative. Conversely,in Model 2 the linear coefficient is negativeand the squared coefficient is positive. Thus,as hypothesized, networks composed of an

integrated mix of embedded ties and arm’s-length ties increase access to capital and re-duce capital costs relative to networks com-posed predominantly of arm’s-length ties orembedded ties. These results suggest that thecomplementarity of different types of ties ina network produce optimal benefits relativeto networks that lack complementarity.

This conclusion is strengthened by the ef-fects of network size on credit access andcredit costs. In Model 1, network size in-creases the propensity for credit accessibil-ity. This is consistent with my argument andfield data that indicate that arm’s-length tiesincrease a firm’s knowledge of market inno-vations and the availability of different loansand pricing. Taken at face value, a larger net-work is better than a smaller network for ac-cess to credit. However, if price is a factor,the null effect for network size in Model 2indicates that network size does not reducepricing. This suggests that increases in thenumber of arm’s-length ties to banks can helpfirms shop the market for loan availability butappear to ineffectively motivate banks to in-corporate rivals’ prices or to cut prices on aloan. Consistent with this inference, severalRMs recounted situations in which a firmlinked to them by an arm’s-length tie appliedfor a loan, and because of the RMs’ desire tobe competitive with the firm’s other banksthey offered a loan but at a high, unattractiveprice (Uzzi 1999). Another effect that sup-ports this line of inference emerged from apost hoc analysis of the employment sizes offirms with networks high in complementarity.The analysis showed that firms with inte-grated networks were larger than firms withembedded networks, but did not differ in sizefrom firms with large arm’s-length networks.This further suggests that the effect of net-work structure is not a proxy for firm size,but is an effect of the social organization ofthe ties in the firm’s network. I infer that largenetworks of arm’s-length ties effectively gar-ner public market data but are comparativelyless effective than embedded ties at promot-ing the trust and reciprocity that facilitatedeal-making and innovation. Thus, networkshigh in complementarity seem to provide pre-mium benefits by providing a bridge for inte-grating the public information found in mar-kets with the private resources of particularrelationships.


Figure 1 illustrates the combined effects ofembeddedness in a three-dimensional surfaceplot that was generated holding the other sig-nificant variables in Model 2 at their means.The plot shows the interplay between rela-tional embeddedness, structural embedded-ness, and the magnitudes of three effects re-lated to the cost of capital. First, firms thatmaintain integrated mixed-mode networks(the trough of the curve area) lower their costof capital, an effect that increases linearly asthe duration of the relationship between thebank and firm increases. Moreover, a properlevel of network complementarity lowers thecost of capital below that of either embed-ded-tie networks or arm’s-length networksby up to 3.0 percentage points, or 300 basispoints (the difference between the U-shapedcurve’s right-hand tails and the trough). Interms of dollars and cents, this difference issubstantial: a $100,000 loan at 8.5 percentinterest and a 10-year maturity has a finalcost of $226,098. In contrast, the same loanwith an interest rate of 11.5 percent has a fi-nal cost of $296,994. Thus, a 3-percent or300 basis point difference in the interest rateresults in savings of $70,896 over the life ofthe loan. Second, consistent with my argu-ment, the difference in the right and left tailsof the U-shaped curve indicates that while

neither an embedded-tie network nor anarm’s-length network is as beneficial as amixed-mode network, an embedded networkis better than an arms-length network, atleast for small firms. Third, the estimated in-terest rate does not fall below about 9.1 per-cent, yet the average prime rate over this pe-riod is 8.5 percent. This suggests that banksand clients share the benefits created by so-cial attachments and social networks, yet onaverage this does not result in loan costs be-low prime—the bank’s cost of capital.Rather, the results suggest that embedded-ness prompts the bank and the firm to sharethe potential profits that could be made fromthe loan. A reasonable conjecture is that thesocial capital created by embeddedness inmodern capitalist markets is both closelyaligned with performance and less particular-istic than typically assumed in classical fi-nancial theory.

DISCUSSION

Bankers have an adage, “a relationship isworth a basis point.” This saying reflectstheir belief that value is measured the old-fashioned and unambiguous way—at thecash register—yet belies the real value theycredit to relationships. Using an embedded-

12.5

12.0

11.5

11.0

10.5

10.0

9.5

9.0

12.5

12.0

11.5

11.0

10.5

10.0

9.5

9.00

.2.4

.6

.8

010

2030

40

50

Duration of Tie (in Years) Network Complementarity

Inte

rest

Rat

e on

Loa

n (%

)

EmbeddedNetworks

IntegratedNetworks

Arm’s LengthNetworks

Figure 1. Social Embeddedness and the Firm’s Cost of Financial Capital: U.S. Nonagricultural Firms,1989


ness approach, I have found that commercialtransactions between firms and banks thatare embedded in relationships increasefirms’ access to capital and lower their bor-rowing costs net of other determinants oflending. Ethnographic and statistical evi-dence showed that the more commercialtransactions between a firm and the bank itborrows from are embedded in social attach-ments, the more expectations of trust andreciprocity shape transacting, thereby pro-moting governance benefits and transfers ofprivate resources that are inaccessiblethrough market ties. The governance benefitsof embedded ties make investments inunique solutions to lending problems morepredictable to banks and firms, while thehigh level of private resource transfer pro-vides the substantive material for creatingdistinctive solutions to a firm’s financingneeds. These findings are important becausemost studies of financial market behavior fo-cus on access to public information and theuse of formal governance mechanisms suchas written contracts. Embeddedness is there-fore a conduit to resources and governancearrangements that are difficult to emulatethrough other exchange mechanisms.

These benefits of embeddedness are alsoenhanced when the firm’s network of ties toall its banks—those banks from which itborrows and those at which it conductsother banking business—consists of acomplementary mix of embedded ties andarm’s-length ties. Firms that have networkscomposed of a complementary mix of tiesoptimize the benefits of embeddedness be-cause the characteristics of different typesof ties offset each other’s weaknesses whilepreserving their strengths. In networks highin complementarity, arm’s-length ties effec-tively search for and broker differences inloan structures among banks, while embed-ded ties facilitate the partnering that isneeded to successfully synthesize diversemarket information and unique private re-sources into innovative, low-cost loan struc-tures. In contrast, networks of only arm’s-length ties can effectively broker marketdifferences but lack arrangements that fa-cilitate partnering, while networks of onlyembedded ties promote partnering yet havelimited facilities for brokering resources be-tween disconnected actors. I refer to this

portfolio-like property of a network as net-work complementarity and view it as oneway network structure optimizes an actor’sbrokering and partnering benefits vis-à-visother actors.

Apart from the substantive ramifications ofthe findings, implications for the sociologyof markets and organizations are also evi-dent. In economic sociology, there are twogeneral theories for understanding how so-cial relations and networks create economicand social benefits. The weak-tie approachargues that a large, nonredundant network ofarm’s-length ties is most advantageous; thestrong-tie approach argues that a closed,tightly knit network of embedded ties is mostadvantageous (Sandefur and Laumann 1998).How can these opposing approaches to tiequality (weak versus strong) and networkstructure (“holely” versus dense) be recon-ciled? My analysis of arm’s-length ties sup-ports the weak-tie thesis, demonstrating thatweak ties are superior at “shopping” the mar-ket for publicly available information. Simi-larly, my analysis of embedded ties supportsthe strong-tie approach by showing that em-bedded ties are superior at “plugging” actorsinto the unique collective resources of densenetwork clusters. Thus, I suggest that embed-ded ties and arm’s-length ties are comple-mentary rather than cannibalistic when theyare combined within the same network, be-cause one type of tie helps overcome thelimitations of the other type while enlarginginformation and governance benefits. WhileI have addressed these portfolio-like proper-ties of network complementarity, the conceptshould be further developed and fully con-firmed. Future research should be directed ataccumulating additional evidence along thelines of this study on network complemen-tarity and build on current work that includesstudies of research productivity, learning, in-dustrial change, and direct sales marketing(Etzkowitz, Kemelgor, and Uzzi forthcom-ing; Gabbay 1997; Talmud and Mesch 1997;Uzzi 1996).

The results also suggest that the theoreti-cal distinction between instrumental versusexpressive interests may be moot becauseembeddedness changes actors’ motivesrather than treats them as immutable. WhileRMs may build networks to gain access toprivate information, enacting a relationship


also attenuates the narrow aims that mayhave motivated it originally because suchties tend to create expectations that normallyaccompany noneconomic attachments. Asmy data show, even in a business culture thatuses the yardstick of money to gauge value,bankers and clients develop expressive bondsthat affect their economic decisions. Conse-quently, there is no tradeoff between selfishinterests and an exchange partner’s interestsbecause valuing an exchange partner’s inter-est is appropriate for roles that are linkedthrough embedded ties.

A class of theoretical implications concernthe similarities and differences between theembeddedness approach and other ap-proaches to transacting across organizationalboundaries, particularly transaction cost eco-nomics, which also conceptualizes exchangein terms of information access and gover-nance (Granovetter 1985; Uzzi 1996; 1997).In this paper, I have presented three factorsthat distinguish the ways in which informa-tion and governance problems are treated inthe embeddedness and transaction cost eco-nomic approaches. Two differences concernthe mechanisms by which information andgovernance problems are resolved (i.e., pri-vate information transfer and networks) andone difference concerns the motivations ofactors (i.e., an embedded logic of exchange).

First, the embeddedness and transactioncost approaches differ in their emphases onthe degree to which private information orasset specificity—the main explanatory vari-able of the transaction cost approach (alsosee Blau 1964:160 for an early treatment ofasset-specificity’s effect on exchange)—af-fect the value of relationships. While theconcepts of private information and assetspecificity are not necessarily in conflictwith one another, private information and as-set specificity are not synonymous. I haveshown how private information is unlike as-set-specific information in that private infor-mation need not be relationship-specific ordecrease in value if redeployed in anothertransaction. Rather, it provides value throughrelationships and therefore does not neces-sarily lose value when redeployed in a newrelationship or when used in several ex-change relationships simultaneously. In thissense, private information shares with rela-tionship specific investments the quality of

value creation, but it has fewer restrictionson how widely it can be used.

Second, non-contractual social relation-ships and network structure also play a keyrole in embeddedness by regulating anactor’s ability to broker and partner in ex-changes—features of social structure that aregiven short shrift in the transaction cost eco-nomics focus on dyads.

Third, transaction cost economics holdsthat trust muddies the waters of individualcalculativeness because “the only reliablehuman motive is avariciousness” (William-son 1996:50). As such, the transaction costmodel focuses on the use of formal gover-nance mechanisms such as contracts, fran-chise agreements, or hostage-taking. In con-trast, I have presented numerous examplesand statistical findings that showed how so-cial trust can offer distinctive governancebenefits that translate into lower costs ofcapital and improved access to capital. An-other important consequence of this distinc-tion is that the governance arrangements ofsocial embeddedness appear to come before,rather than follow from, the attributes oftransactions. In this way, embeddedness in-verts the logic of transaction costs by show-ing how self-enforcing governance arrange-ments pave the way for the transfer of pri-vate information and integrative bargainingrather than follow the attributes of transac-tions as purported in transaction cost eco-nomics.

While the triangulation of methods and in-dependent data sources provides more robustinferences than does a single source of data,the cross-sectional properties of these mate-rials suggests that the framework’s confirma-tion requires longitudinal study of the origin,change, and scope of embeddedness in mar-kets (Uzzi 1999). A specific provisional find-ing of this study is that women-managed andminority-managed firms are less likely thanother firms to access credit. One reason forthese discrepancies may be that the “scripts”that white male RMs use to forge ties withwhite male entrepreneurs are “coded” differ-ently by minorities and women (e.g., anevening of dinner and the theatre becomescomparable to a “date”) because relation-ship-building involves contextually definedactivities. These differences may thereforeunintentionally hamper the formation of em-


bedded ties between groups that use alterna-tive scripts. Thus, one tentative conclusion isthat prejudices against an out-group may ex-plain only part of the discrepancies in lend-ing because collaboration among in-groupmembers improves access for in-group mem-bers, even if out-group bias does not exist.Thus, if these provisos are correct they sug-gest that in-group effects may be as impor-tant as out-group effects in explaining mar-ket stratification. They also suggest that thesystems lenders use to select and train RMsin relational practices can improve minori-ties’ access to credit, as well as lenders’ abil-ity to attract the business of undervaluedfirms.

In summary, my key substantive conclu-sion is that social structure stratifies marketoutcomes by influencing both who getscredit and what that credit costs. My broadfinding is that the ability to meet financialselection criteria is a product of a firm’scharacteristics as well as the socially ar-ranged opportunity structures within which itis embedded. Firms with embedded relationsand high network complementarity are morelikely to be deemed credit eligible and to re-ceive lower cost financing. While it has re-cently been recognized that particular net-works can benefit firms in competitive envi-ronments, less attention has been paid to asecond, commensurately important outcomeof networks: that social and network rela-tions also can encourage a firm to put forthextra effort, increasing its future perfor-mance beyond what it would have been had

it been embedded in a more limiting set ofrelations. In this sense, embeddedness canpromote both individual and social welfarein markets in the same way that advantagedsocial positions, independent of personal at-tributes, help actors get ahead and also moti-vate them to achieve. Conversely, disadvan-taged positions stall mobility and reduce as-piration and motivation. Thus, market-mak-ing—or the creation of exchanges for mutualbenefit—depends on social relations and net-works, which are themselves likely to gener-ate premium benefits for firms and econo-mies when they provide a bridge for integrat-ing the public resources of markets with theprivate resources of relationships.

Brian Uzzi is Associate Professor of Organiza-tion Behavior at the Kellogg Graduate School ofManagement and Associate Professor of Sociol-ogy at Northwestern University. He is also direc-tor of the Ph.D. program in Sociology and Orga-nizations. In the 1999–2000 academic year, hewill be a Fellow at the Institute for Policy Re-search at Northwestern and then a Visiting Pro-fessor at INSEAD, Fountainbleu France. His re-search interests include the economic sociologyof organizations and markets with a current fo-cus on the role of embeddedness in financial mar-ket making and organizational status processes.Two forthcoming projects are, Athena Unbound:Social Capital and the Advancement of Womenin Science (with Henry Etzkowitz and CarolKemelgor, Cambridge University Press), andEmbedded Organizations: Sociological Concep-tions on Organizational Evolution in ThreeTransitioning Economies (with Rueyling Tzeng,Stanford University Press).

Appendix A. Fieldwork Methodology

I obtained interviewees’ names from each bank’sCEO. The CEO’s name was acquired through theBanking Resource Center, a research institute onbanking located within the Kellogg Graduate Schoolof Management. Banks sampled ranged in size fromsmall community banks (assets < $100,000,000) tohigh-end midmarket banks (assets < $225 billion).

At each bank, I interviewed “relationship manag-ers,” a widely used title that displaced the formertitle of “lending officer” in the early 1980s. The ti-tle “Relationship Manager” (RM) is sociologicallyinteresting in that it connotes the social nature andidentity of this role in banks and the manner inwhich it is enacted with corporate clients. RMs nor-mally attain the rank of vice president, a status thatreflects seniority, success, and decision-makingpower. Of the 24 interviewees, 19 were RMs, three

were CEOs actively involved in lending, and twowere bad-debt collectors. Bad-debt collectors arepresumably more skeptical of social ties than RMsgiven that their typical interactions are with personsand firms that default and defraud on loan agree-ments. Total interview time amounted to 26 hours,and the average number of years of experience ofinterviewees was about 10 years. The average num-ber of firms that each interviewee managed rangedfrom 9 to 50. The gender and race demographics ofthe RM sample approximate the population demo-graphics of the banking industry: five were womenand one was an African American male.

I used Miles and Huberman’s (1994) data collec-tion and analysis methods. I recorded all interviewson tape and then transcribed them to create a behav-ioral record for each interviewee. In some cases,


long narrative passages were broken into stanzas,which consisted of an uninterrupted series of sen-tences on a single topic (e.g., transactions, trust,market-making). In these cases and when narrativeswere reported in their entirety, the lack of fluencythat is typical of spoken English was edited to in-crease comprehension. Questions were open-endedand moderately directive. Questions focused on thenature of the credit decision—especially access tocapital (who qualifies) and the cost of capital. Fol-low-up questions focused on the nature, function,and dynamics of bank-client ties. Thus, there wasan active attempt to use the interviews to discoverinteresting and surprising relationships, rather thanto use them as a proxy for survey data. For exam-ple, some typical questions were: “How does thebank assess the creditworthiness of a corporate bor-rower?” “What types of things do you discuss witha client in order to assess their creditworthiness?”“What do you typically do when you meet clients?”“What is the basis of a good relationship with a cli-ent?” “How do relationships between you and theclient develop?” To probe sensitive issues and avoid

directiveness, responses were postscripted withphrases such as: “Can you tell me more about that?I am interested in those kinds of details,” “Is thereanything else?” “Would you consider this typical oratypical?”

Data analysis was a two-step procedure. First, Iformed an understanding of the patterns in the data.This task centered on a content analysis and frequen-cy count of the interviewees’ data in which their re-sponses were compiled into different factors that de-composed the range of responses (i.e., the variance)into its major components. Second, I worked backand forth between theory and the emerging frame-work. In this step, evidence was added, dropped, orrevised as my working formulation took shape. Mypurpose was to explain how social structure influ-ences economic behavior, which in this context con-siders how relationships and network ties conditiona firm’s access to capital and the cost of capital.Like psychometric and econometric models, this for-mulation aims to accurately illustrate the sources ofvariation in the data rather than explain all the vari-ance.

REFERENCES

Abolafia, Mitchell Y. 1997. Making Markets:Opportunism and Restraint on Wall Street.Boston, MA: Harvard University Press.

Arrow, Kenneth J. 1998. “What Has Economicsto Say about Racial Discrimination.” Journalof Economic Perspectives 12(2):91–100.

Baker, Wayne E. 1990. “Market Networks andCorporate Behavior.” American Journal of So-ciology 6:589–625.

———. 1994. Networking Smart: How to BuildRelationships for Personal and OrganizationalSuccess. New York: McGraw-Hill.

Baker, Wayne E., Robert R. Faulkner, and GeneA. Fisher. 1998. “Hazards of the Market: TheContinuity and Dissolution of Interorganiza-tional Market Relationships.” American Socio-logical Review 63:147–77.

Berger, Allen N., Anil K. Kashyap, and JosephM. Scalise. 1995. “The Transformation of theU.S. Banking Industry: What a Long StrangeTrip It’s Been.” Brookings Papers on Eco-nomic Activity 1:55–218.

Blau, Peter M. 1964. Exchange and Power in So-cial Life. New York: Wiley.

Blau, Peter M. and Joseph E. Schwartz. 1984.Crosscutting Social Circles. New York: Aca-demic.

Bollen, Kenneth A. and Pamela Paxton. 1998.“Detection and Determinants of Bias in Sub-jective Measures.” American Sociological Re-view 63:465–78.

Borgatti, Stephan P. 1997. “Structural Holes: Un-packing Burt’s Redundancy Measures.” Con-nections 20:35–8.

Burt, Ronald S. 1992. Structural Holes: The So-

cial Structure of Competition. Boston, MA:Harvard University Press.

Carruthers, Bruce G. 1996. City of Capital: Poli-tics and Markets in the English FinancialRevolution. Princeton, NJ: Princeton Univer-sity Press.

Cole, Rebel A. and John D. Wolken. 1995. “Fi-nancial Services Used by Small Businesses:Evidence from the 1993 National Survey ofSmall Business Finances.” Federal ReserveBulletin (July):629–67.

Coleman, James S. 1988. “Social Capital in theCreation of Human Capital.” American Jour-nal of Sociology 94(supp.):S95–S120.

Davis, Gerald F. 1991. “Agents without Prin-ciples: The Spread of the Poison Pill throughthe Intercorporate Network.” AdministrativeScience Quarterly 36:583–613.

Davis, Gerald F. and Mark S. Mizruchi. 1999.“The Money Center Cannot Hold: CommercialBanks in the U.S. System of Corporate Gover-nance.” Administrative Science Quarterly44:215–39.

DiMaggio, Paul and Hugh Louch. 1998. “SociallyEmbedded Consumer Transactions: For WhatKinds of Purchases Do People Most Often UseNetworks.” American Sociological Review63:619–37.

Eccles, Robert G. and Dwight B. Crane. 1988.Doing Deals: Investment Banks at Work. Bos-ton, MA: Harvard Business School Press.

Ettin, Edward C. 1994. “The Evolution of theNorth American Banking System.” Preparedfor the Experts’ Meeting on Structural Changesin Financial Markets: Trends and Practices,


Organization for Economic Cooperation andDevelopment, Paris, France.

Etzkowitz, Henry, Carol Kemelgor, and BrianUzzi. Forthcoming. Athena Unbound: SocialCapital and the Advancement of Women in Sci-ence. London, Enngland: Cambridge Univer-sity Press.

Fama, Eugene. 1985. “What’s Different aboutBanks?” Journal of Monetary Economics 15:29–36.

FDIC. 1998. 1998 Bank and Thrift Branch OfficeData Book: Office Deposits and Addresses ofFDIC Insured Institutions—Summary of De-posits Central Region. Federal Deposit Insur-ance Corportation: Washington, DC.

Fligstein, Neil. 1996. “Markets as Politics: A Po-litical-Cultural Approach to Market Institu-tions.” American Sociological Review 61:656–73.

Fukuyama, Francis. 1995. Trust, Social Virtuesand the Creation of Prosperity. New York:Free Press.

Gabbay, Shaul. 1997. Social Capital in the Cre-ation of Financial Capital: The Case of Net-work Marketing. Champaign, IL: Stipes.

Gorton, Gary and Richard Rosen. 1995. “Corpo-rate Control, Portfolio Choice, and the Declineof Banking.” Journal of Finance 50:1377–420.

Granovetter, Mark. 1973. “The Strength of WeakTies.” American Journal of Sociology 78:1360–80.

———. 1985. “Economic Action and SocialStructure: The Problem of Embeddedness.”American Journal of Sociology 91:481–510.

Gulati, Ranjay. 1994. “Does Familiarity BreedTrust? The Implications of Repeated Ties forContractual Choice in Alliances.” Academy ofManagement Journal 38(4):85–112.

Haunschild, Pamela R. 1994. “How Much is ThatCompany Worth? Interorganizational Relation-ships, Uncertainty, and Acquisition Premi-ums.” Administrative Science Quarterly 39:391–411.

Hawley, Clifford B. and Edwin T. Fujii. 1991.“Discrimination in Consumer Credit Markets.”Eastern Economic Journal 17(1):21–30.

Heckman, James. 1976. “The Common Structureof Statistical Models of Truncation, SampleSelection, and Limited Dependent Variablesand a Simple Estimator for Such Models.” TheAnnals of Economic and Social Measurement5:475–92.

Hirschman, Albert O. 1977. The Passions and theInterests: Political Arguments for Capitalismbefore its Triumph. Princeton, NJ: PrincetonUniversity Press.

———. 1982. “Rival Interpretations of MarketSociety: Civilizing, Destructive, or Feeble?”Journal of Economic Literature 20:1463–84.

Iacobucci, Dawn and Amy Ostrom. 1996. “Com-

mercial and Interpersonal Relationships: Usingthe Structure of Interpersonal Relationships toUnderstand Individual-to-Individual, Indi-vidual-to-Firm, and Firm-to-Firm Relation-ships in Commerce.” International Journal ofResearch in Marketing 13:53–72.

King, Gary, Robert O. Keohane, and SidneyVerba. 1994. Designing Social Inquiry: Scien-tific Inference in Qualitative Research.Princeton, NJ: Princeton University Press.

Kollock, Peter. 1994. “The Emergence of Ex-change Structures: An Experimental Study ofUncertainty, Commitment, and Trust.” Ameri-can Journal of Sociology 100:313–45.

Kunda, Ziva. 1990. “The Case for Motivated Rea-soning.” Psychological Bulletin 108:480–98.

Larson, Andrea. 1992. “Network Dyads in Entre-preneurial Settings: A Study of the Governanceof Exchange Processes.” Administrative Sci-ence Quarterly 37:76–104.

Lazerson, Mark. 1995. “A New Phoenix: ModernPutting-Out in the Modena Knitwear Industry.”Administrative Science Quarterly 40:34–59.

Levinthal, Daniel A. and Mark Fichman. 1988.“Dynamics of Interorganizational Attachments:Auditor-Client Relationships.” AdministrativeScience Quarterly 33:345–69.

Lie, John. 1997. “Sociology of Markets.” AnnualReview of Sociology 23:341–60.

Lin, Nan, Walter Ensel, and Jan Vaughn. 1981.“Social Resources and Strength of Ties: Struc-tural Factors in Occupational Status Attain-ment.” American Sociological Review 46:393–405.

Lummer, Scott and John McConnell. 1989. “Fur-ther Evidence on the Bank Lending Processand the Capital Market Response to Bank LoanAgreements.” Journal of Financial Economics25:99–112.

Macneil, Ian. Forthcoming. “Relational ContractTheory: Challenges.” Northwestern UniversityLaw Review.

March, James G. 1994. A Primer on DecisionMaking: How Decisions Happen. New York:Free Press.

Marsden, Peter V. 1981. “Introducing InfluenceProcesses into a System of Collective Deci-sions.” American Journal of Sociology 86:1203–35.

Marsden, Peter V. and Karen E. Campbell. 1984.“Measuring Tie Strength.” Social Forces 63:482–501.

McPherson, Miller J., Pamela A. Popielarz, andSonja Drobnic. 1992. “Social Networks andOrganizational Dynamics.” American Socio-logical Review 57:153–70.

Miles, Matthew B. and Michael Huberman. 1994.Qualitative Data Analysis. Newbury Park, CA:Sage.

Mintz, Beth and Michael Schwartz. 1985. The


Power Structure of American Business. Chi-cago, IL: University of Chicago Press.

Mizruchi, Mark S. and Linda B. Stearns. 1994a.“A Longitudinal Study of Borrowing by LargeAmerican Corporations.” Administrative Sci-ence Quarterly 39:118–40.

———. 1994b. “Money, Banking, and FinancialMarkets.” Pp. 313–41 in Handbook of Eco-nomic Sociology, edited by N. J. Smelser andR. Swedberg. Princeton, NJ: Princeton Univer-sity Press.

Montgomery, James D. 1998. “Toward a Role-Theoretic Conception of Embeddedness.”American Journal of Sociology 104:92–125.

Munnell, Alicia H., Lynn E. Browne, JamesMcEneaney, and Geoffrey M. B. Tootell. 1992.“Mortgage Lending in Boston: InterpretingHMDA Data.” Working Paper 92-7, FederalReserve Bank of Boston, Boston, MA.

Padgett, John F. and Christopher K. Ansell. 1993.“Robust Action and the Rise of the Medici,1400–1434.” American Journal of Sociology98:1259–319.

Petersen, Mitchell A. and Raghuram G. Rajan.1994. “The Benefits of Lending Relationships:Evidence from Small Business Data.” TheJournal of Finance 49(1):3–37.

———. 1995. “The Effect of Credit Market Com-petition on Lending Relationships.” QuarterlyJournal of Economics 110:407–43.

Podolny, Joel M. 1993. “A Status-Based Modelof Market Competition.” American Journal ofSociology 98:829–72.

Portes, Alejandro and Julia Sensenbrenner. 1993.“Embeddedness and Immigration: Notes on theSocial Determinants of Economic Action.”American Journal of Sociology 98:1320–50.

Powell, Walter W. 1990. “Neither Market nor Hi-erarchy: Network Forms of Organization.” Pp.295–336 in Research in Organizational Behav-ior, edited by B. Staw and L. L. Cummings.Greenwich, CT: JAI Press.

Powell, Walter W., Kenneth W. Koput, and Lau-rel Smith-Doerr. 1996. “InterorganizationalCollaboration and the Locus of Innovation:Networks of Learning in Biotechnology.” Ad-ministrative Science Quarterly 41:116–45.

Romo, Frank P. and Michael Schwartz. 1995.“Structural Embeddedness of Business Deci-sions: A Sociological Assessment of the Migra-

tion Behavior of Plants in New York State Be-tween 1960 and 1985.” American SociologicalReview 60:874–907.

Sandefur, Rebecca L. and Edward O. Laumann.1998. “A Paradigm for Social Capital.” Ratio-nality and Society 10:481–501.

Seabright, Mark A., Daniel A. Levinthal, andMark Fichman. 1992. “Role of Individual At-tachments in the Dissolution of Interorgan-izational Relationships.” The Academy of Man-agement Journal 35(1):122–60.

Smelser, Neil J. and Richard Swedberg. 1994.“The Sociological Perspective on the Economy.”Pp. 3–26 in The Handbook of Economic Sociol-ogy, edited by N. J. Smelser and R. Swedberg.Princeton, NJ: Princeton University Press.

Spulber, Daniel F. 1998. The Market Makers.New York: McGraw-Hill.

Talmud, Ilan and Gustavo Mesch. 1997. “MarketEmbeddedness and Corporate Instability: TheEconology of Inter-Industrial Networks.” So-cial Science Research 26:419–41.

Uzzi, Brian. 1996. “The Sources and Conse-quences of Embeddedness for the EconomicPerformance of Organizations.” American So-ciological Review 61:674–98.

———. 1997. “Social Structure and Competitionin Interfirm Networks: The Paradox ofEmbeddedness.” Administrative Science Quar-terly 42:35–67.

———. 1999. “Governance Benefits through Em-bedded Ties and Network Complementarity:The Case of Banks Making Corporate Loans.”Working Paper 99-15. Institute for PolicyAnalysis, Northwestern University, Evanston,IL.

Uzzi, Brian and James J. Gillespie. 1999. “Inter-firm Ties and the Organization of the Firm’sCapital Structure in the Middle Financial Mar-ket.” Pp. 107–26 in Research in the Sociologyof Organizations, edited by D. Knoke and S.Andrews. Stanford, CT: JAI Press.

Weber, Max. [1922] 1946. “Class Status andParty.” Pp. 180–95 in From Max Weber: Es-says in Sociology, edited by H. H. Gerth andC. W. Mills. Reprint, New York: Oxford Uni-versity Press.

Williamson, Oliver. 1996. “Economic Organiza-tion: The Case for Candor.” The Academy ofManagement Review 21(1):48–57.

Documents

EMBEDDEDNESS IN THE MAKING OF FINANCIAL CAPITAL: HOW ... · EMBEDDEDNESS IN THE MAKING OF FINANCIAL CAPITAL: HOW SOCIAL RELATIONS AND NETWORKS BENEFIT FIRMS SEEKING FINANCING* Brian