value creation, competition, and performance in ... · value creation, competition, and supplier...

Strategic Management JournalStrat. Mgmt. J. (2010)

Published online EarlyView in Wiley InterScience (www.interscience.wiley.com) DOI: 10.1002/smj.864

Received 25 October 2007; Final revision received 20 April 2010

VALUE CREATION, COMPETITION, ANDPERFORMANCE IN BUYER-SUPPLIERRELATIONSHIPS

OLIVIER CHATAIN*Management Department, The Wharton School, University of Pennsylvania,Philadelphia, Pennsylvania, U.S.A.

The value-based approach to strategy argues that a firm’s ability to capture value depends onthe extent of its added value. In this paper, I empirically test the link between added value andvalue capture using a longitudinal dataset of United Kingdom law firm performance, capabilities,and client relationships. In this setting, competitors relevant for defining a firm’s added valueare those that share a client with the firm. Further, within a client relationship, value creation,and hence added value, can be decomposed in two parts: product-line capability and client-specific scope economies. I find that added value, measured at the level of each buyer-supplierrelationship, is a driver of relationship stability and supplier profitability. This suggests thatsuppliers with similar capabilities might enjoy different economic returns depending on thecomposition of their set of relevant competitors. These findings shed light on the conditionsunder which firms can appropriate returns from their capabilities. They indicate that conceptsfrom cooperative games can be fruitfully applied to empirical studies of firm performance andto the elaboration of insights from the resource-based view of the firm. Copyright 2010 JohnWiley & Sons, Ltd.

INTRODUCTION

Recent research has used formal modeling tech-niques to study how value creation and compe-tition interact to shape firm performance (Bran-denburger and Stuart, 1996, 2007; Lippman andRumelt, 2003; MacDonald and Ryall, 2004). Bran-denburger and Stuart (1996) posited that the con-cept of added value, the increase in total valuecreation when a firm is added to a strategic interac-tion, can be used to gauge ability to capture value.Subsequently, studies applied to specific issues

Keywords: value-based strategies; added value; client-specific scope economies; buyer-supplier relationships;professional service firms*Correspondence to: Olivier Chatain, Management Department,The Wharton School, University of Pennsylvania, 2000 Stein-berg Hall-Dietrich Hall, 3620 Locust Walk, Philadelphia, PA19104-6370, U.S.A. E-mail: [email protected]

have used these formal methods to analyze therole of demand in pursuing sustainable advantage(Adner and Zemsky, 2006), the choice betweengeneralist and specialist strategies (Chatain andZemsky, 2007), and the role of central networkpositions in value capture (Ryall and Sorenson,2007).

This line of inquiry presents an opportunity toelaborate contributions from the resource-basedview of the firm (RBV). Empirical studies in theRBV tradition have shown capabilities to mat-ter to firm performance (e.g., Helfat, 1997; Hen-derson and Cockburn, 1994; Miller and Shamsie,1996), and the contexts in which and mechanismswhereby capabilities create value have begun tobe investigated (e.g., Ethiraj et al., 2005). Yetone question has scarcely been addressed: howdoes competition in the product market influencethe relationship between capabilities and product

Copyright 2010 John Wiley & Sons, Ltd.

O. Chatain

market performance? I argue that the value-basedapproach is ideally suited to explore the intersec-tion of capability heterogeneity and competition inthe product market.

Extant research in the RBV suggests that compe-tition matters to value capture. Competition amongstakeholders, for example, shapes the distributionof the value captured by a firm (Blyler and Coff,2003; Coff, 1999), and competition in factor mar-kets determines the gains that accrue to resourceowners (Barney, 1986). But in product markets, inwhich value is created and captured before beingredistributed to stakeholders and resource owners,empirical studies in the RBV tradition have rarelyaccounted explicitly for competition.

This is due in part to the reliance of theoreticaltreatments of the RBV on a Ricardian model ofmarket competition whereby firms with superiorcapabilities (irrespective of the cost of acquisitionor development) are able to capture value in theproduct market in the face of competition (Peterafand Barney, 2003). This way of modeling compe-tition is conceptually powerful and parsimonious,reducing the need to explicitly model competitionin the product market, and thereby enabling thetheory to focus on other, more distinctive aspects.But it is also somewhat restrictive, presuming themarket to clear at a unique price (adjusted for qual-ity), and implying that all suppliers face preciselythe same competitive pressure.

The value-based approach provides a holistictreatment of how the blending of competition andcapability heterogeneity generates performance dif-ferentials. It retains one of the fundamental insightsof the RBV, that capability heterogeneity is a rootcause of performance differentials in the productmarket, while giving the opportunity to work out-side the domain of the Ricardian model of mar-ket competition and to expand the applicability ofcapability-based analysis of competitive outcomes.

In this paper, I use the value-based frameworkto examine the competitive implications of theexistence of highly client-specific value creation.Client-specific value creation results when, forinstance, the knowledge a supplier acquires abouta client is instrumental to delivering a service toor customizing a product for that client, but is notuseful for serving other clients. I show that whenthe client-specific component of total value cre-ation is high relative to other, non-client-specificcomponents, the set of relevant competitors canbe dramatically reduced. We can thus trace the

competitive pressures faced by a supplier to thesmall number of competitors with a high level ofclient-specific value creation ability. It is to thesesuppliers that a buyer is more likely to threaten toturn to when trying to negotiate better terms withan existing supplier.

My analyses being based on a formal frame-work, I develop empirical hypotheses about thedrivers of supplier performance. I argue that ifclient-specific value creation is important, the setof relevant competitors can be defined with ref-erence to buyers’ existing supplier relationships.The hypotheses focus on two dimensions of valuecreation, expertise advantage and client-specificeconomies of scope, and their link to value captureunder competition. The concept of added value isused to relate value creation and value capture. Theuse of a formal framework supports a tight connec-tion between extant theoretical developments in thevalue-based literature and the empirical applica-tion. The empirical analysis uses fine-grained dataon law firm expertise, client base, and performancein the United Kingdom corporate legal market.

I find that new client needs are much more likelyto be fulfilled by law firms that are already provid-ing other legal services to that client. This suggeststhat client-specific knowledge is a large componentof value creation. I analyze the determinants of twodependent variables linked to value capture: clientrelationship stability and law firm profitability. Ifind that a law firm’s level of expertise relativeto the set of competitors sharing the same clientis significantly related to the stability of the clientrelationship. Moreover, the absolute level of lawfirm expertise (i.e., expertise level vis-a-vis allcompetitors) is not related to relationship stability.I also find that client relationship stability dependsboth on a law firm’s client-specific scope and oncompetitors’ client-specific scope. The analysis oflaw firm profitability reveals similar patterns. Thissuggests that a supplier’s value capture varies withits added value, measured by its level of expertiseand client-specific scope relative to a well-definedset of competitors.

The paper aims to make the following contribu-tions. First, it aims to contribute to our understand-ing of the relationship between industry structureand firm performance. The paper explores the com-petitive implications of the existence of high lev-els of client-specific, relative to non-client-specific,value creation. In the presence of client-specificvalue creation, competition in a market can be less

Copyright 2010 John Wiley & Sons, Ltd. Strat. Mgmt. J. (2010)DOI: 10.1002/smj

Value Creation, Competition, and Supplier Performance

intense than might be suggested by the number ofcompetitors. Even in the presence of a large num-ber of competitors, effective competition for theopportunity to serve a given buyer might originatefrom a small set of relevant competitors, whichimplies that the advantage associated with a firm’scapabilities should be evaluated in relation only torelevant competitors, not necessarily to the overallbest competitors in the market.

Second, these findings speak to the literaturesthat link firm capabilities to performance. The find-ings suggest that an accurate understanding of therelationship between superior capabilities and per-formance may need to rely in some cases on afine-grained understanding of competition in theproduct market. Defining the set of relevant com-petitors can thus be key to understanding howcapabilities affect performance. The RBV uses aRicardian setup to model competition in the prod-uct market (Peteraf and Barney, 2003). This setupassumes that all suppliers can compete equally forall buyers. In this study, however, suppliers can-not compete equally for all buyers in the marketbecause of the large value created by client-specificknowledge. The implication is that similar capabil-ities among suppliers in the same market mightgenerate different economic returns, contrary towhat a Ricardian model of product market compe-tition would suggest. Differences in the sets of rele-vant competitors across firms can thus be a sourceof performance heterogeneity within a market inaddition to firm-level differences in capabilities.This may, in turn, provide suppliers with differentincentives for developing their capabilities.

Moreover, considering a client relationship as aresource in the RBV sense, that is, as a hard-to-imitate or acquire enhancement to value creation,does not fully account for the changes that suchrelationship create in the competitive landscape.To be sure, a supplier’s client relationship andthe accompanying client-specific knowledge giveit an advantage over competitors without suchrelationships. But the existence of the relationshipis changing the competitive landscape for thosecompetitors that have a relationship with the client.Their value capture is now affected by the focalsupplier’s capabilities. This competitive pressureonly concerns a subset of all competitors in themarket and thus cannot be conceptualized withinthe frame of the RBV because of its assumptionof Ricardian competition.

Third, the paper speaks as well to the formalliterature on value creation and value capture byproposing a way to translate its insights into anempirical study. The value-based approach hasthe advantage of jointly considering the effectsof firm heterogeneity and competition on perfor-mance. This study suggests that this theory canbe used successfully to generate empirical predic-tions, buttressing the relevance of the value-basedapproach for theory development in the strategyresearch field.

Fourth, recent empirical (Siggelkow, 2003) andtheoretical (Chatain and Zemsky, 2007) work hassuggested the importance of client-specific knowl-edge and client-specific economies of scope tosupplier performance.1 This paper extends theseideas to show how high client-specific knowl-edge can influence the structure of within-marketcompetition by making it highly localized. Thispaper also contributes to the existing literatureon the performance implications of buyer-supplierrelationships (e.g., Levinthal and Fichman, 1988;Martin, Swaminathan, and Mitchell, 1998) andprovides further empirical evidence of the impor-tance of client-specific economies of scope to valuecreation.

The rest of the paper is organized as follows.I introduce a formal model of value creation bysuppliers of highly customized products or servicesand use the model to derive empirical hypotheses.I then test these hypotheses using data from theUnited Kingdom corporate legal market. The paperconcludes with a discussion of the results.

A FRAMEWORK FOR VALUECREATION AND VALUE CAPTURE INBUYER-SUPPLIER RELATIONSHIPS

The literature on value-based strategies has beenfocused on developing theory using formal model-ing rather than translating the logic from the mod-els into empirical applications. In this section, Iuse a formal framework to bridge the gap betweentheoretical concepts from the value-based approach

1 Economies of scope are usually understood to lower productioncosts, irrespective of the identity of the client (Panzar andWillig, 1981; Teece, 1980). Client-specific economies of scope,in contrast, refer to increases in value creation due to higherwillingness to pay or lower costs, or both, thanks to the bundlingof different products or services.


O. Chatain

and the data that can be used in an empirical anal-ysis. Developing such a framework before formu-lating empirical hypotheses has two benefits, (1) itnecessitates the use of a formal language simi-lar to that used in the theoretical literature, whichfacilitates building on existing work (using, forexample, notation similar to that used in Chatainand Zemsky [2007]), thereby tightening the linkbetween theory and empirical application, and(2) it clarifies the additional assumptions needed tomove from the theoretical model to the empiricalapplication and associated boundary conditions,which is more easily accomplished via a formalframework than verbal argument. Thus, the focusof the formal framework in this paper is not on thedevelopment of a full-fledged theory, but rather onbridging extant formal theory and applied empiri-cal work.

I rely on the concept of added value advancedby Brandenburger and Stuart (1996) to concep-tually combine value creation, competition, andvalue capture. Added value in a competitive inter-action is the amount of value creation that wouldbe lost should a given player withdraw. It also rep-resents the maximum amount of value a player cancapture, and is a good indicator of the competitiona firm faces. Intuitively, the existence of alternative(i.e., substitute) players will tend to diminish andscarcity of alternative players will tend to enhancethe ability to capture value. The concept of addedvalue is a straightforward way to articulate thisidea.

For empirical applications, two ingredients areneeded to compute proxies of a supplier’s addedvalue, (1) an estimate of its ability to create value,and (2) the set of relevant competitors and theirability to create value. I review the components ofvalue creation in buyer-supplier relationships andthen incorporate these elements into a model ofvalue creation that I use to delineate, on the basisof criteria determined by the formal analysis, theset of competitors that matter to determining addedvalue.

Components of value creation inbuyer-supplier relationships

The three components of interest in value creationin buyer-supplier relationships are service or prod-uct line capability, client-specific knowledge, andclient-specific economies of scope. Product linecapability is the baseline ability of a supplier of

services or products to deliver value to a cus-tomer absent client-specific knowledge and client-specific economies of scope. It represents a sup-plier’s level of general capability. Client-specificknowledge augments product line capability byenabling suppliers to tailor products or services toa buyer’s particular needs and idiosyncrasies. Sup-pliers can increase this knowledge by using pub-licly available information, but direct interactionwith the buyer provides a unique channel of acqui-sition. Direct interaction is relevant for acquiringan understanding of confidential, or simply unad-vertised, aspects of a client’s business. In the con-text of legal advice, this might include knowledgeof the details of important contracts with third par-ties, and in a manufacturing context, knowledgeof distinctive production operations. Also relevantto project execution is practical knowledge thatcan guide interaction with a buyer’s staff. Anotherenabler of the creation of additional value for buy-ers is the provision of ‘one-stop-shopping,’ whichaffords suppliers two benefits: (1) client-specificknowledge acquired through work in one area canbe shared across other areas, and (2) managingmultiple projects for the same buyer presents anopportunity to create value through coordination.

I term the cost savings and additional ben-efits that accrue to providing services to thesame buyer in different areas ‘client-specific scopeeconomies’ (Chatain and Zemsky, 2007). Client-specific economies of scope differ from economiesof scope as usually construed (Panzar and Willig,1981) in that they arise only when different prod-ucts are made for, or services are delivered to,the same buyer. Production-side economies ofscope, in contrast, arise when different goodsare produced by the same supplier regardless ofwhether they are sold to the same buyer. Client-specific scope economies are akin to the demand-side economies of scope analyzed in Siggelkow’s(2003) study of mutual fund performance. Thedefinition I use, which follows Chatain and Zem-sky (2007), is broader than Siggelkow’s (2003) inadmitting the possibility of benefits to the relation-ship beyond the sharing of shopping costs (Klem-perer, 1992), that is, the costs of using more thanone supplier.

Value creation and added value

In this section, I present a simple model that drawson Chatain and Zemsky’s (2007) model of value



capture in buyer-supplier relationships. I eschewan examination of a fully dynamic model in whichagents consider the intertemporal implications oftheir decisions in favor of a simpler model that isconsistent with the idea that agents are engagedin optimizing, albeit myopic, behavior (Levinthal,2008).2 The value created by supplier i for buyerj in the set of areas Aij can be broken downinto the three elements mentioned above, the valuecreated by (1) service or product line capability(E), (2) client-specific knowledge (R), and (3) thecoordination of different services, which is a func-tion of the set of areas (C(Aij )). To this I add anextra term νij , which represents a shock to valuecreation that is idiosyncratic to the area and buyer-supplier relationship. I assume it to have a meanof zero and to be identically and independentlydistributed. The term νij allows for unobserved het-erogeneity in a client’s preferences around a meangiven by the observed elements of value creation(E, R, and C). In this manner, deviations of clienttastes from a market consensus given by E, R, andC can be modeled.

The value created by supplier i for buyer j isthus, with a denoting a legal area:

Vij =∑

a∈ Aij

(Eia + Rija + C(Aij ) + νija

)

To calculate the added value of supplier i, weneed to compare its value creation ability to thatof the best available alternative supplier. Denotingby −i the set of suppliers that are not supplier I

gives:

AVij = max(0, Vij − max−i V−ij )

In other words, the added value of supplier i

in its relationship with buyer j is the differencebetween its value creation and the value creationability of the next best alternative if the difference

2 I leave aside in this paper the very interesting question of pos-sible collusion among suppliers. Collusive behavior could befacilitated by repeated interaction between suppliers over timeacross common clients in a mechanism similar to that of mul-timarket contacts (Gimeno and Woo, 1996b). Formally incorpo-rating these aspects in the formal framework would substantiallycomplicate the analysis. Unreported empirical analyses yieldedambiguous results regarding the impact of multimarket contactson value capture but also showed that the other empirical resultswere robust to their inclusion in the analysis. I thank an anony-mous referee for pointing to this potential extension that warrantsfurther study.

is greater than zero, and equal to zero otherwise.A supplier has positive added value only if itcan create more value than any of the availablealternative suppliers. It can also be readily seenthat added value weakly increases in the valuecreated by the supplier, and weakly decreases withthe maximum value created by competitors.

Set of relevant competitors

We can build on this formalization to refine the setof relevant competitors, defined as the set of com-petitors that can affect a supplier’s added value.Narrowing this set can enable sharper predictionsof the impact of competition on value capture. I usethe client-specific knowledge element of knowl-edge creation to characterize the set of relevantcompetitors.

It is reasonable to assume current suppliers tobe more knowledgeable than other suppliers abouta client. Assuming, for simplicity, that firms in abuyer’s current supplier base can create an excess�R in value on top of a baseline R that is sharedby all suppliers, we have the following proposi-tion3 (all proofs are provided in Appendix 2):

Proposition 1: Denote as S R the supplierscurrently serving client j that are not thefocal supplier, and as S N the suppliers notcurrently serving the client. One can checkwhether a supplier serving the client hasstrictly positive added value by consideringonly suppliers that currently have a relation-ship with the client if and only if:

�R > maxl∈SN

(El + Clj ) − maxk∈SR

(Ek + Ckj).

This proposition implies that if current suppliersare much more knowledgeable than outside sup-pliers about a client, restricting attention to insidesuppliers is enough to evaluate a supplier’s addedvalue. Simply put, if �R is sufficiently high, oneneed not worry about suppliers not currently serv-ing the buyer. This gives the intuitive corollary toProposition 1.

Corollary 2: There exists a threshold, T, suchthat if �R > T one need only consider suppliers

3 Throughout this development, I distinguish between proposi-tions and corollaries, derived within the formal framework, andempirical hypotheses, which will be tested on the data.


O. Chatain

with a current relationship to check whether asupplier has strictly positive added value in itsrelationship with a buyer.

We need a test to evaluate the strength of �Rrelative to the other components of value creation.Such a test can be devised by considering the sit-uation of a buyer who needs to fulfill a new needand can choose between suppliers currently sup-plying it other services or new suppliers. Underthe assumption that the buyer is keeping its exist-ing suppliers, one can show:4

Proposition 3: The higher �R, the more likelya buyer will use a supplier from its currentsupplier base over a supplier outside this basewhen new needs arise.

Proposition 3 is crucial for testing the extentto which the set of relevant competitors can berestricted to suppliers currently serving the buyer.To gauge the effect of incumbent suppliers’ result-ing advantage in value creation, we need to exam-ine a situation in which a buyer can choosebetween incumbent suppliers and suppliers withwhich it has not previously worked, a clear casebeing the fulfillment of a new need. If client-specific knowledge is higher for incumbents, then,holding other observable characteristics relevant tovalue creation such as expertise and client-specificscope, current suppliers should be disproportion-ately likely to be chosen over outsiders, whichsuggests the following empirical hypothesis:

Empirical Hypothesis 1: Incumbent suppliershave a higher probability to be selected to fulfillnew needs, controlling for other factors thatinfluence value creation.

If this hypothesis holds, and the magnitude of theeffect is high, it will make sense to approximatethe true set of potential competitors with the set ofsuppliers that currently serve a given client.

4 Underlying this proposition is the idea that, holding the othersupplier relationships constant, the supplier that maximizes valuecreation in the new area is also maximizing value capture foritself and for the buyer. Hence studying absolute levels of valuecreation is enough to understand the probability of relationshipcreation. When we analyze variations in value capture later inthis section, we will look at relative levels of value creationbecause the amount of value captured depends both on the valuecreated by the focal firm and on the value created by the nextbest alternative.

Added value and value capture

I assume assumptions (A1), (A2), and (A3)—formal definitions are provided in Appendix 1—detailed in Chatain and Zemsky (2007: 554) andStuart (2004) to hold. These assumptions state,respectively, that buyers and suppliers are mem-bers of distinct sets, there are no externalities inconsumption or production, and a supplier’s addedvalue does not increase when other suppliers areadded to the game. The implication that the valuea supplier creates for a buyer is independent ofwhat other buyers and suppliers are doing rulesout capacity constraints. The value a supplier cre-ates for a buyer is also independent of the numberof buyers with which the supplier is dealing. From(A2) (no externalities), it is immediately clear thata supplier’s added value across all buyers is equalto the sum of its added value for each buyer. Thatis, AVi = ∑

j AVij .Under (A1) and (A2), the following proposition

can be demonstrated (Lemma 1 in Stuart, 2004;Proposition 1 in Chatain and Zemsky, 2007).

Proposition 4: Assume that (A1) and (A2) hold.Then:

(i) The core exists.(ii) The allocation of value received by a sup-

plier belongs to the core if and only if theallocation is in [0, AVij].

This result is important because (1) it ensuresthe existence of the core of the cooperative game,5

which means that players can always share thevalue they create together in a way that makes itunprofitable for any group of players to leave thegrand coalition, and (2) it implies that a supplier’sprofit is a direct function of its added value.

Going further, we can apply Brandenburger andStuart’s (2007) biform game framework to relatecore allocation to expected profits. Under assump-tions (A1) and (A2), the relationship between theadded value of supplier i with buyer j (AVij )and the supplier’s expected value capture in thisrelationship (�ij ) is linear (Chatain and Zem-sky, 2007). It depends on the supplier’s subjectiveassessment of its ability to bargain, denoted by the

5 Note that, in general, the core is not guaranteed to exist(Myerson, 1991: 429).



parameter αi , so:

�ij = αiAVij

With assumption (A3), it is possible to furtherinterpret αi as a supplier bargaining power index.Moreover, the following corollary can be demon-strated (Corollary 1 in Chatain and Zemsky, 2007):

Corollary 5: Consider a biform game in whichsuppliers are denoted by i and buyers by j. Withαi = α for all i’s and αj = (1 − α) for all j’s,the expected profits are as follow:

�i = αiAVi

�j = AVj − αi

∑

i

AVij

To link the expected value captured and actualprofit in period t in an econometric model, it isuseful to specify a supplier-specific fixed effect µi

and error term ε that accounts for other unobservedfactors, and allow the added value and error termto vary over time. Thus:

�ijt = αiAVijt + µi + εijt

Implications for buyer-supplier relationshipstability

I now extend the model to formulate propositionsrelated to the stability of buyer-supplier relation-ships. Being able to predict relationship termina-tion is useful when it is not possible to directlyobserve value capture at the level of the relation-ship, as factors that affect value capture might alsoaffect tie termination. This can be accomplished byamending the initial model to account for a formof optimizing, albeit myopic, behavior on the partof the supplier. Specifically, I assume that the sup-plier terminates at the end of period t when itsprofit from the relationship drops below a threshold(normalized to zero) that represents an unmod-eled opportunity cost for the resources the supplierdevotes to providing the product or service.

Proposition 6: The higher a supplier’s addedvalue, the less likely the buyer-supplier relation-ship is to be terminated.

According to the theory, a relative capabilityadvantage (over the other suppliers that belong

to a buyer’s supplier base) should affect a sup-plier’s ability to sustain a profitable relationshipindependent of the absolute level of the supplier’scapability. In other words, relationship terminationcan be explained in terms of whether a supplierenjoys an advantage over the set of relevant com-petitors (in this paper, competitors that work forthe same client), controlling for the position of thesupplier in the broader market.

A buyer-supplier relationship in which eitherparty is unable to capture sufficient value rela-tive to alternative suppliers is more likely to beterminated, the logic being that a supplier thatdoes not capture enough value from a relation-ship is less likely to expend additional effort orrelinquish value to satisfy the client. Conversely,a client might terminate a relationship with a sup-plier that is unwilling to relinquish sufficient valuerelative to competitors. This suggests the followinghypothesis.

Empirical Hypothesis 2: Controlling for abso-lute capability level, relationships of supplierswith a capability advantage over the set of rele-vant competitors are less likely to be terminated.

The same reasoning gives rise to hypothesesabout the effect on relationship termination oftwo other components of added value, the client-specific scope of the focal supplier and of com-petitors.

Empirical Hypothesis 3: The more areas inwhich a supplier works for a client, the lesslikely the relationship is to be terminated.

Empirical Hypothesis 4: The larger the maxi-mum number of areas in which a single com-petitor is supplying services to a buyer, the morelikely the focal buyer’s relationship is to beterminated.

Implications for supplier performance

Finally, taking advantage of (A2), it is possibleto make a clear statement about how drivers ofprofitability at the level of the buyer translate intoaggregated supplier profits.

Proposition 7: Profitability at the supplier levelis positively related to the weighted average


O. Chatain

of added value measured at the dyadic buyer-supplier level.

The logic for aggregation is the following: if fac-tors that enhance value capture in a single relation-ship become more prevalent in a supplier’s clientbase, overall supplier profitability should increaseto the extent that there are no externalities betweenthe individual buyer-supplier relationships.6 Con-versely, if these factors become less prevalent in asupplier’s client base, overall supplier profitabilityshould decrease. We can first express this with ahypothesis about relative capability advantage:

Empirical Hypothesis 5: Controlling for abso-lute capability level, the greater a supplier’s rel-ative capability advantage across its client base,the higher the supplier’s overall profitability.

One can apply the same reasoning to the effectof client-specific economies of scope onprofitability.

Empirical Hypothesis 6: The more lines of ser-vice a supplier sells to clients, the higher thesupplier’s overall profitability.

Empirical Hypothesis 7: The more lines of ser-vice a supplier’s competitors sell to the sup-plier’s clients, the lower the supplier’s overallprofitability.

In summary, I have developed a model of valuecreation in buyer-supplier relationships that sug-gests that the set of potential competitors thatdetermines a supplier’s added value can, undersome circumstances, be restricted to the set ofsuppliers that currently serve the same buyer.From the model, one can infer an empirical testfor the strength of client-specific knowledge andperformance implications for suppliers relative torelationship termination and overall profitability.This model, and the empirical hypotheses that arederived from it, are combining features that arestandard in value-based applications (e.g., the roleof expertise) and others that are more specific tothis context, especially value from client-specificknowledge and client-specific scope economies.

6 In the formal framework, such externalities are ruled out byassumption A2.

The benefits from using the model extendbeyond the tight link beyond theory and empirics.Most importantly, the analysis of the model bringsto the fore that client-specific knowledge has aqualitatively different impact on value capture thanother sources of value creation. For instance, dif-ferences in levels of expertise between two firmsonly matters for performance if both firms havea sufficiently high level of client-specific knowl-edge. Without the analysis of this model, it wouldhave been tempting to enter product-line expertiseand client relationships as two separate variablesin a regression analysis of supplier performancealthough differences in product-line expertise mat-ter only if a relationship exists in the first place.The more complex interplay between the variablessuggested by the model would not have been con-sidered. Another example derives from the value-based approach insistence on paying attention athow much value creates the best available alterna-tive available to a player. This suggests construct-ing variables with the maximum expertise andscope across competitors rather than, for instance,looking at the average across competitors. Onevirtue of the model is thus to provide a clear struc-ture that empirical analysis can lean upon, suggest-ing in some cases how to compute theoreticallyrelevant independent variables.

The value of the model also manifests itselfin the systematic consideration of competitors’value creation as a factor affecting value capture.For instance, competitors’ client-specific scope istreated symmetrically with a firm’s own scope asfar as value capture is concerned. An analysisderived exclusively from verbal theory may haveonly looked at the focal firm’s scope and levelsof expertise without explicitly measuring competi-tors’ capabilities.

Finally, laying out the model allowed uncover-ing the many ancillary assumptions that need tobe made to lead to the hypotheses, which is typi-cally less transparent with purely verbal theorizing.For instance, that assumption (A2) matters for bothlinking added value to value appropriation and foraggregating supplier profits across buyers is notintuitive. Extending the model to include relation-ship termination shows that adding behaviorallysensible assumptions about buyer decision makingconserves the insights from the basic value-basedframework. All this contributes to making explicitthe boundary conditions of the theory, which is



valuable in addition to the content of the empiricalhypotheses.

DATA OVERVIEW

Empirical setting and levels of analysis

The theory is tested with the United Kingdomlegal market using data on client-law firm rela-tionships and law firm profitability for the period2002–2005. Suppliers are drawn from the list ofthe 100 largest United Kingdom, and 30 largestLondon offices of major U.S. corporate law firmsincluded in the annual survey conducted by indus-try monthly Legal Business. Buyers are among theUnited Kingdom’s large, publicly traded corpora-tions (the top 250 largest market capitalizations)as surveyed by Chambers and Partners in ClientReport. The data thus includes the major playerson two sides of a large business-to-business mar-ket of high value services, and, being longitudinal,makes it possible to control, as necessary, for time-invariant, unobserved heterogeneity.

The setting is an appropriate one in which totest the theory because buyer-supplier relation-ships matter to the players involved. Services canbe customized to, and depend on knowing, thebuyers, not just the business, and buyers and sup-pliers alike are knowledgeable about the services.Articles in trade publications and interviews withactors of the market suggest that buyers are will-ing and able to negotiate the fees they pay theirlawyers, and their general counsels are usuallylawyers, sometimes with previous experience inlaw firms. Suppliers take great care to nurture andmanage their relationships with buyers, and buy-ers, in turn, to manage their supplier base so as toavoid excessive dependence on a single supplier.

A major empirical matter is to find appropri-ate proxies for the value capture construct thatis central to the theory. Ideal data would includethe profitability of each buyer-supplier relation-ship, but collecting such fine-grained, highly con-fidential data across so many players (more thana hundred suppliers) was impractical. The longi-tudinal nature of the data, however, enables thestudy of tie termination as an alternative depen-dent variable that, as argued above in the formalframework, is related to relationship profitability.The theory also has implications for the overallprofitability of each supplier, which is by defi-nition the sum of the profitability of each client

relationship. Overall supplier profitability figuresbeing available, analysis is also conducted at thislevel, and the results triangulated with those of therelationship-level analyses.

Analyses are conducted at three levels. The first,and finest, level of analysis is supplier-buyer-area.Consider, for example, a datum that records thatthe law firm Addleshaw Goddard supplies BritishAirways with advice on employment law. At thislevel of analysis, I examine whether buyers exhibita preference for hiring current suppliers whennew needs arise. This gives an indication of thestrength of buyer-supplier relationships, which, inturn, provides a test for the relevance of focusingon competition and added value among suppliersthat currently have a relationship with a buyer.

The second level of analysis is the entiresupplier-buyer relationship; for example, Addle-shaw Goddard working for British Airways acrossseveral areas. Here, I analyze the determinants ofrelationship termination, which, as argued above,can be interpreted as a proxy for insufficient valuecapture.

The third level of analysis is the supplier and itsoverall profitability. Aggregation from profitabilityat the relationship level to profitability at the sup-plier level is key to this analysis. The data beinglongitudinal, it is possible to relate variation overtime in a supplier’s competitive position across itsrelationships to its ability to capture value for itsshareholders. In the corporate legal market, a lawfirm that sees an increase in the proportion of rela-tionships in which it has an advantage relative torelevant competitors should, as a whole, be moreprofitable and its partners capture more value.

Data sources and main features

I now present the characteristics of the data thatare common to all levels of analysis.

Buyer-supplier relationships. Annual surveysconducted by Chambers Client Report, an industrytrade magazine published quarterly and targeted atgeneral counsels, is the source of data on relation-ships. The survey queries general counsels of firmsincluded among the top 250 largest market capital-izations in the London Stock Exchange about theirmain legal advisers and the legal areas in whichthey are used. The response rate exceeds 90 per-cent, and the set of firms represents in excess of


O. Chatain

80 percent of the total market capitalization of theLondon Stock Exchange.

Supplier expertise—absolute. To assess the levelof expertise of law firms in all legal areas, I use rat-ings provided in Chambers and Partners (the pub-lisher of Chambers Client Report) annual guidesChambers UK: A Client’s Guide to the UK LegalProfession. The guide covers more than 60 areasof law divided into subcategories, and lists rec-ommended corporate law firms in as many as sixtiers, within each of which firms are deemed to beof comparable expertise level.7 This expertise levelcan be construed as a measure of the market con-sensus regarding a supplier’s expertise.8 Withoutfurther transformation, this information represents,by comparing it to that of the best supplier inthe market irrespective of whether both suppliershave relationships with the same buyers, a sup-plier’s absolute level of expertise. Rankings fromthe immediate preceding year are used throughoutthis paper so that they reflect the same informationthat was available to the players at the momentof the interaction and alleviate issues of reversecausality.

Supplier expertise—relative to that of the set ofrelevant competitors. I also constructed measuresof supplier expertise relative to that possessedby the competitors that comprise the narrowerset that constitutes a credible threat given currentrelationships. I argue below that in this market thevalue of client-specific knowledge acquired frompast interactions is so high that the set of relevantcompetitors can be approximated by the set ofsuppliers that currently serves the buyer.

Client-specific scope. A supplier’s ability to reapclient-specific scope economies can be approxi-mated by the extent of the scope of the activitiesit performs for a given client. In practice, this canbe measured by counting the number of areas in

7 The rankings reflect, according to Chambers and Partners Website (http://www.chambersandpartners.com/Rankings-Ex-plained), the law firm’s ‘technical legal ability, professionalconduct, client service, commercial astuteness, diligence, com-mitment, and other qualities most valued by the client.’ (accessed19 April 2010).8 Individual client may differ from this consensus and the formalframework incorporates this possibility by introducing an errorterm of mean zero in the value creation function. In the empiricalanalysis, deviations from the consensus are captured in the errorterm of the regressions.

which a supplier is providing services to a client,or by a dummy variable that indicates whether asupplier is selling services to a client in more thanone area.

Client-specific scope of competitors. The extentof client-specific economies of scope generatedby competitors, which matters to the extent thatit provides a client with strong alternatives, ismeasured as the maximum number of areas inwhich a given competitor is selling services to agiven client.

Other control variables. I also account for otherimportant drivers of performance. In most analy-ses, I control for supplier size (headcount, numberof clients in the Financial Times Stock Exchange[FTSE] 250) and geographical scope (in Londonand in the English provinces).

SELECTION OF SUPPLIERS FOR NEWAREAS OF WORK

The first set of empirical analyses examines thedeterminants of buyers’ choice of supplier whennew needs arise. Suppliers that currently serve abuyer’s other needs would be more likely to bechosen if being part of the supplier base providesan advantage for value creation.

Dataset construction and variables

The main source of data is the aforementionedChambers and Partners Client Report annual sur-veys, which match observations of buyers to sup-pliers at the level of area of legal expertise. Icompare the areas of legal advice buyers list fromyear to year over four years of observation, iden-tifying when a buyer first procured services in anarea of legal expertise not previously required aswell as the supplier selected to fill this new need.The unit of analysis is the client-area-supplier. Foreach instance, I created a risk set consisting of allpotential suppliers for which I had a minimum offirm-level information (i.e., size and scope). Theannual survey having been conducted each yearfrom 2002 to 2005, I was able to construct risk setsfor the years 2003 to 2005. The sample comprises38,293 yearly spells and 1,806 new area creationevents.



Buyer choice of supplier for newly needed area(dependent variable). The dependent variable wascoded 1 for the supplier selected by a buyer to filla new need, and 0 otherwise.

Supplier worked for buyer in the previous year.The main independent variable of interest is adummy variable coded 1 for suppliers that workedfor the buyer during the previous year, and 0otherwise.

Expertise in the area of interest. I controlled forabsolute level of expertise in the area of interestby transforming the Chambers and Partners guideranking into a continuous variable that takes avalue of 1 if the supplier is in the top level and 0 ifthe supplier is not ranked. I also created a dummyvariable to account for the case of a supplier notbeing ranked.

Client-specific scope. The theory suggests that,because it might be affected by adding a new area,the level of client-specific scope economies mattersto the choice of supplier.

Competitor scope. I included the maximum num-ber of areas in which services were sold by a singlecompetitor to capture the potential influence ofcompetitor entrenchment on the selection process.

Other control variables. I controlled for suppliersize (measured as headcount), supplier scope (mea-sured as the number of areas in which a ranking isgiven in the Chambers and Partners guide), numberof clients in the FTSE 250, and whether a clienthad changed its general counsel. I also includedyear dummies.

Methods and results

The dependent variable being categorical (1 forbeing selected, 0 for not being selected), I useda logistic regression. To address the possibilityof heteroskedasticity and autocorrelation amongobservations pertaining to the same supplier, robuststandard errors, clustered by supplier, were esti-mated (Froot, 1989; Rogers, 1993; White, 1980).9

9 Other methods used as robustness checks included a fixed-effect (conditional) logit with groups defined at the supplierlevel to control for time-invariant unobserved supplier-levelheterogeneity, and a rare event logit (King and Zeng, 2001),

Summary statistics and pairwise correlations arepresented in Table 1, results of the regression anal-ysis in Table 2. Model 1 of Table 2 presents theregression results with only the controls and thevariables for level of expertise. Models 2 and3 introduce separately the scope variables anddummy for already being a supplier. Model 4 isthe full model. The table shows the client-specificscope of a firm and its competitors to significantlyaffect the probability of selection. The dummy forsuppliers already in the supplier base has a posi-tive sign and is highly significant (p < 0.001). Asexpected, the variable for expertise level also has apositive and highly significant sign. Hypothesis 1,that suppliers with a current relationship and higherexpertise levels are more likely to be selected, isthus supported. Other, unreported regressions thatinclude a dummy for whether the general counselhas been changed and supplier-level measures ofcapability yield similar results.

To get a sense of the magnitude of these effects,I computed the marginal effects for selected vari-ables at their median value. The results of thesecalculations are reported in Table 3, panel A. Thebaseline probability of selection is 0.0079; beingin the client’s supplier base adds 0.2544 to this. Bycomparison, an extra standard deviation in exper-tise level adds only 0.0052 to the probability ofselection. Panel B presents predicted probabilityof selection under various scenarios. For instance,a median insider has a probability selection of0.3464. These analyses suggest that suppliers thatbelong to a buyer’s current supplier base are muchmore likely than outsiders to be selected to fillneeds for new services, controlling for other rele-vant factors such as current levels of general exper-tise and client-specific scope economies.

The results are suggestive about how to thinkabout competition in this market and how it affectsvalue capture. There are two sources of compe-tition for suppliers already in the supplier base.The most direct competition is from other sup-pliers that currently have a relationship with thebuyer, less direct competition from other suppliers,a theme further developed in the remainder of theempirical analysis. Variables relevant to competi-tion are computed with respect to two referencegroups, that of the whole market (consisting of

there being few creation events relative to the size of the riskset. Both methods yielded results similar to those presented inthe text.


O. Chatain

Tabl

e1.

Des

crip

tive

stat

istic

san

dco

rrel

atio

nsfo

ran

alys

isof

new

area

sele

ctio

n

Mea

nS.

D.

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

(10)

(11)

(12)

(13)

(14)

(15)

(1)

Are

ase

lect

ed0.

030.

181.

00(2

)E

xper

tise

inar

ea0.

600.

260.

121.

00(3

)N

otra

nked

inar

ea0.

030.

16−0

.02

−0.3

71.

00(4

)L

ever

age

ratio

6.04

2.38

0.05

0.09

−0.0

21.

00(5

)H

eadc

ount

ofsu

pplie

r66

6.4

718.

60.

130.

25−0

.06

0.26

1.00

(6)

No.

area

sL

ondo

n12

.53

11.2

20.

110.

07−0

.05

−0.0

60.

701.

00(7

)N

o.ar

eas

prov

ince

s9.

4311

.18

0.05

0.27

−0.0

40.

360.

08−0

.37

1.00

(8)

At

leas

ton

ecl

ient

inFT

SE25

00.

880.

330.

070.

14−0

.03

0.06

0.26

0.32

0.13

1.00

(9)

No.

clie

nts

inFT

SE25

012

.59

14.7

90.

170.

26−0

.07

0.18

0.81

0.68

0.11

0.31

1.00

(10)

Yea

r20

030.

370.

480.

020.

03−0

.01

0.05

0.02

0.01

0.05

0.06

−0.0

21.

00(1

1)Y

ear

2004

0.40

0.49

0.00

−0.1

00.

070.

06−0

.01

−0.0

5−0

.01

−0.0

6−0

.02

−0.6

21.

00(1

2)Y

ear

2005

0.23

0.42

−0.0

10.

08−0

.07

−0.1

2−0

.02

0.05

−0.0

50.

010.

04−0

.42

−0.4

51.

00(1

3)Su

pplie

rw

orke

dfo

rbu

yer

the

prev

ious

year

0.05

0.21

0.48

0.08

−0.0

20.

040.

200.

170.

040.

080.

260.

000.

02−0

.02

1.00

(14)

No.

area

sso

ldto

clie

ntby

foca

lsu

pplie

r0.

070.

360.

430.

07−0

.01

0.05

0.16

0.14

0.04

0.07

0.23

−0.0

20.

020.

000.

881.

00

(15)

Max

.no

.ar

eas

sold

byco

mpe

titor

1.42

1.33

−0.0

6−0

.03

0.01

−0.0

10.

000.

00−0

.02

−0.0

20.

00−0

.27

0.22

0.05

0.07

0.09

1



Table 2. Analysis of new area selection

Model (1) (2) (3) (4)Method Logit Logit Logit Logit

Control variablesNot ranked in area 0.401 0.244 0.494 0.322

[0.387] [0.511] [0.413] [0.483]Leverage ratio 0.042 0.036 0.060∗ 0.064∗

[0.026] [0.029] [0.031] [0.033]Headcount of supplier −0.000∗∗∗ −0.000∗∗∗ −0.000∗∗∗ −0.000∗∗∗

[0.000] [0.000] [0.000] [0.000]No. areas London 0.041∗∗∗ 0.038∗∗∗ 0.035∗∗∗ 0.038∗∗∗

[0.010] [0.010] [0.011] [0.012]No. areas provinces 0.024∗∗∗ 0.021∗∗∗ 0.018∗∗ 0.019∗∗

[0.007] [0.007] [0.008] [0.008]At least one client in FTSE 250 2.358∗∗∗ 2.156∗∗∗ 2.177∗∗∗ 2.076∗∗∗

[0.481] [0.486] [0.494] [0.495]No. clients in FTSE 250 0.035∗∗∗ 0.015∗∗∗ 0.009 0.007

[0.006] [0.005] [0.008] [0.008]Year 2003 −8.391∗∗∗ −7.852∗∗∗ −8.670∗∗∗ −8.139∗∗∗

[0.543] [0.559] [0.556] [0.569]Year 2004 −8.506∗∗∗ −7.683∗∗∗ −8.887∗∗∗ −7.873∗∗∗

[0.535] [0.571] [0.554] [0.584]Year 2005 −8.835∗∗∗ −8.217∗∗∗ −8.933∗∗∗ −8.110∗∗∗

[0.552] [0.582] [0.571] [0.581]

Variables of interestNot ranked in area 0.401 0.244 0.494 0.322

[0.387] [0.511] [0.413] [0.483]Expertise in area 2.156∗∗∗ 2.442∗∗∗ 2.453∗∗∗ 2.482∗∗∗

[0.328] [0.342] [0.349] [0.354]No. areas sold to client by focal supplier 2.255∗∗∗ 0.460∗∗

[0.256] [0.218]Max. no. areas sold by competitor −0.918∗∗∗ −0.983∗∗∗

[0.132] [0.099]Supplier worked for buyer the previous year [H1] 3.532∗∗∗ 3.804∗∗∗

[0.196] [0.238]

Observations 38293 38293 38293 38293Log pseudolikelihood −5038 −3856 −3815 −3468

∗∗∗ p < 0.01, ∗∗ p < 0.05, ∗ p < 0.1; two-tailed tests.Robust standard errors, clustered by supplier, in brackets.

all competitors), and that of the current supplierbase (consisting only of competitors that currentlyserve the client). Subsequent empirical analysesapply this logic at the level of the buyer-supplierrelationship, and of the supplier by considering itsportfolio of client relationships.

TERMINATION OF BUYER-SUPPLIERRELATIONSHIPS

The second set of empirical analyses focuses on thedeterminants of relationship termination. The levelof analysis is the relationship between supplier and

buyer. The theoretical argument for examining thisdependent variable is that relationships that cre-ate less value relative to the others in which asupplier is engaged are more vulnerable to pres-sures from competitors and therefore are morelikely to be terminated. The key to this anal-ysis is to introduce explanatory variables thataccount for the focal firm’s level of value cre-ation relative to that of firms in the set of rel-evant competitors. Based on the analysis in theprevious section, I define the set of relevant com-petitors as those with which a buyer is currentlyworking.


O. Chatain

Table 3. Marginal effects for selection of supplierPanel A: Marginal effects for selected variables

Predicted probability of selection at median values 0.0079

Variable Change in variable Change in predictedprobability

In previous year’s supplier base Zero to one (dummy) 0.2544Expertise level One s.d. around median 0.0052Client-specific scope One s.d. around median 0.0013Competitor client-specific scope One s.d. around median −0.0109

Note: All other variables are held at their median. Calculations are based on estimates of Model 4 in Table 2.

Panel B: Predicted probabilities of selection in selected scenarios

Scenario Predicted probability ofselection

Median outsider 0.0076Most competent outsider (maximum expertise level) 0.0213Median insider 0.3464Least competent insider (minimum client-specific scope and expertise level) 0.1733


I used the Chambers Client Report surveys toidentify in which years buyer-supplier relation-ships were terminated. I constructed 1,878 yearlyspells with 219 termination events. Terminationsare observed in 2003, 2004, and 2005. For eachspell, I created the following variables.

Relationship termination (dependent variable).The dependent variable was coded 0 if a relation-ship was still observed the following year, and 1if it was not.

Number of areas with superior expertise relative tothe set of competitors. The variable is a count ofthe number of areas in which a focal supplier’slevel of expertise is strictly superior to that of anysupplier belonging to the set of relevant competi-tors. This is to reflect the idea that these competi-tors are construed to be alternatives by the client.

Client-specific scope. This is a count of the num-ber of areas in which the focal supplier serves thebuyer.

Competitors’ client-specific scope. This is themaximum client-specific scope among competitorsof the focal firm that serve the buyer.

Control variables. The most important controlvariable is the average expertise rating within theareas a supplier sells to a client. This rating, whichcompares the focal supplier to all other suppliersin the market, measures the supplier’s expertiserelative to that of the widest set of competitorsincluding those outside the set of relevant com-petitors. Variables for leverage and organizationalsize and scope are also included, as in the analysisin the previous section. Table 4 presents the cor-relations and summary statistics for all variables.

Methods and results

Relationships termination being observed annu-ally, a discrete-time transition model (Cameronand Trivedi, 2005: 602) is appropriate. I first esti-mate a logistic model, allowing for a nonpara-metric baseline hazard by using a different inter-cept for each year, and use robust standard errors(Froot, 1989; Rogers, 1993; White, 1980) thatare clustered at the supplier level to correct forpossible heteroskedasticity and autocorrelation.10

10 Because there may be unobserved heterogeneity, making somerelationships intrinsically more or less stable, I also estimateda gamma-distributed discrete time proportional hazard model(Meyer, 1990) (unreported) which produced results comparableto those given by the discrete time logit. I also used a conditionallogit (unreported), with groups constituted with observationsbelonging to a similar firm, again showing very similar results.



Tabl

e4.

Des

crip

tive

stat

istic

san

dco

rrel

atio

nsfo

ran

alys

isof

rela

tions

hip

term

inat

ion

Mea

nS.

D.

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

(10)

(11)

(12)

(13)

(14)

(1)

Clie

nttie

term

inat

ion

0.11

0.32

1.00

(2)

Ave

rage

expe

rtis

ew

ithin

rela

tions

hip

area

s1.

671.

03−0

.04

1.00

(3)

Lev

erag

e6.

012.

170.

03−0

.38

1.00

(4)

Hea

dcou

nt11

40.2

911.

9−0

.05

0.00

0.18

1.00

(5)

No.

area

sin

Lon

don

20.3

712

.07

−0.1

00.

36−0

.05

0.68

1.00

(6)

No.

area

sou

tsid

eL

ondo

n7.

2211

.18

0.08

−0.6

60.

38−0

.12

−0.5

41.

00(7

)N

o.cl

ient

sin

FTSE

250

26.3

618

.31

−0.1

2−0

.01

0.07

0.69

0.67

−0.1

51.

00(8

)Y

ear

2002

0.28

0.45

0.05

−0.1

10.

060.

020.

03−0

.02

−0.0

41.

00(9

)Y

ear

2003

0.35

0.48

−0.0

50.

010.

040.

03−0

.01

−0.0

20.

03−0

.46

1.00

(10)

Yea

r20

040.

370.

480.

000.

09−0

.10

−0.0

6−0

.01

0.03

0.01

−0.4

8−0

.56

1.00

(11)

Lef

t-ce

nsor

ed0.

340.

480.

06−0

.09

0.05

0.02

0.01

−0.0

2−0

.06

0.86

−0.3

3−0

.47

1.00

(12)

No.

area

sw

ithsu

peri

orex

pert

ise

inco

mpe

titor

set

0.77

0.96

−0.1

2−0

.21

0.07

0.16

0.05

0.14

0.22

−0.0

2−0

.03

0.05

−0.0

11.

00

(13)

Clie

nt-s

peci

ficsc

ope

(no.

area

s)1.

620.

95−0

.12

−0.0

40.

03−0

.01

−0.0

10.

030.

03−0

.13

0.00

0.11

−0.1

20.

571.

00

(14)

Com

petit

orcl

ient

-spe

c.sc

ope

(max

.no

.ar

eas)

2.10

1.13

0.02

0.01

−0.0

6−0

.03

−0.0

40.

01−0

.03

−0.1

9−0

.01

0.19

−0.2

10.

060.

271.

00


O. Chatain

Table 5. Determinants of relationship termination

Model (1)Discrete

logit

(2)Discrete

logit

(3)Discrete

logit

(4)Discrete

logit

(5)Discrete

logit

Control variablesAvg. expertise 0.051 −0.011 0.035 0.055 0.004

[0.096] [0.103] [0.096] [0.097] [0.101]Leverage 0.006 −0.007 0.013 0.007 0.004

[0.042] [0.043] [0.041] [0.042] [0.042]Headcount 0.000∗∗ 0.000∗∗∗ 0.000∗∗ 0.000∗∗ 0.000∗∗

[0.000] [0.000] [0.000] [0.000] [0.000]No. areas in London −0.005 −0.008 −0.006 −0.005 −0.006

[0.012] [0.012] [0.012] [0.012] [0.012]No. areas outside London 0.019∗ 0.022∗∗ 0.018∗ 0.020∗ 0.021∗∗

[0.010] [0.010] [0.010] [0.010] [0.010]No. clients in FTSE 250 −0.032∗∗∗ −0.028∗∗∗ −0.032∗∗∗ −0.032∗∗∗ −0.028∗∗∗

[0.007] [0.008] [0.007] [0.007] [0.008]Year 2002 −2.122∗∗∗ −1.811∗∗∗ −1.406∗∗∗ −2.309∗∗∗ −1.665∗∗∗

[0.450] [0.462] [0.475] [0.474] [0.495]Year 2003 −2.187∗∗∗ −1.850∗∗∗ −1.412∗∗∗ −2.373∗∗∗ −1.680∗∗∗

[0.369] [0.386] [0.393] [0.391] [0.416]Year 2004 −1.811∗∗∗ −1.450∗∗∗ −1.009∗∗ −2.017∗∗∗ −1.304∗∗∗

[0.361] [0.377] [0.396] [0.404] [0.432]Left-censored 0.501∗ 0.568∗∗ 0.481∗ 0.532∗ 0.585∗∗

[0.282] [0.285] [0.283] [0.283] [0.288]

Variables of InterestNo. areas with superior expertise in competitor set [H2] −0.532∗∗∗ −0.384∗∗∗

[0.101] [0.124]Client-specific scope (no. areas) [H3] −0.546∗∗∗ −0.438∗∗∗

[0.113] [0.129]Competitor client-spec. scope (max. no. areas) [H4] 0.074 0.139∗∗

[0.064] [0.065]

Observations 1878 1878 1878 1878 1878Log pseudolikelihood −649.4 −636.0 −635.5 −648.8 −628.6

∗∗∗ p < 0.01, ∗∗ p < 0.05, ∗ p < 0.1; two-tailed tests.Robust standard errors, clustered by supplier, in brackets.

Results of the analysis that relates added value andcross-selling to relationship termination are pre-sented in Table 5.

Model 1 includes only control variables. Amongthe most notable result related to control vari-ables is that number of clients is negatively andsignificantly (p < 0.01) related to relationship ter-mination, which suggests that firms that have manyrelationships are particularly good at maintainingthem. In Models 2 to 4, in which the dependentvariables of interest are introduced individually,the coefficients all show the expected signs. Thecoefficients for both number of areas with supe-rior expertise relative to competitor set and client-specific scope are negative and highly significant inModels 2 and 3, respectively. Maximum scope ofcompetitors has the expected positive sign, but is

not significant in Model 4. In Model 5, which treatsall variables of interest, the variables for numberof areas with superior expertise in the competitorset and client-specific scope remain negative andsignificant (p < 0.01), and the coefficient on localcompetitor scope increases, becoming positive andsignificant (p < 0.05). Thus Hypotheses 2, 3, and4 are supported.

To make sense of these estimates, their eco-nomic as well as statistical significance needs tobe considered. The marginal effects of the theoret-ical variables on the predicted probability of rela-tionship termination are shown in Table 6, panelA. The baseline yearly probability of termina-tion, holding variables at the median, is 0.063.This suggests a relatively high level of stability



Table 6. Marginal Effects and Economic SignificancePanel A: Marginal Effects on Probability of Relationship Termination

Baseline for Dependent Variable: Predicted Probability ofRelationship Termination, Holding All Variables at the Median

0.063

Variable Change in Predicted Probability (Changeof One S.D. around Median)

No. Areas with Superior Expertise in Competitor Set −0.021Client-specific Scope (No. Areas) −0.023Competitor Client-spec. Scope (Max. No. Areas) 0.009

Note: All other variables are held at their median. Calculations are based on estimates of Model 5 in Table 5.

Panel B: Marginal Effects on Profits per Equity Partner

Baseline for Dependent Variable : Mean Profit per Equity Partner(’000 GBP, 2005)

397.4

Variable Change in PEP, as Variable Changes byOne Standard Deviation

Expertise Advantage in Relevant Competitor Set 10.44Client-specific Scope 8.90Competitors’ Client-specific Scope −8.45

consistent with the earlier finding that clients preferto use existing suppliers to fill new needs.

A change in the number of areas with an advan-tage in expertise from half a standard deviationbelow the median to half a standard deviationabove the median, a reduction equal to one-thirdof the baseline probability, reduces the probabil-ity of termination by −0.021 points. The impactof a change in the number of areas sold is of thesame order (−0.023). Conversely, an increase ofone standard deviation in the maximum number ofareas sold by competitors increases the probabilityof termination by 0.009 points.

Overall, the results show that variables that mea-sure competition within the defined set of rele-vant competitors (i.e., those that work with thesame buyer), namely, expertise and client-specificeconomies of scope, are both statistically and eco-nomically significant for predicting relationshiptermination. Variables that are not defined withregard to the set of relevant competitors performedless well in predicting tie termination. Absoluteexpertise, for example, does not seem to be linkedto relationship termination, although we foundearlier that it was related to supplier selection.In the next section, instead of individual buyer-supplier relationships we examine portfolios of

such relationships and how their characteristicsaffect buyer profitability.

SUPPLIER PROFITABILITY

In this section, evidence is presented at the sup-plier level that is consistent with the theory andfindings on relationship termination. Results con-sistent with different dependent variables at dif-ferent levels of analysis improve confidence in theempirical findings, and in this empirical setting thedependent variable closest to the construct of valuecapture (partner profit) is observed only at the sup-plier level. We thus trade a purer measure of thedependent variable for a more aggregated set ofindependent variables.


The data sources are law firms operating in theUnited Kingdom legal market in 2002–2005, andfor supplier profitability, the British weeklies LegalBusiness and The Lawyer. According to Proposi-tion 6, supplier profits vary with the weighted aver-age of a supplier’s added value with respect to eachclient. In this empirical analysis, I weight clientsby the log of their assets, and give equal weight toeach area for which a client buys services.


O. Chatain

Profit per equity partner (dependent variable).Because the law firms in the sample are orga-nized as partnerships and distribute their profits(receipts from clients minus costs, including staffand associates’ salaries) among equity partners,the dependent variable is the profit per equitypartner (PEP). PEP is the analogue of return onequity in the context of a professional firm (Mais-ter, 1993).11 Moreover, law firms are managedby their owners, and PEP, rather than total salesor profit, is the figure owner-managers want tomaximize (Muller and Warneryd, 2001). Being,thus, the commonly employed measure of prof-itability within the United Kingdom and UnitedStates legal industries, PEP has been used as adependent variable in a number of academic stud-ies of the performance of law firms in the UnitedStates market (e.g., Kor and Leblebici, 2005), andis taken for granted by legal scholars to be theappropriate measure of profitability (e.g., Gilsonand Mnookin, 1989; Samuelson and Jaffe, 1990;Henderson, 2006).12 The variable is expressed herein thousand British pounds, and nominal mone-tary figures were translated into their year 2005equivalents.

Supplier expertise relative to the set of relevantcompetitors. This variable aggregates the posi-tion of a supplier relative to the set of relevantcompetitors for each of its buyers. I computedfor each supplier the proportion of areas of ser-vice provided to a buyer in which the supplier hadan advantage (defined as a strictly higher capabil-ity ranking) relative to suppliers in the relevantset of competitors. In accordance to the formalframework, the next step was to weight each buyerproportional to the natural logarithm of its assets asa proxy for the amount of business the buyer repre-sented. The final measure can take values between0 (no superior expertise in any area of service for

11 According to Maister (1993: 31), “‘Profit per partner” shouldbe viewed as the professional firm equivalent of “return onequity.” The time and efforts of the partners (who have a claimon the profits of the firm) can be seen as the firm’s equityinvestment (. . .). The total assets employed in the business arethe sum of the (partner) equity investment, and the nonpartnerstaff, whose salaries are comparable to assets financed by debt,at a fixed interest rate.’12 It is possible to use firm profit instead of PEP as a dependentvariable and obtain results comparable to those shown in thissection. However, heteroskedasticity becomes a more pressingissue as the absolute firm profits vary with firm size, whichis very widely distributed, and would require a more specificeconometric treatment.

any given client relative to the relevant set of com-petitors) and 1.13

Client-specific scope. For each law firm I calcu-lated the proportion of clients for which it providedservices in more than one area of law. This vari-able takes the value 0 if the law firm was not listedby a single client in the Chambers Client Reportsurvey, and 1 otherwise.

Competitor’s client-specific scope. For each lawfirm I calculated for the proportion of clients forwhich at least one supplier in the set of relevantcompetitors provided services in more than onearea of law. This variable takes the value 0 if thelaw firm was not listed by a single client in theChambers Client Report survey, and 1 otherwise.

I also used the following control variables in theanalysis.

Supplier expertise (absolute). I controlled foreffects due to variation in a firm’s overall level ofexpertise, as distinct from variation in competitivepressure at the client level, by including a seriesof variables equal to the firm’s average ratings inthe six areas of legal expertise most mentionedby clients in the Chambers Client Report sur-vey, specifically, corporate finance, property, liti-gation, employment, banking, and capital markets.The rankings were reversed to facilitate interpreta-tion: higher figures correspond to higher levels ofexpertise.

Other control variables. Finally, I controlled, asin the previous empirical analyses, for leverageratio, firm size, firm scope, and number of clientsin the FTSE 250.

Methods and results

The 453 firm-year observations from 128 firmsover four years yield an average of 3.54 obser-vations per firm. With cross-sectional and longi-tudinal data, panel regression techniques can beused. I used a fixed effect regression14 with correc-tions for time invariant unobserved heterogeneity,

13 Conducting the analyses with equal weight applied to eacharea yielded similar results.14 A Hausman test indicated that firm effects were likely to becorrelated with the other regressors. A fixed effect estimator isconsistent in this case.



and, to account for heteroskedascity and within-group autocorrelation, robust estimates of the stan-dard errors (White, 1980) clustered by firm (Froot,1989; Rogers, 1993).15

Table 7 presents the descriptive statistics andcorrelation matrix. Inspection of Table 7 showedsome variables to be highly correlated. I checkedfor the presence of multicolinearity and found thatit was not likely to be a problem in this dataset.

Results of the regressions that analyze law firmprofitability are presented in Table 8. I first ranregressions that included only control variables(Model 1), and then introduced the variables ofinterest sequentially before putting them togetherin Model 4. The variable for expertise advantagewithin the set of relevant competitors, introducedin Model 2, has the expected positive coefficient,but is not significant. The variable for client-specific scope, introduced in Model 3, has a posi-tive sign and is statistically significant (p < 0.01).The variable that represents competitors’ cross-selling has the expected negative coefficient, butthe p-value does not cross the five percent thresh-old (p = 0.128).

In Model 4, which brings all the variables ofinterest together, the signs of the coefficient esti-mates are unchanged, but their absolute valuesincrease and their significance is improved. Theexpertise added value variable becomes signifi-cant (p = 0.06) and the client-specific scope vari-able remains significant and below the one percentthreshold (p < 0.01). Moreover, the significance ofthe variable related to competitors’ client-specificscope remains below the 10 percent level (p =0.088). The improved significance in this modelof all three variables could be due to the combi-nation of a positive correlation among themselvesand opposites signs in the regression. For instance,the absence of the competitor cross-selling vari-able may result in underestimating the absolutevalue of other two variables. In sum, Hypothesis 6is strongly supported, as are Hypotheses 5 and 7,albeit more weakly.

To make better sense of these results, it isuseful to consider the economic significance of theestimated coefficients. Panel B of Table 6 showsthe marginal effect calculations. Changes of one

15 An alternative estimation (unreported) with panel correctedstandard errors to account for potential autocorrelation, andincluding fixed effects, produced coefficients of similar signsand comparable magnitude but with narrower standard errors.

standard deviation in the theoretical variables arelinked to changes in the value of the profit perequity partner of between − 8,450 GBP (GreatBritain pound sterling) and + 10,440 GBP. Theseare not trivial sums, although one must keep inmind that the baseline profit per equity partner is395,000 GBP.

Overall, these results are consistent with the the-ory and in line with the results of the analysis ofrelationship termination. The analysis at the sup-plier level is, however, subject to a number oflimitations. First and foremost is the potential formeasurement error arising from information beingaggregated from lower to higher levels of anal-yses. It is well known that measurement errorstend to reduce the absolute value of coefficients(Hausman, 2001). Moreover, the sample is rela-tively small, which limits the statistical power ofthe analysis given the use of fixed effects. Butbecause the theory is fundamentally about valuecapture, it is reassuring to find a consistent patternwhen the dependent variable measures value cap-ture (as profit per partner) directly, in contrast tovariables such as relationship termination that arerelated to, but distinct from, value capture.

DISCUSSION

The empirical analyses suggest that, in this setting,suppliers that do not currently provide services toa buyer are at a severe disadvantage with respectto doing business with that buyer. This has impli-cations for competition among suppliers. A sup-plier’s most relevant competitors are those withwhich it shares a buyer. A competitor with aver-age expertise that currently provides services toa common buyer represents a greater threat thana highly competent competitor not currently pro-viding services to the buyer. The reason is that abuyer’s threat to create a relationship with a newsupplier is not very credible, considering that thissupplier will have to incur significant set up andlearning costs.

A sharply delineated set of relevant competitorshas implications for our understanding of competi-tive outcomes. The empirical analyses of both rela-tionship termination and supplier profitability sug-gest that expertise advantage relative to the set ofrelevant competitors matters more to value capturethan expertise advantage relative to all competitorsin the market. Keeping everything else constant, a


O. Chatain

Tabl

e7.

Des

crip

tive

stat

istic

san

dco

rrel

atio

nsof

anal

ysis

ofpr

ofit

per

equi

typa

rtne

r

Mea

nS.

D.

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

(9)

(10)

(11)

(12)

(13)

(14)

(15)

(16)

(1)

Profi

tpe

req

uity

part

ner

397.

424

7.5

1.00

(2)

Lev

erag

e5.

62.

640.

081.

00(3

)H

eadc

ount

368.

051

2.6

0.22

0.17

1.00

(4)

No.

clie

nts

inFT

SE25

06.

3511

.10.

330.

090.

811.

00

(5)

At

leas

ton

ecl

ient

inFT

SE25

00.

740.

440.

090.

020.

250.

341.

00

(6)

No.

ofar

eas

ofla

wpr

actic

edin

Lon

don

9.01

10.1

0.28

−0.0

20.

730.

750.

321.

00

(7)

No.

ofar

eas

ofla

wpr

actic

edou

tsid

eL

ondo

n

6.23

8.54

−0.3

60.

240.

050.

060.

16−0

.21

1.00

(8)

Cor

p.fin

ance

0.43

0.36

0.06

0.11

0.45

0.54

0.44

0.34

0.41

1.00

(9)

Prop

erty

0.41

0.36

−0.0

90.

110.

440.

470.

370.

30.

470.

721.

00(1

0)L

itiga

tion

0.41

0.35

−0.1

40.

080.

440.

450.

380.

260.

550.

710.

771.

00(1

1)E

mpl

oym

ent

0.33

0.35

−0.1

10.

100.

360.

40.

290.

290.

360.

580.

580.

641.

00(1

2)B

anki

ng0.

290.

350.

060.

090.

450.

460.

350.

280.

380.

730.

60.

620.

561.

00(1

3)C

apita

lm

arke

ts0.

080.

20.

56−0

.03

0.66

0.61

0.12

0.53

−0.2

70.

250.

130.

140.

130.

291.

00(1

4)A

dded

valu

efo

rmex

pert

ise

0.23

0.32

−0.0

30.

030.

350.

460.

430.

210.

340.

630.

550.

580.

490.

620.

151.

00

(15)

Cro

ssse

lling

0.09

0.18

0.02

−0.0

10.

100.

130.

30.

090.

050.

200.

180.

140.

170.

210.

030.

191.

00(1

6)C

ompe

titor

scr

oss

sell

ing

0.27

0.33

0.09

−0.0

60.

040.

080.

490.

090.

090.

160.

060.

140.

160.

120.

070.

190.

191.

00



Tabl

e8.

Ana

lysi

sof

Supp

lier

Profi

tabi

lity

Mod

el(1

)(2

)(3

)(4

)(5

)M

etho

dFi

xed

Eff

ects

Fixe

dE

ffec

tsFi

xed

Eff

ects

Fixe

dE

ffec

tsFi

xed

Eff

ects

Con

trol

Var

iabl

esL

ever

age

8.32

7[5

.660

]8.

437

[5.6

45]

8.35

2[5

.665

]8.

541

[5.6

64]

8.81

6[5

.643

]H

eadc

ount

0.00

1[0

.081

]0.

000

[0.0

80]

0.00

3[0

.081

]0.

002

[0.0

81]

0.00

4[0

.078

]N

o.C

lient

sin

FTSE

250

−2.3

22[1

.867

]−2

.403

[1.8

78]

−2.0

35[1

.945

]−2

.434

[1.8

41]

−2.2

45[1

.950

]A

tL

east

One

Clie

ntin

FTSE

250

−23.

883

[27.

183]

−26.

787

[28.

208]

−26.

343

[27.

145]

−13.

768

[25.

796]

−20.

884

[26.

892]

No.

ofA

reas

ofL

awPr

actic

edin

Lon

don

1.79

7[1

.910

]1.

890

[1.9

11]

1.80

2[1

.899

]1.

969

[1.8

66]

2.17

9[1

.842

]N

o.of

Are

asof

Law

Prac

ticed

outs

ide

Lon

don

1.51

7[3

.548

]1.

563

[3.5

48]

1.79

2[3

.585

]1.

677

[3.5

12]

2.13

1[3

.562

]C

orp.

Fina

nce

19.8

33[2

6.32

1]19

.614

[26.

631]

15.1

90[2

6.13

1]15

.998

[25.

984]

9.22

6[2

5.73

2]Pr

oper

ty34

.593

[30.

737]

33.4

10[3

0.92

9]37

.389

[31.

052]

28.3

71[3

0.16

1]28

.675

[30.

842]

Liti

gatio

n18

.974

[23.

119]

17.2

97[2

3.25

7]20

.750

[23.

196]

11.9

81[2

3.58

5]9.

956

[23.

671]

Em

ploy

men

t38

.607

∗∗[1

6.78

5]38

.243

∗∗[1

6.51

8]39

.603

∗∗[1

6.89

5]38

.551

∗∗[1

6.72

2]39

.090

∗∗[1

6.41

0]B

anki

ng3.

007

[32.

900]

2.99

8[3

3.28

8]−2

.638

[32.

954]

7.69

4[3

3.25

3]1.

335

[33.

479]

Cap

ital

Mar

kets

65.1

89[1

13.5

60]

63.6

07[1

14.0

74]

68.1

19[1

13.0

44]

68.3

24[1

11.7

82]

69.4

19[1

11.3

18]

The

oret

ical

Var

iabl

esE

xper

tise

Adv

anta

gein

Rel

evan

tC

ompe

titor

Set

[H5]

17.2

40[1

6.47

3]32

.920

∗[1

7.21

6]C

lient

-spe

cific

Scop

e[H

6]38

.742

∗∗∗

[14.

275]

48.2

94∗∗

∗[1

5.70

3]C

ompe

titor

s’C

lient

-spe

cific

Scop

e[H

7]−2

2.65

4[1

4.80

1]−2

6.00

7∗[1

5.10

5]Fi

rmE

ffec

tsY

ES

YE

SY

ES

YE

SY

ES

Yea

rE

ffec

tsY

ES

YE

SY

ES

YE

SY

ES

Obs

erva

tion

s45

345

345

345

345

3R

-squ

ared

(Tot

al)

0.97

0.97

0.97

0.97

0.97

Num

ber

ofFi

rms

128

128

128

128

128

Rob

ust

stan

dard

erro

rs,

clus

tere

dby

supp

liers

,in

brac

kets

∗∗∗

p<

0.01

,∗∗

p<

0.05

,∗ p

<0.

1;tw

o-ta

iled

test

s


O. Chatain

moderately expert supplier competing directly withweak suppliers is better off than a highly expertsupplier competing directly with similarly expertsuppliers.

In the same vein, the analyses suggest thatdrivers of value creation that pertain to the levelof the client relationship matter for value capture.Broader client-specific scope facilitates a focalsupplier’s value capture, but negatively affectscompetitors’ ability to capture value. Buyers canuse the higher value creation that attends a sup-plier’s broader client-specific scope as a threatwhen negotiating with other suppliers.

These findings speak to theories of competi-tive advantage. The RBV has relied on the con-cept of Ricardian rent to explain the link betweenheterogeneity in value creation and heterogeneityin value capture (Peteraf and Barney, 2003). Thefindings reported in this paper suggest that defin-ing the set of relevant competitors is crucial toaccurately mapping value creation to value cap-ture. An important implication is that returns toresources and capabilities might differ within thesame market. If sets of relevant competitors aresmall and sticky, returns from similar resourcesand capabilities might be quite different in theface of different sets of competitors. This is consis-tent with models that suggest that frictions in theproduct market, such as strong client relationships,matter for understanding investments in capabili-ties and the origin of firm heterogeneity (Chatainand Zemsky, 2009). More generally, this suggeststhat to understand the returns from capabilitiesand resources, it might sometimes be necessary toaccount explicitly for heterogeneity in competitionin addition to heterogeneity in capabilities. From astructure-conduct-performance perspective (Porter,1980), the market studied in this paper presenteda puzzle in that it combines low concentration andwhat seems to be high profitability. Here the puz-zle is solved by recognizing that instead of a largemarket with many buyers and suppliers, there arelarge numbers of smaller competitive arenas iso-lated by significant barriers to entry.

The theory and operationalization in this paperwere driven by concepts from value-based strate-gies (Brandenburger and Stuart, 1996, 2007), mostnotably the concept of added value. The paperspeaks to this stream of research by being anexample of how applied models in this line ofwork (e.g., Chatain and Zemsky, 2007; Ryall and

Sorenson, 2007) can be adapted and used to pro-vide insights into the design of empirical work(see also, for instance, Adegbesan and Higgins,2009). The applicability of these methods can,hence, be broadened to include models gearedto help refine empirical work as well as modelsthat have purely theoretical objectives. This seemsespecially promising in relationship to the RBV.The detailed understanding of firm heterogeneityfrom the RBV can be combined with the fine-grained analysis of competition to be found in thecooperative game theory models proposed by thevalue-based approach.

These findings are also of interest to researchconcerned with buyer-supplier relationships, asthey suggest that a supplier’s ability to capturevalue from relationships depends not only on itsown characteristics (e.g., level of expertise) butalso on how its clients exploit competition (Baker,1990), as by choosing suppliers based not only ontheir intrinsic competencies but also on the com-petition they provide to others. This represents oneof many ways in which relationships with buyersinfluence performance, a question of great interestto strategy scholars (e.g., Dyer and Singh, 1998;Levinthal and Fichman, 1988; Martin et al., 1998;Mitchell and Singh, 1996).

This study also has relevance for social networkstudies. The situation of a buyer playing differentsuppliers against one another and an agent ben-efiting from being positioned in a structural hole(Burt, 1992) are analogous. Like brokers in a struc-tural hole, buyers with access to other suppliersoffering superior alternatives gain an advantageover a single supplier. This mechanism closelyresembles the control benefits that emanate fromstructural holes; here, the tertius gaudens is thebuyer that plays one supplier against another, verymuch in the same vein as in Simmel’s originalformulation of the concept (Simmel, 1950: 156).This study also emphasizes a related but differentaspect. In this paper, the individual characteristics(value creation ability, generic and client-specific)of the nodes of the network matter for determiningthe extent of the benefits that accrue to a cen-tral position. For a buyer, the value of a tie toa supplier can depend on the characteristics of thefocal supplier and other suppliers. However, thetheory presented in the paper does not rely on thebenefits that structural holes may provide in termsof superior access to diverse information sourcesas also emphasized by Burt (1992). Moreover,



the paper examines a two-mode network (buy-ers and suppliers) that precludes direct ties amongagents of the same category, which, by assump-tion, creates structural holes. The study reportedin the paper also highlights the role of relation-ship multiplexity, that is, the existence of multipleties between actors (Gimeno and Woo, 1996a), ofwhich client-specific scope is an instance, as rela-tionships between buyers and suppliers can spanmultiple areas. Finally, this paper examines a formof power dependence among actors within thesame industry in a way that is related to stud-ies of power dependence across industries (e.g.,Piskorski and Casciaro, 2006).

How generalizable are these findings beyondthe United Kingdom corporate legal market? Thequestion of how value is shared between buy-ers and suppliers is not specific to law firms,but concerns virtually all business-to-business mar-kets. Similarly, although client-specific economiesof scope are especially salient in professional ser-vice markets, they feature equally in some indus-trial business-to-business markets. These consider-ations suggest that the results of this research applywell beyond the confines of the United Kingdomcorporate legal market.

A number of assumptions made in this workcould be relaxed in future research. This line ofinquiry could be pursued, for example, in con-nection with exploration of the process by whichbuyers and suppliers are matched. Recent researchin this area suggests interesting directions. Anandand Galetovic (2006) argue that high value clientsconnect with highly skilled suppliers of servicesto create strong relationships, whereas lower valueclients and suppliers engage in a transactional mar-ket. Further work could explicitly account for theendogeneity of the matching between buyers andsuppliers.

In conclusion, this research provides evidencethat competition affects the outcome of value shar-ing between buyers and suppliers of legal servicesin the United Kingdom. It shows that both firmcapabilities, in the guise of legal expertise andclient-specific scope economies and competitionwithin a set of relevant suppliers, are among thefactors that make client-supplier relationships moreor less valuable to suppliers. The findings haveimplications for our understanding of competitionin vertical relationships, and also provide evidencethat concepts from formal models of strategy canbe successfully applied in empirical studies.

ACKNOWLEDGEMENTS

I thank two anonymous reviewers for their detailedfeedback and suggestions that helped to signifi-cantly improve the paper. I thank seminar audi-ences at EM Lyon, Emory University, ESSEC,HEC Paris, IESE, INSEAD, London BusinessSchool, National University of Singapore, Singa-pore Management University, University of Michi-gan, University of North Carolina, SouthernMethodist University, and Wharton for their com-ments on previous versions of this paper. I thankRoxana Barbulescu, Matthew Bidwell, CorentinCurchod, Gary Dushnitsky, Javier Gimeno, RahulKapoor, Felipe Monteiro, Evan Rawley, MetinSengul, and Govert Vroom for their feedback andhelp at various stages of this project. I also thankthe lawyers and executives from several law firmsas well as the legal industry journalists and consul-tants who helped me understand the United King-dom legal market. In particular, I thank Chambersand Partners for making available to me the maindataset I use in this paper as well as HildebrandtInternational and The Lawyer for kindly assistingme during data collection. Finally, I acknowledgethe financial support of INSEAD’s PhD Office.A summary of a previous version of this paperappeared in the preceedings of the 2007 Academyof Management Meetings.

REFERENCES

Adegbesan JA, Higgins MJ. 2009. The intra-alliancedivision of value created through collaboration. IESEBusiness School Working paper no. 667. University ofNavarra, Barcelona, Spain.

Adner R, Zemsky P. 2006. A demand-based perspectiveon sustainable competitive advantage. StrategicManagement Journal 27(3): 215–239.

Anand BN, Galetovic A. 2006. Relationships, competi-tion and the structure of investment banking markets.Journal of Industrial Economics 54(2): 151–199.

Baker WE. 1990. Market networks and corporatebehavior. American Journal of Sociology 96(3):589–625.

Barney JB. 1986. Strategic factor markets: expectations,luck, and business strategy. Management Science32(10): 1231–1241.

Blyler M, Coff R. 2003. Dynamic capabilities, socialcapital, and rent appropriation: ties that split pies.Strategic Management Journal 24(7): 677–686.

Brandenburger AM, Stuart HW. 1996. Value-based busi-ness strategy. Journal of Economics and ManagementStrategy 5(1): 5–24.


O. Chatain

Brandenburger A, Stuart HW. 2007. Biform games.Management Science 53(4): 537–549.

Burt RS. 1992. Structural Holes . Harvard UniversityPress: Cambridge, MA.

Cameron AC, Trivedi PK. 2005. Microeconometrics:Methods and Applications . Cambridge UniversityPress: Cambridge,UK.

Chatain O, Zemsky P. 2007. The horizontal scope ofthe firm: organizational tradeoffs vs. buyer-supplierrelationships. Management Science 53(4): 550–565.

Chatain O, Zemsky P. 2009. Value creation and valuecapture with frictions. INSEAD Working paperno. 2009/35/ST/ACGRE. INSEAD, Fontainebleau,France.

Coff RW. 1999. When competitive advantage doesn’tlead to performance: the resource-based view andstakeholder bargaining power. Organization Science10(2): 119–133.

Dyer JH, Singh H. 1998. The relational view: cooperativestrategy and sources of interorganizational competitiveadvantage. Academy of Management Review 23(4):660–679.

Ethiraj SK, Kale P, Krishnan MS, Singh JV. 2005.Where do capabilities come from and how do theymatter? A study in the software services industry.Strategic Management Journal 26(1): 25–45.

Froot KA. 1989. Consistent covariance matrix estimationwith cross-sectional dependence and heteroskedastic-ity in financial data. Journal of Financial and Quanti-tative Analysis 24: 333–355.

Gilson RJ, Mnookin RH. 1989. Coming of age in acorporate law firm: the economics of associate careerpatterns. Stanford Law Review 41: 567–595.

Gimeno J, Woo CY. 1996a. Economic multiplexity: thestructural embeddedness of cooperation in multiplerelations of interdependence. Advances in StrategicManagement 13: 323–261.

Gimeno J, Woo CY. 1996b. Hypercompetition in amultimarket environment: the role of strategicsimilarity and multimarket contact in competitive de-escalation. Organization Science 7(3): 322–341.

Hausman J. 2001. Mismeasured variables in econometricanalysis: problems from the right and problems fromthe left. Journal of Economic Perspectives 15(4):57–68.

Helfat CE. 1997. Know-how and asset complementarityand dynamic capability accumulation: the case ofR&D. Strategic Management Journal 18(5): 339–360.

Henderson R, Cockburn I. 1994. Measuring competence?Exploring firm effects in pharmaceutical research.Strategic Management Journal , Winter Special Issue15: 63–84.

Henderson WD. 2006. An empirical study of single-tierversus two-tier partnership in the Am Law 200. NorthCarolina Law Review 84(5): 1691–1750.

King G, Zeng L. 2001. Logistic regression in rare eventsdata. Political Analysis 9(2): 137–163.

Klemperer P. 1992. Equilibrium product lines: competinghead-to-head may be less competitive. AmericanEconomic Review 82(4): 740–755.

Kor YY, Leblebici H. 2005. How do interdependenciesamong human-capital deployment, development,

and diversification strategies affect firms’ financialperformance? Strategic Management Journal 26(10):967–985.

Levinthal DA. 2008. The scientific value of gametheory. (DRUID Conference Debate, 20 June), DRUIDconference, Copenhagen, Denmark. http://www.druid.dk/streaming/ds2008/fredag/4/msh.htm accessed 14November 2008.

Levinthal DA, Fichman M. 1988. Dynamics of interor-ganizational attachments: auditor-client relationships.Administrative Science Quarterly 33(3): 345–369.

Lippman SA, Rumelt RP. 2003. A bargaining perspectiveon resource advantage. Strategic Management Journal24(11): 1069–1086.

MacDonald G, Ryall MD. 2004. How do value creationand competition determine whether a firm appropriatesvalue? Management Science 50(10): 1319–1333.

Maister DH. 1993. Managing the Professional ServiceFirm. Free Press: New York.

Martin X, Swaminathan A, Mitchell W. 1998. Organi-zational evolution in the interorganizational envi-ronment: incentives and constraints on internationalexpansion strategy. Administrative Science Quarterly43(3): 566–601.

Meyer BD. 1990. Unemployment insurance and unem-ployment spells. Econometrica 58(4): 757–782.

Miller D, Shamsie J. 1996. The resource-based view ofthe firm in two environments: the Hollywood filmstudios from 1936 to 1965. Academy of ManagementJournal 39(3): 519–543.

Mitchell W, Singh K. 1996. Survival of businessesusing collaborative relationships to commercializecomplex goods. Strategic Management Journal 17(3):169–195.

Muller HM, Warneryd K. 2001. Inside versus outsideownership: a political theory of the firm. Rand Journalof Economics 32(3): 527–541.

Myerson RB. 1991. Game Theory: Analysis of Conflict .Harvard University Press: Cambridge, MA.

Panzar JC, Willig RD. 1981. Economies of scope.American Economic Review 71(2):): 268–272.

Peteraf MA, Barney JB. 2003. Unraveling the resource-based tangle. Managerial and Decision Economics24(4): 309–323.

Piskorski MJ, Casciaro T. 2006. When more powermakes actors worse off: turning a profit in theAmerican economy. Social Forces 85(2): 1011–1036.

Porter ME. 1980. Competitive Strategy: Techniques forAnalyzing Industries and Competitors . Free Press:New York.

Rogers WH. 1993. Regression standard errors in clusteredsamples. Stata Technical Bulletin 13: 19–23.

Ryall MD, Sorenson O. 2007. Brokers and competitiveadvantage. Management Science 53(4): 566–583.

Samuelson SS, Jaffe LJ. 1990. A statistical analysis oflaw firm profitability. Boston University Law Review70: 185–211.

Siggelkow N. 2003. Why focus? A study of intra-industryfocus effects. Journal of Industrial Economics 51(2):121–150.



Simmel G. 1950. The triad. In The Sociology of GeorgSimmel , Wolff KH (ed., trans.). Free Press: New York;145–168.

Stuart HW. 2004. Efficient spatial competition. Gamesand Economic Behavior 49: 345–362.

Teece DJ. 1980. Economies of scope and the scope ofthe enterprise. Journal of Economic Behavior andOrganization 1(3): 223–247.

White H. 1980. A heteroskedasticity-consistent covari-ance matrix estimator and a direct test for het-eroskedasticity. Econometrica 48: 817–830.


O. Chatain

APPENDIX 1: MODEL ASSUMPTIONS

Per cooperative game theory terminology (see,for instance, Myerson [1991]), the characteristicfunction v maps sets of players to the value theycan create together. This function is assumed tobe super modular, that is, adding players cannotreduce the value created.

Assumption 1(A1):The set of players N can be split into two,

nonempty, disjoint sets S and B, such that

v(S) = 0 and v(B) = 0.

Assumption 2 (A2):

v(N) =∑

j∈Bv({j} ∪ S)

Assumption 2 states that the ability of suppliers tocreate surplus with a given buyer is independentof the surplus created with other buyers.

Assumption 3 (A3):

If s ′ ⊂ s ⊆ S then AVi({j, s}) ≤ AVi({j, s ′})for all i ∈ s ′ and j ∈ B.

Assumption 3 means that adding suppliers cannotincrease the added value of any supplier.

APPENDIX 2: PROOFS

Proof of Proposition 1: The focal supplier’s addedvalue is determined by the next best alternativeavailable to the client. This alternative is either thebest supplier with a current relationship, excludingthe focal supplier, or the best supplier without acurrent relationship. It is the best supplier with acurrent relationship rather than a supplier withouta relationship if and only if:

maxk∈SR

(Ek + Ckj + �R) > maxl∈SN

(El + Clj )

This is equivalent to the condition:

�R > maxl∈SN

(El + Clj ) − maxk∈SR

(Ek + Ckj )

Proof of Corollary 2: Define T = maxl∈SN

(El + Clj ) −maxk∈SR

(Ek + Ckj ). By Proposition 1, if �R > T ,

then the next best alternative to the focal supplierbelongs to SR . Hence, the added value of a sup-plier is entirely determined by the value creationof suppliers belonging to SR .Proof of Proposition 3: Assume that firms in thebuyer’s current supplier base can create an excess�R in value on top of a baseline R that is sharedby all suppliers, and identical across areas.

Holding the current supplier base constant, sup-pliers with higher added value are more likely toagree to work in a new area because they canexpect to capture more value. Conversely, a buyerwill prefer to work with a supplier with the high-est added value because the value appropriated bythe buyer is equal to the value created by the nextbest supplier plus a share of the supplier’s addedvalue. Moreover, in this situation, choosing a sup-plier with the highest added value is equivalent tochoosing a supplier with the highest value creation.

Denote Vij (b) the incremental value created bya supplier when area b is added to its portfo-lio of activities. We have Vij (b) = (Eib + R +1i∈Sj

c · �R + �Cijb + νijb), where �Cijb is equalto C(Aij ∪ {b}) − C(Aij ), the increment in client-specific scope economies due to the addition ofarea b to the supplier’s current portfolio Aij and1i∈Sj

c equal to one when i is part the set Sjc of

current suppliers of client j and zero otherwise.The probability that supplier i creates the highestamount of value in area b for supplier j is:

Pr(Vij (b) > Vkj(b), ∀k = i)

= Pr(Eib + R + 1i∈Sjc · �R + �Cijb

+ νijb > Ekb + R + 1k∈Sjc · �R + �Ckjb

+ νkjb, ∀k = i)

= Pr(νkjb − νijb < Eib − Ekb + �Cijb

− �Ckjb + 1i∈Sjc · �R − 1k∈Sj

c

· �R, ∀k = i)

This probability is increasing in 1i∈Sjc · �R.

Thus, belonging to Sjc increases the probability of

creating the largest amount of value for the seller,which is the probability of being selected.

Proof of Proposition 4: See Stuart (2004, Lemma1) or Chatain and Zemsky (2007, Proposition 1).Proof of Corollary 5: See Chatain and Zemsky(2007, Corollary 1).



Proof of Proposition 6: Assume that supplier i hasan alternative opportunity that allows the captureof the amount OPi of value should it terminate therelationship. Denote G the cumulative distributionfunction of the εijt . Supplier i stays in the currentrelationship as long as:

Pr(�ijt > OPi) = 1 − G(OPi − αiAVijt − µij )

As G is monotone increasing, this probability ismonotone decreasing in AVijt .

Proof of Proposition 7: Denote as Bit the set ofbuyers that use supplier i in period t , and wjt thesize of the engagement for each supplier. A sim-ple accounting identity relates client-level actualprofitability �ijt to supplier total profitability �it :

�it =∑

j∈Bit

wjt�ijt

To express this identity in terms of added value,we need to be able to decompose a supplier’sadded value into the added value calculated at eachbuyer level. Assumption (A2) allows us to do thisbecause it implies that added values determined atthe client level are independent from each other.Therefore:

�it =∑

j∈Bit

wjt�ijt

=∑

j∈Bit

wjt [αAVijt + µij + εijt ]

=∑

j∈Bit

αwjtAVijt +∑

j

µij +∑

j

wjtεijt

This implies that the profits of supplier i varywith

∑j∈Bit

αwjtAVijt .


value creation, competition, and performance in ... · value creation, competition, and supplier...

Documents