Strategic Management Journal

Strat. Mgmt. J., 23: 707–725 (2002)

Published online 1 May 2002 in Wiley InterScience ([Link]). DOI: 10.1002/smj.249

DO FORMAL CONTRACTS AND RELATIONAL

GOVERNANCE FUNCTION AS SUBSTITUTES OR
COMPLEMENTS?
LAURA POPPO1 * and TODD ZENGER2
1
Pamplin College of Business, Virginia Tech, Blacksburg, Virginia, U.S.A.
2
John M. Olin School of Business, Washington University, St Louis, Missouri, U.S.A.

Relational exchange arrangements supported by trust are commonly viewed as substitutes

for complex contracts in interorganizational exchanges. Many argue that formal contracts
actually undermine trust and thereby encourage the opportunistic behavior they are designed to
discourage. In this paper, we develop and test an alternative perspective: that formal contracts
and relational governance function as complements. Using data from a sample of information
service exchanges, we find empirical support for this proposition of complementarity. Managers
appear to couple their increasingly customized contracts with high levels of relational governance
(and vice versa). Moreover, this interdependence underlies their ability to generate improvements
in exchange performance. Our results concerning the determinants of these governance choices
show their distinct origins, which further augments their complementarity in practice. Copyright
 2002 John Wiley & Sons, Ltd.

Transaction cost economics (TCE) has emerged This view recognizes that in many industries man-
as a common framework for understanding how agers engage in complex, collaborative market
managers craft governance arrangements. The gen- exchanges that involve rather high levels of asset
eral proposition of this literature is that managers specificity and that are characterized by other
align the governance features of interorganiza- known hazards (Hill, 1990; Dyer, 1997). Rela-
tional relationships to match known exchange haz- tional norms, such as trust, are viewed in this liter-
ards, particularly those associated with specialized ature as substitutes for complex, explicit contracts
asset investments, difficult performance measure- or vertical integration (Granovetter, 1985; Bern-
ment, or uncertainty (Williamson, 1985, 1991). In heim and Whinston, 1998; Bradach and Eccles,
response to exchange hazards, managers may craft 1989; Dyer and Singh, 1998; Gulati, 1995b; Uzzi,
complex contracts that define remedies for foresee- 1997; Adler, 2001). Based on this reasoning, trust
able contingencies or specify processes for resolv- and its underlying normative behaviors operate as
ing unforeseeable outcomes. When such contracts a self-enforcing safeguard that is a more effec-
are too costly to craft and enforce, managers may tive and less costly alternative to both contracts
choose to vertically integrate. and vertical integration (Hill, 1990; Uzzi, 1997).
Many have argued, however, that transaction Indeed, some contend that formal contracts may
cost economics overstates the desirability of either even undermine a firm’s capacity to develop rela-
integration or explicit contractual safeguards in tional governance. Formal contracts may signal
exchange settings commonly labeled as hazardous.
distrust of your exchange partner and by undermin-
ing trust, encourage, rather than discourage, oppor-
Key words: outsourcing; trust; transaction cost eco- tunistic behavior (Ghoshal and Moran, 1996: 24,
nomics; IT; contracts; relational governance
*Correspondence to: L. Poppo, Pamplin College of Business, 27; Macaulay, 1963: 64; Fehr and Gachter, 2000).
Virginia Tech, Blacksburg, VA 24061, U.S.A. Taken together, these critiques view relational

Copyright  2002 John Wiley & Sons, Ltd. Received 3 February 2000
Final revision received 12 December 2001
708 L. Poppo and T. Zenger

governance as a substitute for formal contracts. In long-term supply relations with their vendors; oth-
the presence of relational governance, formal con- ers spent considerable time and money customiz-
tracts are at best an unnecessary expense and at ing formal contracts. Still others appear to have
worst counter-productive. done both. Using this variation in the structure
In this paper, we advance and test an alter- of outsourcing relationships, we empirically test
native argument: that formal contracts and rela- the relationship between formal contracts and rela-
tional governance function as complements. Rather tional governance and their effects on exchange
than hindering or substituting for relational gover- performance.
nance, well-specified contracts may actually pro-
mote more cooperative, long-term, trusting ex-
change relationships. Well-specified contracts nar- FORMAL CONTRACTS
row the domain and severity of risk to which an
exchange is exposed and thereby encourage coop- Formal contracts represent promises or obligations
eration and trust. In addition, well-crafted con- to perform particular actions in the future (Mac-
tracts promote longevity in exchanges by increas- neil, 1978). The more complex is the contract, the
ing the penalties that accompany severing an greater is the specification of promises, obligations,
exchange relationship. As discussed in the trans- and processes for dispute resolution. For example,
action cost literature, contracts also provide cus- complex contracts may detail roles and respon-
tomized approaches and mutually agreed upon sibilities to be performed, specify procedures for
policies and procedures for dealing with necessary monitoring and penalties for noncompliance, and,
adaptations in an exchange (Williamson, 1991). most importantly, determine outcomes or outputs
This complementary relationship may also func- to be delivered. According to the logic of transac-
tion in reverse. The continuity and cooperation tion cost economics, the manager’s task is to craft
encouraged by relational governance may generate governance arrangements with minimal cost that
contractual refinements that further support greater ensure the delivery of the desired quantity, price,
cooperation. Relational governance may heighten and quality of a supplier’s services. The man-
the probability that trust and cooperation will safe- ager, therefore, crafts governance arrangements to
guard against hazards poorly protected by the con- match the exchange conditions that accompany
tract. Finally, relational governance may help over- various services. As exchange hazards rise so must
come the adaptive limits of contracts: a bilateral contractual safeguards (Williamson, 1985; Klein,
commitment to ‘keep-on-with-it’ despite the unex- Crawford, and Alchian, 1978), which act to min-
pected complications and conflicts. imize the costs and performance losses arising
We empirically test whether relational gover- from such hazards (Joskow, 1988; Macneil, 1978;
nance and formal contracts operate as comple- Heide, 1994). Because crafting a complex contract
ments or substitutes using data on outsourcing is costly, parties undertake such a cost only when
relationships in information services during the the consequences of a contractual breach are con-
early 1990s. The data were collected from sur- siderable.
veys of senior managers regarding their sourc- Transaction cost economics scholars commonly
ing of various information services, such as data point to three categories of exchange hazards
entry, software application development, data net- that necessitate contractual safeguards (or vertical
work design, and network maintenance. During the integration): asset specificity, measurement diffi-
time period of this survey, outsourcing of infor- culty, and uncertainty. Asset specificity emerges
mation services was escalating, accompanied by when sourcing relationships require significant
considerable debate in the popular press about the relationship-specific investments in physical and/or
merits of outsourcing and the merits of various out- human assets. The presence of these specific assets
sourcing arrangements (Clemons and Row, 1991; transforms an exchange from a world of classical
McFarlan, 1990; McFarlan and Nolan, 1995; Lac- contracting in which the ‘identity of parties is irrel-
ity, Willcocks, and Feeny, 1995, 1996). Perhaps evant’ into a world of neoclassical contracting in
as a consequence of such debate and as a con- which the identity of exchange partners is of crit-
sequence of variation in the underlying exchange ical importance (Williamson, 1991). For example,
conditions, the structure of outsourcing arrange- an information service (IS) provider may need to
ments varied widely. Some managers developed customize a service offering to the clients’ work
Copyright  2002 John Wiley & Sons, Ltd. Strat. Mgmt. J., 23: 707–725 (2002)
 Substitutes or Complements 709

setting. Similarly, the client may need to develop access to such capabilities through the specifica-
a unique understanding of the provider’s proce- tion of clauses and procedures that facilitate nego-
dures, approach, and language to effectively utilize tiations that invariably arise from technological
their services. In such circumstances, the continu- changes.
ity of an exchange becomes vital to its effective- High levels of uncertainty in conjunction with
ness. Severing the relationship results in the forfei- measurement difficulty or asset specificity ren-
ture of the value of these specialized investments. der contracting even more hazardous (Williamson,
Through threats to terminate the relationship, one 1985). High uncertainty may, for instance, par-
or both contractual parties may seek to appro- ticularly discourage a supplier from making spe-
priate returns from these specialized investments. cialized asset investments absent appropriate safe-
To safeguard against such hold-up behavior, man- guards. Similar reasoning applies to measurement
agers adopt neoclassical contracts, which promote difficulty. If the underlying technology is rapidly
the longevity of relationships by specifying not changing, difficulty in measuring the performance
only required actions and conditions of contractual of the services rendered with the technology may
breach, but also a framework for resolving unfore- stretch the limits of contracting. Williamson (1985:
seen disputes. Empirical work demonstrates the 80) proposes that under these combinations of haz-
predicted relationship: asset specificity increases ards, ‘transaction[s] may “flee” to one of the polar
the complexity of contracts (Joskow, 1988). extremes as the degree of uncertainty increases.’
Difficulty in measuring the performance of ex- Thus, firms may choose to vertically integrate or
change partners also generates market hazards. seek to alter the exchange to a shorter-term, more
Markets succeed when they can effectively link arm’s-length exchange. TCE maintains that ver-
rewards to productivity—that is, they can measure tical integration, which has access to fiat, better
productivity and pay for it accordingly (Alchian information disclosure, and alternative incentive
and Demsetz, 1972). When performance is diffi- mechanisms, may cope more effectively with such
cult to measure, parties have incentives to limit combinations of hazards.
their efforts toward fulfilling the agreement. Man- In sum, asset specificity, measurement difficulty,
agers have two choices. They can realize lower and technological uncertainty create exchange haz-
performance because of their inability to measure ards that encourage more complex contracts. Asso-
performance, or expend resources to improve per- ciated with these more complex contracts are
formance measurement by creating more complex additional costs of contracting. Ignoring these
contracts that specify delivered service levels or added costs, crafting complex contracts as an
facilitate the monitoring of a supplier’s behaviors. efficient response to hazardous exchange settings
For example, clauses may specify third party moni- should enhance exchange performance (Masten,
toring, disclosure of necessary documents to justify 1993, 1996). More complex contracts deter behav-
work done, and, if possible, the use of bench- iors that could compromise the performance of a
marks to gauge the performance of the work done. buyer–supplier exchange. Thus:
Thus, as measurement becomes more difficult, we
expect managers to develop more complex con- Hypothesis 1: Increases in exchange hazards
tracts, which enable them to accurately measure encourage more complex contracts.
and reward productivity.
Uncertainty, a third hazard, also challenges an
exchange by requiring the parties to adapt to RELATIONAL GOVERNANCE
problems raised from unforeseeable changes. In
this study, we focus on uncertainty arising from Many scholars, including transaction cost econ-
rapidly changing technology. In general, markets omists, have observed that the governance of
are a marvel at autonomous adaptation, particularly interorganizational exchanges involves more than
when prices serve as sufficient statistics to induce formal contracts. Interorganizational exchanges are
changes in supply and demand (Williamson, 1991: typically repeated exchanges embedded in social
287). However, for more complex forms of adapta- relationships. Governance emerges from the values
tion that require coordination among parties, sim- and agreed-upon processes found in social relation-
ple market governance is not adequate as it lacks ships (Macneil, 1978, 1980; Noordewier, John, and
coordinating capabilities. Contracts, however, have Nevin, 1990; Heide and John, 1992), which may
Copyright  2002 John Wiley & Sons, Ltd. Strat. Mgmt. J., 23: 707–725 (2002)
710 L. Poppo and T. Zenger

minimize transaction costs as compared to formal they are expected to behave in a trustworthy fash-
contracts (Dyer, 1996; Dyer and Singh, 1998). ion in the future. For economists, the trustworthy
For such relationally-governed exchanges, the status is conditional upon the benefits that accrue
enforcement of obligations, promises, and expec- from trustworthy status over time (e.g., repeated
tations occurs through social processes that pro- exchange) contrasted with the benefits that accrue
mote norms of flexibility, solidarity, and informa- from self-interested moves that break from the
tion exchange. Flexibility facilitates adaptation to trustworthy status (Klein, 1996). This logic, com-
unforeseeable events. Solidarity promotes a bilat- mon to game theory, argues that expectations of
eral approach to problem solving, creating a com- pay-offs from future cooperative behavior encour-
mitment to joint action through mutual adjustment. age cooperation in the present (Baker, Gibbons and
Information sharing facilitates problem solving and Murphy, 2002). Williamson (1996: 97) concludes
adaptation because parties are willing to share pri- that the term trust is misleading, given the above
vate information with one another, including short- economic logic, arguing that ‘because commercial
and long-term plans and goals. As the parties com- relations are invariably calculative, the concept of
mit to such norms, mutuality and cooperation char- calculated risk (rather than calculated trust) should
acterize the resultant behavior. be used to describe commercial transactions.’
Through these social processes and the resulting Nonetheless, there is considerable overlap in
norms, relational governance may function to mit- the arguments of sociologists and economists sur-
igate the precise exchange hazards targeted by for- rounding trust and cooperation and we will there-
mal contracts—hazards associated with exchange- fore not attempt to offer any systematic distinction.
specific asset investments, difficult performance Both sociologists and economists, for instance,
measurement, and uncertainty. The expectations of argue that repeated exchange encourages effec-
continuity that accompany relational governance tive exchange, and that repeated exchange pro-
generate incentives to invest in exchange-specific vides information about the cooperative behavior
investments. These investments are protected by of exchange partners that may allow for informed
the mutually-imposed costs of termination. Simi- choices of who to ‘trust’ and who not to trust. In
larly, expectations of longevity minimize the need addition, while the mechanism may differ slightly,
for precise performance measurement in the short both economists and sociologists emphasize that
run. Parties to the exchange expect that short- reputations for trustworthy behavior are rewarded
term inequities will be corrected in the long term. and reputations for untrustworthy behavior pun-
Finally, norms of cooperation and mutual adapta- ished in the broader network of potential exchange
tion provide the flexibility to cope with inevitable partners.
uncertainties that arise in an exchange. As widely Empirical work generally shows that relational
discussed in the management literature, norms of governance is associated with trust and that trust
flexibility, information sharing, and commitment improves the performance of interorganizational
thus help circumvent the potentially high costs of exchanges (Palay, 1984; Heide and John, 1990;
exchange hazards (Macaulay, 1963; Dore, 1983; Zaheer and Venkatraman, 1995; Mohr and
Palay, 1984; Granovetter, 1985, 1992; Bradach and Spekman, 1994; Saxton, 1997; Zaheer, McEvily,
Eccles, 1989; Jones, Hesterly and Borgatti, 1997; and Perrone, 1998). Yet, the development and
Adler, 2001). maintenance of relational governance with its
The mechanisms through which relational gov- dense network of social ties may involve
ernance attenuates exchange hazards are both eco- considerable cost in terms of time and resource
nomic and sociological in nature. Economists em- allocation (Larson, 1992). Furthermore, dense
phasize the rational, calculative origins of rela- social ties in economic exchanges may restrict
tional governance, emphasizing particularly expec- firms from new information and new opportunities
tations of future exchanges that prompt coopera- (Uzzi, 1997; Gargiulo and Benassi, 2000). This
tion in the present. Sociologists emphasize socially reasoning suggests that firms should invest
derived norms and social ties that have emerged in the development of relational governance
from prior exchange (Uzzi, 1997: 45). Trust is only when significant hazards are present.
therefore considered a trait that becomes embedded Absent these hazards, incurring the costs of
in a particular exchange relation. In essence, once relational governance may not be warranted.
an exchange partner is granted ‘trustworthy’ status, Nonetheless, the pervasive logic in the relational
Copyright  2002 John Wiley & Sons, Ltd. Strat. Mgmt. J., 23: 707–725 (2002)
 Substitutes or Complements 711

governance literature is that relational governance undermine the formation of relational governance.
positively affects exchange performance (Dyer, Ghoshal and Moran (1996) argue that the use of
1996; Saxton, 1997; Zaheer et al., 1998). Among rational, formal control has a pernicious effect on
contractual exchanges (those neither integrated cooperation.1 They contend that for those parties
nor managed through spot market contracts), we being controlled
hypothesize the following:
. . .the use of rational control signals that they are
Hypothesis 2: Increases in exchange hazards neither trusted nor trustworthy to behave appro-
will lead to more relational governance. priately without such controls. . . . For the con-
troller, negative feelings arise from what Strickland
(1958) described as ‘the dilemma of the supervi-
sor’ viz., the situation when the use of surveillance,
RELATIONAL GOVERNANCE AND monitoring, and authority led to management’s dis-
FORMAL CONTRACTS AS trust of employees and perceptions of an increased
SUBSTITUTES need for more surveillance and control . . . (Ghoshal
and Moran, 1996: 24)
Academic research in economics and sociology has
generally viewed relational governance and formal Similarly, Macaulay (1963: 64) contends ‘Not only
contracts as substitutes—the presence of one gov- are contracts and contract law not needed in many
ernance device (relational governance, in particu- situations, their use may have, or may be thought to
lar) obviates the need for the other (Larson, 1992; have, undesirable consequences. . . . Detailed nego-
Gulati, 1995b; Dyer and Singh, 1998; Macaulay, tiated contracts can get in the way of creating good
1963). In particular, trust reduces transaction costs exchange relationships between business units.’ He
by ‘replacing contracts with handshakes’ (Adler, further argues that some firms discourage the use
2001). Thus, Dyer and Singh (1998) argue that of an elaborate contract because it ‘indicates a
informal self-enforcing agreements which rely on lack of trust and blunts the demands of friendship,
trust and reputation ‘often supplant’ the formal turning a cooperative venture into an antagonistic
controls characteristic of formal contracts. Gulati horsetrade’ (Macaulay, 1963: 64). Consistent with
(1995b: 93) is quite explicit in arguing that con- this logic, Bernheim and Whinston (1998) develop
tracts and trust function as substitutes: a formal model and show that making contracts
more explicit may encourage opportunistic behav-
. . .trust avoids contracting costs, lowers the need ior surrounding actions that cannot be specified
for monitoring, and facilitates contractual adapta- within contracts. Taken together, these scholars
tion. Trust counteracts fears of opportunistic behav- view relational governance and formal contracts
ior and as a result, is likely to limit the transaction
costs associated with an exchange. . . . In other as substitutes, which operate through one of two
words, trust can substitute for hierarchical contracts mechanisms. Either relational governance elimi-
in many exchanges . . . (emphasis added) nates the need for formal contracts and vice versa,
or formal contracts directly hinder the formation
Similarly, Uzzi (1997) argues that the embedded- of relational governance.
ness of exchanges within social structures circum- In light of this predicted substitution, the net
vents and thus economizes on time otherwise spent effects of formal contracts and relational gover-
in costly contract renegotiations. Finally, Larson nance on exchange performance are ambiguous.
(1992: 98) argues that formal contracts are rather While relational governance and formal contracts
unimportant in the exchange agreements she exam- may have positive direct effects on exchange per-
ined. Informal social controls push these formal formance, because they function as replacements
contracts to the background. A common underly- for one another (or in the case of formal con-
ing rationale for substitution emerges: if one party tracts causally damage the other), the net effect on
trusts the other, there is simply little need for exchange performance is, at a minimum, reduced
contractually specifying actions. Relational gover- and potentially negative. Evidence of a substitution
nance lowers transaction costs and facilitates adap-
tive responses. 1
Interestingly, Williamson (1996a: 271) makes a similar argu-
Still other scholars suggest an additional rea- ment, but restricts its application to purely social, noneconomic
son for substitution: formal contracts may actually relationships.

Copyright  2002 John Wiley & Sons, Ltd. Strat. Mgmt. J., 23: 707–725 (2002)
712 L. Poppo and T. Zenger

Previous Business
Relations
Relational
Governance

+ or − Exchange
Exchange Performance
Hazards
+ or −

Customized
Contracts
IT Size
Managerial Experience

Notes:
+ = Positive relationship, support for Complements (H4a, H4b)
− = Negative relationship, support for Substitutes (H3a, H3b)

Figure 1. Empirical Model

effect on performance exists if both relational gov- choice in isolation. The presence of clearly artic-
ernance and contract complexity positively influ- ulated contractual terms, remedies, and processes
ence performance, but negatively influence one of dispute resolution as well as relational norms of
another. flexibility, solidarity, bilateralism, and continuance
Following from the above arguments, we hypo- may inspire confidence to cooperate in interorga-
thesize (see Figure 1): nizational exchanges.
We noted earlier that economic models of rela-
Hypothesis 3a: Increases in contractual com- tional governance (Klein, 1996; Baker et al.,
plexity discourage the formation of relational 2002) highlight the role of simple repeated ex-
governance. change in motivating long-term cooperation. In
such models the expected pay-offs from a pattern
Hypothesis 3b: Increases in relational gover- of future exchange deters the pursuit of short-
nance discourage the use of complex contracts. run gains that undermine the longevity of the
relationship. Contracts not only have this source
Hypothesis 3c: Contractual complexity and rela- of advantage because of their formal specifica-
tional governance will function as substitutes in
tion of a long-term commitment to exchange, but
explaining exchange performance.
through clearly articulated clauses that specify
punishments they also limit the gains from oppor-
RELATIONAL GOVERNANCE AND tunistic behavior. This reduction in short-run gains
FORMAL CONTRACTS AS heightens comparatively the gains from cooper-
COMPLEMENTS ating in the exchange relationship. By contrast,
failing to contractually specify elements of the
Despite compelling arguments for viewing rela- exchange that are easily specified merely height-
tional governance and contractual complexity as ens incentives for short-run cheating and low-
substitutes, the logic for viewing them as com- ers expectations of cooperation (Baker, Gibbons
plements appears equally compelling. In settings and Murphy, 1994). Thus, the specification of
where hazards are severe, the combination of for- contractual safeguards promotes expectations that
mal and informal safeguards may deliver greater the other party will behave cooperatively and
exchange performance than either governance thus complements the informal limits of relational
Copyright  2002 John Wiley & Sons, Ltd. Strat. Mgmt. J., 23: 707–725 (2002)
 Substitutes or Complements 713

governance. Cooperative behavior in the present change and conflict arise (Macneil, 1978: 876,
then reinforces an expectation of cooperation in 883). The relational value of solidarity figures
the future. Supportive of this logic, empirical work prominently in promoting exchange into the future:
suggests that past success in contracting with a par- it ensures a ‘keep on with it’ attitude such that each
ticular exchange partner yields greater success in party desires to and is able to depend on the other
the present (Larson, 1992). Formal contracts help (Macneil, 1980: 92). Thus, managers choose rela-
ensure that the early, more vulnerable stages of tional governance, as contracts become increas-
exchange are successful. ingly customized, to increase the odds of continu-
Long-term contracts are also explicitly drafted ance, and thereby further safeguard specific invest-
with provisions to promote the longevity of ments from premature and costly termination.
exchanges. Unexpected disturbances may place Relational governance may also promote the
considerable strain on an exchange relationship refinement (and hence increased complexity) of
(Williamson, 1991: 271–273). Contracts that shift formal contracts. As a close relationship is devel-
from merely specifying deliverable outcomes to oped and sustained, lessons from the prior period
providing frameworks for bilateral adjustments are reflected in revisions of the contract. Exchange
may facilitate the evolution of highly cooperative experience, patterns of information sharing, and
exchange relations. In addition, the process of evolving performance measurement and monitor-
contracting may itself promote expectations of ing may all enable greater specificity (and com-
cooperation consistent with relational governance. plexity) in contractual provisions. As a conse-
The activity of creating complex contracts quence, relational exchanges may gradually devel-
requires parties to mutually determine and op more complex formal contracts, as mutually
commit to processes for dealing with unexpected agreed upon processes become formalized.
changes, penalties for noncompliance, and other In sum, the argument for complementarity
joint expectations of trade. Thus, the process suggests positive reciprocal relationships between
of developing complex contracts in response relational governance and formal contracts (see
to exchange hazards positively affects future Figure 1). Formal contracts promote relational
exchange performance through the development of governance in exchange settings and relational
social relations (i.e., relational governance) as well governance enables the refinement of contracts
as complements relational governance through the and promotes stability in interorganizational
formal specification of limits and expectations. exchanges. Further, due to this complementarity,
The complementary relationship between rela- the combination of relational governance and
tional governance and formal contracts may work formal contracts should generate higher exchange
in reverse, as well. Regardless of the duration of performance than either governance mechanism in
an exchange, vast dimensions of the exchange may isolation. Thus, we hypothesize:
prove impossible to contractually specify; man-
agers are clearly constrained in their capacity to Hypothesis 4a: Increases in contractual com-
foresee and contractually resolve potential future plexity will increase the level of relational gov-
contingencies. As a result, when unforeseen dis- ernance.
turbances arise, contracts in and of themselves are
unable to maintain the continuity of the relation- Hypothesis 4b: Increases in relational gover-
ship. Formally specified processes for adapting to nance will enhance contractual complexity.
change promote longevity in the exchange, but do
not guarantee continuance or a mutually accept- Hypothesis 4c: Contractual complexity and rela-
able, bilateral resolution. Thus, contracts alone tional governance will function as complements
may serve simply to facilitate termination of an in explaining exchange performance.
exchange as courts use it to review the broken
aspects of the contract and then allocate assets
between the parties on some basis deemed equi- DATA AND METHODS
table (Macneil, 1978: 875).
Data collection
Relational governance becomes a necessary
complement to the adaptive limits of contracts We obtained data for this study through a sur-
by fostering continuance and bilateralism when vey instrument. Respondents to this survey were
Copyright  2002 John Wiley & Sons, Ltd. Strat. Mgmt. J., 23: 707–725 (2002)
714 L. Poppo and T. Zenger

key informants, IS executives who held one of Narus, 1990; Mohr and Spekman, 1994).3 The sur-
two positions: (1) the senior corporate IS man- vey requested information on nine commonly used
ager who provided overall guidance and planning information services for which there is variance in
for information services, or (2) the manager who exchange attributes: data entry, data center oper-
had control over major data-processing facilities in ations, network design, network operations (data),
operating departments, divisions, and subsidiaries. network operations (voice), end user support, train-
These individuals both managed and reviewed out- ing and education, applications development, and
sourced IS activities. Our list of key informants applications maintenance. For purposes of data
came from the Directory of Top Computer Exec- analysis, the unit of analysis is the outsourced
utives. This directory, which has been in exis- service, resulting in a total sample of 345 obser-
tence since 1972, included top computer execu- vations. Due to missing data, the core sample for
tives of Fortune 500 companies and any other data analysis is 285.4
companies with an annual data-processing bud- To test for a potential response bias in our sam-
get of $250,000 or more. Unlike previous studies, ple, we compared the industries and the geographic
which constrained their samples to single indus- locations represented in the sample to the popu-
tries (Mohr and Spekman, 1994; Goodman et al., lation. We found that manufacturing firms were
1995; Zaheer and Venkatraman, 1995), we sought underrepresented in the sample by about 15 per-
to enhance the external validity of this study by cent. One explanation for this bias is that manu-
using a broader population of key informants. facturing companies tended to be larger and there-
Obtaining survey responses from corporate-level fore more bureaucratic than service companies.
managers is rather problematic, and response rates Authorization is more of a challenge to obtain
among IS executives are particularly low.2 A in a bureaucratic organization, leading to lower
common technique in surveying executives is to response rates (Tomaskovic-Devey, Leither, and
define populations and response rates based on Thompson, 1994). The sample and population did
those who will precommit to respond. Although not appear to differ by geographic location.
not necessarily reported (Heide and John, 1990, We also used a procedure suggested by Arm-
1992; Mohr and Spekman, 1994), such precom- strong and Overton (1977) to further test for
mitment rates appear to be quite low, about 10 a nonresponse bias. We compared early-returned
percent (Anderson and Narus, 1990), while the questionnaires to late-returned questionnaires on a
response rates based on the precommitted sample number of variables: respondent position, company
are at acceptable levels, greater than 40 percent. size, industry, IS attributes, and performance. The
Partly to avoid the potential for sampling bias assumption of this analysis is that late respondents
share similar characteristics and response biases
from this method, we instead chose to mail sur-
with nonrespondents. Analyses indicated that no
veys to a randomly selected set of 3000 names
significant mean differences existed between early
from the Directory of Top Computer Executives.
and late respondents. Our data on the exchange
We obtained 181 responses and 152 of these were
relations are from the buyers’ perspective. While
usable. To gauge comparability with studies that
informant bias is possible, buyers’ and suppliers’
use a precommitment technique, we performed a
perceptions of exchanges appear to be quite consis-
supplemental telephone survey of 300 names from
tent (Anderson and Narus, 1990; Heide and John,
the Directory soliciting completion precommit- 1990, 1992; Zaheer et al., 1998). Hence, we found
ments. Eleven percent responded that they would no evidence of obvious response bias in the sam-
complete surveys. Extrapolating this number to ple, other than the underrepresentation of manu-
the broader population of names suggests that facturing companies.
our response rate is quite consistent with studies
that use precommitment techniques (Anderson and 3
One of the advantages of a precommitment approach is that
bad addresses caused by turnover, location changes, and general
organizational changes are avoided. While the publishers of the
2 Directory of Top Computer Executives are quite thorough in
Rapid technological change, considerable investment in infor-
mation technology, and widespread interest in outsourcing had updating their database on a yearly basis, they estimate that more
made IS managers a common target of surveys, particularly from than half have some change to their addresses each year.
vendors. Several industry contacts indicated that at the time of 4
Note that most firms source the majority of these activities
the survey (1992) IS executives were receiving three to five internally. These internal sourcing arrangements were not part
surveys a week. of this study.

Copyright  2002 John Wiley & Sons, Ltd. Strat. Mgmt. J., 23: 707–725 (2002)
 Substitutes or Complements 715

While we believe our sample is without response 1997; Artz and Brush, 2000). Yet, to incorporate
bias, there remains a potential problem of sam- both production and governance efficiency,
ple selection bias (Heckman, 1979; Masten, 1993; we examine overall satisfaction with exchange
Poppo and Zenger, 1998; Artz and Brush, 2000). performance, rather than governance costs (see
The choice as to whether a particular exchange is Poppo and Zenger, 1998). This composite measure
internalized or outsourced is determined theoreti- is consistent with previous measurements of
cally by the independent variables (i.e., hazards) alliance performance found in the strategy
in our model. Thus, if theory holds, our sam- literature (Mohr and Spekman, 1994; Saxton,
ple of customized contracts and relational gover- 1997). The underlying logic of our composite
nance as compared to vertical integration should measure is that satisfaction is a focal consequence
have a restricted range on such variables. For of a working partnership. It is not only a close
example, exchanges with very high levels of asset proxy for concepts such as perceived effectiveness,
specificity or measurement difficulty will simply but is also predictive of future actions by partner
be internalized and therefore not exist in our firm managers (Gladstein, 1984). At the same
sample of outsourced exchanges. Calculating and time, however, by not assessing governance costs
including an inverse Mills ratio is a common directly, we are constrained in our capacity to
method for correcting sample selection bias in assess optimal levels of governance. Thus, absent
OLS models. Our model is a three-stage least costs, more relational governance or contractual
squares procedure and therefore not fully suited complexity should be preferred.
for this correction procedure. However, includ- Previous empirical studies measure partnership
ing the inverse Mills ratio may provide a good satisfaction as the level of buyers’ satisfaction
first approximation in correcting for sample selec- with exchange performance (Anderson and Narus,
tion bias. The inverse Mills ratio is calculated 1990; Mohr and Spekman, 1994; Saxton, 1997).
from a probit model predicting whether a par- The degree of satisfaction was measured using a
ticular observation is included or excluded from 7-point scale in which ‘1’ represented ‘dissatis-
the sample; in this case, whether the exchange is fied’ and ‘7’ represented ‘satisfied.’ We measured
internalized or outsourced. Fortunately, our data satisfaction with: (1) the overall cost of the ser-
include both vertically integrated and market-based vice (Proposition 1); (2) the quality of the output
exchanges and therefore permit us to calculate or service (Proposition 2); and (3) the vendor’s
this common adjustment factor, and a previous responsiveness to problems or inquiries (Proposi-
paper using this data set (Poppo and Zenger, tion 3) (Cronbach alpha = 0.84).
1998) focuses precisely on this modeling tech-
nique. The inclusion of this ratio in our three-stage
least squares model generates results consistent Relational governance
with those presented and provides some confi-
dence that such bias is not dramatically altering In this study, we view relational governance as
our results. a composite factor with the following underly-
ing norms and dimensions: open communication
and sharing of information, trust, dependence, and
Measurement cooperation. This specification is consistent with
Questionnaire items, unless stated otherwise, were previous measurement (Macneil, 1978; Anderson
measured using a 7-point scale in which ‘1’ rep- and Narus, 1990). We modified three indicators
resented ‘low degree’ and ‘7’ represented ‘high from these empirical studies and asked the key
degree.’ Table 1 presents the means and correla- informant to indicate their degree of agreement
tions for each of the measures in the study. with each of the following statements using a
7-point scale (1 = strongly disagree, 7 = strongly
agree): (1) the buyer has an extremely collabora-
Performance
tive relationship with the vendor (RG1); (2) both
When measuring exchange performance, most parties share long- and short-term goals and plans
work in transaction cost economics focuses (RG2); and (3) the buyer can rely on the ven-
on governance efficiency (Masten, Meehan and dor to keep promises (RG3) (Cronbach alpha =
Snyder, 1991; Walker and Poppo, 1991; Uzzi, 0.78).
Copyright  2002 John Wiley & Sons, Ltd. Strat. Mgmt. J., 23: 707–725 (2002)
 716

Table 1. Descriptive Statistics

AS1 AS2 SW CH1 CH2 Meas Long Tenure Budget Manu Bank Insur Contract RG1 RG2 RG3 P1 P2 P3

Copyright  2002 John Wiley & Sons, Ltd.

Correlations
AS1 1.000
L. Poppo and T. Zenger

AS2 0.701 1.000

SW 0.473 0.549 1.000
CH1 0.263 0.350 0.374 1.000
CH2 0.320 0.389 0.429 0.698 1.000
Meas −0.081 −0.062 −0.069 0.121 0.009 1.000
Long −0.015 −0.097 −0.084 0.026 0.035 −0.074 1.000
Tenure 0.037 0.078 0.192 0.037 0.059 0.069 0.065 1.000
Budget 0.123 0.223 0.226 0.138 0.125 −0.084 0.079 0.176 1.000
Manu −0.166 −0.238 −0.163 0.032 −0.070 0.127 0.046 0.070 −0.090 1.000
Banking 0.155 0.107 0.089 0.049 −0.019 0.027 0.104 −0.039 0.127 −0.212 1.000
Insur 0.164 0.137 0.115 0.072 0.075 0.071 −0.016 −0.033 0.053 −0.103 −0.037 1.000
Contract 0.181 0.249 0.173 0.009 0.072 −0.098 0.069 0.260 0.373 −0.256 −0.017 −0.037 1.000
RG1 0.025 0.048 0.064 0.120 0.084 −0.113 0.470 0.025 −0.028 0.047 0.013 −0.070 0.002 1.000
RG2 0.089 0.122 0.185 0.091 0.125 −0.092 0.346 0.140 0.163 −0.136 0.031 0.015 0.252 0.628 1.000
RG3 −0.030 −0.019 0.006 −0.005 0.004 −0.078 0.247 0.126 0.105 −0.054 −0.019 −0.145 0.043 0.586 0.456 1.000
P1 −0.196 −0.198 −0.166 −0.155 −0.176 −0.185 0.117 −0.035 −0.004 0.054 −0.203 −0.135 −0.080 0.359 0.250 0.429 1.000
P2 −0.080 −0.125 −0.100 −0.054 −0.089 −0.038 0.206 −0.024 0.121 −0.013 −0.038 −0.003 −0.034 0.471 0.319 0.567 0.565 1.000
P3 −0.078 −0.140 −0.164 −0.053 −0.049 −0.004 0.170 0.021 0.180 −0.009 −0.019 −0.111 0.055 0.418 0.290 0.583 0.536 0.740 1.000
Mean 4.377 4.474 3.967 4.476 4.439 2.442 3.818 12.062 15.475 0.364 0.077 0.026 2.963 4.532 3.854 4.983 4.667 4.948 4.707
S.D. 1.654 1.705 1.689 1.645 1.552 1.349 1.684 8.444 1.596 0.482 0.267 0.158 1.838 1.532 1.728 1.362 1.442 1.294 1.493

Note: p ≤ 0.05 for |ρ| ≥ 0.10

Strat. Mgmt. J., 23: 707–725 (2002)

 Substitutes or Complements 717

Contractual complexity Technological change

Following Macneil (1978), we measured the de- In general, information services have had high lev-
gree to which the parties created a complex con- els of technological change. Still, there is likely to
tract to deal with future contingencies by ask- be variation in the level of technological stability
ing key informants to indicate their level of across industries, firms, and services. We measured
agreement (Contract, 1 = strongly disagree, 7 = the degree of change in both skills and technol-
strongly agree) with the following statement: the ogy using two items: (1) To what degree are the
formal contract is highly customized and required underlying skills associated with this IS function
considerable legal work. The measurement of this (or set of applications) rapidly changing? (CH1);
construct by a single item is limiting. To further and (2) To what degree is the optimal configuration
validate this measure, our survey also requested of hardware and software required to perform this
respondents to indicate the length of the contract function (or set of applications) rapidly changing?
(in pages), which previous work has shown as (CH2) (Cronbach alpha = 0.84).
an indicator of contractual complexity (Joskow,
1988). While a lower response rate for this ques-
tion precluded its use as a second indicator of a Longevity of relationship
complex contract, the correlation between the two
items was significant (ρ = 0.65). A key determinant of relational governance is the
duration of time that the two parties have worked
with one another because a history of trade is
Asset specificity necessary for the development of relational norms
(Macneil, 1978, 1980). We measured this construct
Firm-specific assets were defined by human assets,
using a 7-point scale (Long, 1 = strongly disagree,
physical assets, and company-specific routines and
7 = strongly agree) and the following item: the
knowledge that were not redeployable to alterna-
buyer has worked with the vendor for years and
tive uses (Williamson, 1985). Since human capital
years. This measure is used as an identifying
is a critical component of information services,
instrument in our equation predicting relational
our measurement focused primarily on specialized
governance.
human assets, such as knowledge and skills. We
used three items to measure the degree to which the
assets used to produce an information service were
Tenure and IS budget
custom-tailored to the buying firm: (1) To what
degree must individuals acquire company-specific We use measures of the firm’s IS budget and
or division-specific information to adequately per- the IS director’s experience as identifying instru-
form the IS function? (AS1); (2) To what degree ments in the contractual complexity equation. We
is your approach to this function (or set of appli- expect that firms with greater IS resources will
cations) custom-tailored to the company? (AS2); develop more customized contracts. In our study,
and (3) How costly would it be to switch outsourc- top managers’ experience in IS proxies for firm
ing vendors? (Consider the time required to locate, knowledge about outsourcing. We expect man-
qualify, train, make investments, conduct testing, agers will heighten the complexity of contracts
and develop a working relationship) (SW) (Cron- as experience increases. In addition, we expect
bach alpha = 0.83). firms with larger IS budgets are also likely have
in-house legal counsel with experience in draft-
ing complex contracts. We have no theoretical
Measurement difficulty
basis for believing either construct is related to
Our measurement focused on the level of ease in relational-governance. Our measure of resource
measuring worker performance: To what degree scale is the natural log of size of the operating bud-
is it easy to measure the collective performance get for IS (Budget). Our measure of experience is
of those individuals who perform this function? the job tenure in years (ln) of the IS executive
(Meas, 1 = very difficult, 7 = very easy). We (Tenure). These measures are used as identify-
reverse scored this item to create a measure of ing instruments in the equation predicting contract
measurement difficulty. complexity.
Copyright  2002 John Wiley & Sons, Ltd. Strat. Mgmt. J., 23: 707–725 (2002)
718 L. Poppo and T. Zenger

Controls from a two-stage least squares estimation of each

structural equation. Finally, Stage 3 performs a
Because significant variance in industry exists for
GLS-type estimation using the covariance matrix
our sample of firms that outsource IS, we control
from Stage 2 and with the instrumented values
for potential differences that might exist in out-
replacing all endogenous right-hand side variables.
sourcing practices. We also use dummy measures
Using this system of equations, we can test
to control for the primary business in which the
for a pattern of substitution or complementar-
firm operates: banking/financial services (Bank-
ity between relational governance and contractual
ing), manufacturing (Manu), and insurance (Insur).
complexity. Primary support for complementarity
exists if relational governance positively affects
Method contractual complexity and contractual complex-
ity positively affects relational governance. Sup-
Our hypotheses require that we test a system of port for the substitution hypotheses exists if these
equations simultaneously, since both contractual relationships are negative. We also test for the
complexity and relational governance are endoge- presence of a complementary effect of contractual
nously determined. First, we must test whether complexity and relational governance on exchange
contractual complexity is influenced by the scope
performance by examining the direction of these
of exchange hazards and by the extent of relational
relationships in conjunction with the performance
governance. Second, and simultaneously, we must
equation (see Figure 1). Since our measure of per-
test whether relational governance is influenced by
formance is a satisfaction measure, which does
the scope of exchange attributes and by contractual
not directly encompass the costs of generating
complexity. Finally, we must examine the effects
performance, our prediction is that, net of costs,
of relational governance and contractual complex-
increases in contractual refinement and relational
ity on performance, while modeling the effects
governance increase exchange performance. Our
these constructs have on one another and while
controlling for exchange hazards. Specifying and test in the performance equation is thus: after
testing these three equations independently would, controlling for the direct effects of hazards on
of course, introduce significantly biased estimates exchange performance, does more relational gov-
due to the endogeneity of key independent mea- ernance, encouraged in part by greater contractual
sures in all equations and due to common distur- complexity, increase exchange performance? Sim-
bances across equations. More importantly, by cor- ilarly, after controlling for the direct effects of
rectly specifying this system of equations, we can hazards on exchange performance, does more con-
test for a pattern of substitution or complementar- tractual complexity, encouraged in part by greater
ity between relational governance and contractual relational governance, increase exchange perfor-
complexity. mance?
Our econometric approach is a simultane- An alternative test of complementarity or substi-
ous equation estimation using a three-stage least tution in the performance equation is to specify an
squares method (Greene, 1997; Stata, 1999; Nick- interaction term between relational governance and
erson, Hamilton, and Wada, 2001). This method contractual complexity in a simple OLS regression
uses instrumental variables to produce consistent model of exchange performance. This specification
estimates and generalized least squares (GLS) to allows us to test whether the marginal effects of
account for correlation in the disturbances across relational governance (or contractual complexity)
equations (Stata, 1999). Stage 1 of this proce- rise as contractual complexity (or relational gov-
dure can be thought of as producing instrumented ernance) increases. Using this method, we found a
values of all endogenous variables: in this case, strong positive and significant interaction effect,
relational governance and contractual complexity. consistent with the prediction of complementar-
These instrumented values are essentially predicted ity (Hypothesis 4a–c). However, this specifica-
values generated by the regression of each endoge- tion is problematic, because both relational gov-
nous variable on all exogenous variables in the ernance and contractual complexity are endoge-
system. Stage 2 produces a consistent estimate of nously determined. Nonetheless, the strength of
the covariance matrix of the equation disturbances. the empirical result and its consistency with the
Estimates are obtained from the residuals produced findings of the reported three-stage least squares
Copyright  2002 John Wiley & Sons, Ltd. Strat. Mgmt. J., 23: 707–725 (2002)
 Substitutes or Complements 719

model provides some confidence that our finding For relational governance (Hypothesis 2), we
of complementarity is robust. find that greater levels of relational norms accom-
pany exchanges with greater technological change
(p < 0.01). The effect of asset specificity on rela-
RESULTS tional governance is more complex. While the
main effect is not significant (see Equation 1:
Table 2 presents the results of our estimation. We M1), once an interaction term with technological
show three different model specifications. Our base change is added to the model, asset specificity now
model (see M1) contains no interaction terms for has a positive and significant effect on relational
the exchange hazards, whereas the other two model governance, and its interaction now has a posi-
specifications, M2 and M3, estimate interaction tive and negative effect on relational governance
terms. Model fit is acceptable with significant chi- (Equation 1: M2 and M3). This result suggests
square values (p < 0.001) and R 2 values ranging that relational governance has adaptive limits when
from 0.16 to 0.26 for all specifications. Interpreta- faced with coordinating changes involving both
tion of the results follows. specialized and technologically uncertain assets.
Finally, we find that measurement difficulty is not
associated with greater relational governance.
The determinants of contract customization Finally, the instruments for contract complex-
and relational governance ity, tenure of the IS Director and the IS budget,
We first examine whether increases in hazard are both significantly related to contract complex-
levels are accompanied by an increase in the ity. Similarly, the instrument for relational gover-
level of contract customization (Hypothesis 1) nance, the longevity of the relationship, is strongly
and relational governance (Hypothesis 2). The significant.
results show some support for Hypothesis 1 and
weak support for Hypothesis 2. Consistent with The complementary relationship between
Hypothesis 1, managers appear to craft more cus- contractual complexity and relational
tomized contracts as asset specificity increases (see governance
Equation 2: M3, p < 0.05).5 The effects of tech-
nological change and measurement difficulty on The critical test of the relationship, as comple-
contractual complexity are more complex. While ments or substitutes between relational governance
the two main effects and interaction terms are sig- and contractual complexity, hinges on the sign and
nificant, their signs are not consistent with Hypoth- significance of coefficients for relational gover-
nance and contractual complexity in the first two
esis 1. Consistent with Hypothesis 1, measure-
equations. Negative coefficients in Equations 1 and
ment difficulty has a significant and positive effect
2 for relational governance and contract complex-
on contract customization. Technological change,
ity, respectively, would support a substitute rela-
however, has a significant and negative effect on
tionship between relational governance and con-
contract customization. Furthermore, the interac-
tractual complexity. Positive coefficients suggest
tion of technological change and measurement dif-
a complementary relationship in which greater
ficulty has a significant and negative effect on
relational governance predicts greater contractual
contract customization. This result suggests that
complexity and greater contractual complexity pre-
managers may lose confidence in contracts, as
dicts greater relational governance. Consistent with
hazards become particularly severe (through their
our hypothesis of a complementary relationship
interaction).
and inconsistent with the substitution hypotheses
(Hypothesis 3a and b), we find that increases in the
5
Note that for M2 (Equation 2), which specifies both interaction level of relational governance are associated with
terms, the effect of asset specificity on contract customization greater levels of contractual complexity (Hypoth-
is not significant. Yet for M3 (Equation 2), which drops the
insignificant interaction term containing asset specificity and esis 4b, see Equation 2) and that increases in the
technological change, the parameter estimate for asset specificity level of contractual complexity are associated with
on contract customization is significant. We suggest that asset greater levels of relational governance (Hypothesis
specificity * technological change captures some of the same
variance as the main effect, and this accounts for the above 4a, see Equation 1). These results are robust across
results. all model specifications and suggest that managers
Copyright  2002 John Wiley & Sons, Ltd. Strat. Mgmt. J., 23: 707–725 (2002)
 Table 2. Assessing the determinants and complementarity of relational governance and contracts

Labels Determinants of Determinants of contractual Determinants of 720

relational governance complexity performance
Equation 1 Equation 2 Equation 3
M1 M2 M3 M1 M2 M3 M1 M2 M3

Contract 0.742∗∗ 0.666∗∗ 0.665∗∗ 0.675∗∗∗ 0.623∗∗∗ 0.616∗∗∗

(0.328) (0.304) (0.304) (0.256) (0.237) (0.234)
Relational 0.083∗ 0.103∗∗ 0.096∗∗ 0.256∗∗∗ 0.230∗∗ 0.241∗∗∗
Governance (0.055) (0.057) (0.052) (0.096) (0.098) (0.089)
Asset −0.020 0.407∗∗ 0.386∗∗ 0.075∗∗ 0.051 0.074∗∗ −0.259∗∗∗ −0.222∗ −0.254∗∗∗

Copyright  2002 John Wiley & Sons, Ltd.

Specificity (0.088) (0.185) (0.172) (0.035) (0.085) (0.034) (0.063) (0.136) (0.060)
Technological 0.165∗∗ 0.943∗∗∗ 0.916∗∗∗ −0.065 −0.487∗∗∗ −0.459∗∗∗ −0.012 0.553∗∗ 0.509∗∗
L. Poppo and T. Zenger

Change (0.100) (0.345) (0.334) (0.044) (0.161) (0.127) (0.074) (0.276) (0.218)
Measurement −0.128 −0.507 −0.509 −0.024 0.563∗∗∗ 0.563∗∗∗ −0.031 −0.817∗ −0.814∗∗
Difficulty (0.167) (0.438) (0.438) (0.074) (0.193) (0.193) (0.119) (0.231) (0.320)
Measurement 0.047 0.047 −0.081∗∗∗ −0.081∗∗∗ 0.108∗∗∗ 0.108∗∗∗
Difficulty × (0.058) (0.058) (0.025) (0.025) (0.042) (0.041)
Technological
Change
Asset −0.060∗∗ −0.057∗∗∗ 0.003 −0.005
Specificity × (0.023) (0.021) (0.011) (0.018)
Technological
Change
Longevity of 0.891∗∗∗ 0.850∗∗∗ 0.858∗∗∗
Relationship (0.124) (0.123) (0.121)
Tenure 0.034∗∗∗ 0.035∗∗∗ 0.035∗∗∗
(0.011) (0.011) (0.010)
Budget 0.359∗∗∗ 0.381∗∗∗ 0.383∗∗∗
(0.061) (0.062) (0.061)
Manu −0.543 −0.539 −0.547 −0.830∗∗∗ −0.793∗∗∗ −0.795∗∗∗ −0.141 −0.236 −0.240
(0.580) (0.561) (0.561) (0.244) (0.241) (0.240) (0.403) (0.389) (0.388)
Insur −1.526∗ −1.453∗∗ −1.460∗∗ 0.061 0.032 0.027 −1.385∗∗∗ −1.352∗∗∗ −1.344∗∗∗
(0.674) (0.655) (0.654) (0.301) (0.298) (0.297) (0.482) (0.468) (0.465)
Banking −0.884 −0.792 −0.802 −0.824∗ −0.973∗∗ −0.976∗∗ −0.844 −0.698 −0.701
(0.956) (0.944) (0.944) (0.410) (0.408) (0.407) (0.663) (0.654) (0.651)
Constant 6.810∗∗∗ 1.672 1.826 −4.090∗∗∗ −1.644 −1.772 8.621∗∗∗ 5.002∗∗∗ 5.182∗∗∗
(1.384) (2.723) (2.679) (1.083) (1.323) (1.251) (1.281) (1.925) (1.784)
N 285 285 285 285 285 285 285 285 285
χ2 75.06 85.51 85.57 98.73 111.45 111.76 47.51 54.38 54.81
P -value 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
R2 0.16 0.19 0.19 0.24 0.19 0.26 0.16 0.19 0.20

∗∗ ∗∗∗
Note: LSE (S.E.). One-tail t-test for hypothesized effects. ∗ p < 0.10; p < 0.05; p < 0.01

Strat. Mgmt. J., 23: 707–725 (2002)

 Substitutes or Complements 721

may complement their use of one governance tool others argue more strongly that the combined use
with the other. of relational governance and formal contracts is
fundamentally problematic, since formal controls
The effect of complements on exchange signal distrust and relational governance is based
performance on trust (Macaulay, 1963; Ghoshal and Moran,
1996; Bernheim and Whinston, 1998). Contrary
We next examine the effects of the complemen-
to this substitution position, our data are consis-
tary relationship of relational governance and con-
tent with a conclusion that contracts and relational
tractual complexity on exchange performance. Our
governance function as complements. At a mini-
system of equations takes into account the comple-
mum, these results suggest a need to explore more
mentary relationship between the two governance
carefully and predict more cautiously the rela-
forms and indicates that both relational governance
and contractual complexity deliver higher levels of tionship between formal contracts and relational
satisfaction with exchange performance (p ≤ 0.01 governance.
for each governance type). Given these positive Consistent with the notion of complements, our
effects of relational governance and contract com- results show that managers tend to employ greater
plexity on performance, and the positive effects levels of relational norms as their contracts become
that relational governance and contract complex- increasingly customized, and to employ greater
ity appear to have on one another, our system contractual complexity as they develop greater lev-
of equations confirms a complementary relation- els of relational governance. We suggest that cus-
ship with performance (Hypothesis 4c). Relational tomized contracts narrow the domain around which
governance and contractual complexity appear to parties can be opportunistic. Customized contracts
function as complements in influencing satisfaction specify contingencies, adaptive processes, and con-
with exchange performance. trols likely to mitigate opportunistic behavior and
Interpretation of our coefficients (see our base thereby support relational governance. However,
model, M1) shows that a one standard devia- customized contracts do not guarantee the intent of
tion change in contract complexity changes rela- mutuality, bilateralism, and continuance when con-
tional governance by 0.33 of a standard deviation; flict arises. Relational governance complements
whereas, a one standard deviation change in rela- such adaptive limits of contracts by fostering con-
tional governance changes contractual complexity tinuance of the exchange and entrusting both par-
by 0.18 of a standard deviation. We further cal- ties with mutually agreeable outcomes.
culate the impact of these two governance mod- Our results also suggest that contractual com-
els on performance. We find that a one standard plexity and relational governance function as com-
deviation change in relational governance changes plements in explaining satisfaction with exchange
exchange performance by 0.35 of a standard devi- performance. The system of three equations sug-
ation, and a one standard deviation change in con- gests that relational governance and contract cus-
tract changes exchange performance by 0.42 of a tomization both directly and indirectly increase
standard deviation. exchange performance as measured by satisfac-
tion with the cost, quality, and responsiveness of
the outsourced service. Contract complexity indi-
DISCUSSION AND CONCLUSION rectly increases exchange performance by increas-
ing relational governance, which in turn increases
Relational governance and contracts function exchange performance. Similarly, greater relational
as complements governance appears to positively affect contrac-
In the field of strategic management, most empir- tual complexity, which in turn increases exchange
ical and theoretical work on relational governance performance. Evidence of these indirect effects
couches it as a self-enforcing mechanism. Within supports the predicted complementary relationship.
this tradition some ignore the role of formal con- Note, however, that our performance measure gen-
tracts (Mohr and Spekman, 1994; Saxton, 1997), erally ignores the costs associated with increasing
while others view formal contracts as a more the complexity of contracts and developing rela-
costly substitute for relational governance (Gulati, tional norms, which is necessary to thoroughly test
1995b; Uzzi, 1997; Dyer and Singh, 1998). Still the substitution argument.
Copyright  2002 John Wiley & Sons, Ltd. Strat. Mgmt. J., 23: 707–725 (2002)
722 L. Poppo and T. Zenger

The determinants of relational governance and court ‘picking up the pieces’ and resolving ter-
customized contracts mination issues (Macneil, 1978). This result con-
firms the nontriviality of specialized assets and its
We advanced that familiarity, based on years of importance in the governance decision. We further
personal relationships, is necessary to develop rela- find, however, that information services that are
tionally governed exchanges. We infer that over associated with difficult performance measurement
time parties develop, test, observe, and confirm the and technological change are less likely to use
existence of trust and the other requisite norms for customized contracts; that technological change
cooperative relationship, and empirically confirm decreases the use of customized contracts; and that
that parties with such a social history increasingly difficult performance measurement increases the
rely on relational norms to govern their exchange. use of customized contracts. Finally, our results
This result is consistent with both social theorists confirm that both the tenure of the IS director and
(Granovetter, 1985, 1992; Gulati, 1995a, 1995b; the magnitude of the IS budget are associated with
Uzzi, 1997), and legal scholars (Macneil, 1978, increased contractual complexity.
1980). Overall, these results suggest that relational gov-
Our results are less consistent on whether man- ernance and contractual complexity have unique
agers select relational governance in response to origins, and such a finding confirms the comple-
exchange hazards. For the exchange hazard of mentarity between the governance forms, rather
asset specificity, our results are complex and not than their substitution. If they both had common
wholly consistent with the transaction cost logic. origins, then they would likely deliver common
Managers do not appear to select relational gov- functionality; and hence function as substitutes.
ernance in response to increasing levels of spe- Yet, our results suggest their distinct origins and
cialized assets; yet, to the contrary, when the thus roles in promoting exchange performance.
model specifies the interaction of specialized assets
and technological change, asset specificity now Limitations and extensions
leads to greater levels of relational governance.
This study has important limitations that imply
The sign on this interaction term, however, is
caution in generalizing the findings. First, the
negative, and suggests a declining relational gov-
broader institutions of nations and their legal sys-
ernance capability for such complex exchanges.
tems are likely to alter the effectiveness of formal
Interestingly, we find that technological change
contracts as governance devices. Countries with-
prompts greater levels of relational governance.
out an enforceable system of property rights can-
Managers appear to use relational governance to not rely on contracts to enforce expectations and
encourage continuance amidst the rapidly chang- promises (North and Weingast, 1989). Thus, our
ing IS technology. This result is also broadly notion of complements is not likely to generalize
consistent with the Crocker and Masten (1991) to countries that lack a cultural and legal commit-
finding that firms adopt relational contracts when ment to the use of formal contracts.
uncertainty is high. It is also consistent with the- Second, the institutional environment in which
ory: relational norms are necessary to facilitate these exchanges are embedded may be evolv-
adjustments to highly consequential disturbances, ing. At the time of this study, many outsourced
which are likely to occur with high levels of relationships in IS were considered to be in rel-
technological change (Macneil, 1978; Williamson, atively early stages of development and conse-
1991). quently the institutional ‘rules of the game’ that
Consistent with the transaction cost logic, our support trust and other informal mechanisms may
results confirm that asset specificity generates not have emerged (McFarlan and Nolan, 1995;
greater levels of contractual complexity. Because Lacity et al., 1995, 1996). The importance of con-
specialized assets have no value in alternative uses, tracts may thus decline with time, as trust emerges
if an exchange relationship is prematurely termi- in an exchange. Thus, contracts may perform a
nated, managers are not likely to have recouped critical role in the early stages of an exchange,
their investment. Through contractual complexity, but thereafter decrease in significance as patterns
managers specify penalties for early termination, of cooperative behavior and reputation emerge.
and this formal record facilitates, if need be, the We advance that the theoretical mechanisms that
Copyright  2002 John Wiley & Sons, Ltd. Strat. Mgmt. J., 23: 707–725 (2002)
 Substitutes or Complements 723

drive a complementary relationship between rela- ACKNOWLEDGEMENTS

tional governance and contractual complexity are
dynamic in nature. To fully test the dynamics of The authors particularly thank Jackson Nicker-
this complementarity, we would examine exchange son and Bart Hamilton for helpful suggestions in
relationships over time, noting the evolving nature developing this research. Joel Baum, Don Hatfield,
of the institutions that embed these relationships Bill Hesterly, Sergio Lazzarini, Julia Liebeskind,
and the pathways they pursue in seeking high per- Michael Lubatkin, Will Mitchell, Robert Gibbons,
formance (see, for example, Poppo and Lacity, Scott Stern, Zheng Zhou, two anonymous review-
2002). Our cross-sectional sample design clearly ers, and participants at the Harvard–MIT Seminar
constrains our capacity to carefully and fully exam- in Organization Economics, the BYU-Utah Win-
ine the nature of the complementary relationship ter Strategy Conference, and the LINK Seminar at
that we observe. Thus longitudinal data or experi- Copenhagen Business School have also provided
mental methods are needed to fully test the dynam- valuable comments.
ics of this complex relationship (Lazzarini, Miller,
and Zenger, 2002).
Third, our study is also limited by an inability REFERENCES
to distinguish between economic and sociological
explanations for relational governance. Disentan- Adler P. 2001. Market, hierarchy, and trust: the
gling sociologists’ more emotive, and backward- knowledge economy and the future of capitalism.
looking concept of trust from economists’ more Organization Science 12(2): 214–234.
Alchian A, Demsetz H. 1972. Production, information
calculative, forward-looking concept remains an costs, and economic organization. American Economic
important and presently unresolved empirical ques- Review 62: 777–795.
tion. Equally important, our empirical specification Anderson JC, Narus JA. 1990. A model of the
of contract customization is limited. We did not distributor’s firm and manufacturer firm working
measure, for example, the types of clauses that partnerships. Journal of Marketing 54: 42–58.
Armstrong JS, Overton T. 1977. Estimating nonresponse
managers used in their contract. Certainly, more bias in mail surveys. Journal of Marketing Research
detailed and expansive measures would improve 14: 396–402.
model specification, and the reliability and validity Artz K, Brush T. 2000. Asset specificity, uncertainty
of an empirical study. Finally, our empirical study and relational norms: an examination of coordination
costs in collaborative strategic alliances. Journal of
fails to measure a central aspect of the substitution
Economic Behavior and Organization 41: 337–362.
argument: the costs of customizing contracts and Baker G, Gibbons R, Murphy KJ. 1994. Subjective
using contracts to adapt to disturbances far exceed performance measures in optimal incentive contracts.
those of building and using relational norms for Quarterly Journal of Economics 109: 1125–1156.
adjustments. Future work needs to capture and dis- Baker G, Gibbons R, Murphy KJ. 2002. Relational
contracts and the theory of the firm. Quarterly Journal
entangle the underlying costs of such governance of Economics 117: 39–84.
structures. Balakrishnan S, Wernerfelt B. 1986. Technical change,
In sum, our study is the first to empirically competition and vertical integration. Strategic Man-
explore a complementary relationship between agement Journal 7(4): 347–359.
relational governance and formal contracts. Our Bernheim BD, Whinston MD. 1998. Incomplete con-
tracts and strategic ambiguity. American Economic
arguments and empirical conclusion depart sharply Review 88: 902–932.
from most current work in strategic management, Bradach J, Eccles R. 1989. Price, authority, and trust:
which highlights relational governance and trust as from ideal types to plural forms. Annual Review of
a substitute for contracts. Further work is clearly Sociology 15: 97–118.
Clemons EK, Row MC. 1991. Sustaining IT advantage:
needed to explore the relationship among the fea- the role of structural differences. MIS Quarterly,
tures of governance that support interorganiza- September: 275–292.
tional relations. In particular, we see a need to Crocker KJ, Masten SE. 1991. Pretia ex machina? Prices
explore the relationship between relational gov- and process in long-term contracts. Journal of Law and
ernance and formal contracts over the life of an Economics 34: 69–99.
Dore R. 1983. Goodwill and the spirit of market
exchange relationship, and to develop more pre- capitalism. British Journal of Sociology 34: 459–482.
cise measures of contractual clauses and of the Dyer J. 1996. Does governance matter? Keiretsu
underlying costs of these governance structures. alliances and asset specificity as sources of Japanese
Copyright  2002 John Wiley & Sons, Ltd. Strat. Mgmt. J., 23: 707–725 (2002)
724 L. Poppo and T. Zenger

competitive advantage. Organization Science 7: Joskow P. 1988. Asset specificity and the structure of
649–666. vertical relationships: empirical evidence. Journal of
Dyer J. 1997. Effective interfirm collaboration: how firms Law, Economics, and Organization 4: 95–118.
minimize transaction costs and maximize transaction Klein BA. 1996. Why hold-ups occur: the self-enforcing
value. Strategic Management Journal 18(7): 535–556. range of contractual relationships. Economic Inquiry
Dyer J, Singh H. 1998. The relational view: cooperative 34: 444–463.
strategy and sources of interorganizational competitive Klein BA, Crawford A, Alchian A. 1978. Vertical
advantage. Academy of Management Review 23: integration, appropriable rents, and the competitive
660–679. contracting process. Journal of Law, Economics, and
Fehr E, Gachter S. 2000. Do incentive contracts crowd Organization 7: 972–991.
out voluntary cooperation? Working paper no. 34, Lacity M, Willcocks L, Feeny D. 1995. IT outsourcing:
Institute for Empirical Research in Economics, maximize flexibility and control. Harvard Business
University of Zurich. Review 73(3): 84–93.
Gargiulo M, Benassi M. 2000. Trapped in your own Lacity M, Willcocks L, Feeny D. 1996. The value of
net? Network cohesion, structural holes, and the selective IT sourcing. Sloan Management Review
adaptation of social capital. Organization Science Spring: 13–32.
11(2): 183–196. Larson A. 1992. Network dyads in entrepreneurial
Ghoshal S, Moran P. 1996. Bad for practice: a critique of settings: a study of governance of exchange
the transaction cost theory. Academy of Management relationships. Administrative Science Quarterly 37:
Review 21: 13–47. 76–104.
Gladstein DL. 1984. Groups in context: a model of task Lazzarini SG, Miller GJ, Zenger TR. 2002. Order with
group effectiveness. Administrative Science Quarterly a little bit of law: complementarity vs. substitution
29: 499–517. between formal and informal arrangements. Working
Goodman P, Fichman M, Lerch FJ, Snyder P. 1995. paper, Olin School of Business, Washington Univer-
Customer-firm relationships, involvement, and cus- sity.
tomer satisfaction. Academy of Management Journal Macaulay S. 1963. Non-contractual relations in business:
38(5): 1310–1324. a preliminary study. American Sociological Review
Granovetter M. 1985. Economic action and social 28: 55–69.
structure: the problem of embeddedness. American Macneil IR. 1978. Contracts: adjustment of long-term
Journal of Sociology 91(3): 481–510. economic relations under classical, neoclassical and
Granovetter M. 1992. Problems of explanation in relational contract law. Northwestern University Law
economic sociology. In Networks and Organizations: Review 72: 854–905.
Structure, Form, and Actions, Nohria N, Eccles RG Macneil IR. 1980. The New Social Contract: An Inquiry
(eds). Harvard Business School Press: Boston, MA; into Modern Contractual Relations. Yale University
25–56. Press: New Haven, CT.
Greene WH. 1997. Econometric Analysis (3rd edn). Masten S. 1993. Transaction costs, mistakes, and
Prentice-Hall: Upper Saddle River, NJ. performance: assessing the importance of gover-
Gulati R. 1995a. Social structure and alliance formation nance. Managerial and Decision Economics 14:
patterns: a longitudinal analysis. Administrative 119–129.
Science Quarterly 40: 619–652. Masten S. 1996. Empirical research in transaction cost
Gulati R. 1995b. Does familiarity breed trust? The economics: challenges, progress, directions. In Trans-
implications of repeated ties for contractual choice action Cost Economics and Beyond , Grenewege J
in alliances. Academy of Management Journal 38: (ed.). Kluwer Academic: Boston, MA; 43–64.
85–112. Masten S, Meehan J, Snyder E. 1991. The costs
Heckman J. 1979. Sample selection bias as a specification of organization. Journal of Law, Economics, and
error. Econometrica 47: 153–161. Organization 7: 1–25.
Heide J. 1994. Interorganizational governance in market- Mayer K. 2000. The evolution of complex buyer-supplier
ing channels. Journal of Marketing 58: 71–85. relationships: contracts as repositories of inter-
Heide J, John G. 1990. Alliances in industrial purchasing: organizational learning in buyer–supplier relationship
the determinants of joint action in buyer-supplier in high technology industries. PhD dissertation, Haas
relationships. Journal of Marketing Research 27: School of Business, University of California at
24–36. Berkeley.
Heide J, John G. 1992. Do norms matter in marketing McFarlan W. 1990. The 1990s: the information decade.
relationships? Journal of Marketing 56: 32–44. Business Quarterly 55: 73–79.
Hill C. 1990. Cooperation, opportunism, and the invisible McFarlan W, Nolan R. 1995. How to manage an
hand: implications for transaction cost theory. IT outsourcing alliance. Sloan Management Review
Academy of Management Review 15: 500–513. Winter: 9–23.
Jones C, Hesterly W, Borgatti S. 1997. A general theory Mohr J, Spekman R. 1994. Characteristics of partnership
of network governance: exchange conditions and success: partnership attributes, communication behav-
social mechanisms. Academy of Management Review ior, and conflict resolution techniques. Strategic Man-
22: 911–945. agement Journal 15(2): 135–142.
Copyright  2002 John Wiley & Sons, Ltd. Strat. Mgmt. J., 23: 707–725 (2002)
 Substitutes or Complements 725
Nickerson JA, Hamilton BH, Wada T. 2001. Market Stata. 1999. Reference Manual: Release 6. Stata: College
position, resource profile, and governance: linking Station, TX.
Porter and Williamson in the context of international Strickland LH. 1958. Surveillance and trust. Journal of
courier and small package services in Japan. Strategic Personality 26: 200–215.
Management Journal 22(3): 251–273. Tomaskovic-Devey D, Leither J, Thompson S. 1994.
Noordewier T, John G, Nevin J. 1990. Performance Organizational survey nonresponse. Administrative
outcomes of purchasing arrangements in industrial Science Quarterly 39: 439–457.
buyer-vendor relationships. Journal of Marketing 54: Uzzi B. 1997. Social structure and competition in
80–93. interfirm networks: the paradox of embeddedness.
North D, Weingast B. 1989. Constitutions and commit- Administrative Science Quarterly 42: 35–67.
ment: the evolution of institutions governing pub- Walker G, Poppo L. 1991. Profit centers, single-
lic choice in seventeenth-century England. Journal of source suppliers, and transaction costs. Administrative
Economic History 49: 803–832. Science Quarterly 42: 35–67.
Palay T. 1984. Comparative institutional economics: the Williamson O. 1985. The Economic Institutions of
governance of rail freight contracting. Journal of Legal Capitalism. Free Press: New York.
Studies 13: 265–287. Williamson O. 1991. Comparative economic organiza-
Poppo L, Lacity M. 2002. The normative value of tion: the analysis of discrete structural alternatives.
transaction cost economics: what managers have Administrative Science Quarterly 36: 269–296.
learned about TCE principles in the IT context. Williamson O. 1996. Transaction cost economics and
In Information Systems Outsourcing in the New organization theory. In The Handbook of Economic
Economy, Hirschheim R, Heinzl A, Dibbern J (eds). Sociology, Smelser NJ, Swedberg R (eds). Princeton
Springer-Verlag: New York; 253–276. University Press: Princeton, NJ; 77–107.
Poppo L, Zenger T. 1998. Testing alternative theories of Zaheer A, Venkatraman N. 1995. Relational governance
the firm: transaction cost, knowledge-based, and mea- as an interorganizational strategy: an empirical test
surement explanations for make-or-buy decisions in of the role of trust in economic exchange. Strategic
information services. Strategic Management Journal Management Journal 16(5): 373–392.
19(9): 853–877. Zaheer A, McEvily B, Perrone V. 1998. Does trust
Saxton T. 1997. The effects of partner and relationship matter? Exploring the effects of interorganizational
characteristics on alliance outcomes. Academy of and interpersonal trust on performance. Organization
Management Journal 40: 443–461. Science 9: 141–159.

Copyright  2002 John Wiley & Sons, Ltd. Strat. Mgmt. J., 23: 707–725 (2002)