Description
This paper develops a structural model to simultaneously analyze firms’ partner selection activities and governance
design for managing interfirm relationships. The model predicts these choices to be influenced by appropriation concerns,
coordination requirements, dependence and prior experiences with the exchange partner. An analysis of 817
information technology transactions shows that appropriation concerns, coordination requirements and dependence
influence governance design, while buyers’ partner selection effort relates mainly to appropriation concerns
Partner selection and governance design
in inter?rm relationships
Henri C. Dekker
*
Vrije Universiteit Amsterdam, Department of Accounting, De Boelelaan 1105, 1081 HV Amsterdam, The Netherlands
Abstract
This paper develops a structural model to simultaneously analyze ?rms’ partner selection activities and governance
design for managing inter?rm relationships. The model predicts these choices to be in?uenced by appropriation con-
cerns, coordination requirements, dependence and prior experiences with the exchange partner. An analysis of 817
information technology transactions shows that appropriation concerns, coordination requirements and dependence
in?uence governance design, while buyers’ partner selection e?ort relates mainly to appropriation concerns. Partner
experience developed during prior ties with the exchange partner is found to moderate the e?ects of these control prob-
lems. In addition, partner selection and formal governance are found to be used as complements instead of substitutes
in coping with inter?rm control problems.
Ó 2007 Elsevier Ltd. All rights reserved.
Introduction
Recognizing the importance of adequate gover-
nance structures for the management, control and
success of inter?rm relationships, organizational
researchers have devoted considerable e?ort to
explaining ?rms’ governance choices. Informed
by multiple theoretical perspectives, these studies
indicate the importance of adapting governance
design to di?erent conditions of the inter?rm
relationship, such as to transaction and task char-
acteristics. Designing governance structures, how-
ever, is not the only way in which ?rms attempt
to achieve control over their cooperative activities.
Several studies, for instance, point to the impor-
tance of experiences and trust from prior ties as
alternative social control mechanisms for coping
with transaction risks (e.g., Batenburg, Raub, &
Snijders, 2003; Dekker, 2004; Gulati, 1995; Gulati
& Singh, 1998; Tomkins, 2001). Another way that
?rms can mitigate control problems in inter?rm
relationships is by selecting an appropriate part-
ner. Some recent studies suggest that selecting a
‘good’ partner limits the need for governance
0361-3682/$ - see front matter Ó 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.aos.2007.02.002
*
Tel.: +31 20 598 6066; fax: +31 20 598 9870.
E-mail address: [email protected]
www.elsevier.com/locate/aos
Available online at www.sciencedirect.com
Accounting, Organizations and Society 33 (2008) 915–941
(Buskens, Batenburg, & Weesie, 2003), and that it
is essential for the e?ective management and suc-
cess of inter?rm relationships (Hitt, Dacin, Lev-
itas, Arregle, & Borza, 2000; Ireland, Hitt, &
Vaidyanath, 2002). In addition, the partner selec-
tion phase can strongly in?uence later stages of
the collaboration, including the use of governance
arrangements (Dekker, 2004). Despite its impor-
tance for the control of inter?rm relationships,
however, the partner selection process has received
scant attention in empirical studies. Recognizing
this void, this paper aims to: (1) provide an under-
standing of how ?rms use both partner selection
and governance arrangements to cope with poten-
tial control problems when engaging in a transac-
tion, and (2) examine how prior transaction
experiences in?uence these selection and design
choices. The study thus focuses on how social con-
trol mechanisms complement, or even substitute,
the use of governance arrangements in governing
and coordinating transactions between exchange
partners (Ouchi, 1980). Fig. 1 summarizes the
key ideas developed in this paper and suggests that
anticipated control problems in?uence both part-
ner selection e?ort and use of governance arrange-
ments, while prior experiences with the exchange
partner in?uence these associations.
The contribution of this study to the literature
on governance and control in inter?rm relation-
ships is threefold. First, the structural model that
is developed and empirically examined provides
an explanation of the extensiveness of the gover-
nance structure used by ?rms engaging in a trans-
action. As shown in Fig. 1, this explanation
employs the concepts of appropriation concerns,
coordination requirements, dependence and partner
experience. Appropriation concerns originate from
transaction characteristics, such as asset speci?c-
ity, uncertainty and frequency, and essentially
relate to the risk of opportunism by transaction
partners. Coordination requirements result from
task characteristics, in particular task interdepen-
dence, which creates a need for mutual adaptation
and coordination between partners. Dependence is
a factor that extends beyond the transaction
between exchange partners and relates to their eco-
nomic relationship, which may also involve other
(current and future) transactions. Jointly, these
factors generate a need for governance to mitigate
and manage the control problems that they create.
Second, the study investigates the e?ects of partner
selection on the use of governance arrangements,
arguing that partner selection contributes to miti-
gating concerns about the control problems of
appropriation, coordination and dependence. In
addition, it is put forward that a more extensive
partner selection process is also associated with
learning about the transaction and exchange part-
ner, which enhances the design of a more extensive
governance structure. As Fig. 1 illustrates, the
structural model provides an explanation of the
e?ort invested by the buyer in supplier selection
following from these control problems, and of its
consequent e?ects on the use of governance
arrangements. Third, the study investigates how
prior partner experiences in?uence the association
between control problems and both the e?ort
invested in partner selection and the use of gover-
nance arrangements. Partner experience essentially
relates to familiarity between exchange partners
generated during prior ties, which may have a sub-
stantial impact on governance choices. While most
prior studies have emphasized a direct e?ect of
prior ties on the use of governance arrangements,
arguing that experience and trust generated during
prior ties allow the use of less extensive governance
structures, this study provides support for a mod-
erating e?ect of partner experience. This moderat-
ing e?ect suggests that the e?ects of control
problems on partner selection e?ort and on gover-
nance extensiveness vary with the extent to which
exchange partners have prior experiences. More
speci?cally, the ?ndings indicate that the increased
familiarity from prior ties enables a reduction in
the use of governance arrangements for managing
Control problems
•Appropriation concerns
•Coordination requirements
•Dependence
Partner
experience
Governance
extensiveness
Selection
effort
Fig. 1. Key constructs of the study.
916 H.C. Dekker / Accounting, Organizations and Society 33 (2008) 915–941
concerns about dependence and appropriation,
while it simultaneously helps to develop more elab-
orate governance arrangements for managing
coordination requirements. This also implies that
partner experience gains in importance for the
management of inter?rm transactions as potential
control problems increase in magnitude.
The structural model is examined using detailed
survey data of information technology (IT) trans-
actions between small-to-medium enterprises and
their suppliers. The empirical tests focus in partic-
ular on the e?ort invested by buyers in the selec-
tion of an adequate partner and the use of
(contractual) agreements between exchange part-
ners. Studies of IT outsourcing report that buyers
in these relationships frequently experience sub-
stantial control problems, such as opportunistic
behavior, low supplier capabilities, high costs,
poor quality and service, loss of control and exper-
tise, and over-dependence on the supplier (e.g.,
Barthelemy, 2001). These problems relate to the
complicated and uncertain nature of IT, providing
conditions for market failure (Ouchi, 1980; Wil-
liamson, 1975). Software development, for
instance, typically involves information asymme-
try between developer and user, di?culties in mon-
itoring progress and outcomes, and a need for
intensive cooperation and coordination between
developer and user in order to understand and pri-
oritize issues which a?ect the software’s quality
and success (Wang, Baron, & Seidman, 1997).
Hence, the interdependencies between exchange
partners and the uncertainty in determining the
value of exchanged products and services can
entail considerable transaction costs (Ouchi,
1980). Selecting a reliable and competent partner
and developing an adequate governance structure
are important ways for ?rms to curb potential
opportunism and to facilitate coordination across
?rm boundaries.
The next section develops the research model by
reviewing prior studies of inter?rm governance
design. This review discusses the in?uence of
appropriation concerns and coordination require-
ments on the design of governance structures, the
in?uence of the relational context between ?rms
and the in?uence of the partner selection process
on governance design. The subsequent sections
discuss the design, data and results of the empirical
study. The ?nal section concludes the paper by
proposing directions for future research, based
on the ?ndings and limitations of the study.
The governance of inter?rm relationships
Literature on inter?rm governance describes
two key ways in which ?rms can cope with control
problems that arise in inter?rm relationships: (1)
by selecting a ‘good’ partner, and (2) by designing
an appropriate governance structure (Dekker,
2004). First, after determining that cooperation
with an outside entity is preferable to internal pro-
duction or vertical integration, ?rms need to
choose an appropriate partner (Ireland et al.,
2002). Having selected a partner, ?rms then need
to design and implement a governance structure
that includes arrangements to adequately govern
the collaboration. While the partner selection
phase essentially focuses on mitigating control
problems by ?nding a reliable and competent part-
ner, the design of a governance structure focuses
on both mitigation and management of control
problems. Fig. 2 presents the structural model
and research hypotheses that are developed in
Uncertainty
Task
interdependence
Dependence
Selection
effort
Governance
extensiveness
H7a/b
Partner
experience
H1a
Size
Asset
specificity
H1b
H1c
H2
H3
H8a
H8b
H8c
H9
H4-H6
H11-H13
H10
Fig. 2. The structural model.
H.C. Dekker / Accounting, Organizations and Society 33 (2008) 915–941 917
the next sections, following the framework in Dek-
ker (2004). The ?gure shows how variables which
generate appropriation concerns, coordination
requirements and dependence are expected to
in?uence both the e?ort invested in partner selec-
tion and the extensiveness of the governance struc-
ture for an inter?rm transaction. Furthermore, it
shows the expectation that partner experience will
moderate the in?uence of these control problems
on selection e?ort and on governance extensive-
ness. Finally, it shows that the e?ort invested in
partner selection is expected to a?ect governance
extensiveness.
The antecedents of inter?rm governance:
appropriation concerns and coordination
requirements
Governance structures in inter?rm relationships
can comprise a broad set of control mechanisms,
varying from contractual arrangements to non-
contractual organizational mechanisms. These
mechanisms enable control by ?at, by aligning
partners’ incentives, and by monitoring and
rewarding outcomes and behaviors (Dekker,
2004).
1
Similar to Anderson and Dekker (2005),
in its analysis of governance extensiveness this
paper examines the extent to which exchange part-
ners in IT transactions use contractual agreements
to govern their transaction. These agreements
relate to issues such as product speci?cation, pric-
ing, service levels, decision rights and responsibili-
ties, and con?ict resolution. In addition, the
analysis of governance extensiveness also includes
the use of non-contractual agreements between
exchange partners regarding these issues, which
are often used to complement (incomplete) con-
tractual agreements.
Theoretically, prior studies of inter?rm gover-
nance design have predominantly used transaction
cost economics (TCE) to explain ?rms’ governance
choices (e.g., Anderson & Dekker, 2005; Ander-
son, Glenn, & Sedatole, 2000; Dekker, 2003,
2004; Gulati, 1995; Gulati & Singh, 1998; Lang-
?eld-Smith & Smith, 2003; Van der Meer-Kooistra
& Vosselman, 2000). TCE posits that exchange
hazards arise as a function of transaction charac-
teristics and that increasing concerns about these
exchange hazards induce ?rms to use governance
structures with more elaborate problem solving
abilities. Asset speci?city refers to the degree to
which resources dedicated to a transaction can be
re-deployed in other transactions without loss of
value or high cost (Williamson, 1985). Because of
the potential costly consequences of opportunism,
increasing asset speci?city generates a need for
more extensive governance to promote incentive
alignment, e?cient adaptation and transaction
cost mitigation (Masten, 2000). The transaction
size augments exchange hazards by since a larger
transaction increases the investing ?rm’s exposure
to opportunistic holdup, and increases the part-
ner’s incentives to behave opportunistically (Blum-
berg, 2001a, 2001b; Buskens, 2002). Firms facing a
larger potential loss are therefore likely to prefer
more extensive governance structures in order to
safeguard their investment (Batenburg et al.,
2003).
Uncertainty is a third key ingredient of appro-
priation concerns, since it creates information
asymmetry between exchanges partners and gener-
ates a need to adapt to unforeseeable changes.
Important causes of uncertainty are the environ-
ment, such as market and technological ?uctua-
tions, and the complexity and unpredictability of
tasks. These uncertainties cause di?culties in mon-
itoring performance and generate a need for gover-
nance to moderate the risk that the transaction
partner will exploit these monitoring problems
(Anderson & Dekker, 2005; Batenburg et al.,
2003; Blumberg, 2001a, 2001b; Buskens, 2002;
1
Gulati and Singh (1998), for instance, discussed ?ve
important types of hierarchical control mechanisms which are
frequently used in inter?rm alliances: command structures and
authority systems, incentive systems, standard operating pro-
cedures, dispute resolution procedures and non-market pricing
systems. Typically, empirical studies on inter?rm governance
choices are conducted at a high level of analysis, such as the
choice between market, hybrid and hierarchy, or between
equity and non-equity alliances, which are argued to di?er in
the use of control mechanisms. However, a variety of structures
is also used within these governance types, which di?er in their
mechanisms and ability to coordinate. A focus on the actual
content of governance structures, of which contract agreements
are essential constituents, therefore o?ers an important way to
enhance theory development and empirical testing (Poppo &
Zenger, 2002).
918 H.C. Dekker / Accounting, Organizations and Society 33 (2008) 915–941
Leiblein & Miller, 2003; Poppo & Zenger, 2002).
Although uncertainty warrants a greater reliance
on governance mechanisms in order to manage
the associated appropriation concerns, by its own
nature it also reduces the possibilities to design
and use controls based on monitoring (Blumberg,
2001a). Poppo and Zenger (2002) therefore argue
that when measurement di?culties in IT outsourc-
ing increase, managers are likely to spend more
resources on developing more complex contracts
which specify required service levels or facilitate
behavior monitoring, and which enable to measure
and reward output and productivity. Following
these TCE arguments, it is hypothesized that gov-
ernance extensiveness (i.e., the extent to which
exchange partners use contractual and non-con-
tractual agreements) increases with increasing
exchange hazards:
H1: The greater the appropriation concerns for an
inter?rm transaction, as indicated by its (a)
size, (b) asset speci?city, and (c) uncertainty,
the more extensive its governance structure.
While a vast amount of empirical evidence sup-
ports the view that exchange hazards contribute to
explaining inter?rm governance structures (e.g.,
Geyskens, Steenkamp, & Kumar, 2006; Masten,
2000), managing potential opportunism is not the
only purpose for which governance structures are
designed. Gulati and Singh (1998) found support
for the in?uence of a second important control
problem to be managed in strategic alliances: the
coordination of interdependent tasks. A need for
coordination arises when, in the pursuit of value
creation, ?rms engage in collaborative activities
involving interdependent tasks. A higher level of
task interdependence increases ?rms’ reliance
upon each other in the performance of tasks and
creates a need for coordination and mutual adap-
tation (Thompson, 1967). Consistent with this,
Gulati and Singh (1998) found that strategic alli-
ances which performed more interdependent activ-
ities used more complex governance structures
suited to facilitating this coordination. Similarly,
Dekker (2004), Gulati, Lawrence, and Puranam
(2005) and White and Lui (2005) found that, in
addition to con?icting interests and incentive
problems, coordination requirements in?uence
governance and control choices in inter?rm rela-
tionships.
2
Accordingly, it is hypothesized that:
H2: The greater the coordination requirements for
an inter?rm transaction, as indicated by its
task interdependence, the more extensive its
governance structure.
The in?uence of relationship characteristics on
inter?rm governance: dependence and partner
experience
While transaction and task characteristics gen-
erate concerns about potential appropriation and
requirements for coordination, the economic and
social context of the relationship between
exchange partners can signi?cantly alter their
response to these control problems. Two impor-
tant contextual variables relating to this economic
and social relationship are dependence and partner
experience. First, in their design of a governance
structure, transacting ?rms may seek to mitigate
concerns of dependence and power asymmetries
(Gulati & Singh, 1998; Pfe?er & Salancik, 1978).
While resource considerations and complementar-
ity are important drivers of relationship forma-
tion, they can also lead to increasing economic
dependence of ?rms upon others. The dependence
of one ?rm upon another, and the associated
power di?erences, relate to the relationship
between exchange partners, not to a particular
transaction per se; nevertheless, they are likely to
in?uence speci?c transactions between them. For
instance, even when a transaction comprises non-
critical and non-speci?c resources, the buyer may
still be highly dependent, because of other
(planned) critical transactions with the supplier.
This dependence position gives the supplier a
power advantage, and is likely to induce the buyer
to employ more safeguards for the focal transac-
tion. Hence, it is hypothesized that:
2
Similar to its in?uence on appropriation concerns, task
uncertainty can augment the coordination problem by creating
monitoring problems and hindering e?cient coordination
(Leiblein & Miller, 2003).
H.C. Dekker / Accounting, Organizations and Society 33 (2008) 915–941 919
H3: The greater the dependence of a ?rm on a
partner, the more extensive the governance
structure for an inter?rm transaction with this
partner.
Second, the social context in which transactions
take place can lead to familiarity and informal
controls between exchange partners, which can
signi?cantly in?uence their governance choices
(e.g., Batenburg et al., 2003; Buskens et al., 2003;
Dekker, 2003, 2004; Lang?eld-Smith & Smith,
2003; Gulati, 1995; Gulati & Gargiulo, 1999; Tom-
kins, 2001; Uzzi, 1997). As in earlier studies (e.g.,
Gulati, 1995; Gulati & Singh, 1998), this study
focuses on prior ties between exchange partners,
during which a level of partner experience is gener-
ated by directly experiencing each other’s behav-
iors and performance. Repeated interactions over
time can contribute strongly to learning about
partners, since they provide ?rms with unique
?rst-hand information about the other’s reliability
and competence (Blumberg, 2001b; Gulati & Gar-
giulo, 1999). The possibility of terminating the
relationship and withdrawing future business fur-
ther promotes ‘calculative trust’ based on deter-
rence (Gulati & Singh, 1998).
Prior studies which examined the e?ects of part-
ner experience focused mainly on the development
of social ties during interactions, often referred to
as ‘relational trust’, which is argued to either sub-
stitute or complement the use of governance
arrangements (e.g., Das & Teng, 1998; Dekker,
2004; Gulati, 1995; Poppo & Zenger, 2002). Sub-
stitution can occur since enhanced social ties are
associated with reduced goal con?ict, limiting the
need for governance to deter opportunism. In
addition, use of governance arrangements, such
as contractual agreements, may also signal distrust
and potentially damage the relationship (Das &
Teng, 1998; Dekker, 2004). A complementary rela-
tionship implies that increased familiarity and gov-
ernance lead to an overall higher level of control,
and it is suggested that more extensive governance
can actually enhance social ties by narrowing the
domain and severity of risk (Poppo & Zenger,
2002) and by providing objective information
about the partner’s performance, behaviors and
skills (Das & Teng, 1998). Empirically, several
studies found that prior ties between exchange
partners are associated with the use of less exten-
sive governance structures (e.g., Batenburg et al.,
2003; Buskens, 2002; Gulati, 1995; Gulati & Singh,
1998). However, Poppo and Zenger (2002) found a
positive association between trust-based relational
governance and the customization of contracts for
information services exchanges. By providing evi-
dence for both perspectives, these studies point
to the need to analyze under which circumstances
a substitutive or complementary relationship
between partner experience and governance exten-
siveness is more likely.
Whereas most prior studies predict and test a
direct in?uence of prior ties and trust on gover-
nance design, Dirks and Ferrin (2001) argued that
in many organizational settings trust has moderat-
ing e?ects on organizational outcome variables. In
in?uencing governance design, a moderating e?ect
of partner experience seems more plausible than a
direct e?ect, as it is not the (lack of) familiarity or
(dis)trust which generates a need for governance,
but the potential control problems associated with
a transaction (Dekker, 2004). In the absence of
any control problems, partner experience is unli-
kely to in?uence governance design at all, since
there is no need for governance. Thus only in the
presence of potential control problems can partner
experience be expected to a?ect the use of gover-
nance arrangements, by altering ?rms’ responses
to control problems. Consequently, the association
between control problems and governance exten-
siveness is expected to vary with partner experi-
ence. A substitutive e?ect then implies that in
relationships with increased familiarity, control
problems have a weaker e?ect on the use of gover-
nance arrangements than in newly established rela-
tionships in which partners are unfamiliar with
each other. This leads to the expectation that
appropriation concerns have a weaker in?uence
on governance extensiveness when exchange part-
ners have prior mutual experience.
A second issue concerning how partner experi-
ence a?ects governance design relates to the nature
of the control problems to be managed (i.e., the
purposes of control; cf. Dekker, 2004). While prior
studies focused mainly on how partner experience
in?uences the use of governance arrangements for
920 H.C. Dekker / Accounting, Organizations and Society 33 (2008) 915–941
managing appropriation concerns, increased
familiarity with the exchange partner may also
in?uence the e?ects of coordination requirements
and dependence. Gulati and Singh (1998), for
instance, argued that trust and learning from prior
ties can strongly in?uence coordination costs.
Similarly, Uzzi (1997) found that the embedded
relationships he studied were characterized by
?ne-grained information transfer and joint prob-
lem solving arrangements, facilitating informal
modes of coordination. Poppo and Zenger (2002)
further argued that as relationships are developed
and sustained, partners will include lessons learned
in the past in their contract, thereby formalizing
experiences, mutually agreed processes, informa-
tion sharing, and performance measurement and
monitoring practices. These arguments thus indi-
cate that increased familiarity may not only reduce
the need for coordination mechanisms, but may
also enhance partners’ abilities to jointly design
and implement such mechanisms, leading to more
extensive governance (Dekker, 2004). They also
again point to moderating e?ects of partner expe-
rience, which is likely to alter the in?uence of coor-
dination requirements on governance design. In
addition, they imply the possibility of both substi-
tutive and complementary e?ects of partner experi-
ence occurring simultaneously, when the increased
familiarity reduces the need for governance to
manage appropriation concerns, while it enhanced
the design of coordination mechanisms. Finally,
similar to its e?ect on appropriation concerns,
partner experience is also likely to weaken the
e?ects of dependence on governance extensiveness,
since ?rms who are more familiar with each other
will expect their partner not to use their power
advantage.
Hypotheses 4, 5 and 6 describe the expected
moderating e?ects of partner experience on the
in?uence of appropriation concerns, coordination
requirements and dependence respectively. In
Fig. 2 these moderating e?ects are shown by the
dotted line crossing the e?ects of control problems
on governance extensiveness.
H4: Partner experience weakens the association
between the appropriation concerns for an
inter?rm transaction, as indicated by its (a)
size, (b) asset speci?city, and (c) uncertainty,
and the extensiveness of the governance struc-
ture of the inter?rm transaction.
H5: Partner experience strengthens the association
between the coordination requirements for
an inter?rm transaction, as indicated by its
task interdependence, and the extensiveness
of the governance structure of the inter?rm
transaction.
H6: Partner experience weakens the association
between the dependence of a ?rm on a partner
and the extensiveness of the governance struc-
ture for an inter?rm transaction with this
partner.
The in?uence of partner selection on inter?rm
governance
While governance structures are important for
the management of appropriation concerns, coor-
dination requirements and power asymmetries, the
partner selection phase which precedes governance
design can have a substantial in?uence by mitigat-
ing potential control problems. Ireland et al.
(2002) therefore argued that e?ective alliance man-
agement starts with the selection of the ‘right’ part-
ner. Selecting an appropriate partner is a critical
decision for the success of a relationship; not only
for gaining access to critical and complementary
resources (e.g., Hitt et al., 2000), but also for e?ec-
tive management of the cooperation. Conse-
quently, instead of attempting to explain
governance choices, a few recent empirical studies
focused on a di?erent dependent variable: the
searching for partners with whom to transact
(Blumberg, 2001b; Buskens et al., 2003). The selec-
tion of a reliable and competent partner can
reduce ?rms’ concerns about appropriation and
coordination for the transaction in which they
wish to engage and consequently reduce the need
to design and implement a governance structure
(Buskens et al., 2003; Dekker, 2004). Provided that
buyers have a choice among potential partners,
investing more e?ort in evaluating and comparing
potential partners may increase con?dence in the
goodwill and capabilities of the chosen partner,
thus reducing the need for governance. Alterna-
tively, during a more extensive search process,
H.C. Dekker / Accounting, Organizations and Society 33 (2008) 915–941 921
?rms also acquire knowledge about the chosen
partner and product, which can facilitate the
development of more detailed agreements for
managing the transaction. This argument suggests
that the selection process functions as a learning
process which enables the design of enhanced gov-
ernance structures. Similar to the e?ects of partner
experience, these arguments raise the question
whether partner selection e?ort and use of gover-
nance arrangements are substitutes or comple-
ments for coping with inter?rm control
problems. Despite its potential for mitigating con-
trol problems and in?uencing governance design,
the partner selection behavior of ?rms has not
received much attention in empirical research.
Given the ambiguity about whether partner selec-
tion substitutes or complements governance exten-
siveness, two competing hypotheses are for-
mulated:
H7a: The greater the e?ort spent on partner selec-
tion for an inter?rm transaction, the less
extensive its governance structure.
H7b: The greater the e?ort spent on partner selec-
tion for an inter?rm transaction, the more
extensive its governance structure.
The antecedents of partner selection e?ort:
appropriation concerns, coordination requirements
and dependence
In a similar way to governance design, investing
time and resources in partner selection gains in
importance when potential control problems for
a transaction increase in magnitude (Batenburg
et al., 2003; Blumberg, 2001b; Buskens et al.,
2003; Dekker, 2004). Consistent with this argu-
ment, Blumberg (2001b) found that with increas-
ing exchange hazards, as re?ected by transaction
size, asset speci?city and environmental uncer-
tainty, ?rms increased their search e?orts to ?nd
a reliable partner for technology alliances. Simi-
larly, increasing coordination requirements are
likely to induce ?rms to invest more e?ort in
search and selection of a competent partner, in
order to avoid coordination costs and perfor-
mance problems (Dekker, 2004). Although an
increasing level of interdependence requires more
coordination e?ort, this e?ort may be less exten-
sive when cooperating with a more competent
partner who requires less guidance and monitor-
ing. In addition, cooperating with a more compe-
tent partner may result in coordination which is
not only more e?cient but also more e?ective, thus
producing superior outcomes.
The e?ect of concerns about power asymmetries
and dependence on partner selection has not
received much attention in prior studies, and seems
more complicated than its e?ect on governance
design. On the one hand, when a ?rm is highly
dependent on another, it may lack alternatives
and consequently limit partner search e?ort for
new transactions. On the other hand, before
engaging in a new transaction with the more pow-
erful partner, the ?rm may actually prefer to invest
more e?ort in partner search as a ‘dependence
avoidance strategy’ (Pfe?er & Salancik, 1978),
with the purpose of identifying alternative ?rms
to change to should the partner exploit its power
advantage. Although existing theory seems insu?-
cient to make clear predictions about the e?ect of
dependence on selection e?ort, in the empirical
analyses the possibility of dependence avoidance
strategies is explored, thus expecting an increasing
level of dependence to be associated with increased
search e?ort. To summarize these arguments, the
control problems hypothesized to in?uence gover-
nance extensiveness can also be expected to in?u-
ence ?rms’ partner selection e?ort for an
inter?rm transaction.
H8: The greater the appropriation concerns for
an inter?rm transaction, as indicated by its
(a) size, (b) asset speci?city, and (c) uncer-
tainty, the more e?ort is invested in partner
selection.
H9: The greater the coordination requirements of
an inter?rm transaction, as indicated by its
task interdependence, the more e?ort is
invested in partner selection.
H10: The greater the dependence of a ?rm on a
partner, the more e?ort is invested in partner
selection for an inter?rm transaction with
this partner.
922 H.C. Dekker / Accounting, Organizations and Society 33 (2008) 915–941
The in?uence of partner experience on partner
selection e?ort
The ?nal model predictions relate to the e?ect
of partner experience on partner selection e?ort.
While prior studies have devoted considerable
attention to how partner experience a?ects gover-
nance choices, its e?ect on partner selection has
been largely unexplored. A few recent studies sug-
gested that increased familiarity allows ?rms to
economize not only on contracting costs but also
on search costs. Gulati and Gargiulo (1999)
argued that prior ties reduce uncertainty with
respect to future relationships and increase the
probability of forming new alliances, indicating a
preference for known partners. Consistent with
this argument, Blumberg (2001b) found that ?rms
invested less e?ort in partner selection for technol-
ogy alliances when the partners had prior ties. Bus-
kens et al. (2003) further argued that the in?uence
of trust on supplier search activities increases with
transaction size and importance, implying that
trust moderates the relation between transaction
characteristics and supplier search. Thus, the e?ect
of transaction characteristics on the e?ort exerted
by ?rms in evaluating a potential partner and
searching among alternative partners is likely to
di?er when the ?rm has transacted with the poten-
tial partner before. More speci?cally, for a given
level of control problems, ?rms are likely to spend
less e?ort on partner selection when the ?rm is
already familiar with a potential partner than
when this partner is new. For instance, a buyer
who knows a supplier from prior exchanges may
be less concerned about a substantial and highly
speci?c new transaction compared to a buyer with-
out such prior experience, and will consequently
invest less time in searching among alternative
suppliers. Thus, while appropriation concerns,
coordination requirements and dependence are
expected to directly in?uence partner selection
e?ort, their e?ect is predicted to vary with the
extent to which the buyer and the chosen supplier
have had prior experiences (cf. Dekker, 2004).
Hypotheses 11–13 describe these moderating
e?ects of partner experience. In Fig. 2 these e?ects
are depicted by the dotted line from Partner expe-
rience crossing the e?ects of the left-hand variables
on Selection e?ort.
H11: Partner experience weakens the association
between the appropriation concerns for an
inter?rm transaction, as indicated by its (a)
size, (b) asset speci?city, and (c) uncertainty,
and the e?ort invested in partner selection
for this inter?rm transaction.
H12: Partner experience weakens the association
between the coordination requirements for
an inter?rm transaction, as indicated by its
task interdependence, and the e?ort invested
in partner selection for this inter?rm
transaction.
H13: Partner experience weakens the association
between the dependence of a ?rm on a part-
ner and the e?ort invested in partner selec-
tion for an inter?rm transaction with this
partner.
Empirical design
For empirical testing of the hypotheses, struc-
tural equation modeling (SEM) is applied. SEM
provides the ability to concurrently assess the
validity and reliability of parameter estimates. A
recommended practice for assessing the predictive
ability of a model is to compare it with competing
models or with models built on alternative formu-
lations of the underlying theory (Bollen, 1989;
Jo¨ reskog & So¨ rbom, 1993). In the empirical anal-
yses, four models are estimated and compared in
terms of parameter estimates and ?t with the data
(see Fig. 3, Panels A and B). Model 1 relates clo-
sely to prior studies of the joint e?ects of appropri-
ation concerns and coordination requirements on
governance choice, by specifying direct e?ects of
size, asset speci?city, uncertainty, task interdepen-
dence and dependence on governance extensive-
ness. To facilitate comparisons with prior studies,
this model also includes a direct e?ect of partner
experience. Model 2 includes the buyer’s partner
selection e?ort, in order to test whether increasing
selection e?ort in?uences governance extensive-
ness. This model thus adds indirect e?ects of the
H.C. Dekker / Accounting, Organizations and Society 33 (2008) 915–941 923
explanatory variables on governance extensive-
ness, via selection e?ort. Models 3 and 4 re-specify
the e?ect of partner experience in Models 1 and 2
to moderate the e?ects of the explanatory variables
on governance extensiveness and partner selection
e?ort. Model 4, the ‘target model’, is the complete
speci?cation of all expectations (see also Fig. 2).
Data and sample description
For the empirical analyses, a dataset is used
called The External Management of Automation
1995 (Batenburg & Raub, 1995).
3
Using a multi-
purpose survey, in 1995 data were collected from
a large sample of Dutch small-to-medium enter-
prises (SMEs; 5–200 employees) relating to inde-
pendent transactions with IT suppliers. The data
set includes transactional relationships which vary
from simple market transactions with few associ-
ated control problems, to extensive and compli-
cated transactions in which the buyer and
supplier are closely intertwined and face potentially
large control problems. Typically, SMEs lack the
expertise and resources for in-house production
and are receptive to outsourcing their IT needs.
Because of the size of the ?rms involved, the IT
transactions often represent substantial investment
decisions. Moreover, SMEs often lack the knowl-
edge to adequately specify desired outcomes and
monitor performance ex post, creating serious con-
cerns about opportunism and posing coordination
di?culties. The rapid growth of the IT market in
the 1990s, with its continuous technological
changes, added to these problems by increasing
the di?culties for buyers to keep informed about
developments and enlarging information asymme-
try with suppliers (Buskens et al., 2003).
A database from Directview (a ?rm specialized
in IT marketing data of Dutch ?rms) containing
information about the automation of approxi-
mately 80% of all Dutch ?rms with more than ?ve
employees, was used as sampling frame. In the sam-
pling design (described in Batenburg et al., 2003),
the researchers used strati?cation on three criteria
in order to ensure su?cient variation between
Panel A Direct effects models 1 and 2 Panel B Moderated effects models 3 and 4
-
[M3/4]
+
+
+
+
+
+
+
+
+
+
Asset
specificity
Uncertainty
Task
intrdependencee
Dependence
Selection
effort [M2]
Governance
extensiveness
+/-
Partner
experience
+
+
+
+
+
+
+
+
+
+
-
-
Size
Asset
specificity
Uncertainty
Task
intrdependencee
Dependence
Selection
effort [M4]
Governance
extensiveness
Partner
experience
Size
Fig. 3. Overview of the estimated models. Note. M2 and M4 indicate the addition of selection e?ort to Models 1 and 3 respectively.
M3/4 indicates that in Model 3 and 4 the e?ect of partner experience is re-speci?ed as a moderating e?ect.
3
The data set The External Management of Automation 1995
has been collected as part of the NWO-Pionier program. The
Management of Matches (PGS 50-370) and is available from
the Steinmetz-archive (Study no. P1512).
924 H.C. Dekker / Accounting, Organizations and Society 33 (2008) 915–941
transactions on these aspects: type of product
exchanged (hardware–software and standard-com-
plex), network embeddedness (weak–medium–
strong, as estimated by industry experts), and buyer
expertise (low–medium–high, as indicated by the
number of IT specialists employed). These criteria
resulted in 36 cells (4 · 3 · 3), for which randomiza-
tion procedures were used until per cell at least 15
cases were selected. This main sample was extended
with observations on more innovative and complex
IT products in sectors where SMEs typically use
such products (i.e., food, metal, transportation
equipment, wholesale trade, and road transporta-
tion), using only the third criterion (buyer exper-
tise). Computer Assisted Telephone Interviews
(CATI) were employed to request cooperation from
a key informant in the buying ?rm, usually an IT
manager responsible for automation, and to ran-
domly select an IT investment made in the last ?ve
years that met the ?rst criterion (product type). The
questionnaire focused on this transaction, for
which the informant was usually responsible or
had been intensively involved with.
A total of 1798 ?rms were approached by tele-
phone, of which 463 were not suitable for data col-
lection (i.e., the ?rm was too large/small, had
ceased to exist or had no independent IT invest-
ments). Of the remaining 1335 ?rms, 902 partici-
pated in the CATI and 788 agreed to ?ll out the
survey, representing a ?eld response rate of 59%.
A member of the ?eldwork team subsequently vis-
ited the key informant of the ?rm with a question-
naire for the transaction that had been selected.
The questionnaire was sent by mail to 132 ?rms
unwilling to receive a visit. In addition, 183
respondents were willing to provide data about a
second IT transaction, in most cases ful?lled by a
di?erent supplier. This resulted in a total sample
of 971 transactions.
4
On average respondents had
worked at their ?rm for over 10 years, had over
10 years IT experience (almost 8 years of which
at their ?rm) and were well educated. Non-
response analysis indicated that non-participating
?rms did not di?er from participating ?rms on
?rm characteristics such as size, industry or region,
and a CATI question also indicated no di?erences
in their satisfaction with IT suppliers (reported in
Batenburg, 1997).
In the survey, respondents were asked to re?ect
on the time periods before, around and after the
transaction. This study employs data relating to
the periods before and around the transaction,
relating to the content of the transaction, the
selection of the supplier and prior transactions,
and the agreements made by the exchange
partners.
Variable measurement
For each of the model constructs at least one
indicator is available in the dataset, and in order
to maximize construct validity, indicators were
selected which are most similar to the construct.
In addition, in the selection of construct measures
use is made of prior studies which employ the
dataset to analyze governance choices (Anderson
& Dekker, 2005; Batenburg et al., 2003; Buskens,
2002).
Governance extensiveness is measured by the
number of agreements made by the exchange part-
ners for governing the transaction, based on a list
of 24 ?nancial, legal and operational issues (see
Table 1, Panel A). An increasing number of agree-
ments re?ects an increasingly extensive governance
structure. Respondents could indicate whether
issues were not arranged, were arranged orally,
which for scale construction were given a weight
of one, or were arranged in a written document,
which were given a weight of two (cf. Buskens,
2002). The Governance extensiveness measure is
constructed by summing the weighted scores on
the 24 items (oral = 1, written = 2), and thus
accounts for both contractual and non-contractual
elements of governance. Since written (contrac-
tual) agreements (mean = 10.8) were made sub-
stantially more often than oral agreements
(mean = 2.9), the number of written agreements
4
The mean (median) elapsed time between the transaction
and the administering of the survey is 3.25 (2) years. Consistent
with the expectation that buyers engage less frequently in larger
and more complex transactions, correlation analyses indicate
that this time span increases with higher values of the
independent variables. Although the dataset records transac-
tions anonymously, Batenburg and Raub (1995) reported that
943 transactions related to 602 di?erent suppliers, while in 28
cases the supplier identity was missing.
H.C. Dekker / Accounting, Organizations and Society 33 (2008) 915–941 925
correlates strongly with the weighted measure of
Governance extensiveness (r = 0.94).
5
Selection e?ort is measured by the time spent in
person-days on comparing and selecting potential
suppliers and products. This item best represents
the e?ort that the buyer exerted to ?nd a ‘good’
supplier. Several additional items are available to
assess the adequacy of the indicator. First, a
Mann–Whitney test con?rms that buyers that
had formed a team to manage the transaction
spent substantially more time on supplier selection
(N = 310, median = 15 days) than buyers that had
not formed a team (N = 596, median = 3 days).
Likely, such teams are formed only for larger
and more complex transactions, requiring a careful
supplier selection process. In addition, selection
time correlates signi?cantly with the number of
suppliers approached for information (r
s
= 0.39),
the number of quotations asked from suppliers
(r
s
= 0.47), and the number of di?erent informa-
tion sources consulted to ?nd a supplier
(r
s
= 0.44).
6
The distribution of the indicator
ranges from 0 to 300 days and is highly peaked
and skewed, as most buyers spent only a few or
a moderate number of days on supplier selection,
Table 1
Descriptive statistics of governance agreements and transaction contents
a
Item Written Oral Item Written Oral
Panel A – Frequencies of agreements made to manage the transaction
Price determination 637 116 Restrictions on supplier 247 117
Price-level 712 70 Non-disclosure 114 69
Price changes 296 136 Insurance supplier 182 62
Payment terms 644 58 Duration of service 592 90
Sanctions for late payment 205 44 Reservation of spare parts 131 92
Delivery time 567 171 Duration of maintenance 507 67
Liability supplier 438 109 Updating 362 121
Force majeure 251 99 Arbitration 210 62
Warranties supplier 643 72 Calculation R&D costs 147 74
Quality (norms) 294 162 Joint management of transaction 169 292
Intellectual property 395 50 Technical speci?cations 555 90
Piracy protection 229 137 Termination 241 59
Type of product/service Frequency Scale value Type of product/service Frequency Scale value
Panel B – Frequencies of products and services included in transactions
Standard software 494 1 Branch-speci?c software 168 3
Personal computers 388 1 Training 181 3
Work stations 174 1 Instruction 310 3
Side equipment 337 1 Documentation 284 3
Cabling 311 1 Customized software 151 4
Network con?guration 281 2 Consulting 240 4
Mini system 149 2 Accompaniment 304 4
Mainframe 36 2 Design 70 5
Computer-controlled machines 18 2 Tailor-made software 171 5
a
N = 817 (after listwise deletion based on all indicators used in the study).
5
Note that this measure does not focus on the nature of
agreements and relationships among them. See Anderson and
Dekker (2005) for a detailed analysis of contract structure and
its relation with management control dimensions.
6
Although supporting criterion-related validity, these items
are not used for measurement since they are more distant from
the construct and may also relate to buyers’ e?orts to ?nd
competitive prices for standard products. This is unrelated to a
focus on suppliers’ goodwill and capabilities. Analyses includ-
ing the items for measurement support this: when using a four-
item construct, the items number of suppliers approached and
number of quotations asked load strongest and selection time
loads weakest, indicating that the nature of the construct
changes. In addition, including the items reduces model ?t of
the structural models estimated, although the direction and
signi?cance of estimated coe?cients are largely the same.
926 H.C. Dekker / Accounting, Organizations and Society 33 (2008) 915–941
while a few spent a substantial amount of time. A
double square root transformation was conducted
to obtain a better approximation of a normal
distribution.
Size is measured by the sum of money paid to
the supplier for the transaction, using a ?ve-point
scale (cf. Buskens, 2002). Using an approximate
currency conversion of 1.65 Dutch guilders per
US dollar at the time of the survey administration,
the cuto? values for the ?ve price categories are:
less than $15,000, $30,000, $60,000 and $120,000,
and more than $120,000. This indicator correlates
signi?cantly with measures of the importance of
the transaction for the automation of the buyer
(r = 0.35), for the pro?t of the buyer (r = 0.27),
and for the pro?t of the supplier (r = 0.31), indi-
cating that larger transactions were also more
important for both exchange partners.
Asset speci?city has three indicators re?ecting
the costs of switching to another supplier and prod-
uct (Anderson & Dekker, 2005; Batenburg et al.,
2003). Speci?cally, questions were asked about
the estimated cost or damage in time and resources
that the buyer would have incurred should the
product have failed to function and needed to be
replaced (1 = very small, 5 = very large). These
costs relate to the re-education of personnel, re-entry
of data and idle production of other ?rm units, and
can be interpreted as a direct consequence of asset
speci?city, since more speci?c products will involve
higher switching costs. The re-education costs
relate to the buyer’s investments in speci?c knowl-
edge to use the product, which would be of no use
for other IT products (i.e., human asset speci?city).
IT transactions often involve substantial human
asset speci?city, in particular when systems and
applications are custom-tailored (Poppo & Zenger,
1998). The specialized knowledge and skills
required to develop, use and maintain customized
software applications impede switching to other
suppliers without incurring considerable costs of
obsolete knowledge and new training. Similarly,
the costs of data re-entry and idle production relate
to dedicated assets that cannot easily be replaced.
Jointly, these indicators indicate the magnitude of
the buyer’s ‘holdup problem’.
Uncertainty has three indicators re?ecting the
extent to which the buyer was able to: (1) evaluate
the quality of the product at delivery, (2) compare
the product with other suppliers’ products, and (3)
compare potential suppliers’ price/quality ratio
(1 = very easy, 5 = very di?cult). These items thus
re?ect the di?culty of specifying ex ante and veri-
fying ex post the products and services transacted.
Task interdependence has two indicators relat-
ing to the type of software, hardware and services
exchanged: transaction scope and complexity (cf.
Anderson & Dekker, 2005) (Table 1, Panel B).
The majority of transactions in the sample did
not exist of the purchase of a single product or ser-
vice, but instead of a ‘package’ of di?erent prod-
ucts and services. A response pattern analysis in
PRELIS (Jo¨ reskog & So¨ rbom, 1993) reveals 636
di?erent transaction patterns in the data, with buy-
ers purchasing one to sixteen di?erent products
and services. Scope measures the number of di?er-
ent products and services exchanged and thus
re?ects the diversity of transaction components.
Transactions of broader scope are more diverse
and likely require more coordination between
exchange partners than transactions of narrow
scope.
Complexity measures the highest level of com-
plexity of the products and services transacted,
re?ecting the highest level of interdependence to
be managed by the exchange partners. As shown
in Panel B of Table 1, a ?ve-point scale is used
to categorize all transaction components by di?er-
ent levels of complexity following from customiza-
tion. Increasingly customized products and
services involve higher levels of interdependence
and an increasing need for adaptation between
exchange partners. The purchase of on-shelf soft-
ware, for instance, re?ects a situation of sequential
interdependence with limited coordination require-
ments. Customized design of special purpose soft-
ware, in contrast, re?ects a situation of reciprocal
interdependence, requiring more extensive gover-
nance to facilitate coordination and mutual adjust-
ment between buyer and supplier. The ?ve
categories thus are increasing in the degree to
which coordination and transfer of ?rm-speci?c
knowledge need to take place, and include: (1)
standard products, (2) complex standard products,
(3) branch speci?c software and added services, (4)
customized products and (5) tailor-made products.
H.C. Dekker / Accounting, Organizations and Society 33 (2008) 915–941 927
This Guttman type scale recognizes that higher
levels of interdependence include all lower levels
to be managed (Thompson, 1967).
Consistent with the expectation of White and
Lui (2005) that increasing task complexity requires
managers to spend greater time and e?ort to work
with an alliance partner, transaction scope and
complexity correlate well with three available indi-
cators of coordination e?ort: the duration of deliv-
ery (r
s
= 0.32 and r
s
= 0.47), the time that supplier
employees were present at the buyer for delivery
(r
s
= 0.32 and r
s
= 0.40), and the time that buyer
employees were present at the supplier for delivery
(r
s
= 0.23 and r
s
= 0.25). Similarly, the mean scope
and complexity are signi?cantly broader and
higher for transactions with onsite presence of
employees at the exchange partner than for trans-
actions without onsite presence.
Dependence is re?ected by the buyer’s perceived
dependence on the supplier before the focal trans-
action (1 = very small, 5 = very large). This indi-
cator correlates signi?cantly with the buyer’s
expectation of future transactions with the sup-
plier (r = 0.31), which can be a source of
dependence.
Partner experience is indicated by whether or
not the buyer and supplier had transacted before
(similar to e.g., Batenburg et al., 2003; Buskens,
2002; Gulati, 1995; Gulati & Singh, 1998). In the
sub-sample of exchange partners with prior ties
(N = 404), the median (mean) length of prior busi-
ness relationships was 5 (6.4) years, the median
transaction frequency was ‘regularly’ (4 on a ?ve-
point scale), and the median size of prior business
was ‘reasonably sized’ (3 on a ?ve-point scale).
Most buyers indicated to be reasonably to very
satis?ed with these prior experiences.
7
In addition,
before the focal transaction, buyers with prior ties
with the supplier had higher expectations of future
business with the supplier, and were more inclined
to recommend the supplier to other ?rms.
8
The descriptive statistics in Table 2 show that
the responses to the indicators of both dependent
and independent variables vary substantially.
Most missing values relate to the indicators of
Governance extensiveness (67) and Selection e?ort
(65), while the other indicators have relatively
few missing values. Since the number of missing
values is modest, listwise deletion was used, leav-
ing a main sample size of 817 observations for
analysis. Apart from selection time, the variable
distributions show no important deviations from
normality. Unreported subgroup descriptive statis-
tics indicate that on average, ?rms with prior ties
used less extensive governance structures and
spent less time selecting a partner, while the aver-
ages are also somewhat lower for most exogenous
variables (tests for structural di?erences between
these groups are discussed in the section ‘‘Results
for the moderated e?ect models’’). Table 3 reports
Pearson correlations between the indicators.
Statistical method
PRELIS and LISREL 8.52 were used to pre-
pare and analyze the data (Jo¨ reskog & So¨ rbom,
1993). In the analyses, the measurement model
(MM; relating indicators to latent constructs)
and the structural model (SM; relating latent con-
structs to each other) are estimated simultaneously
using maximum likelihood (ML) estimation. ML
performs well even under conditions of non-nor-
mality, model misspeci?cation and sample size
variation (Boomsma & Hoogland, 2001; Olsson,
Foss, Troye, & Howell, 2000). Robust ML
(RML) standard errors are calculated, which
improve ML standard errors for larger samples
(Boomsma & Hoogland, 2001; Distefano, 2002).
Because RML standard errors are adjusted
upwards, parameters are less likely to become sig-
ni?cant. The main sample, and the subsamples of
7
The sub-sample includes only 14 buyers who, despite
unsatisfactory prior transaction experiences re-used the sup-
plier. Positive experiences were much more common: 109
buyers were ‘reasonably satis?ed’, 283 buyers were ‘satis?ed’
and 66 buyers were ‘very satis?ed’. Buyers who transacted with
known suppliers also reported substantially more often that
former suppliers had been considered as potential partners for
the focal transaction and that they started their supplier search
among prior partners.
8
The standardized mean di?erence between ?rms with and
without prior ties (Cohen’s d) is 0.72 for ‘level of expected future
business’ and 0.68 for ‘would recommend to other ?rms’,
indicating substantial group di?erences.
928 H.C. Dekker / Accounting, Organizations and Society 33 (2008) 915–941
?rms with and without prior ties in the multi-
group analyses that follow, are su?ciently large
for RML estimation. Although the sample sizes
may be too small to obtain equally reliable esti-
mates, weighted least squares (WLS) estimation
based on polychoric correlations is also employed
as a form of triangulation. Converging results pro-
vide additional evidence that parameter estimates
are accurate and that the correct model structure
is identi?ed (Olsson et al., 2000).
Multiple ?t measures are used to assess the extent
to which the estimated models ?t the sample data,
and to increase the probability of rejecting false
models and not rejecting ‘true’ models (Hu & Ben-
tler, 1999). The goodness-of-?t index (GFI),
adjusted GFI (AGFI), standardized root mean
residual (SMSR), and root mean squared error of
approximation (RMSEA) indicate how well the
model reproduces the sample data. The non-
normed ?t index (NNFI) and comparative ?t index
(CFI) compare the discrepancies from a ‘null-
model’ with the discrepancies from the ?tted model
to evaluate the improvement in ?t by specifying
relationships. Recommended cuto? values are 0.08
for SRMR, 0.06 for RMSEA and 0.95 for
(A)GFI, NNFI and CFI, with ‘loosened’ values
for combinations of measures (Fan, Thompson, &
Wang, 1999; Hu & Bentler, 1999). In addition, the
Scaled MLchi-square (SMLv
2
) goodness-of-?t sta-
tistic (correcting v
2
for non-normality; Boomsma &
Hoogland, 2001) and the Akaike Information
Criterion (AIC) are used to compare non-nested
Table 2
Descriptive statistics and measurement model estimates (N = 817)
Descriptive statistics Measurement model estimates
a
Min Max Mean SD Skew Kurt k SE t-Value k
s
Endogenous variables
Governance extensiveness
Number of agreements 2 48 24.42 10.84 0.10 À0.65 1.00 – – 0.95
Selection e?ort
Selection time – person-days 0 300 15.75 28.79 3.98 22.09
Selection time – transformed 0 4.16 1.59 0.71 0.33 0.63 1.00 – – 0.95
Exogenous variables
Size
Financial volume 1 5 2.54 1.45 0.42 À1.21 1.00 – – 0.89
Asset speci?city
Re-education costs (k = 1) 1 5 2.36 1.15 0.45 À0.82 1.00 – – 0.86
Data re-entry cost 1 5 2.99 1.38 À0.01 À1.28 1.01 0.09 11.71
***
0.69
Idle production costs 1 5 2.33 1.32 0.63 À0.85 0.78 0.07 10.92
***
0.54
Uncertainty
Di?culty measuring quality (k = 1) 1 5 2.71 0.99 0.42 À0.42 1.00 – – 0.61
Di?culty comparing products 1 5 2.91 1.09 0.19 À0.70 1.81 0.20 9.24
***
0.89
Di?culty comparing price/quality 1 5 2.96 1.08 0.12 À0.72 1.76 0.19 9.37
***
0.85
Task interdependence
Transaction complexity (k = 1) 1 5 3.32 1.43 À0.46 À1.11 1.00 0.08 11.03
***
0.80
Transaction scope 1 16 5.20 3.44 0.61 À0.45 0.85 – – 0.89
Dependence
Perceived dependence on supplier 1 5 2.59 1.21 0.03 À1.24 1.00 – – 0.89
Partner experience
Previous transacting 0 1 0.49 – 0.02 À2.00 1.00 – – 0.89
a
The measurement model estimates are based on robust ML estimation of Model 2. Reference indicators for multi-item constructs
are denoted by k = 1.
***
Indicates a p value
This paper develops a structural model to simultaneously analyze firms’ partner selection activities and governance
design for managing interfirm relationships. The model predicts these choices to be influenced by appropriation concerns,
coordination requirements, dependence and prior experiences with the exchange partner. An analysis of 817
information technology transactions shows that appropriation concerns, coordination requirements and dependence
influence governance design, while buyers’ partner selection effort relates mainly to appropriation concerns
Partner selection and governance design
in inter?rm relationships
Henri C. Dekker
*
Vrije Universiteit Amsterdam, Department of Accounting, De Boelelaan 1105, 1081 HV Amsterdam, The Netherlands
Abstract
This paper develops a structural model to simultaneously analyze ?rms’ partner selection activities and governance
design for managing inter?rm relationships. The model predicts these choices to be in?uenced by appropriation con-
cerns, coordination requirements, dependence and prior experiences with the exchange partner. An analysis of 817
information technology transactions shows that appropriation concerns, coordination requirements and dependence
in?uence governance design, while buyers’ partner selection e?ort relates mainly to appropriation concerns. Partner
experience developed during prior ties with the exchange partner is found to moderate the e?ects of these control prob-
lems. In addition, partner selection and formal governance are found to be used as complements instead of substitutes
in coping with inter?rm control problems.
Ó 2007 Elsevier Ltd. All rights reserved.
Introduction
Recognizing the importance of adequate gover-
nance structures for the management, control and
success of inter?rm relationships, organizational
researchers have devoted considerable e?ort to
explaining ?rms’ governance choices. Informed
by multiple theoretical perspectives, these studies
indicate the importance of adapting governance
design to di?erent conditions of the inter?rm
relationship, such as to transaction and task char-
acteristics. Designing governance structures, how-
ever, is not the only way in which ?rms attempt
to achieve control over their cooperative activities.
Several studies, for instance, point to the impor-
tance of experiences and trust from prior ties as
alternative social control mechanisms for coping
with transaction risks (e.g., Batenburg, Raub, &
Snijders, 2003; Dekker, 2004; Gulati, 1995; Gulati
& Singh, 1998; Tomkins, 2001). Another way that
?rms can mitigate control problems in inter?rm
relationships is by selecting an appropriate part-
ner. Some recent studies suggest that selecting a
‘good’ partner limits the need for governance
0361-3682/$ - see front matter Ó 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.aos.2007.02.002
*
Tel.: +31 20 598 6066; fax: +31 20 598 9870.
E-mail address: [email protected]
www.elsevier.com/locate/aos
Available online at www.sciencedirect.com
Accounting, Organizations and Society 33 (2008) 915–941
(Buskens, Batenburg, & Weesie, 2003), and that it
is essential for the e?ective management and suc-
cess of inter?rm relationships (Hitt, Dacin, Lev-
itas, Arregle, & Borza, 2000; Ireland, Hitt, &
Vaidyanath, 2002). In addition, the partner selec-
tion phase can strongly in?uence later stages of
the collaboration, including the use of governance
arrangements (Dekker, 2004). Despite its impor-
tance for the control of inter?rm relationships,
however, the partner selection process has received
scant attention in empirical studies. Recognizing
this void, this paper aims to: (1) provide an under-
standing of how ?rms use both partner selection
and governance arrangements to cope with poten-
tial control problems when engaging in a transac-
tion, and (2) examine how prior transaction
experiences in?uence these selection and design
choices. The study thus focuses on how social con-
trol mechanisms complement, or even substitute,
the use of governance arrangements in governing
and coordinating transactions between exchange
partners (Ouchi, 1980). Fig. 1 summarizes the
key ideas developed in this paper and suggests that
anticipated control problems in?uence both part-
ner selection e?ort and use of governance arrange-
ments, while prior experiences with the exchange
partner in?uence these associations.
The contribution of this study to the literature
on governance and control in inter?rm relation-
ships is threefold. First, the structural model that
is developed and empirically examined provides
an explanation of the extensiveness of the gover-
nance structure used by ?rms engaging in a trans-
action. As shown in Fig. 1, this explanation
employs the concepts of appropriation concerns,
coordination requirements, dependence and partner
experience. Appropriation concerns originate from
transaction characteristics, such as asset speci?c-
ity, uncertainty and frequency, and essentially
relate to the risk of opportunism by transaction
partners. Coordination requirements result from
task characteristics, in particular task interdepen-
dence, which creates a need for mutual adaptation
and coordination between partners. Dependence is
a factor that extends beyond the transaction
between exchange partners and relates to their eco-
nomic relationship, which may also involve other
(current and future) transactions. Jointly, these
factors generate a need for governance to mitigate
and manage the control problems that they create.
Second, the study investigates the e?ects of partner
selection on the use of governance arrangements,
arguing that partner selection contributes to miti-
gating concerns about the control problems of
appropriation, coordination and dependence. In
addition, it is put forward that a more extensive
partner selection process is also associated with
learning about the transaction and exchange part-
ner, which enhances the design of a more extensive
governance structure. As Fig. 1 illustrates, the
structural model provides an explanation of the
e?ort invested by the buyer in supplier selection
following from these control problems, and of its
consequent e?ects on the use of governance
arrangements. Third, the study investigates how
prior partner experiences in?uence the association
between control problems and both the e?ort
invested in partner selection and the use of gover-
nance arrangements. Partner experience essentially
relates to familiarity between exchange partners
generated during prior ties, which may have a sub-
stantial impact on governance choices. While most
prior studies have emphasized a direct e?ect of
prior ties on the use of governance arrangements,
arguing that experience and trust generated during
prior ties allow the use of less extensive governance
structures, this study provides support for a mod-
erating e?ect of partner experience. This moderat-
ing e?ect suggests that the e?ects of control
problems on partner selection e?ort and on gover-
nance extensiveness vary with the extent to which
exchange partners have prior experiences. More
speci?cally, the ?ndings indicate that the increased
familiarity from prior ties enables a reduction in
the use of governance arrangements for managing
Control problems
•Appropriation concerns
•Coordination requirements
•Dependence
Partner
experience
Governance
extensiveness
Selection
effort
Fig. 1. Key constructs of the study.
916 H.C. Dekker / Accounting, Organizations and Society 33 (2008) 915–941
concerns about dependence and appropriation,
while it simultaneously helps to develop more elab-
orate governance arrangements for managing
coordination requirements. This also implies that
partner experience gains in importance for the
management of inter?rm transactions as potential
control problems increase in magnitude.
The structural model is examined using detailed
survey data of information technology (IT) trans-
actions between small-to-medium enterprises and
their suppliers. The empirical tests focus in partic-
ular on the e?ort invested by buyers in the selec-
tion of an adequate partner and the use of
(contractual) agreements between exchange part-
ners. Studies of IT outsourcing report that buyers
in these relationships frequently experience sub-
stantial control problems, such as opportunistic
behavior, low supplier capabilities, high costs,
poor quality and service, loss of control and exper-
tise, and over-dependence on the supplier (e.g.,
Barthelemy, 2001). These problems relate to the
complicated and uncertain nature of IT, providing
conditions for market failure (Ouchi, 1980; Wil-
liamson, 1975). Software development, for
instance, typically involves information asymme-
try between developer and user, di?culties in mon-
itoring progress and outcomes, and a need for
intensive cooperation and coordination between
developer and user in order to understand and pri-
oritize issues which a?ect the software’s quality
and success (Wang, Baron, & Seidman, 1997).
Hence, the interdependencies between exchange
partners and the uncertainty in determining the
value of exchanged products and services can
entail considerable transaction costs (Ouchi,
1980). Selecting a reliable and competent partner
and developing an adequate governance structure
are important ways for ?rms to curb potential
opportunism and to facilitate coordination across
?rm boundaries.
The next section develops the research model by
reviewing prior studies of inter?rm governance
design. This review discusses the in?uence of
appropriation concerns and coordination require-
ments on the design of governance structures, the
in?uence of the relational context between ?rms
and the in?uence of the partner selection process
on governance design. The subsequent sections
discuss the design, data and results of the empirical
study. The ?nal section concludes the paper by
proposing directions for future research, based
on the ?ndings and limitations of the study.
The governance of inter?rm relationships
Literature on inter?rm governance describes
two key ways in which ?rms can cope with control
problems that arise in inter?rm relationships: (1)
by selecting a ‘good’ partner, and (2) by designing
an appropriate governance structure (Dekker,
2004). First, after determining that cooperation
with an outside entity is preferable to internal pro-
duction or vertical integration, ?rms need to
choose an appropriate partner (Ireland et al.,
2002). Having selected a partner, ?rms then need
to design and implement a governance structure
that includes arrangements to adequately govern
the collaboration. While the partner selection
phase essentially focuses on mitigating control
problems by ?nding a reliable and competent part-
ner, the design of a governance structure focuses
on both mitigation and management of control
problems. Fig. 2 presents the structural model
and research hypotheses that are developed in
Uncertainty
Task
interdependence
Dependence
Selection
effort
Governance
extensiveness
H7a/b
Partner
experience
H1a
Size
Asset
specificity
H1b
H1c
H2
H3
H8a
H8b
H8c
H9
H4-H6
H11-H13
H10
Fig. 2. The structural model.
H.C. Dekker / Accounting, Organizations and Society 33 (2008) 915–941 917
the next sections, following the framework in Dek-
ker (2004). The ?gure shows how variables which
generate appropriation concerns, coordination
requirements and dependence are expected to
in?uence both the e?ort invested in partner selec-
tion and the extensiveness of the governance struc-
ture for an inter?rm transaction. Furthermore, it
shows the expectation that partner experience will
moderate the in?uence of these control problems
on selection e?ort and on governance extensive-
ness. Finally, it shows that the e?ort invested in
partner selection is expected to a?ect governance
extensiveness.
The antecedents of inter?rm governance:
appropriation concerns and coordination
requirements
Governance structures in inter?rm relationships
can comprise a broad set of control mechanisms,
varying from contractual arrangements to non-
contractual organizational mechanisms. These
mechanisms enable control by ?at, by aligning
partners’ incentives, and by monitoring and
rewarding outcomes and behaviors (Dekker,
2004).
1
Similar to Anderson and Dekker (2005),
in its analysis of governance extensiveness this
paper examines the extent to which exchange part-
ners in IT transactions use contractual agreements
to govern their transaction. These agreements
relate to issues such as product speci?cation, pric-
ing, service levels, decision rights and responsibili-
ties, and con?ict resolution. In addition, the
analysis of governance extensiveness also includes
the use of non-contractual agreements between
exchange partners regarding these issues, which
are often used to complement (incomplete) con-
tractual agreements.
Theoretically, prior studies of inter?rm gover-
nance design have predominantly used transaction
cost economics (TCE) to explain ?rms’ governance
choices (e.g., Anderson & Dekker, 2005; Ander-
son, Glenn, & Sedatole, 2000; Dekker, 2003,
2004; Gulati, 1995; Gulati & Singh, 1998; Lang-
?eld-Smith & Smith, 2003; Van der Meer-Kooistra
& Vosselman, 2000). TCE posits that exchange
hazards arise as a function of transaction charac-
teristics and that increasing concerns about these
exchange hazards induce ?rms to use governance
structures with more elaborate problem solving
abilities. Asset speci?city refers to the degree to
which resources dedicated to a transaction can be
re-deployed in other transactions without loss of
value or high cost (Williamson, 1985). Because of
the potential costly consequences of opportunism,
increasing asset speci?city generates a need for
more extensive governance to promote incentive
alignment, e?cient adaptation and transaction
cost mitigation (Masten, 2000). The transaction
size augments exchange hazards by since a larger
transaction increases the investing ?rm’s exposure
to opportunistic holdup, and increases the part-
ner’s incentives to behave opportunistically (Blum-
berg, 2001a, 2001b; Buskens, 2002). Firms facing a
larger potential loss are therefore likely to prefer
more extensive governance structures in order to
safeguard their investment (Batenburg et al.,
2003).
Uncertainty is a third key ingredient of appro-
priation concerns, since it creates information
asymmetry between exchanges partners and gener-
ates a need to adapt to unforeseeable changes.
Important causes of uncertainty are the environ-
ment, such as market and technological ?uctua-
tions, and the complexity and unpredictability of
tasks. These uncertainties cause di?culties in mon-
itoring performance and generate a need for gover-
nance to moderate the risk that the transaction
partner will exploit these monitoring problems
(Anderson & Dekker, 2005; Batenburg et al.,
2003; Blumberg, 2001a, 2001b; Buskens, 2002;
1
Gulati and Singh (1998), for instance, discussed ?ve
important types of hierarchical control mechanisms which are
frequently used in inter?rm alliances: command structures and
authority systems, incentive systems, standard operating pro-
cedures, dispute resolution procedures and non-market pricing
systems. Typically, empirical studies on inter?rm governance
choices are conducted at a high level of analysis, such as the
choice between market, hybrid and hierarchy, or between
equity and non-equity alliances, which are argued to di?er in
the use of control mechanisms. However, a variety of structures
is also used within these governance types, which di?er in their
mechanisms and ability to coordinate. A focus on the actual
content of governance structures, of which contract agreements
are essential constituents, therefore o?ers an important way to
enhance theory development and empirical testing (Poppo &
Zenger, 2002).
918 H.C. Dekker / Accounting, Organizations and Society 33 (2008) 915–941
Leiblein & Miller, 2003; Poppo & Zenger, 2002).
Although uncertainty warrants a greater reliance
on governance mechanisms in order to manage
the associated appropriation concerns, by its own
nature it also reduces the possibilities to design
and use controls based on monitoring (Blumberg,
2001a). Poppo and Zenger (2002) therefore argue
that when measurement di?culties in IT outsourc-
ing increase, managers are likely to spend more
resources on developing more complex contracts
which specify required service levels or facilitate
behavior monitoring, and which enable to measure
and reward output and productivity. Following
these TCE arguments, it is hypothesized that gov-
ernance extensiveness (i.e., the extent to which
exchange partners use contractual and non-con-
tractual agreements) increases with increasing
exchange hazards:
H1: The greater the appropriation concerns for an
inter?rm transaction, as indicated by its (a)
size, (b) asset speci?city, and (c) uncertainty,
the more extensive its governance structure.
While a vast amount of empirical evidence sup-
ports the view that exchange hazards contribute to
explaining inter?rm governance structures (e.g.,
Geyskens, Steenkamp, & Kumar, 2006; Masten,
2000), managing potential opportunism is not the
only purpose for which governance structures are
designed. Gulati and Singh (1998) found support
for the in?uence of a second important control
problem to be managed in strategic alliances: the
coordination of interdependent tasks. A need for
coordination arises when, in the pursuit of value
creation, ?rms engage in collaborative activities
involving interdependent tasks. A higher level of
task interdependence increases ?rms’ reliance
upon each other in the performance of tasks and
creates a need for coordination and mutual adap-
tation (Thompson, 1967). Consistent with this,
Gulati and Singh (1998) found that strategic alli-
ances which performed more interdependent activ-
ities used more complex governance structures
suited to facilitating this coordination. Similarly,
Dekker (2004), Gulati, Lawrence, and Puranam
(2005) and White and Lui (2005) found that, in
addition to con?icting interests and incentive
problems, coordination requirements in?uence
governance and control choices in inter?rm rela-
tionships.
2
Accordingly, it is hypothesized that:
H2: The greater the coordination requirements for
an inter?rm transaction, as indicated by its
task interdependence, the more extensive its
governance structure.
The in?uence of relationship characteristics on
inter?rm governance: dependence and partner
experience
While transaction and task characteristics gen-
erate concerns about potential appropriation and
requirements for coordination, the economic and
social context of the relationship between
exchange partners can signi?cantly alter their
response to these control problems. Two impor-
tant contextual variables relating to this economic
and social relationship are dependence and partner
experience. First, in their design of a governance
structure, transacting ?rms may seek to mitigate
concerns of dependence and power asymmetries
(Gulati & Singh, 1998; Pfe?er & Salancik, 1978).
While resource considerations and complementar-
ity are important drivers of relationship forma-
tion, they can also lead to increasing economic
dependence of ?rms upon others. The dependence
of one ?rm upon another, and the associated
power di?erences, relate to the relationship
between exchange partners, not to a particular
transaction per se; nevertheless, they are likely to
in?uence speci?c transactions between them. For
instance, even when a transaction comprises non-
critical and non-speci?c resources, the buyer may
still be highly dependent, because of other
(planned) critical transactions with the supplier.
This dependence position gives the supplier a
power advantage, and is likely to induce the buyer
to employ more safeguards for the focal transac-
tion. Hence, it is hypothesized that:
2
Similar to its in?uence on appropriation concerns, task
uncertainty can augment the coordination problem by creating
monitoring problems and hindering e?cient coordination
(Leiblein & Miller, 2003).
H.C. Dekker / Accounting, Organizations and Society 33 (2008) 915–941 919
H3: The greater the dependence of a ?rm on a
partner, the more extensive the governance
structure for an inter?rm transaction with this
partner.
Second, the social context in which transactions
take place can lead to familiarity and informal
controls between exchange partners, which can
signi?cantly in?uence their governance choices
(e.g., Batenburg et al., 2003; Buskens et al., 2003;
Dekker, 2003, 2004; Lang?eld-Smith & Smith,
2003; Gulati, 1995; Gulati & Gargiulo, 1999; Tom-
kins, 2001; Uzzi, 1997). As in earlier studies (e.g.,
Gulati, 1995; Gulati & Singh, 1998), this study
focuses on prior ties between exchange partners,
during which a level of partner experience is gener-
ated by directly experiencing each other’s behav-
iors and performance. Repeated interactions over
time can contribute strongly to learning about
partners, since they provide ?rms with unique
?rst-hand information about the other’s reliability
and competence (Blumberg, 2001b; Gulati & Gar-
giulo, 1999). The possibility of terminating the
relationship and withdrawing future business fur-
ther promotes ‘calculative trust’ based on deter-
rence (Gulati & Singh, 1998).
Prior studies which examined the e?ects of part-
ner experience focused mainly on the development
of social ties during interactions, often referred to
as ‘relational trust’, which is argued to either sub-
stitute or complement the use of governance
arrangements (e.g., Das & Teng, 1998; Dekker,
2004; Gulati, 1995; Poppo & Zenger, 2002). Sub-
stitution can occur since enhanced social ties are
associated with reduced goal con?ict, limiting the
need for governance to deter opportunism. In
addition, use of governance arrangements, such
as contractual agreements, may also signal distrust
and potentially damage the relationship (Das &
Teng, 1998; Dekker, 2004). A complementary rela-
tionship implies that increased familiarity and gov-
ernance lead to an overall higher level of control,
and it is suggested that more extensive governance
can actually enhance social ties by narrowing the
domain and severity of risk (Poppo & Zenger,
2002) and by providing objective information
about the partner’s performance, behaviors and
skills (Das & Teng, 1998). Empirically, several
studies found that prior ties between exchange
partners are associated with the use of less exten-
sive governance structures (e.g., Batenburg et al.,
2003; Buskens, 2002; Gulati, 1995; Gulati & Singh,
1998). However, Poppo and Zenger (2002) found a
positive association between trust-based relational
governance and the customization of contracts for
information services exchanges. By providing evi-
dence for both perspectives, these studies point
to the need to analyze under which circumstances
a substitutive or complementary relationship
between partner experience and governance exten-
siveness is more likely.
Whereas most prior studies predict and test a
direct in?uence of prior ties and trust on gover-
nance design, Dirks and Ferrin (2001) argued that
in many organizational settings trust has moderat-
ing e?ects on organizational outcome variables. In
in?uencing governance design, a moderating e?ect
of partner experience seems more plausible than a
direct e?ect, as it is not the (lack of) familiarity or
(dis)trust which generates a need for governance,
but the potential control problems associated with
a transaction (Dekker, 2004). In the absence of
any control problems, partner experience is unli-
kely to in?uence governance design at all, since
there is no need for governance. Thus only in the
presence of potential control problems can partner
experience be expected to a?ect the use of gover-
nance arrangements, by altering ?rms’ responses
to control problems. Consequently, the association
between control problems and governance exten-
siveness is expected to vary with partner experi-
ence. A substitutive e?ect then implies that in
relationships with increased familiarity, control
problems have a weaker e?ect on the use of gover-
nance arrangements than in newly established rela-
tionships in which partners are unfamiliar with
each other. This leads to the expectation that
appropriation concerns have a weaker in?uence
on governance extensiveness when exchange part-
ners have prior mutual experience.
A second issue concerning how partner experi-
ence a?ects governance design relates to the nature
of the control problems to be managed (i.e., the
purposes of control; cf. Dekker, 2004). While prior
studies focused mainly on how partner experience
in?uences the use of governance arrangements for
920 H.C. Dekker / Accounting, Organizations and Society 33 (2008) 915–941
managing appropriation concerns, increased
familiarity with the exchange partner may also
in?uence the e?ects of coordination requirements
and dependence. Gulati and Singh (1998), for
instance, argued that trust and learning from prior
ties can strongly in?uence coordination costs.
Similarly, Uzzi (1997) found that the embedded
relationships he studied were characterized by
?ne-grained information transfer and joint prob-
lem solving arrangements, facilitating informal
modes of coordination. Poppo and Zenger (2002)
further argued that as relationships are developed
and sustained, partners will include lessons learned
in the past in their contract, thereby formalizing
experiences, mutually agreed processes, informa-
tion sharing, and performance measurement and
monitoring practices. These arguments thus indi-
cate that increased familiarity may not only reduce
the need for coordination mechanisms, but may
also enhance partners’ abilities to jointly design
and implement such mechanisms, leading to more
extensive governance (Dekker, 2004). They also
again point to moderating e?ects of partner expe-
rience, which is likely to alter the in?uence of coor-
dination requirements on governance design. In
addition, they imply the possibility of both substi-
tutive and complementary e?ects of partner experi-
ence occurring simultaneously, when the increased
familiarity reduces the need for governance to
manage appropriation concerns, while it enhanced
the design of coordination mechanisms. Finally,
similar to its e?ect on appropriation concerns,
partner experience is also likely to weaken the
e?ects of dependence on governance extensiveness,
since ?rms who are more familiar with each other
will expect their partner not to use their power
advantage.
Hypotheses 4, 5 and 6 describe the expected
moderating e?ects of partner experience on the
in?uence of appropriation concerns, coordination
requirements and dependence respectively. In
Fig. 2 these moderating e?ects are shown by the
dotted line crossing the e?ects of control problems
on governance extensiveness.
H4: Partner experience weakens the association
between the appropriation concerns for an
inter?rm transaction, as indicated by its (a)
size, (b) asset speci?city, and (c) uncertainty,
and the extensiveness of the governance struc-
ture of the inter?rm transaction.
H5: Partner experience strengthens the association
between the coordination requirements for
an inter?rm transaction, as indicated by its
task interdependence, and the extensiveness
of the governance structure of the inter?rm
transaction.
H6: Partner experience weakens the association
between the dependence of a ?rm on a partner
and the extensiveness of the governance struc-
ture for an inter?rm transaction with this
partner.
The in?uence of partner selection on inter?rm
governance
While governance structures are important for
the management of appropriation concerns, coor-
dination requirements and power asymmetries, the
partner selection phase which precedes governance
design can have a substantial in?uence by mitigat-
ing potential control problems. Ireland et al.
(2002) therefore argued that e?ective alliance man-
agement starts with the selection of the ‘right’ part-
ner. Selecting an appropriate partner is a critical
decision for the success of a relationship; not only
for gaining access to critical and complementary
resources (e.g., Hitt et al., 2000), but also for e?ec-
tive management of the cooperation. Conse-
quently, instead of attempting to explain
governance choices, a few recent empirical studies
focused on a di?erent dependent variable: the
searching for partners with whom to transact
(Blumberg, 2001b; Buskens et al., 2003). The selec-
tion of a reliable and competent partner can
reduce ?rms’ concerns about appropriation and
coordination for the transaction in which they
wish to engage and consequently reduce the need
to design and implement a governance structure
(Buskens et al., 2003; Dekker, 2004). Provided that
buyers have a choice among potential partners,
investing more e?ort in evaluating and comparing
potential partners may increase con?dence in the
goodwill and capabilities of the chosen partner,
thus reducing the need for governance. Alterna-
tively, during a more extensive search process,
H.C. Dekker / Accounting, Organizations and Society 33 (2008) 915–941 921
?rms also acquire knowledge about the chosen
partner and product, which can facilitate the
development of more detailed agreements for
managing the transaction. This argument suggests
that the selection process functions as a learning
process which enables the design of enhanced gov-
ernance structures. Similar to the e?ects of partner
experience, these arguments raise the question
whether partner selection e?ort and use of gover-
nance arrangements are substitutes or comple-
ments for coping with inter?rm control
problems. Despite its potential for mitigating con-
trol problems and in?uencing governance design,
the partner selection behavior of ?rms has not
received much attention in empirical research.
Given the ambiguity about whether partner selec-
tion substitutes or complements governance exten-
siveness, two competing hypotheses are for-
mulated:
H7a: The greater the e?ort spent on partner selec-
tion for an inter?rm transaction, the less
extensive its governance structure.
H7b: The greater the e?ort spent on partner selec-
tion for an inter?rm transaction, the more
extensive its governance structure.
The antecedents of partner selection e?ort:
appropriation concerns, coordination requirements
and dependence
In a similar way to governance design, investing
time and resources in partner selection gains in
importance when potential control problems for
a transaction increase in magnitude (Batenburg
et al., 2003; Blumberg, 2001b; Buskens et al.,
2003; Dekker, 2004). Consistent with this argu-
ment, Blumberg (2001b) found that with increas-
ing exchange hazards, as re?ected by transaction
size, asset speci?city and environmental uncer-
tainty, ?rms increased their search e?orts to ?nd
a reliable partner for technology alliances. Simi-
larly, increasing coordination requirements are
likely to induce ?rms to invest more e?ort in
search and selection of a competent partner, in
order to avoid coordination costs and perfor-
mance problems (Dekker, 2004). Although an
increasing level of interdependence requires more
coordination e?ort, this e?ort may be less exten-
sive when cooperating with a more competent
partner who requires less guidance and monitor-
ing. In addition, cooperating with a more compe-
tent partner may result in coordination which is
not only more e?cient but also more e?ective, thus
producing superior outcomes.
The e?ect of concerns about power asymmetries
and dependence on partner selection has not
received much attention in prior studies, and seems
more complicated than its e?ect on governance
design. On the one hand, when a ?rm is highly
dependent on another, it may lack alternatives
and consequently limit partner search e?ort for
new transactions. On the other hand, before
engaging in a new transaction with the more pow-
erful partner, the ?rm may actually prefer to invest
more e?ort in partner search as a ‘dependence
avoidance strategy’ (Pfe?er & Salancik, 1978),
with the purpose of identifying alternative ?rms
to change to should the partner exploit its power
advantage. Although existing theory seems insu?-
cient to make clear predictions about the e?ect of
dependence on selection e?ort, in the empirical
analyses the possibility of dependence avoidance
strategies is explored, thus expecting an increasing
level of dependence to be associated with increased
search e?ort. To summarize these arguments, the
control problems hypothesized to in?uence gover-
nance extensiveness can also be expected to in?u-
ence ?rms’ partner selection e?ort for an
inter?rm transaction.
H8: The greater the appropriation concerns for
an inter?rm transaction, as indicated by its
(a) size, (b) asset speci?city, and (c) uncer-
tainty, the more e?ort is invested in partner
selection.
H9: The greater the coordination requirements of
an inter?rm transaction, as indicated by its
task interdependence, the more e?ort is
invested in partner selection.
H10: The greater the dependence of a ?rm on a
partner, the more e?ort is invested in partner
selection for an inter?rm transaction with
this partner.
922 H.C. Dekker / Accounting, Organizations and Society 33 (2008) 915–941
The in?uence of partner experience on partner
selection e?ort
The ?nal model predictions relate to the e?ect
of partner experience on partner selection e?ort.
While prior studies have devoted considerable
attention to how partner experience a?ects gover-
nance choices, its e?ect on partner selection has
been largely unexplored. A few recent studies sug-
gested that increased familiarity allows ?rms to
economize not only on contracting costs but also
on search costs. Gulati and Gargiulo (1999)
argued that prior ties reduce uncertainty with
respect to future relationships and increase the
probability of forming new alliances, indicating a
preference for known partners. Consistent with
this argument, Blumberg (2001b) found that ?rms
invested less e?ort in partner selection for technol-
ogy alliances when the partners had prior ties. Bus-
kens et al. (2003) further argued that the in?uence
of trust on supplier search activities increases with
transaction size and importance, implying that
trust moderates the relation between transaction
characteristics and supplier search. Thus, the e?ect
of transaction characteristics on the e?ort exerted
by ?rms in evaluating a potential partner and
searching among alternative partners is likely to
di?er when the ?rm has transacted with the poten-
tial partner before. More speci?cally, for a given
level of control problems, ?rms are likely to spend
less e?ort on partner selection when the ?rm is
already familiar with a potential partner than
when this partner is new. For instance, a buyer
who knows a supplier from prior exchanges may
be less concerned about a substantial and highly
speci?c new transaction compared to a buyer with-
out such prior experience, and will consequently
invest less time in searching among alternative
suppliers. Thus, while appropriation concerns,
coordination requirements and dependence are
expected to directly in?uence partner selection
e?ort, their e?ect is predicted to vary with the
extent to which the buyer and the chosen supplier
have had prior experiences (cf. Dekker, 2004).
Hypotheses 11–13 describe these moderating
e?ects of partner experience. In Fig. 2 these e?ects
are depicted by the dotted line from Partner expe-
rience crossing the e?ects of the left-hand variables
on Selection e?ort.
H11: Partner experience weakens the association
between the appropriation concerns for an
inter?rm transaction, as indicated by its (a)
size, (b) asset speci?city, and (c) uncertainty,
and the e?ort invested in partner selection
for this inter?rm transaction.
H12: Partner experience weakens the association
between the coordination requirements for
an inter?rm transaction, as indicated by its
task interdependence, and the e?ort invested
in partner selection for this inter?rm
transaction.
H13: Partner experience weakens the association
between the dependence of a ?rm on a part-
ner and the e?ort invested in partner selec-
tion for an inter?rm transaction with this
partner.
Empirical design
For empirical testing of the hypotheses, struc-
tural equation modeling (SEM) is applied. SEM
provides the ability to concurrently assess the
validity and reliability of parameter estimates. A
recommended practice for assessing the predictive
ability of a model is to compare it with competing
models or with models built on alternative formu-
lations of the underlying theory (Bollen, 1989;
Jo¨ reskog & So¨ rbom, 1993). In the empirical anal-
yses, four models are estimated and compared in
terms of parameter estimates and ?t with the data
(see Fig. 3, Panels A and B). Model 1 relates clo-
sely to prior studies of the joint e?ects of appropri-
ation concerns and coordination requirements on
governance choice, by specifying direct e?ects of
size, asset speci?city, uncertainty, task interdepen-
dence and dependence on governance extensive-
ness. To facilitate comparisons with prior studies,
this model also includes a direct e?ect of partner
experience. Model 2 includes the buyer’s partner
selection e?ort, in order to test whether increasing
selection e?ort in?uences governance extensive-
ness. This model thus adds indirect e?ects of the
H.C. Dekker / Accounting, Organizations and Society 33 (2008) 915–941 923
explanatory variables on governance extensive-
ness, via selection e?ort. Models 3 and 4 re-specify
the e?ect of partner experience in Models 1 and 2
to moderate the e?ects of the explanatory variables
on governance extensiveness and partner selection
e?ort. Model 4, the ‘target model’, is the complete
speci?cation of all expectations (see also Fig. 2).
Data and sample description
For the empirical analyses, a dataset is used
called The External Management of Automation
1995 (Batenburg & Raub, 1995).
3
Using a multi-
purpose survey, in 1995 data were collected from
a large sample of Dutch small-to-medium enter-
prises (SMEs; 5–200 employees) relating to inde-
pendent transactions with IT suppliers. The data
set includes transactional relationships which vary
from simple market transactions with few associ-
ated control problems, to extensive and compli-
cated transactions in which the buyer and
supplier are closely intertwined and face potentially
large control problems. Typically, SMEs lack the
expertise and resources for in-house production
and are receptive to outsourcing their IT needs.
Because of the size of the ?rms involved, the IT
transactions often represent substantial investment
decisions. Moreover, SMEs often lack the knowl-
edge to adequately specify desired outcomes and
monitor performance ex post, creating serious con-
cerns about opportunism and posing coordination
di?culties. The rapid growth of the IT market in
the 1990s, with its continuous technological
changes, added to these problems by increasing
the di?culties for buyers to keep informed about
developments and enlarging information asymme-
try with suppliers (Buskens et al., 2003).
A database from Directview (a ?rm specialized
in IT marketing data of Dutch ?rms) containing
information about the automation of approxi-
mately 80% of all Dutch ?rms with more than ?ve
employees, was used as sampling frame. In the sam-
pling design (described in Batenburg et al., 2003),
the researchers used strati?cation on three criteria
in order to ensure su?cient variation between
Panel A Direct effects models 1 and 2 Panel B Moderated effects models 3 and 4
-
[M3/4]
+
+
+
+
+
+
+
+
+
+
Asset
specificity
Uncertainty
Task
intrdependencee
Dependence
Selection
effort [M2]
Governance
extensiveness
+/-
Partner
experience
+
+
+
+
+
+
+
+
+
+
-
-
Size
Asset
specificity
Uncertainty
Task
intrdependencee
Dependence
Selection
effort [M4]
Governance
extensiveness
Partner
experience
Size
Fig. 3. Overview of the estimated models. Note. M2 and M4 indicate the addition of selection e?ort to Models 1 and 3 respectively.
M3/4 indicates that in Model 3 and 4 the e?ect of partner experience is re-speci?ed as a moderating e?ect.
3
The data set The External Management of Automation 1995
has been collected as part of the NWO-Pionier program. The
Management of Matches (PGS 50-370) and is available from
the Steinmetz-archive (Study no. P1512).
924 H.C. Dekker / Accounting, Organizations and Society 33 (2008) 915–941
transactions on these aspects: type of product
exchanged (hardware–software and standard-com-
plex), network embeddedness (weak–medium–
strong, as estimated by industry experts), and buyer
expertise (low–medium–high, as indicated by the
number of IT specialists employed). These criteria
resulted in 36 cells (4 · 3 · 3), for which randomiza-
tion procedures were used until per cell at least 15
cases were selected. This main sample was extended
with observations on more innovative and complex
IT products in sectors where SMEs typically use
such products (i.e., food, metal, transportation
equipment, wholesale trade, and road transporta-
tion), using only the third criterion (buyer exper-
tise). Computer Assisted Telephone Interviews
(CATI) were employed to request cooperation from
a key informant in the buying ?rm, usually an IT
manager responsible for automation, and to ran-
domly select an IT investment made in the last ?ve
years that met the ?rst criterion (product type). The
questionnaire focused on this transaction, for
which the informant was usually responsible or
had been intensively involved with.
A total of 1798 ?rms were approached by tele-
phone, of which 463 were not suitable for data col-
lection (i.e., the ?rm was too large/small, had
ceased to exist or had no independent IT invest-
ments). Of the remaining 1335 ?rms, 902 partici-
pated in the CATI and 788 agreed to ?ll out the
survey, representing a ?eld response rate of 59%.
A member of the ?eldwork team subsequently vis-
ited the key informant of the ?rm with a question-
naire for the transaction that had been selected.
The questionnaire was sent by mail to 132 ?rms
unwilling to receive a visit. In addition, 183
respondents were willing to provide data about a
second IT transaction, in most cases ful?lled by a
di?erent supplier. This resulted in a total sample
of 971 transactions.
4
On average respondents had
worked at their ?rm for over 10 years, had over
10 years IT experience (almost 8 years of which
at their ?rm) and were well educated. Non-
response analysis indicated that non-participating
?rms did not di?er from participating ?rms on
?rm characteristics such as size, industry or region,
and a CATI question also indicated no di?erences
in their satisfaction with IT suppliers (reported in
Batenburg, 1997).
In the survey, respondents were asked to re?ect
on the time periods before, around and after the
transaction. This study employs data relating to
the periods before and around the transaction,
relating to the content of the transaction, the
selection of the supplier and prior transactions,
and the agreements made by the exchange
partners.
Variable measurement
For each of the model constructs at least one
indicator is available in the dataset, and in order
to maximize construct validity, indicators were
selected which are most similar to the construct.
In addition, in the selection of construct measures
use is made of prior studies which employ the
dataset to analyze governance choices (Anderson
& Dekker, 2005; Batenburg et al., 2003; Buskens,
2002).
Governance extensiveness is measured by the
number of agreements made by the exchange part-
ners for governing the transaction, based on a list
of 24 ?nancial, legal and operational issues (see
Table 1, Panel A). An increasing number of agree-
ments re?ects an increasingly extensive governance
structure. Respondents could indicate whether
issues were not arranged, were arranged orally,
which for scale construction were given a weight
of one, or were arranged in a written document,
which were given a weight of two (cf. Buskens,
2002). The Governance extensiveness measure is
constructed by summing the weighted scores on
the 24 items (oral = 1, written = 2), and thus
accounts for both contractual and non-contractual
elements of governance. Since written (contrac-
tual) agreements (mean = 10.8) were made sub-
stantially more often than oral agreements
(mean = 2.9), the number of written agreements
4
The mean (median) elapsed time between the transaction
and the administering of the survey is 3.25 (2) years. Consistent
with the expectation that buyers engage less frequently in larger
and more complex transactions, correlation analyses indicate
that this time span increases with higher values of the
independent variables. Although the dataset records transac-
tions anonymously, Batenburg and Raub (1995) reported that
943 transactions related to 602 di?erent suppliers, while in 28
cases the supplier identity was missing.
H.C. Dekker / Accounting, Organizations and Society 33 (2008) 915–941 925
correlates strongly with the weighted measure of
Governance extensiveness (r = 0.94).
5
Selection e?ort is measured by the time spent in
person-days on comparing and selecting potential
suppliers and products. This item best represents
the e?ort that the buyer exerted to ?nd a ‘good’
supplier. Several additional items are available to
assess the adequacy of the indicator. First, a
Mann–Whitney test con?rms that buyers that
had formed a team to manage the transaction
spent substantially more time on supplier selection
(N = 310, median = 15 days) than buyers that had
not formed a team (N = 596, median = 3 days).
Likely, such teams are formed only for larger
and more complex transactions, requiring a careful
supplier selection process. In addition, selection
time correlates signi?cantly with the number of
suppliers approached for information (r
s
= 0.39),
the number of quotations asked from suppliers
(r
s
= 0.47), and the number of di?erent informa-
tion sources consulted to ?nd a supplier
(r
s
= 0.44).
6
The distribution of the indicator
ranges from 0 to 300 days and is highly peaked
and skewed, as most buyers spent only a few or
a moderate number of days on supplier selection,
Table 1
Descriptive statistics of governance agreements and transaction contents
a
Item Written Oral Item Written Oral
Panel A – Frequencies of agreements made to manage the transaction
Price determination 637 116 Restrictions on supplier 247 117
Price-level 712 70 Non-disclosure 114 69
Price changes 296 136 Insurance supplier 182 62
Payment terms 644 58 Duration of service 592 90
Sanctions for late payment 205 44 Reservation of spare parts 131 92
Delivery time 567 171 Duration of maintenance 507 67
Liability supplier 438 109 Updating 362 121
Force majeure 251 99 Arbitration 210 62
Warranties supplier 643 72 Calculation R&D costs 147 74
Quality (norms) 294 162 Joint management of transaction 169 292
Intellectual property 395 50 Technical speci?cations 555 90
Piracy protection 229 137 Termination 241 59
Type of product/service Frequency Scale value Type of product/service Frequency Scale value
Panel B – Frequencies of products and services included in transactions
Standard software 494 1 Branch-speci?c software 168 3
Personal computers 388 1 Training 181 3
Work stations 174 1 Instruction 310 3
Side equipment 337 1 Documentation 284 3
Cabling 311 1 Customized software 151 4
Network con?guration 281 2 Consulting 240 4
Mini system 149 2 Accompaniment 304 4
Mainframe 36 2 Design 70 5
Computer-controlled machines 18 2 Tailor-made software 171 5
a
N = 817 (after listwise deletion based on all indicators used in the study).
5
Note that this measure does not focus on the nature of
agreements and relationships among them. See Anderson and
Dekker (2005) for a detailed analysis of contract structure and
its relation with management control dimensions.
6
Although supporting criterion-related validity, these items
are not used for measurement since they are more distant from
the construct and may also relate to buyers’ e?orts to ?nd
competitive prices for standard products. This is unrelated to a
focus on suppliers’ goodwill and capabilities. Analyses includ-
ing the items for measurement support this: when using a four-
item construct, the items number of suppliers approached and
number of quotations asked load strongest and selection time
loads weakest, indicating that the nature of the construct
changes. In addition, including the items reduces model ?t of
the structural models estimated, although the direction and
signi?cance of estimated coe?cients are largely the same.
926 H.C. Dekker / Accounting, Organizations and Society 33 (2008) 915–941
while a few spent a substantial amount of time. A
double square root transformation was conducted
to obtain a better approximation of a normal
distribution.
Size is measured by the sum of money paid to
the supplier for the transaction, using a ?ve-point
scale (cf. Buskens, 2002). Using an approximate
currency conversion of 1.65 Dutch guilders per
US dollar at the time of the survey administration,
the cuto? values for the ?ve price categories are:
less than $15,000, $30,000, $60,000 and $120,000,
and more than $120,000. This indicator correlates
signi?cantly with measures of the importance of
the transaction for the automation of the buyer
(r = 0.35), for the pro?t of the buyer (r = 0.27),
and for the pro?t of the supplier (r = 0.31), indi-
cating that larger transactions were also more
important for both exchange partners.
Asset speci?city has three indicators re?ecting
the costs of switching to another supplier and prod-
uct (Anderson & Dekker, 2005; Batenburg et al.,
2003). Speci?cally, questions were asked about
the estimated cost or damage in time and resources
that the buyer would have incurred should the
product have failed to function and needed to be
replaced (1 = very small, 5 = very large). These
costs relate to the re-education of personnel, re-entry
of data and idle production of other ?rm units, and
can be interpreted as a direct consequence of asset
speci?city, since more speci?c products will involve
higher switching costs. The re-education costs
relate to the buyer’s investments in speci?c knowl-
edge to use the product, which would be of no use
for other IT products (i.e., human asset speci?city).
IT transactions often involve substantial human
asset speci?city, in particular when systems and
applications are custom-tailored (Poppo & Zenger,
1998). The specialized knowledge and skills
required to develop, use and maintain customized
software applications impede switching to other
suppliers without incurring considerable costs of
obsolete knowledge and new training. Similarly,
the costs of data re-entry and idle production relate
to dedicated assets that cannot easily be replaced.
Jointly, these indicators indicate the magnitude of
the buyer’s ‘holdup problem’.
Uncertainty has three indicators re?ecting the
extent to which the buyer was able to: (1) evaluate
the quality of the product at delivery, (2) compare
the product with other suppliers’ products, and (3)
compare potential suppliers’ price/quality ratio
(1 = very easy, 5 = very di?cult). These items thus
re?ect the di?culty of specifying ex ante and veri-
fying ex post the products and services transacted.
Task interdependence has two indicators relat-
ing to the type of software, hardware and services
exchanged: transaction scope and complexity (cf.
Anderson & Dekker, 2005) (Table 1, Panel B).
The majority of transactions in the sample did
not exist of the purchase of a single product or ser-
vice, but instead of a ‘package’ of di?erent prod-
ucts and services. A response pattern analysis in
PRELIS (Jo¨ reskog & So¨ rbom, 1993) reveals 636
di?erent transaction patterns in the data, with buy-
ers purchasing one to sixteen di?erent products
and services. Scope measures the number of di?er-
ent products and services exchanged and thus
re?ects the diversity of transaction components.
Transactions of broader scope are more diverse
and likely require more coordination between
exchange partners than transactions of narrow
scope.
Complexity measures the highest level of com-
plexity of the products and services transacted,
re?ecting the highest level of interdependence to
be managed by the exchange partners. As shown
in Panel B of Table 1, a ?ve-point scale is used
to categorize all transaction components by di?er-
ent levels of complexity following from customiza-
tion. Increasingly customized products and
services involve higher levels of interdependence
and an increasing need for adaptation between
exchange partners. The purchase of on-shelf soft-
ware, for instance, re?ects a situation of sequential
interdependence with limited coordination require-
ments. Customized design of special purpose soft-
ware, in contrast, re?ects a situation of reciprocal
interdependence, requiring more extensive gover-
nance to facilitate coordination and mutual adjust-
ment between buyer and supplier. The ?ve
categories thus are increasing in the degree to
which coordination and transfer of ?rm-speci?c
knowledge need to take place, and include: (1)
standard products, (2) complex standard products,
(3) branch speci?c software and added services, (4)
customized products and (5) tailor-made products.
H.C. Dekker / Accounting, Organizations and Society 33 (2008) 915–941 927
This Guttman type scale recognizes that higher
levels of interdependence include all lower levels
to be managed (Thompson, 1967).
Consistent with the expectation of White and
Lui (2005) that increasing task complexity requires
managers to spend greater time and e?ort to work
with an alliance partner, transaction scope and
complexity correlate well with three available indi-
cators of coordination e?ort: the duration of deliv-
ery (r
s
= 0.32 and r
s
= 0.47), the time that supplier
employees were present at the buyer for delivery
(r
s
= 0.32 and r
s
= 0.40), and the time that buyer
employees were present at the supplier for delivery
(r
s
= 0.23 and r
s
= 0.25). Similarly, the mean scope
and complexity are signi?cantly broader and
higher for transactions with onsite presence of
employees at the exchange partner than for trans-
actions without onsite presence.
Dependence is re?ected by the buyer’s perceived
dependence on the supplier before the focal trans-
action (1 = very small, 5 = very large). This indi-
cator correlates signi?cantly with the buyer’s
expectation of future transactions with the sup-
plier (r = 0.31), which can be a source of
dependence.
Partner experience is indicated by whether or
not the buyer and supplier had transacted before
(similar to e.g., Batenburg et al., 2003; Buskens,
2002; Gulati, 1995; Gulati & Singh, 1998). In the
sub-sample of exchange partners with prior ties
(N = 404), the median (mean) length of prior busi-
ness relationships was 5 (6.4) years, the median
transaction frequency was ‘regularly’ (4 on a ?ve-
point scale), and the median size of prior business
was ‘reasonably sized’ (3 on a ?ve-point scale).
Most buyers indicated to be reasonably to very
satis?ed with these prior experiences.
7
In addition,
before the focal transaction, buyers with prior ties
with the supplier had higher expectations of future
business with the supplier, and were more inclined
to recommend the supplier to other ?rms.
8
The descriptive statistics in Table 2 show that
the responses to the indicators of both dependent
and independent variables vary substantially.
Most missing values relate to the indicators of
Governance extensiveness (67) and Selection e?ort
(65), while the other indicators have relatively
few missing values. Since the number of missing
values is modest, listwise deletion was used, leav-
ing a main sample size of 817 observations for
analysis. Apart from selection time, the variable
distributions show no important deviations from
normality. Unreported subgroup descriptive statis-
tics indicate that on average, ?rms with prior ties
used less extensive governance structures and
spent less time selecting a partner, while the aver-
ages are also somewhat lower for most exogenous
variables (tests for structural di?erences between
these groups are discussed in the section ‘‘Results
for the moderated e?ect models’’). Table 3 reports
Pearson correlations between the indicators.
Statistical method
PRELIS and LISREL 8.52 were used to pre-
pare and analyze the data (Jo¨ reskog & So¨ rbom,
1993). In the analyses, the measurement model
(MM; relating indicators to latent constructs)
and the structural model (SM; relating latent con-
structs to each other) are estimated simultaneously
using maximum likelihood (ML) estimation. ML
performs well even under conditions of non-nor-
mality, model misspeci?cation and sample size
variation (Boomsma & Hoogland, 2001; Olsson,
Foss, Troye, & Howell, 2000). Robust ML
(RML) standard errors are calculated, which
improve ML standard errors for larger samples
(Boomsma & Hoogland, 2001; Distefano, 2002).
Because RML standard errors are adjusted
upwards, parameters are less likely to become sig-
ni?cant. The main sample, and the subsamples of
7
The sub-sample includes only 14 buyers who, despite
unsatisfactory prior transaction experiences re-used the sup-
plier. Positive experiences were much more common: 109
buyers were ‘reasonably satis?ed’, 283 buyers were ‘satis?ed’
and 66 buyers were ‘very satis?ed’. Buyers who transacted with
known suppliers also reported substantially more often that
former suppliers had been considered as potential partners for
the focal transaction and that they started their supplier search
among prior partners.
8
The standardized mean di?erence between ?rms with and
without prior ties (Cohen’s d) is 0.72 for ‘level of expected future
business’ and 0.68 for ‘would recommend to other ?rms’,
indicating substantial group di?erences.
928 H.C. Dekker / Accounting, Organizations and Society 33 (2008) 915–941
?rms with and without prior ties in the multi-
group analyses that follow, are su?ciently large
for RML estimation. Although the sample sizes
may be too small to obtain equally reliable esti-
mates, weighted least squares (WLS) estimation
based on polychoric correlations is also employed
as a form of triangulation. Converging results pro-
vide additional evidence that parameter estimates
are accurate and that the correct model structure
is identi?ed (Olsson et al., 2000).
Multiple ?t measures are used to assess the extent
to which the estimated models ?t the sample data,
and to increase the probability of rejecting false
models and not rejecting ‘true’ models (Hu & Ben-
tler, 1999). The goodness-of-?t index (GFI),
adjusted GFI (AGFI), standardized root mean
residual (SMSR), and root mean squared error of
approximation (RMSEA) indicate how well the
model reproduces the sample data. The non-
normed ?t index (NNFI) and comparative ?t index
(CFI) compare the discrepancies from a ‘null-
model’ with the discrepancies from the ?tted model
to evaluate the improvement in ?t by specifying
relationships. Recommended cuto? values are 0.08
for SRMR, 0.06 for RMSEA and 0.95 for
(A)GFI, NNFI and CFI, with ‘loosened’ values
for combinations of measures (Fan, Thompson, &
Wang, 1999; Hu & Bentler, 1999). In addition, the
Scaled MLchi-square (SMLv
2
) goodness-of-?t sta-
tistic (correcting v
2
for non-normality; Boomsma &
Hoogland, 2001) and the Akaike Information
Criterion (AIC) are used to compare non-nested
Table 2
Descriptive statistics and measurement model estimates (N = 817)
Descriptive statistics Measurement model estimates
a
Min Max Mean SD Skew Kurt k SE t-Value k
s
Endogenous variables
Governance extensiveness
Number of agreements 2 48 24.42 10.84 0.10 À0.65 1.00 – – 0.95
Selection e?ort
Selection time – person-days 0 300 15.75 28.79 3.98 22.09
Selection time – transformed 0 4.16 1.59 0.71 0.33 0.63 1.00 – – 0.95
Exogenous variables
Size
Financial volume 1 5 2.54 1.45 0.42 À1.21 1.00 – – 0.89
Asset speci?city
Re-education costs (k = 1) 1 5 2.36 1.15 0.45 À0.82 1.00 – – 0.86
Data re-entry cost 1 5 2.99 1.38 À0.01 À1.28 1.01 0.09 11.71
***
0.69
Idle production costs 1 5 2.33 1.32 0.63 À0.85 0.78 0.07 10.92
***
0.54
Uncertainty
Di?culty measuring quality (k = 1) 1 5 2.71 0.99 0.42 À0.42 1.00 – – 0.61
Di?culty comparing products 1 5 2.91 1.09 0.19 À0.70 1.81 0.20 9.24
***
0.89
Di?culty comparing price/quality 1 5 2.96 1.08 0.12 À0.72 1.76 0.19 9.37
***
0.85
Task interdependence
Transaction complexity (k = 1) 1 5 3.32 1.43 À0.46 À1.11 1.00 0.08 11.03
***
0.80
Transaction scope 1 16 5.20 3.44 0.61 À0.45 0.85 – – 0.89
Dependence
Perceived dependence on supplier 1 5 2.59 1.21 0.03 À1.24 1.00 – – 0.89
Partner experience
Previous transacting 0 1 0.49 – 0.02 À2.00 1.00 – – 0.89
a
The measurement model estimates are based on robust ML estimation of Model 2. Reference indicators for multi-item constructs
are denoted by k = 1.
***
Indicates a p value