You are on page 1of 18

A Model of Distributor Firm and Manufacturer Firm Working Partnerships

Author(s): James C. Anderson and James A. Narus


Source: Journal of Marketing, Vol. 54, No. 1 (Jan., 1990), pp. 42-58
Published by: American Marketing Association
Stable URL: http://www.jstor.org/stable/1252172 .
Accessed: 22/06/2014 01:50

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp

.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.

American Marketing Association is collaborating with JSTOR to digitize, preserve and extend access to
Journal of Marketing.

http://www.jstor.org

This content downloaded from 95.149.73.109 on Sun, 22 Jun 2014 01:50:22 AM


All use subject to JSTOR Terms and Conditions
James C. Anderson & James A. Narus

A Model of DistributorFirm
and Manufacturer Firm
Working Partnerships
A model of distributor firm and manufacturer firm working partnerships is presented and is assessed
empirically on a sample of distributor firms and a sample of manufacturer firms. A multiple-informant
research method is employed. Support is found for a number of the hypothesized construct relations
and, in both manufacturer firm and distributor firm models, for the respecification of cooperation as an
antecedent rather than a consequence of trust. Some implications for marketing practice are discussed
briefly.

M ARKETPLACE trends have underscored the commitments (Arthur Andersen & Co. 1987). As a
need for a better understanding of working part- confluent result, manufacturer firms and distributor
nerships between manufacturer firms and distributor firms are involved in fewer, but increasingly signifi-
firms. Sales through wholesaler-distributors have been cant, working partnerships in which better coordina-
growing at a rate faster than the U.S. gross national tion of marketing and technical activities is essential
product (GNP) and topped the $1.4 trillion mark in for their mutual success in the customer marketplace.
1987 (Arthur Andersen & Co. 1987). In another trend, In this article, we report a study of distributor and
significant consolidation is occurring in the whole- manufacturer working partnerships, which we define
sale-distribution industry as larger, professionally as the extent to which there is mutual recognition and
managed firms acquire smaller "mom-and-pop" firms understanding that the success of each firm depends
at a rapid rate. Finally, end-user or customer firms in part on the other firm, with each firm consequently
have signalled through the growth of systems con-
taking actions so as to provide a coordinated effort
tracts and preferred vendor programs an increased de- focused on jointly satisfying the requirements of the
sire to concentrate their purchases with fewer whole- customer marketplace. Aspects of manufacturer and
saler-distributors, to whom they offer longer-term distributor working partnerships have received con-
siderable attention in channels research over the past
20 years (e.g., Frazier 1983a; Frazier and Sheth 1985;
JamesC.Anderson is theWilliamL.FordProfessor of Marketingand Stern and Reve 1980). We attempt to contribute to
WholesaleDistribution andAssociateProfessorof Behavioral
Science this knowledge and research in marketing channels in
in Management,J. L.Kellogg GraduateSchoolof Management, North- at least two ways. First, we develop and present a
westernUniversity.
JamesA. Narusis AssociateProfessor of Manage-
ment,Babcock Graduate Schoolof Management, WakeForestUniver- model of distributor firm and manufacturer firm work-
sity.Theauthorsgratefully acknowledgethe financialsupportof the ing partnerships that is both broad in scope and spec-
MarketingScienceInstituteandthe Distribution
Research & Education ified in a testable form. Ster and Reve (1980) and
the helpfulcomments
Foundation, andsuggestions of ErinAnderson, Frazier and Sheth (1985), in particular, have stressed
F.RobertDwyer,GaryFrazier, GeorgeJohn,Robert Spekman, Barton the need for constructing and testing more compre-
Weitz,andthethreeanonymous JMreviewers, andthe assistance of
Sethuraman withtheanalysis. hensive models of marketing channel phenomena.
Rajagopalan
Second, we report an initial empirical assessment of

Journal of Marketing
42 / Journalof Marketing,January1990 Vol. 54 (January 1990), 42-58

This content downloaded from 95.149.73.109 on Sun, 22 Jun 2014 01:50:22 AM


All use subject to JSTOR Terms and Conditions
this model with research employing multiple infor- ceived dependence on a working relationship, we con-
mants within distribution firms and within manufac- tend that the firm's perception of its dependence relative
turer firms drawn from a cross-section of industries. to its partner's dependence on the relationship is a
Phillips (1981) has demonstrated the need to use mul- construct of greater interest in channels research. It is
tiple informants within each firm to obtain more valid this relative dependence that determines the extent to
estimates of organizational properties, and rarely have which a firm will have influence over, and be influ-
both the manufacturer's and the distributor's perspec- enced by, its partner. Relative dependence can be de-
tives on channel partnerships been assessed empiri- fined as a firm's perceived difference between its own
cally. John and Reve (1982) and Eliashberg and Michie and its partner firm's dependence on the working part-
(1984) studied the joint perspective of manufacturers nership. This relative dependence construct is sup-
and distributors, which has been referred to as a dyadic ported by previous conceptual and empirical work on
perspective (cf. Achrol, Reve, and Stem 1983). In interdependence (Cadotte and Ster 1979; Robicheaux
contrast, we first assess separately the manufacturer and El-Ansary 1976), countervailing power (Etgar
firm and distributor firm perspectives on working 1976), and "power advantage" (Emerson 1962).
partnerships. Then we assess the commonalities and A primary consequence of relative dependence is
differences across perspectives. influence, or power. The conceptualizations of "in-
terfirm power," "potentialfor influence," and "achieved
A Model of Working Partnerships influence" advanced by Frazier (1983b; Frazier and
Summers 1984, 1986; Frazier and Sheth 1985), which
Building on previous work in social exchange theory
are compatible with social exchange theory, appear to
(Kelley and Thibaut 1978; Thibaut and Kelley 1959)
and interorganizational exchange behavior (Anderson offer the most managerially germane theoretical ex-
and Narus 1984; Cadotte and Stem 1979; Frazier 1983a; planation of the channel power phenomenon. Build-
Robicheaux and El-Ansary 1976) and melding this work ing upon this work, we use the constructs influence
with a series of interviews with 20 manufacturer firm over the partnerfirm and influence by the partnerfirm
and distributor firm managers, we constructed a com- to reflect the interdependent nature of working part-
prehensive model of ongoing, working partnerships in nerships and the resultant need to have one's partner
firm take actions that can bring about (directly or in-
marketing channels (Figure 1). Like previous concep-
tual models in this area (e.g., Cadotte and Ster 1979; directly through facilitation) better outcomes for one's
own firm. These constructs representthe extent to which
Dwyer, Schurr, and Oh 1987; Frazier 1983a;
Robicheaux and El-Ansary 1976), this model is meant a firm has applied its power to induce desired actions
to apply to both manufacturerfirm and distributor firm by its partner firm. By doing so, a firm can "shape
its relationship with another firm" (Frazier 1983a). This
perspectives on working partnerships. The rationale is
that both firms are participants in the same exchange conceptualization of the influence constructs also is
consistent with the concept of behavioral control (cf.
relationship and, though they differ in the marketing
functions they perform, symmetry is expected in the Anderson and Narus 1984; Kelley and Thibaut 1978),
behavioral constructs that underlie the relationship. In where by varying its own actions, a firm can make it
in the best interest of the partner firm to covary its
contrast, differences between manufacturers and dis-
tributors are expected in terms of the indicators that actions.
reflect these constructs and the presence or relative We predict a positive link between relative depen-
dence and influence by the partner firm and a negative
strengths of the posited construct relationships. Fi-
link between relative dependence and influence over
nally, the constructs in this model are conceptualized
as firm-level representations of the critical factors in the partner firm. Specifically, we contend that the firm
with greater relative dependence has, by definition,
working partnerships.
relatively greater interest in sustaining the relation-
Dependence and working partnerships. Depen- ship, and that one way to do so is to be more receptive
dence, influence (i.e., power), and conflict are per- to requests and amenable to changes suggested by its
haps the most widely studied aspects of channel work- partner firm. In contrast, the firm with lesser relative
ing partnerships(Frazier 1983a; Heide and John 1988). dependence can use its superior position to request
Dependence on the working partnership is concep- changes of its partner that it believes will either mu-
tualized for each firm as outcomes given comparison tually increase the outcomes of both partners or singly
levelfor alternatives (CLat) (cf. Anderson and Narus increase its own outcomes from the relationship (which
1984; Kelley and Thibaut 1978). CLalt in this context will have the effect, though, of moving the relation-
is a standard that represents the overall quality of out-
ship toward parity in relative dependence). These re-
comes (economic, social, and technical) available to
lationships are supported by substantial conceptual and
the firm from the best alternative exchange relation-
empirical work (e.g., Frazier and Summers 1986; Gaski
ship. Rather than simply considering a firm's per- 1984).

A Modelof DistributorFirmand Manufacturer


FirmWorkingPartnerships
/ 43

This content downloaded from 95.149.73.109 on Sun, 22 Jun 2014 01:50:22 AM


All use subject to JSTOR Terms and Conditions
FIGURE1
Model of Manufacturer and Distributor Working Partnerships

In our model, conflict represents the overall level sessment of the results (rewards obtained minus costs
of disagreement in the working partnership (for a re- incurred) from a given working partnership in com-
view of channel conflict research, see Reve and Ster parison with expectations based on present and past
1979). As such, conflict is determined by the fre- experience with similar relationships, and knowledge
quency, intensity, and duration of disagreements. On of other firms' relationships (Anderson and Narus
the basis of past research (Frazier 1983a; Gaski 1984), 1984).
we predict a positive causal relationship between in- Closely linked with outcomes given comparison
fluence by the partner firm and the level of conflict level is the construct communication, which can be
in the working relationship. Conversely, we posit a defined broadly as the formal as well as informal shar-
negative causal relationship between influence over ing of meaningful and timely information between firms
the partner firm and conflict (Gaski 1984). The greater (Anderson and Narus 1984). This definition has as its
the influence a firm has over its partner firm, the less focus the efficacy of information exchange rather than
conflict the firm encounters because it is in the best the quantity or amount, and the construct inherently
interest of the partner to comply with the firm's re- taps past communications. That is, an informant who
quests for action. is asked to characterize his or her firm's communi-
cation with a partner firm reflects on recent past ex-
Coordinating efforts in working partnerships. In perience between the firms.
this section, we consider constructs that underlie the We predict a positive correlation between out-
process of working together within a partnership. Be- comes given comparison level and communication
cause our modeling efforts focus on ongoing partner- (Anderson and Narus 1984). An explanation for this
ships, we begin with the current evaluation of past prediction is reciprocal facilitation, whereby attaining
relationship outcomes. The construct that captures this outcomes that surpass the comparison level facilitate
evaluation is outcomes given comparison level (CL), good communication between firms and communi-
which is a focal construct of social exchange theory cation facilitates achieving outcomes that surpass the
(cf. Kelley and Thibaut 1978; Thibaut and Kelley comparison level. We also predict a positive corre-
1959). Outcomes given CL is defined as a firm's as- lation between relative dependence and outcomes given

44 / Journalof Marketing,
January1990

This content downloaded from 95.149.73.109 on Sun, 22 Jun 2014 01:50:22 AM


All use subject to JSTOR Terms and Conditions
CL, based on the conceptualizationsof CLalt and CL of communicationas a construct, we posit that at any
given by Thibaut and Kelley (1959), whereby out- one point in time, (inherently past) communication
comes available from the best alternativerelationship causes (present) trust.
are a part of the firm's experience and so contribute Consideringthe consequences of trust, we posit a
to the firm's expectations. positive causal path from trust to cooperation (cf.
Trust in a working relationship and its implica- Anderson and Narus 1986; Dwyer, Schurr, and Oh
tions for a firm's actions have been defined as "the 1987). Cooperationis defined here as similaror com-
firm's belief that another company will perform ac- plementarycoordinatedactions taken by firms in in-
tions that will result in positive outcomes for the firm, terdependent relationships to achieve mutual out-
as well as not take unexpected actions that would re- comes or singularoutcomeswith expectedreciprocation
sult in negative outcomes for the firm. The strength over time. Once trust is established, firms learn that
of this belief may lead the firm to make a trusting coordinated,joint efforts will lead to outcomes that
response or action, whereby the firm commits itself exceed what the firm would achieve if it acted solely
to a possible loss, dependingupon the subsequentac- in its own best interests. In addition, within the con-
tions of the othercompany"(AndersonandNarus1986, text of a trusting relationship, firms sometimes are
p. 326). Truststill will be maintained,however, if the willing to postponetemporarilythe receiptof theirown
firm believes its partnerhas taken the expected ac- outcomes until some later time. Managerial inter-
tions, but that forces beyond its control have negated views underscoredthe importantrole of this notion of
the expected outcomes. Some care must be taken, expected reciprocityin channel partnerships.
though, in generalizing trust as a construct in inter- Another posited consequence of trust is function-
personalrelationsto trust in working partnerships.In ality of conflict, which can be defined as an evaluative
interpersonal relations, participants expose them- appraisalof the results of recent efforts to resolve dis-
selves and theirown resourcesto potentialloss whereas agreements (Rosenberg and Ster 1970; see also
in interorganizationalrelations, it is the firm that po- Robicheauxand El-Ansary 1976; Rosenbloom 1973).
tentially will incur the loss. Trust as a construct in That is, in the minds of a firm's managers, to what
channel partnershipstherefore may entail less inten- extent have recent disagreementswith a partnerfirm
sity and personal commitment. been productively resolved? When partnerfirms use
We posit a positive path from outcomes given disagreementsas a means of "clearingthe air" of po-
comparisonlevel to trust. The rationale is that a re- tentiallyharmfultensionsand ill-will, conflictcan have
view of past results in comparisonwith expectations functional and productive consequences. By making
leads to the firm's predictionor estimate of the extent conflict functional, partnerfirms maintaincordial re-
to which the partnerfirm will follow through on its lations and tend to give each other the "benefitof the
currentpromises. On the basis of these estimates, the doubt"in conflict episodes (Hardyand Magrath1988).
firm engages in or refrainsfrom trustingresponses or We believe that firms that have developed strongtrust
actions.Ample supportfor this causallink can be found in a relationshipare more likely to work out their dis-
in social science researchreviewed by Anderson and agreementsamicably and, in fact, accept some level
Narus (1986), the conceptualwork of Dwyer, Schurr, of conflict as being "just anotherpart of doing busi-
and Oh (1987), and the managerialinterviewswe con- ness." This view is sharedby Dwyer, Schurr,and Oh
ducted. (1987).
Though the link between communicationand trust We furtherhypothesize a negative causal relation-
has received attention (e.g., Bialaszewski and ship between functionality of conflict and conflict.
Giallourakis 1985), researchersdo not agree on the Firms that have the ability to resolve conflicts have a
direction of the relationship. For instance, Dwyer, lower overall level of conflict in their working part-
Schurr, and Oh (1987) hypothesize that trust causes nerships, in part because there are fewer recurring
communication,whereasAnderson,Lodish, and Weitz conflicts. Another rationaleis that partnerfirms that
(1987) contendthatcommunicationleads to trust. Our have functional conflict frame the issues differently
interpretationis that building and sustaining working (Tversky and Kahneman 1981); conflict is seen as
partnershipsis an iterativeprocess. Meaningfulcom- "productivediscussion" rather than an "argument."
municationbetween firms in a working partnershipis Support for this path also is found in the work of
a necessaryantecedentof trust(cf. Andersonand Narus Robicheauxand El-Ansary(1976) and Frazier(1983a).
1986). In subsequent periods, however, this accu-
mulation of trust leads to better communication.Our Sustaining satisfaction in working partnerships.
model is meant to representthe working partnership Satisfaction has been defined as "a positive affective
at one period of time. That is, it is a static model that state resulting from the appraisalof all aspects of a
provides a single-time-periodperspective of ongoing firm's working relationship with another firm"
partnerships,in contrastto a dynamic model or inter- (Anderson and Narus 1984, p. 66). Note that satis-
temporalperspective. Therefore,and given the nature faction is affective and can be contrastedwith an ob-

A Modelof DistributorFirmand Manufacturer


FirmWorkingPartnerships
/ 45

This content downloaded from 95.149.73.109 on Sun, 22 Jun 2014 01:50:22 AM


All use subject to JSTOR Terms and Conditions
jective or "rational" summary assessment of out- than 110 industries. A total of 504 wholesaler-distrib-
comes, as captured by the concept of outcomes given utor firms returned completed, usable research-infor-
comparison level (cf. Thibaut and Kelley 1959). Be- mation forms.
cause of this conceptualization, a positive causal re- The research design called for each distributor firm
lationship is posited from outcomes given CL to sat- to evaluate its working partnership with one manu-
isfaction. facturer firm. Preliminary field interviews revealed that
Satisfaction is the focal consequence of working the working relationships with manufacturer firms that
partnerships in our model, which is consistent with supplied the first- or second-highest-selling product
past interorganizational exchange models (e.g., lines tended to be uniformly positive, whereas rela-
Anderson and Narus 1984; Frazier 1983a; Frazier, tionships with the fourth manufacturerfirms were more
Spekman, and O'Neal 1988). We contend that satis- variable. To avoid potential restriction in range prob-
faction, by its nature, is not only a close proxy for lems (Lord and Novick 1968) and to facilitate obtain-
concepts such as perceived effectiveness, but also may ing variation in the relationships studied, the distrib-
be more predictive of future actions by partner firm utor executives were asked to specify as their
managers. Further, satisfaction has been found to lead manufacturer partner the firm that supplied the fourth-
to the long-term continuation of relationships (Gladstein highest-selling product line and accounted for at least
1984). 5% of their total sales. As the typical distributor may
We posit a positive causal relationship between in- represent more than 100 product lines, however, the
fluence over the partner firm and satisfaction and a fourth relationships were still of considerable impor-
negative causal relationship between influence by the tance to the distributor. If the distributor firm did not
partner firm and satisfaction (Dwyer 1980; Wilkinson carry four lines or if the fourth line did not account
1979). Firms that perceive that they are "calling the for at least 5% of total sales, the executives were in-
shots" in a relationship will experience greater satis- structed to use the manufacturer of the third-highest-
faction. Conversely, given that firms often perceive selling product line, and so on.
some cost in complying with a partner's requests, either The research design also called for multiple in-
in taking some action or in foregoing some other pos- formants for each manufacturer firm and distributor
itive outcomes, a firm will experience lower satisfac- firm. The distributor executives were asked to name
tion to the extent that it perceives that it is influenced the highest-level manager at the manufacturerfirm with
by its partner firm. We further posit a negative causal whom they dealt, as well as the operational-level
link between conflict and satisfaction (cf. Frazier 1983a; manager with whom they interacted most frequently.
Gaski 1984). Disagreements tend to block achieve- Finally, they were asked to name the operational-level
ment of the firm's goals, eliciting frustration, and manager in their own firm who interacted with this
thereby cause feelings of unpleasantness about the manufacturer firm most frequently. Thus, the unit of
partnership. analysis for the study was the firm, represented by
Finally, we predict a positive path from coopera- two informants.
tion to satisfaction. Working together with a partner Of the 504 sets of manufacturerand distributorfirms
firm in pursuit of mutual benefits increases each firm's mailed questionnaires, 16 returned the questionnaires
perceptions of compatibility with its partner firm. This uncompleted (six partnerships had been terminated,
perceived compatibility, as well as the fulfillment as- eight distributor firms had gone out of business, and
sociated with attaining desired outcomes, affords a two manufacturer firms had gone out of business).
strong feeling of "chemistry" and results in satisfac- Therefore, the effective sample size was reduced to
tion with the partnership. This path has received long- 488 sets of firms or 1952 informants.
time support from channels analysts (e.g., Dwyer 1980; A total of 1365 informants completed and returned
Mallen 1963; Sibley and Michie 1982). questionnaires, an overall response rate of 70%; the
response rate was 75% (730) for distributors and 65%
(635) for manufacturers. The average percentage of
Method distributor firm sales accounted for by the evaluated
manufacturer firm was 16.8%. Of primary interest are
Respondent Samples the 253 distributor firms and 217 manufacturer firms
To construct the samples of distributor firms and man- for which both informants in each firm responded.
ufacturer firms, the Distribution Research and Edu- A critical assumption in multiple-informant re-
cation Foundation (DREF) initially mailed letters to search is that random measurement errors across in-
the chief executive officers of 5000 member firms of formants' reports are uncorrelated. This assumption
the National Association of Wholesaler-Distributors will hold if the two informants independently answer
(NAW) inquiring about their firm's interest in partic- the questionnaire. If there is collusion in answering
ipating in the research. These firms represent more the measures or if only one individual answers both

46 / Journalof Marketing,
January1990

This content downloaded from 95.149.73.109 on Sun, 22 Jun 2014 01:50:22 AM


All use subject to JSTOR Terms and Conditions
questionnaires,the errorsare likely to be correlated. posited working partnershipmodel in Figure 1.
To assess this independenceassumption,a correlation Finally, a multiple-sampleanalysis was performed
between informantsin each firm was computedacross to comparethe manufacturerfirm and distributorfirm
responsesto 55 questionnaireitems (four measuresfor perspectives(J6reskog 1971). Figure 2 is an overview
each of the 10 constructs,plus 15 additionalitems that of the model-buildingtask. For furtherexposition of
were common across questionnaires).The two infor- this confirmatory model-building approach, see
mantsfor four distributorfirms and four manufacturer Anderson(1987). The LISRELprogram(J6reskogand
firms had correlationcoefficients greaterthan .99. In- Sorbom 1984) was used throughoutfor model esti-
spection of the two informantreports from each of mation.
these eight firms revealed identicalresponses to all 55
Respecification.Two kinds of respecificationswere
questions in seven cases and only one response dif- made. First, two measures were respecified as being
ferent in the eighth. As the independenceof response related to a single underlying indicator if the corre-
assumptionappeareduntenable for these firms, they lation between their original two indicators was not
were eliminated from consideration. Thus, the final
significantly different from 1.0, suggesting that the
samples for analysis consisted of 249 distributorfirms informantshad not differentiatedthe indicatorsfrom
and 213 manufacturerfirms.
one another, and this finding was supportedby con-
Procedure tent considerations.Supportfor these respecifications
also was provided by the similarity coefficient (cf.
Preliminaryfield interviews were conductedwith ap- Anderson and Gerbing 1982) for the two indicators.
proximately20 manufacturerfirm managersand dis- The other respecificationwas eliminationof mea-
tributorfirm managers. On the basis of these discus-
sures, which was done underthreecircumstances.First,
sions, two questionnaires-one for manufacturerfirms measures were eliminated if perceptual agreement was
and one for distributorfirms-were constructedto be
as parallel as possible. Four measures were included
for each of the 10 constructs in Figure 1. Both ver- FIGURE2
sions of the questionnairewere pretestedwith 15 dis- Overview of the Model-Building Task
tributormanagersand manufacturermanagers. In ad-
dition, the 12-person DREF Board of Directors I. COMPOSITION MODEL - SPECIFIES THE FUNCTIONAL
reviewed the drafts and offered comments on im- RELATIONSHIPBETWEEN MEASUREMENTSTAKEN FROM
MULTIPLEINFORMANTSAND THEIRCORRESPONDING
provements. ORGANIZATION-LEVEL INDICATORS(OF CONSTRUCTS).
Each of the 2016 individuals whose names were PROVIDES AN ASSESSMENT OF THE PRECEPTUAL
AGREEMENTBETWEEN INFORMANTSWITHRESPECT TO
obtainedon the research-information forms was mailed EACH MEASURE.
under separate cover the appropriateversion of the
questionnaire. An envelope addressed to one of the
researcherswas enclosed. Two followup mailingswere
sent to nonrespondinginformants. As a followup to
II. ORGANIZATIONALMEASUREMENTMODEL- RELATESTHE
the research, several field interviews and two focus INDICATORSTO THEIR POSITED, UNDERLYINGORGANIZA-
TIONALCONSTRUCTS. PROVIDES AN ASSESSMENT OF
group interviews were conducted to gain additional CONVERGENTVALIDITYAND DISCRIMINANTVALIDITY.
understandingof the results and to compare the re-
search-basedexplanationswith those from the every-
day experience of managers(cf. Calder 1977).
Analyses
Modeling approach. The approachfor confirmatory III. ORGANIZATIONAL STRUCTURALMODEL - RELATESTHE
CONSTRUCTS TO ONE ANOTHERAS SPECIFIED BY THEORY
measurementand structuralequation modeling of or- AND PAST RESEARCH. PROVIDES AN ASSESMENT OF
NOMOLOGICALVALIDITY.
ganizational properties recently given by Anderson
(1987) was used. First, a confirmatorycomposition
model specified the functional relationship between
measurementstaken from the two informantsper firm
and their correspondingorganization-levelindicators
(of the organizational constructs). A confirmatory IV. MULTIPLE- SAMPLEANALYSIS - GIVES AN ASSESSMENT
OF THE COMMONALITIES AND DIFFERENCES BETWEEN
measurementor factor analysis model (cf. Joreskog DISTRIBUTORFIRMSAND MANUFACTURERFIRMSWITH
and Sorbom 1984, p. 1.9-10) next related the indi- RESPECT TO THE SUBSTANTIVE,THEORETICALMODEL
OF INTEREST.
cators defined in the composition model to their pos-
ited underlying organizationalconstructs. A confir-
matory structuralmodel then was used to assess the

A Modelof DistributorFirmand Manufacturer


FirmWorkingPartnerships
/ 47

This content downloaded from 95.149.73.109 on Sun, 22 Jun 2014 01:50:22 AM


All use subject to JSTOR Terms and Conditions
not significant (cf. James 1982), as revealed by a non- practical sense as indicated by the normed fit index
significant pattern coefficient for one or both infor- (cf. Bentler and Bonett 1980) value of .86. This value
mants on the underlying indicator. Second, multidi- means that 86% of observed-measure covariation is
mensional measures were eliminated. Such measures explained by the composition model (Anderson and
have underlying indicators that exhibit a pattern of Gerbing 1988). Further reinforcing this judgment are
correlations not significantly different from 1.0 with the GFI value of .87 and the standardized RMR value
several indicators of different posited constructs. Fi- of .042, both of which are comparable to values ex-
nally, measures that were not unidimensional were pected from sampling error (cf. Anderson and Gerbing
eliminated; that is, the presence of large normalized 1984). Though two normalized residuals remain that
residuals (Joreskog and Sorbom 1984) signalled that have absolute values greater than two, the largest one
the alternate measures (from the two informants) could is only 2.32. Table 1 is a summary of the final com-
not be related to an underlying indicator in a consis- position model for distributor firms.
tent way (cf. Gerbing and Anderson 1988). All patterncoefficients relating the measures to their
It is important to note that respecifications in the underlying indicators are positive and statistically sig-
composition model, as well as in the organizational nificant, indicating perceptual agreement between in-
measurement model, are made independently of con- formants on each of the organization-level indicators.
siderations of the substantive, structural model. This Note that no hypothesized indicators of conflict
fact is a critical relative strength of staged estimation (CONFL) are able to be defined. The estimated cor-
approaches (cf. Anderson 1987) in comparison with relation/covariance matrix for the indicators then be-
a "one-step" estimation approach whereby the com- comes the input covariance matrix for the organiza-
position, measurement, and structural models are es- tional measurement model.
timated at the same time. In a staged approach, an
Organizational measurement model. After several
adequate goodness of fit for the composition model
respecifications, a final measurement model for dis-
implies nothing about the subsequent goodness of fit
tributor firms was obtained that is judged to provide
(or need for respecification) of a posited measurement
model, just as in turn an adequate goodness of fit for acceptable goodness of fit, though its associated chi
the measurement model implies nothing about the square value remains statistically significant (X13) =
89.27, p < .001). This judgment is based on mean-
subsequent goodness of fit (or need for respecifica-
tion) of a posited structural model. ingful interpretability of the model in terms of content
and theory, an acceptable value of .92 for the normed
fit index, and the absence of normalized residuals with
Results absolute values greater than two. This judgment is
supported further by a GFI value of .95 and an RMR
DistributorFirm Perspective value of .032. The parameter estimates for this mea-
surement model are reported in Table 2. Note that the
Composition model. After a series of respecifications
single indicator hypothesized to be related to satisfac-
guided by goodness-of-fit information in conjunction tion (SATIS) was found to be multidimensional and
with content considerations, a final composition model
was not retained. This final measurement model, which
for the 249 distributor firms was obtained. This model
can be thought of as a "saturated" structural model
retains 42 of the 80 measures and yields 17 organi-
zation-level indicators. Though the chi square good- (cf. Bentler and Bonett 1980) because each construct
is related to every other construct, served as a base
ness of fit is still statistically significant (X64) =
model for assessments of the structural model when
814.74, p < .001), this model is judged to provide
the measurement and structural submodels were es-
acceptable fit.' The bases for this judgment are that timated simultaneously (cf. Anderson and Gerbing
this model has relatively unambiguous meaning in terms
of content and has acceptable goodness of fit in a 1988).
Organizational structural model. The initial struc-
'To provide a basis for comparison, eight-measure models, con- tural model, specified from the posited relationships
sisting of the four posited measures of each construct taken across the in Figure 1, gave relatively poor goodness of fit
two informants, were estimated. These models are comparable to the
ones reported by Phillips (1981, Figure 2). For distributor firms, six (X35) = 341.30, p < .001). Several respecifications
of the 10 "single-construct" models have chi square goodness-of-fit were made that were indicated in terms of goodness
values that are not significant (p > .05), two that have significant
values for chi square have acceptable values for GFI (.987 and .983) of fit and made theoretical sense. Of perhaps greatest
and RMR (.028 and .046), and two are nonconvergent because of a interest is the respecification of cooperation (COOPER)
lack of external consistency of the two informants on one measure. as causally antecedent to trust (TRUST) rather than
The point is that what constitutes "acceptable" goodness of fit and
what amounts of respecification are to be expected should be framed being consequent to it. Though causal ordering is dif-
by considerations of the size and complexity of the model (cf. Bentler ficult to ascertain with cross-sectional data, this res-
and Chou 1987).
pecified ordering does provide estimates that better re-

48 / Journalof Marketing,January1990

This content downloaded from 95.149.73.109 on Sun, 22 Jun 2014 01:50:22 AM


All use subject to JSTOR Terms and Conditions
TABLE 1
Summary of Final Composition Model for Distributor Firms
Constructa Indicators Example Measure
Influence b)y ,T ManufacturerX has considerable latitude in deciding how much field sales
partner firrn assistance and technical support they give to our firm for their product line
(7-point scale: strongly disagree/strongly agree)
Relative T?, T3 Computed as the difference between: (a) In your judgment, the total costs to
dependenc ;e your firm in switching to a competing manufacturer's product line would be
and (b) In your judgment, the total costs to ManufacturerX in
replacing your firm with another distributor in your trade area would be
(5-point scales: prohibitive/negligible)
Communication ManufacturerX lets our firm know as soon as possible of any unexpected
problems with things such as lead times, delivery schedules, or product
quality (7-point scale: strongly disagree/strongly agree)
Functionality of Disagreements between ManufacturerX and our firm have the
conflict productivity of our working relationship (5-point scale: considerably
increased/considerably decreased)
Trust Based upon your past and present experience, how would you characterize
the level of trust your firm has in its working relationship with Manufacturer
X? (7-point scale: don't trust ManufacturerX/trust ManufacturerX
completely)
Cooperation T11-T13 Computed as the sum of: (a) Our firm helps out Manufacturer X in whatever
ways they ask and (b) ManufacturerX helps our firm out in whatever ways
we ask (7-point scales: strongly disagree/strongly agree)
Satisfaction T14 Our firm's working relationship with ManufacturerX has been an unhappy
one (7-point scale: strongly disagree/strongly agree)
Outcomes given T15 The financial returns our firm gets from ManufacturerX's product line are
comparison level what we look for in distributing a product line (5-point scale: greatly
above/greatly below)
Influence over T16, 717 To what extent does ManufacturerX follow whatever recommendations your
partner firm firm makes regarding the marketing and selling of their product line? (7-
point scale: not at all/to a great extent)
aPosited as underlying the given organizational indicator(s).

produce the construct correlations, given the pattern the remaining explanatory ability of the 19 paths that
of posited relationships of TRUST and COOPER with were constrained to zero is of little practical conse-
their antecedent constructs and consequent constructs. quence, as indicated by a decrement in the normed fit
Consistent with this change in causal sequence, paths index of only .046 for this structural model from the
were added from communication (COMMUN) and measurement model (from .919 to .873). Second,
outcomes given CL (CL) to COOPER. Further, paths though adding the next path (INFLBY -> TRUST) that
were added from influence over the partner firm (IN- was indicated and could potentially be justified the-
FLOV) to influence by the partner firm (INFLBY) and oretically produces a statistically significant differ-
to functionality of conflict (FUNCON), and from ence in goodness of fit (X2(l) = 8.2, p < .005), this
COOPER to FUNCON. path is judged not to be of practical significance in
The hypothesized correlation between relative de- that the magnitude of the estimate is small (332 = -.13)
pendence (RELDEP) and CL and the predicted path and the increment in the normed fit index it provides
from TRUST to FUNCON were not significant and is only .007.
thereafter were not estimated. This respecified struc-
tural model for distributor firms is shown in Figure 3.
Manufacturer Firm Perspective
Though this structural model still has a significant
chi square difference value in comparison with the
Composition model. As with the distributorfirm model,
measurement model (X2(9) = 51.30, p < .001), it is after a series of respecifications guided by goodness-
judged to provide acceptable goodness of fit in a prac- of-fit information in conjunction with content consid-
tical sense. There are two bases for this judgment. erations, a final composition model for the 213 man-
First, this structural model provides adequate expla- ufacturer firms was obtained. This model also retains
nation of the construct covariances; put differently, 44 of the 80 measures and yields 17 organization-level

A Modelof DistributorFirmand Manufacturer


FirmWorkingPartnerships
/49

This content downloaded from 95.149.73.109 on Sun, 22 Jun 2014 01:50:22 AM


All use subject to JSTOR Terms and Conditions
TABLE2
Final Organizational Measurement Model for Distributor Firms
Constructa
Indicator
Indicator INFLBY RELDEP COMMUN FUNCON TRUST COOPER CL INFLOV Specificity
L1 .995 (.00)b 0c 0 0 0 0 0 0 .01 (.09)
12 0 .995 (.00) 0 0 0 0 0 0 .01 (.09)
15 0 0 .82 (.06) 0 0 0 0 0 .33 (.04)
L6 0 0 .88 (.05) 0 0 0 0 0 .24 (.04)
17 0 0 0 .995 (.00) 0 0 0 0 .01 (.09)
9L 0 0 0 0 .995 (.00) 0 0 0 .01 (.09)
11 0 0 0 0 0 .83 (.05) 0 0 .32 (.04)
L12 0 0 0 0 0 .90 (.05) 0 0 .19 (.03)
L15 0 0 0 0 0 0 .995 (.00) 0 .01 (.09)
16 0 0 0 0 0 0 0 .995 (.00) .01 (.09)
Organizational Construct Estimated Correlation (4D)d
INFLBY RELDEP COMMUN FUNCON TRUST COOPER CL INFLOV
INFLBY 1.000 .066 .069 .065 .065 .068 .064 .066
RELDEP .258 1.000 .070 .064 .064 .068 .064 .066
COMMUN -.126 -.028 1.000 .063 .063 .033 .069 .069
FUNCON -.120 -.051 .563 1.000 .071 .060 .065 .066
TRUST -.211 -.067 .558 .493 1.000 .058 .066 .064
COOPER -.098 -.089 .839 .667 .733 1.000 .067 .068
CL -.074 .080 .178 .153 .245 .328 1.000 .064
INFLOV .213 -.252 .153 .253 -.035 .085 .094 1.000
aParameterestimate with its standard error given in parentheses.
bParameterfor constructs with single indicator fixed at .995.
CParameterfixed at zero.
dOrganizationconstruct factor correlations are below the diagonal and standard errors are above the diagonal.

FIGURE3
Model of Working Partnerships from the Distributor Firm Perspective

Influence over
Partner Firm
-.25

Relative
Dependence .33 .20
) Influence by
Partner Firm

Communication 79
Cooperation .65 \
CoopeA \ Functionality
.88 / .83 .91 of Conflict
A~.82
.82 .88
.18
5 16 .18 ~ 11
73
15/ 16 12

Outcomes Given
Comparison
Levels Trust

Note: Indicators and their pattern


-- coefficients r are -depicted
- onlv for constriarte
--..., .W. %,WI..IOLI. U with
*111 miltinp ;
Th rlr-- -
were each defined by a.-.-single indicator with
-. -- VV iL u i LIinr;inn trr
i i : inuicaiuiao ir ri e em ining constructs
the pattern coefficient set at .995 and the indicator specificity set at .01. All
parameter estimates are statistically significant (p < .05). Standardized estimates are given.

50 / Journalof Marketing,
January1990

This content downloaded from 95.149.73.109 on Sun, 22 Jun 2014 01:50:22 AM


All use subject to JSTOR Terms and Conditions
indicators. Though the chi square goodness of fit is trix for the manufacturer organizational measurement
still statistically significant (X(722) = 1022.65, p < model.
.001), this model is judged to provide acceptable fit.2
This judgment is based on a value of .82 for the normed Organizational measurement model. After several
fit index, a GFI value of .83, and a standardized RMR respecifications, a final measurement model for man-
ufacturer firms was obtained that is judged to provide
value of .046. Though four normalized residuals re-
main that have absolute values greater than two, the adequate goodness of fit, though its corresponding chi
largest one is only 2.40. Table 3 is a summary of the square value remains statistically significant (X22) =
final composition model for manufacturer firms. 187.29, p < .001). This judgment is based on mean-
ingful interpretability in terms of content and theo-
Again, all pattern coefficients relating the mea- retical considerations, an acceptable value of .88 for
sures to their underlying indicators are positive and
the normed fit index, a GFI value of .88, and a stan-
statistically significant, indicating perceptual agree- dardized RMR value of .049. Only a single normal-
ment between informants for each of the organization-
ized residual has an absolute value greater than two
level indicators. Note that no hypothesized indicators
of the constructs influence over the partner firm (IN- (-2.76). The parameter estimates for this final mea-
surement model are reported in Table 4.
FLOV) or functionality of conflict (FUNCON) are able
to be defined. The estimated correlation/covariance Organizational structural model. The initial struc-
matrix for the indicators then becomes the input ma- tural model, again specified from the posited rela-
tionships in Figure 1, gave relatively poor goodness
of fit (X41) = 334.52, p < .001). Several respecifi-
2As before, to provide a basis for comparison, eight-measure models, cations were made that were indicated from goodness
consisting of the four posited measures of each construct taken across of fit information and made substantive sense. As with
the two informants, were estimated. For manufacturer firms, seven
of the 10 "single-construct" models have chi square values that are the distributor firm perspective, COOPER was res-
not significant (p > .05) and three are nonconvergent, two because pecified as causally antecedent to TRUST, rather than
of a lack of significant perceptual agreement between the two infor-
mants on one measure and the other because of a lack of external being consequent to it. Consistent with this change in
consistency of the two informants on one measure. Note that Phillips causal sequence, paths were added from COMMUN
(1981) also obtained some nonconvergent solutions. and CL to COOPER. However, in this model, a direct

TABLE3
Summary of Final Composition Model for Manufacturer Firms
a.^ _ .m . ..
construct- Indicators Example
-- - - - - - u-- -- Measure
- - --- -1 -
Influence by T1, T2 Firm X exerts influence over the way our company markets our
partner firm product line through their firm (5-point scale: a great deal of/next to no)
Relative 73, 74 Computed as the difference between: (a) There are other manufacturers
dependence available to Firm X who sell product lines comparable to those of our
company and (b) There are other distributors in Firm X's trading area who
could provide comparable distribution for our company's products (7-point
scales: strongly disagree/strongly agree)
Communication Firm X lets our company know as soon as possible of any unexpected
problems they are experiencing with such things as poor cash flow or other
financial difficulties (7-point scale: strongly disagree/strongly agree)
Conflict Firm X and our company have significant arguments in our working
relationship (7-point scale: strongly disagree/strongly agree)
Trust Based upon your past and present experience, how would you characterize
the level of trust your company has in its working relationship with Firm X?
(7-point scale: don't trust Firm X/trust Firm X completely)
Cooperation T13, T14 Computed as the sum of: (a) Our company helps out Firm X in whatever
ways they ask and (b) Firm X helps our company out in whatever ways we
ask (7-point scales: strongly disagree/strongly agree)
Outcomes given T15, T16 Overall, how would you characterize the results of your company's working
comparison level relationship with Firm X? (7-point scale: it has fallen far short of
expectations/it has greatly exceeded our expectations)
Satisfaction T17 Our company's working relationship with Firm X has been an
unhappy one
(7-point scale: strongly disagree/strongly agree)
aPosited as underlying the given organizational indicator(s).

A Modelof DistributorFirmand Manufacturer


FirmWorkingPartnerships
/ 51

This content downloaded from 95.149.73.109 on Sun, 22 Jun 2014 01:50:22 AM


All use subject to JSTOR Terms and Conditions
TABLE4
Final Organizational Measurement Model for Manufacturer Firms
Constructa Indicator
Indicator INFLBY RELDEP COMMUN CONFL TRUST COOPER CL SATIS Specificity
L1 .995 (.0)b C 0 0 0 0 0 .01 (.10)
L4 0 .995 (.00) 0 0 0 0 0 0 .01 (.10)
L6 0 0 .63 (.07) 0 0 0 0 0 .61 (.07)
L7 0 0 .91 (.07) 0 0 0 0 0 .17 (.07)
L8 0 0 0 .995 (.00) 0 0 0 .01 (.10)
L9 0 0 0 0 .95 (.05) 0 0 0 .10 (.02)
,11 0 0 0 0 .88 (.05) 0 0 0 .22 (.03)
L12 0 0 0 0 .81 (.06) 0 0 0 .35 (.04)
113 0 0 0 0 0 .995 (.00) 0 0 .01 (.10)
L15 0 0 0 0 0 0 .995 (.00) 0 .01 (.10)
i17 0 0 0 0 0 0 0 .995 .01 (.10)

Organizational Construct Estimated Correlation (()d


INFLBY RELDEP COMMUN CONFL TRUST COOPER CL SATIS
INFLBY 1.000 .078 .074 .069 .071 .076 .077 .070
RELDEP .520 1.000 .072 .070 .071 .073 .073 .071
COMMUN .295 .499 1.000 .073 .052 .072 .072 .073
CONFL .019 -.183 -.414 1.000 .059 .076 .076 .082
TRUST .144 .196 .668 -.778 1.000 .066 .062 .064
COOPER .456 .349 .452 -.437 .548 1.000 .086 .073
CL .481 .336 .475 -.456 .670 .731 1.000 .076
SATIS .079 .214 .348 -.644 .628 .347 .459 1.000
aParameterestimate with its standard error given in parentheses.
bParameterfor constructs with single indicator fixed at .995.
CParameterfixed at zero.
dOrganizationconstruct factor correlations are below the diagonal and standard errors are above the diagonal.

path from COMMUN to TRUST was still indicated. working partnerships, two alternative multiple-sample
Paths were added from CL to INFLBY, TRUST to models were estimated. First, we estimated a model
CONFL, and INFLBY to COOPER; also, a correla- having no constraints across samples on the structural
tion parameter (421) was added between RELDEP and parameters. As this model gives the lowest chi square
COMMUN. The hypothesized direct path from IN- goodness-of-fit value that could be obtained with any
FLBY to SATIS was not significant and thereafter was model that has equality constraints, it provides the ba-
not estimated. sis for judging the acceptability of models in which
This respecified structural model provides an ad- equality constraints are specified. We then estimated
equate explanation of the construct covariances, though a model in which the five relationships that the dis-
its goodness-of-fit difference from the measurement tributor firm and manufacturer firm structural models
model is statistically significant (X2() = 27.81, p < have in common were constrained to be equal across
.01). As in the distributor firm results, there is no loss the two samples: RELDEP -- INFLBY, COMMUN
in explanation of practical consequence from con- --->COOPER, CL -- COOPER, COOPER -> TRUST,
straining the remaining 13 paths to zero, as indicated and the correlation between COMMUN and CL.
by a decrement in the normed fit index of only .017 The unconstrained model gives a chi square value
(from .885 to .868).3 The parameter estimates for this of 355.67 (d.f. = 67, p < .001). Note that this chi
final model are reported in Figure 4. square value and degrees of freedom are equal to the
respective sums for the structural models estimated
Multiple-Sample Analysis separately for the two samples. The model with equal-
ity constraints on the five common relationships pro-
To assess the commonalities and differences in man- vides a chi square value of 416.29 (d.f. = 72, p <
ufacturer firm and distributor firm perspectives on
.001). Of most interest here, though, is the rejection
of the hypothesis that these five relationships are in-
3Nonetheless, to provide another comparison model (as was done with variant across the two samples (X5) = 60.62, p <
distributor firms), the manufacturer firm structural model that repre-
sented the next-most-likely theoreticalalternativewas estimated. Adding .001). Relaxing the equality constraints on the COM-
a direct path from COOPER to SATIS did not result in even a sta- MUN -> COOPER and COOPER --- TRUST paths,
tistically significant improvement in goodness of fit (X2 = 1.59, in turn, results in a nonsignificant chi square differ-
p > .05). Hence, the more parsimonious respecified structuralmodel
was judged to give the best representation of the manufacturer firm ence value (Xd(3) = 2.44, p > .25). Hence, the REL-
perspective. DEP -> INFLBY and CL -> COOPER paths and the

52 / Journalof Marketing,January1990

This content downloaded from 95.149.73.109 on Sun, 22 Jun 2014 01:50:22 AM


All use subject to JSTOR Terms and Conditions
FIGURE4
Model of Working Partnerships from the Manufacturer Firm Perspective

Note: Indicators and their pattern coefficients are depicted only for constructs with multiple indicators. The remaining constructs
were each defined by a single indicator with the pattern coefficient set at .995 and the indicator specificity set at .01. All
parameter estimates are statistically significant (p < .05). Standardized estimates are given.

correlation between COMMUN and CL can be con- plications for marketing practice.
sidered invariant across the distributor firm and man-
ufacturer firm samples. The structural parameter es- Implications for Channels Research
timates for this constrained model are reported in Table Theory development. Substantively, the final respe-
5. Note that the COMMUN -> COOPER and COOPER cified structural models represent acceptable expla-
TRUST paths differ only in magnitude across the nations of the construct covariances for the distributor
two samples and may reflect the differences in oper- firm perspective and the manufacturer firm perspec-
ationalizations of the constructs across samples. tive. Outcomes given comparison level and relative
dependence, constructs adapted from social exchange
Discussion theory (Kelley and Thibaut 1978; Thibaut and Kelley
1959), along with the construct of communication ap-
We have attempted to make several contributions to pear to be critical exogenous constructs in the expla-
the knowledge and study of channel relationship phe- nation of manufacturer and distributor working part-
nomena. We have provided a more comprehensive, nerships. Cast within a nomological network that
testable model of manufacturer firm and distributor encompasses the understudied constructs of trust, co-
firm working partnerships. Substantively, though the operation, and satisfaction and the often-researched
research findings must be viewed as tentative because constructs of influence and conflict, this social ex-
of the number of respecifications that were necessary, change perspective appears to offer the basis for a
they provide a rich basis for further theory develop- comprehensive model of channel working partner-
ment in this area. We also have applied a research ships. The construct relationships present in these
method that enables more comprehensive assessments models contribute to channel knowledge in two ways:
of models composed of firm-level constructs, such as support is gained for the robustness and generality of
those that typify channels of distribution. We first dis- relationships predicted from past research and the un-
cuss the implications of our research findings for anticipated relationships provide a basis for further
channels research, then briefly speculate on some im- theory development and research in this area. We limit

A Modelof DistributorFirmand Manufacturer


FirmWorkingPartnerships
/ 53

This content downloaded from 95.149.73.109 on Sun, 22 Jun 2014 01:50:22 AM


All use subject to JSTOR Terms and Conditions
TABLE 5
Final Structural Model for Distributor and Manufacturer Firms
Distributor Firms Manufacturer Firms
Path Estimate S.E. Estimate S.E.
RELDEP- INFLOV (y11) -.25 .06 .00 .00
RELDEP - INFLBY (Y21)a .37 .04 .37 .04
CL - INFLBY (723) .00 .00 .36 .06
COMMUN -> TRUST (Y32) .00 .00 .50 .07
COMMUN - COOPER (Y42) .80 .07 .32 .06
CL - COOPER (Y43)a .17 .04 .17 .04
CL - SATIS (Y73) .00 .00 .20 .05
INFLOV - INFLBY (P21) .31 .06 .00 .00
COOPER - TRUST (134) .73 .06 .34 .06
INFLBY - COOPER (142) .00 .00 .29 .06
INFLBY - CONFL (352) .00 .00 .15 .05
TRUST CONFL (353) .00 .00 -.77 .05
INFLOV FUNCON (P61) .20 .05 .00 .00
COOPER - FUNCON (P64) .65 .06 .00 .00
TRUST - SATIS (173) .00 .00 .25 .08
CONFL -, SATIS (175) .00 .00 -.41 .08
RELDEP, COMMUN (b21) .00 .00 .46 .06
RELDEP, CL (b31) .00 .00 .33 .06
COMMUN, CL (432)a .24 .05 .24 .05
aParameterestimate constrained to be equal between distributorfirm and manufacturer firm samples.

our discussion to the supported predictions and re- actions, such as closer bonding with end-user firms.
specifications of greatest theoretical interest. Relative dependence may be a critical antecedent to
The relationship of relative dependence to influ- the enactment of these dependence-balancing opera-
ence by the partner firm found for both manufacturer tions.
firm and distributor firm perspectives and the rela- The respecification of greatest interest in terms of
tionship of relative dependence to influence over the theory development is the respecification of cooper-
partner firm found for the distributor firm perspective ation as being causally antecedent, rather than con-
support our hypotheses. A contribution to theory de- sequent, to trust. This respecification provided better
velopment is the use of relative rather than simple de- explanation of the construct covariances for both per-
pendence on the working relationship. Relative de- spectives. The reason may be the way in which man-
pendence, as we conceptualize it, seems to capture ufacturer and distributor informants think about co-
more fully the interdependent nature of channel work- operation and trust. When asked about their perceptions
ing partnerships. Further, drawing upon Emerson of their firm's trust in a working relationship, infor-
(1962), we believe relative dependence captures any mants give a present state report; that is, they an-
extant imbalance in the "relationship state" of depen- swered on how much their firm trusts the partner firm
dence. Consequently, "individual behavior" is under- at the current point in time. In contrast, informant re-
taken by each firm to restore balance, which is rep- ports on cooperation implicitly tap past cooperation
resented in our model as two constructs, influence over in that, when asked about their perceptions of their
the partner firm and influence by the partner firm. A firm's cooperation with the partner firm, informants
contribution from the model for distributor firms is considered recent past experiences when the two firms
that these two constructs appear to be distinct, though had worked together. That past cooperation is a nec-
related, constructs rather than opposite ends of a sin- essary antecedent of present trust is supported by the
gle continuum, supporting the notion of "individual research of Matthews and Shimoff (1979), which found
behavior" by each partner firm. that trust in two-person exchange relationships devel-
An intriguing possibility for further theory devel- oped slowly over a number of cooperative exchanges,
opment and channels research is the linkage between even when a contingency was present that made low
relative dependence and the dependence-balancing levels of trust costly. Support also is provided by the
operations proposed by Heide and John (1988). Based work of Frazier (1983a) and Dwyer and Lagace (1986).
on transaction cost analysis, the premise of their work, Iteratively, cooperation leads to trust which, in turn,
which was supported empirically, is that the more de- leads to a greater willingness to cooperate in the fu-
pendent firm in a working relationship needs to pro- ture, which then generates greater trust, and so on.
tect its transaction-specific assets by taking various Thus, in a static model of working partnerships, co-

54 / Journalof Marketing,January1990

This content downloaded from 95.149.73.109 on Sun, 22 Jun 2014 01:50:22 AM


All use subject to JSTOR Terms and Conditions
operation tentatively appears to be causally antecedent There is an implicit tradeoff between making models
to trust. To our knowledge, the relationship between of some set of phenomena, such as channel working
cooperation and trust has not previously been assessed partnerships, more comprehensive and the subsequent
empirically in marketing channels. Given the critical ability to obtain "acceptable" goodness of fit for such
nature of these constructs, further channels research "large" models. In discussing this tradeoff, Bentler
is needed in which these constructs again are embed- and Chou (1987) suggest the heuristic that in areas
ded in a nomological network. Naturally, longitudinal where knowledge about the variables is limited (as in
research or laboratory studies would better enable re- marketing channels), models be limited to, at most,
searchers to make inferences about the causal ordering 20 variables. Hence, with three or preferably four
of cooperation and trust. measured variables per construct, a model would con-
As a consequence of this change in causal se- tain about five or six constructs. Note, though, that
quence for cooperation and trust, several related re- the use of multiple informants further accentuates this
specifications were made. Direct paths from com- tradeoff.
munication and from outcomes given comparison level Future work in channel partnerships might ad-
to cooperation were substituted for those leading to vance in two complementary ways. First, greater
trust, though a direct path from communication to trust knowledge of channels variables can be gained through
for the manufacturer firm perspective was retained. more rigorous construct definition, pretesting, and scale
The classical work of Deutsch (1958, 1960, 1962) on development (cf. Angleitner, John, and Lohr 1986).
the conditions that promote trust and cooperation sup- Second, longitudinal research would make great con-
ports these direct paths. tributions to our understanding of these channels vari-
Finally, the five relationships that the distributor ables. Longitudinal research could be directed at sets
firm and manufacturer firm structural models have in of "core" constructs, making possible better infer-
common can be thought of as the "core" set of sub- ences about both their development over time and their
stantive findings. The multiple-sample analysis re- causal sequence.
vealed that distributor firms and manufacturer firms In this research approach, multiple informants
share similar perspectives on the positive effect of rel- within each firm each respond to several measures for
ative dependence on influence by the partner firm, the each construct. This affords a more precise delinea-
positive effect of outcomes given comparison level (CL) tion of firm-level constructs from potential sources of
on cooperation, and the positive correlation between measurement error contained in multiple-informant
communication and outcomes given CL. Also, though reports (cf. Anderson 1987). The crucial component
there are differences in magnitude across perspec- in this representation is informant bias; the extent of
tives, manufacturers and distributors appear to have its presence determines whether multiple informants
similar perspectives on the positive effect of com- are needed or a single informant is sufficient. Firm-
munication on cooperation and the positive effect of level representation of marketing channel constructs
cooperation on trust. Further study of these relation- necessitates the use of multiple informants unless one
ships, either in greater depth within smaller models or can justify the assumption that informant bias is of no
again embedded within a larger nomological network, practical consequence. In other words, the individual
is particularly recommended for future empirical re- informant's perceptions of the firm must be reasona-
search in marketing channels. bly veridical. To enable channel researchers to make
a better informed decision on the usage of multiple
Research approach. We used a structural equation versus single informants, further research is needed to
modeling approach in conjunction with a multiple-in- determine the research contexts and marketing chan-
formant method. Structural equation models com- nel phenomena for which informant bias is likely to
posed of a "large" number of constructs have some be substantial, thus necessitating multiple informants.
comparative strengths in terms of their ability to con- Several considerations in employing multiple in-
tribute to knowledge and understanding. Though two- formants certainly make the research process more
or three-construct studies, which characterize much of difficult. First, a multiple-informant method involves
past channels research, contribute to our knowledge the independent participation of two or more individ-
of marketing channels, many such studies would have uals at each firm, exacerbating the problem of ob-
to be done to investigate systematically all of the con- taining adequate sample sizes. In addition, as dem-
struct interrelationships specified by a more compre- onstrated in our research, the assumption that the
hensive model. Moreover, comprehensive testing of multiple informants are answering the questionnaires
such models enables one to make more meaningful independently can be untenable for some firms, es-
assessment of nomological validity (Campbell 1960), pecially when mail questionnaires are used. Finally,
an essential part of construct validity (Cronbach and the cost of conducting key informant research in-
Meehl 1955). creases at a rate greater than simply the number of

A Modelof DistributorFirmand Manufacturer


FirmWorkingPartnerships
/ 55

This content downloaded from 95.149.73.109 on Sun, 22 Jun 2014 01:50:22 AM


All use subject to JSTOR Terms and Conditions
informants per firm because of incomplete firm par- planning seems a worthwhile mechanism for bringing
ticipation. these essential constructs together in practice. Such a
planning process would entail mutually defining per-
Implications for Marketing Practice formance objectives for each partner, thereby estab-
lishing expectations with which outcomes can be
Research findings and the followup qualitative re- compared; specifying the cooperative efforts each firm
search suggest several ways in which both manufac- requires of its partner to attain its objectives; periodic
turer firms and distributor firms can actively manage communication on progress toward the performance
their working partnerships. Relative dependence is objectives, including discussion of any changes in
found to be a significant antecedent of influence in marketplace or production conditions that would sig-
the working partnership, which suggests that a less nificantly affect them; and finally a formal retrospec-
dependent firm can enact successfully the various in- tive review of each partner's performance. Meeting or
fluence strategies (e.g., requests, information ex- exceeding the performance objectives through coop-
change) proposed by Frazier and Summers (1984, eration leads to trust and satisfaction with the working
1986). A more dependent firm, in contrast, should seek partnership.
ways to add value (or reduce cost) to the exchange Even firms in successful partnershipswould readily
for the partner firm at a relatively small cost to itself acknowledge that disagreements are inevitable. Rather
(cf. Narus and Anderson 1987, 1988). At the same than allowing these conflicts to run their course ca-
time, as suggested by the work of Heide and John priciously, however, adroit partner firms develop
(1988), the more dependent firm can protect its trans- mediating mechanisms to defuse and settle their dif-
action-specific assets through various dependence- ferences rapidly. Three such mechanisms can be sug-
balancing actions. gested. First, firms can train their boundary-spanning
Another implication for marketing practice comes personnel (e.g., salespersons, purchasing managers)
from consideration of the two social exchange con- to be sensitive to inherent channel relationship trouble
structs, which can be expressed managerially as out- spots (e.g., warranty claims handling). These indi-
comes compared with expectations and outcomes viduals also should be given appropriate authority to
compared with alternatives. In our experience, firms expedite problem resolution before conflicts escalate.
typically have only a coarse understanding of how the Second, a distributorcouncil can be established. When
outcomes they are providing to partner firms compare managed successfully, a council serves as a forum for
with these two standards. Hence each firm needs to airing complaints, discussing emerging channel prob-
gain and periodically update its understanding of its lems, and developing mutually acceptable solutions.
partner firm's requirements or expectations and the al- Finally, an office of distributor ombudsman can be
ternative outcomes competitors are offering. This un- created (Jacobs 1985). An ombudsman might be a re-
derstanding then can be translated into responsive pro- tired distributor, employed by the manufacturer, who
grams and systems that enable the partnership to be would have credibility with both parties. Serving as a
mutually satisfying over time. "friend of the distributor," the ombudsman listens to
Trust is a core construct in our model, with an- distributor concerns, draws upon experience to assist
tecedents of outcomes given comparison level, com- the manufacturer firm in finding equitable solutions,
munication, and cooperation. Meaningful joint annual and then helps explain those solutions to distributors.

REFERENCES
Achrol, Ravi S., Torger Reve, and Louis Stern (1983), "The Channels," Journal of Marketing Research, 24 (February),
Environment of Marketing Channel Dyads: A Framework 254-62.
for ComparativeAnalysis," Journal of Marketing, 47 (Fall), Anderson, James C. (1987), "An Approach for Confirmatory
55-67. Measurement and Structural Equation Modeling of Orga-
Angleitner, Alois, Oliver P. John, and Franz-JosefLohr (1986), nizational Properties," Management Science, 33 (April),
"It's What You Ask and How You Ask It: An Itemmetric 525-41.
Analysis of Personality Questionnaires," in Personality As- and David W. Gerbing (1982), "Some Methods for
sessment via Questionnaires, A. Angleitner and J. S.
Respecifying Measurement Models to Obtain Unidimen-
Wiggins, eds. Berlin: Springer-Verlag, 61-108. sional Construct Measurement," Journal of Marketing Re-
Anderson, Erin, Leonard M. Lodish, and Barton A. Weitz search, 19 (November), 453-60.
(1987), "Resource Allocation Behavior in Conventional and (1984), "The Effect of Sampling Er-

56 / Journalof Marketing,January1990

This content downloaded from 95.149.73.109 on Sun, 22 Jun 2014 01:50:22 AM


All use subject to JSTOR Terms and Conditions
ror on Convergence, ImproperSolutions, and Goodness-of- Etgar, Michael (1976), "Channel Domination and Counter-
Fit Indices for Maximum Likelihood Confirmatory Factor vailing Power in Distribution Channels," Journal of Mar-
Analysis," Psychometrika, 49 (June), 155-73. keting Research, 13 (August), 254-62.
and (1988), "Structural Equation Mod- Frazier, Gary L. (1983a), "InterorganizationalExchange Be-
eling in Practice: A Review and Recommended Two-Step havior in Marketing Channels: A Broadened Perspective,"
Approach," Psychological Bulletin, 103 (May), 411-23. Journal of Marketing, 47 (Fall), 68-78.
and James A. Narus (1984), "A Model of the Dis- (1983b), "On the Measurement of Interfirm Power
tributor's Perspective of Distributor-ManufacturerWorking in Channels of Distribution," Journal of Marketing Re-
Relationships," Journal of Marketing, 48 (Fall), 62-74. search, 20 (May), 158-66.
and (1986), "Toward a Better Under- and Jagdish N. Sheth (1985), "An Attitude-Behav-
standing of Distribution Channel Working Relationships," ior Framework for Distribution Channel Management,"
in Industrial Marketing: A German-American Perspective, Journal of Marketing, 49 (Summer), 38-48.
K. Backhaus and D. Wilson, eds. Berlin: Springer-Verlag, , Robert E. Spekman, and Charles R. O'Neal (1988),
Inc., 320-36. "Just-in-Time Exchange Relationships in Industrial Mar-
Arthur Andersen & Co. (1987), Facing the Forces of Change: kets," Journal of Marketing, 52 (October), 52-67.
Beyond Future Trends in Wholesale Distribution. Wash- and John O. Summers (1984), "InterfirmInfluence
ington, DC: Distribution Research and Education Founda- Strategies and Their Application Within Distribution Chan-
tion. nels," Journal of Marketing, 48 (Summer), 43-55.
Bentler, P. M. and D. G. Bonett (1980), "Significance Tests and (1986), "Perceptions of Interfirm
and Goodness-of-Fit in the Analysis of Covariance Struc- Power and Its Use Within a Franchise Channel of Distri-
tures," Psychological Bulletin, 33 (May), 588-606. bution," Journal of Marketing Research, 23 (May), 169-
and Chih-Ping Chou (1987), "Practical Issues in 76.
Structural Modeling," Sociological Methods & Research, Hardy, Kenneth G. and Allan J. Magrath (1988), Marketing
16 (August), 78-117. Channel Management: Strategic Planning and Tactics.
Bialaszewski, Dennis and Michael Giallourakis (1985), "Per- Glenview, IL: Scott, Foresman and Company.
ceived Communication Skills and Resultant Trust Percep- Heide, Jan B. and George John (1988), "The Role of Depen-
tions Within the Channel of Distribution," Journal of the dence Balancing in Safeguarding Transaction-Specific As-
Academy of Marketing Science, 13 (Spring), 206-17. sets in Conventional Channels," Journal of Marketing, 52
Cadotte, Ernest R. and Louis W. Stern (1979), "A Process (January), 20-35.
Model of InterorganizationalRelations in Marketing Chan- Gaski, John F. (1984), "The Theory of Power and Conflict in
nels," in Research in Marketing, 2, Jagdish N. Sheth, ed. Channels of Distribution," Journal of Marketing, 48 (Sum-
Greenwich, CT: JAI Press, Inc., 127-58. mer), 9-29.
Calder, Bobby J. (1977), "Focus Groups and the Nature of Gerbing, David W. and James C. Anderson (1988), "An Up-
Qualitative Marketing Research," Journal of Marketing Re- dated Paradigm for Scale Development Incorporating Uni-
search, 14 (August), 353-64. dimensionality and Its Assessment," Journal of Marketing
Campbell, Donald T. (1960), "Recommendationsfor APA Test Research, 25 (May), 186-92.
Standards Regarding Construct, Trait, or Discriminant Va- Gladstein, Debra L. (1984), "Groups in Context: A Model of
lidity," American Psychologist, 15, 546-53. Task Group Effectiveness," Administrative Science Quar-
Cronbach, Lee J. and Paul E. Meehl (1955), "Construct Va- terly, 29 (December), 499-517.
lidity in Psychological Tests," Psychological Bulletin, 52 Jacobs, Bruce A. (1985), "Tell It to Bob," Industry Week, 226
(July), 281-302. (September 16), 62-3.
Deutsch, Morton (1958), "Trustand Suspicion," Conflict Res- James, Lawrence R. (1982), "Aggregation Bias in Estimates
olution, 11 (December), 265-79. of Perceptual Agreement," Journal of Applied Psychology,
(1960), "The Effect of MotivationalOrientationUpon 67 (May), 219-29.
Trust and Suspicion," Human Relations, 13 (May), 123- John, George and Torger Reve (1982), "The Reliability and
39. Validity of Key InformantData From Dyadic Relationships
(1962), "Cooperation and Trust: Some Theoretical in Marketing Channels," Journal of Marketing Research,
Notes," in Nebraska Symposium on Motivation, 10, M. R. 19 (January), 517-24.
Jones, ed. Omaha: University of Nebraska Press, 275-319. Joreskog, K. G. (1971), "Simultaneous Factor Analysis in
Dwyer, F. Robert (1980), "Channel-Member Satisfaction: Several Populations," Psychometrika, 36 (December), 409-
Laboratory Insights," Journal of Retailing, 56 (Summer), 26.
45-65. and Dag Sorbom (1984), LISREL VI: Analysis of
and Rosemary R. Lagace (1986), "On the Nature Linear Structural Relationships by the Method of Maximum
and Role of Buyer-Seller Trust," in AMA Summer Edu- Likelihood. Chicago: National Educational Resources, Inc.
cators' Conference Proceedings, Series 52, Terence Shimp Kelley, H. H. and J. W. Thibaut (1978), Interpersonal Re-
et al., eds. Chicago: American Marketing Association, lations: A Theoryof Interdependence.New York: John Wiley
40-5. & Sons, Inc.
, Paul H. Schurr, and Sejo Oh (1987), "Developing Lord, Frederic M. and Melvin R. Novick (1968), Statistical
Buyer-Seller Relationships," Journal of Marketing, 51 Theories of Mental Test Scores. Reading, MA: Addison-
(April), 11-27. Wesley Publishing Company.
Eliashberg, Jehoshua and Donald A. Michie (1984), "Multiple Mallen, Bruce (1963), "A Theory of Retailer-Supplier Con-
Business Goals Sets as Determinants of Marketing Channel flict, Control and Cooperation," Journal of Retailing, 39
Conflict: An Empirical Study," Journal of Marketing Re- (Summer), 24-32, 51-2.
search, 21 (February), 75-88. Matthews, Byron A. and Eliot Shimoff (1979), "Expansion of
Emerson, Richard M. (1962), "Power-Dependence Rela- Exchange: Monitoring Trust Levels in Ongoing Exchange
tions," American Sociological Review, 27 (February), 31- Relations," Journal of Conflict Resolution, 23 (Septem-
41. ber), 538-60.

A Modelof DistributorFirmand Manufacturer


FirmWorkingPartnerships
/ 57

This content downloaded from 95.149.73.109 on Sun, 22 Jun 2014 01:50:22 AM


All use subject to JSTOR Terms and Conditions
Narus, James A. and James C. Anderson (1987), "Distribution Rosenbloom, Bert (1973), "Conflict and Channel Efficiency:
Contributions to Partnerships With Manufacturers," Busi- Some Conceptual Models for the Decision Maker," Journal
ness Horizons, 30 (September-October), 34-42. of Marketing, 37 (July), 26-30.
and (1988), "Strengthen Distributor Per- Sibley, Stanley D. and Donald A. Michie (1982), "An Ex-
formance Through Channel Positioning," Sloan Manage- ploratory Investigation of Cooperation in a Franchise Chan-
ment Review, 31 (Winter), 31-40. nel," Journal of Retailing, 58 (Winter), 23-45.
Phillips, Lynn W. (1981), "Assessing Measurement Error in Stem, Louis W. and Torger Reve (1980), "DistributionChan-
Key Informant Reports: A Methodological Note on Orga- nels as Political Economies: A Frameworkfor Comparative
nizational Analysis in Marketing," Journal of Marketing Analysis," Journal of Marketing, 44 (Summer), 52-64.
Research, 18 (November), 395-415. Thibaut, John W. and Harold Kelley (1959), The Social Psy-
Reve, Torger and Louis W. Ster (1979), "Interorganizational chology of Groups. New York: John Wiley & Sons, Inc.
Relations in Marketing Channels," Academy of Manage- Tversky, Amos and Daniel Kahneman (1981), "The Framing
ment Review, 4 (3), 405-16. of Decisions and the Psychology of Choice," Science, 211
Robicheaux, Robert A. and Adel I. El-Ansary (1976), "A (January), 453-8.
General Model for Understanding Channel Member Be- Wilkinson, Ian F. (1979), "Power and Satisfaction in Chan-
havior," Journal of Retailing, 52 (Winter), 13-30, 93-4. nels of Distribution," Journal of Retailing, 55 (Summer),
Rosenberg, Larry J. and Louis W. Stem (1970), "Toward the 79-94.
Analysis of Conflict in Distribution Channels: A Descrip-
tive Model," Journal of Marketing, 34 (October), 40-6. Reprint No. JM541102

58 / Journalof Marketing,January1990

This content downloaded from 95.149.73.109 on Sun, 22 Jun 2014 01:50:22 AM


All use subject to JSTOR Terms and Conditions

You might also like