1. A Transaction Cost Analysis Model of Channel Integration in International Markets
Author(s): Saul Klein, Gary L. Frazier, Victor J. Roth
Source: Journal of Marketing Research, Vol. 27, No. 2 (May, 1990), pp. 196-208
Published by: American Marketing Association
Stable URL: http://www.jstor.org/stable/3172846
Accessed: 03/03/2010 15:14
Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at
http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless
you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you
may use content in the JSTOR archive only for your personal, non-commercial use.
Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at
http://www.jstor.org/action/showPublisher?publisherCode=ama.
Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.
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 Research.
http://www.jstor.org
2. GARY FRAZIER, VICTOR ROTH*
SAULKLEIN, L. and J.
The authorsdevelop a transaction cost analysismodel designed to explain the
Lhornnel integrationchoices of firmsin international markets.In a test with data
collected from a group of Canadianexport firms, the model receivessignificant
support,suggestingthat an importunt contingency whendecidingon Lhunnel struc-
ture in a foreigncountryis the abilityof the marketto limitthe opportunistic ten-
dencies of outside intermediaries. When the enforcement contractual
of arran,ge-
mentscannotbe relieduponin the market,different degrees of forwardintegration
are feasible alternatives.Other empiricalresultssuggest that the firmmay prefer
use of intermediaries a foreignmarketwithhigh environmental
in in
diversity order
to cope with its inherentcomplexity and maintain flexibility.Channelvolume,the
use of sharedcliuninels, country
and destination
also are shownto affect the nature
of integration Jiunnels in international
in markets.
A Transaction Cost Analysis Model of Channel
Integration in International Markets
I
Doing business in international markets is a difficult amples of such intermediate options are the use of
challenge for any firm (cf. Cavusgil 1980; Root 1987). commission agents to serve foreign markets, joint ven-
Especially difficult is deciding what levels of integration tures, and cooperative marketing arrangements.
the firm should use within its channels of distribution in Stern and El-Ansary (1988) stress that the channel in-
various foreign markets. At one extreme, the firm can tegration decision is a critical component of any firm's
perform all marketing and distribution functions itself. marketingchannel strategy.The decision is of even greater
At the other extreme, the firm can choose not to perform significance in an internationalcontext. Choosing the right
any of the necessary functions, instead using outside level of channel integration can make the difference be-
merchants who take title to the firm's goods for resale tween success and failure in a foreign market, as it rep-
to other middlemen and final buyers. Between these ex- resents a point of vulnerability for the firm in terms of
tremes, a continuum of market hierarchy options is usu- both marketresponse and opportunitylosses (cf. Keegan
ally available (cf. Anderson and Gatignon 1986).1 Ex- 1984; Root 1987). Moreover, the correct decision must
be made early, because initial commitments may not be
easy to terminate. The question that must be answered
'In relation to this diversity of choices, the available data (U.S. is what degree of forwardintegrationfirms should choose
Bureau of the Census) show that industries vary considerably in their in foreign markets and why.
degree of forward integration. Further, as John and Weitz (1988) point Most empirical research in the channels literature has
out, such diversity is likely to extend to firms within industries as
well. centered on the management of ongoing dyadic channel
relationships ratherthan on the structureof the channel.
*Saul Klein is Assistant Professor of Marketing, College of Busi- Fortunately,interestin channel integrationissues has been
ness Administration, Northeastern University. Gary L. Frazier is As- rising among channels researchersin recent years, in part
sociate Professor of Marketing, School of Business Administration, because of the development of transaction cost analysis
University of Southern California. Victor J. Roth is Associate Pro-
fessor of Marketing, Department of Consumer Studies, University of by Williamson (1975, 1985; see Anderson and Weitz
Guelph. 1986). Important empirical studies on channel integra-
The authors are indebted to George S. Day for his support and en- tion have been performed by Lilien (1979), Anderson
couragement in carrying out the research. They also thank the Editor (1985), Coughlan (1985), Dwyer and Welsh (1985), An-
and three anonymous JMR reviewers for their comments on previous
versions of the article.
derson and Coughlan (1987), and John and Weitz (1988).
Coughlan (1985) and Anderson and Coughlan (1987) fo-
196
Journal of Marketing Research
Vol. XXVII (May 1990), 196-208
3. MODEL
COSTANALYSIS
TRANSACTION 197
cused on the natureof channel integrationin foreign associatedfixed costs can be spreadover a largevolume
markets, Anderson
and (1985) andJohnandWeitz(1988) of business.Furthermore, the volumeof businessin-
as
developed and tested models based on transaction cost creases, firms are able to specializein the performance
analysis. These studiesrepresentan excellent start, but of marketing-distribution functionsandreapthe benefits
a varietyof important research issues remainunresolved. of economiesof scale. Thoughproduction costs appear
For example, Reid (1983) has advocated application
the important,especially in terms of "efficiency"as op-
of transaction cost analysis to questionsof firm inter- posedto "control," they aloneare insufficient explain
to
nationalization. variationsin channel integration.For example, a pro-
The purposeof our articleis to expandunderstanding ductioncost explanation cannot accountfor the use of
of the reasonsunderlying firms' channelintegration de- marketexchangesby largefirmsor differentdegreesof
cisionsin international markets. the firsttime, a model
For integration firmsof smallersize and less experience.
by
based on transaction cost analysisis developedand ap- Transaction analysis(Williamson1975, 1985)of-
cost
plied to explainlevels of forward integration withindis- fers another perspective help us understand
to betterthe
tribution channelsin foreignmarkets.2 Production costs forces shapingchannelstructure. The basic premiseof
also are treatedin the model. The datanecessaryto test transaction cost analysis(TCA) is that the firm will in-
the model were gatheredfrom a groupof Canadian ex- ternalizeactivitiesthatit is able to perform lower cost
at
port firms. and will rely on the marketfor activitiesin which other
Our study is unique in two other respects. First, we providers have an advantage. TCAis builton a microan-
examinefour differentintegration choices (i.e., market alyticframework with strongbehavioral reality.Channel
exchange, intermediate exchange, and two forms of hi- membersare assumedto be subjectto boundedration-
erarchicalexchange). With the exception of John and ality. Furthermore, least some actorsare assumedto
at
Weitz' (1988) work, the otherempiricalstudieson chan- be opportunistic (i.e., having a tendencyto cheat other
nel integration focusedon only two choices,whether
have parties)if given the chance. Imperfect,or asymmetric,
the channelis director indirect.Second, we divide "ex- information may give such actorsan exploitableadvan-
into
ternaluncertainty" two dimensions(environmental tage in theirdealingswith otherparties.
volatility anddiversity),each being arguedto have a dif- Transaction costs (i.e., the costs of governingthe sys-
ferentialimpacton the natureof channelintegration in tem)tendto be low in highlycompetitive markets, thereby
foreignmarkets.Previousresearchbased on transaction providinglittle or no incentiveto substitute internalor-
cost analysis has treatedexternaluncertainty unidi-as ganization market
for exchange.In contrast,when faced
mensional. Prestudyinterviewswith approximately 10 with an inabilityof marketsto impose behavioralcon-
Canadian exportersguided the developmentof the the- straints enforcesimplecontracts,
and firmsare expected
oreticalmodel and the operational measures. to internalize transactions reducecosts of exchange.
to
Afterbrieflyexaminingthe basic theoretical rationale A limit on integration the fact that organizations
is are
for the impactof production costs and transaction costs not perfectand transaction costs also are presentwithin
on channelintegration,we review in greaterdetail the them.
empiricalresearchon verticalintegration channelsof
in ThoughTCA tends to downplaythe impact of pro-
distribution.We then develop the researchhypotheses ductioncosts on forwardintegration, objectiveis to
the
and describethe methodsused to collect the data, de- minimizethe sum of transaction production
and costs in
velop the measures,andtest the hypotheses.Finally,im- making forward integration decisions(see JohnandWeitz
plicationsand limitationsof the study are discussed. 1988 andWilliamson1985, p. 92-94, 129). The higher
the costs of contracting externally,the greateris the in-
PRODUCTION COSTS AND TRANSACTIONCOSTS centive to internalize transactions.
To what extent should a firm performall marketing- Unlikeproduction costs, transaction costs areverydif-
distributionfunctionsinternallyinsteadof relyingheavily ficult to measurebecause they representthe potential
on outside intermediaries? traditional
The answerin the consequencesof alternative decisions. Researchers ex-
marketing literature been basedon a production
has cost amining transaction cost issues almostnever attemptto
argument (cf. Rosenbloom 1987; Stern and El-Ansary measuresuch costs directly,but rathertest whetheror-
1988).The assumption beenthatall firmsdesiremore
has ganizational relationsalign with the attributes trans-
of
control, which leads to a preference integration,
for but actionsas predicted transaction reasoning(Wil-
by cost
that such arrangements not be feasible unless the
will liamson1985). "Thelevel of specializedassets required
to support exchange,the uncertainty
the surrounding the
exchange,and the frequencyof exchangeare identified
as the principalfactorsthat make market-mediated ex-
changeinefficient" (JohnandWeitz 1988, p. 121-122).
2Anderson Coughlan(1987) examinea numberof constructs
and Asset specificity is the extent to which specializedin-
suggestedby severaldifferentperspectivesand do not study uncer- vestmentsare neededto supportan exchange, whereas
tainty.Gatignonand Anderson(1988) examinethe degreeof control
multinational
corporationshave over foreignsubsidiaries do not
and uncertainty reflectsthe abilityto predictrelevantcontin-
examinevariations channelintegration
in choicesper se. gencies, both internaland externalto the firm. "Fre-
4. 198 JOURNAL MARKETING
OF MAY
RESEARCH, 1990
quency" refers to the distinction between one-time and firm's products and perform most necessary functions
recurrentexchange.3 within the foreign country. Intermediateexchange is re-
flected by the use of independent organizations that per-
LITERATURE REVIEW form only the selling function (e.g., commission agents),
The predictions of transaction cost analysis in terms other necessary functions being performed largely by the
of asset specificity and internal uncertainty (i.e., per- focal organization. Given the international focus of our
formance evaluation) have been supported in studies by study, two hierarchical modes are relevant-the estab-
Anderson (1985) and John and Weitz (1988), and An- lishment of a wholly owned foreign sales subsidiary or
derson and Coughlan (1987) find support for the rela- serving the foreign market directly from the home coun-
tionship between asset specificity and channel integra- try of the firm. Though the firm performs a relatively
tion. Asset specificity and internal uncertaintyhave been large number of functions itself in either case, the mode
shown to be related positively to the level of integration of operation is different across the two. When the sub-
in the channel in these studies. sidiary option is used, the firm establishes a physical
Results relating to the external uncertainty construct presence in the foreign country, with salespeople being
have been mixed. Only John and Weitz (1988) have found housed and significant inventory being carried therein.
a significant, positive relationship between external un- Under the other hierarchical option, salespeople travel
certainty and the level of channel integration. Unpre- and products are shipped from the firm's home base di-
dictability (Anderson 1985) and heterogeneity (Dwyer rectly to foreign customers.
and Welsh 1985) have been found to be unrelated to
channel integration, though Anderson (1985) did find a Channel Volume
significant interaction between asset specificity and ex- Production costs are the costs of actually performing
ternal uncertainty. marketing-distributionfunctions. If channel volume for
Results have also been mixed on constructs relating to a product line is relatively low in a foreign country, the
production costs. Several empirical findings of Lilien firm is likely to prefer a market exchange whereby in-
(1979) support an economies of scale interpretation. termediaries can handle the product line more econom-
However, Anderson (1985) found no significant produc- ically throughassembling productlines from many firms.
tion cost effects, and John and Weitz (1988) found only In contrast, when channel volume is high, the production
one of two contructs reflecting production costs to be cost advantage of the market mode can be reduced vir-
significant. Transaction cost constructs dominated pro- tually to zero. As Williamson (1985, p. 94) states, "The
duction cost constructs in both of the latter studies. firm is simply better able to realize economies of scale
A variety of other constructs have been found to be as its own requirementsbecome larger in relation to the
related to channel integration in previous empirical re- size of the market." Economies of scale are facilitated
search. For example, in a study of channels in interna- by the potential for internal specialization and division
tional markets, Anderson and Coighlan (1987) found that of labor, which lead to reduced production costs. More
shared channels and country destination were important. specifically, Anderson (1985) argues that high product
volume facilitates economies of scale in finding, hold-
RESEARCHHYPOTHESES
ing, and utilizing management skills, enabling the firm
Because theory suggests that firms act in such a way to get more benefit from its expenditures on a field sales-
as to minimize the sum of transaction and production force and other marketing instruments.
costs, both types of cost must be incorporated in any H1: The greater channelvolumefor a product in
the line
conceptualization. The model developed and tested here? a foreignmarket,the greater the degreeof channel
is
is based on choices between alternative channel struc- integration.
tures for a specific product line in a specific foreign mar-
ket as determined by constructs related to both types of In terms of the two hierarchical options, the estab-
costs. lishment of foreign subsidiaries is expected to be facil-
Within this study, the differences between market ex- itated by relatively high levels of channel volume for the
change, intermediate exchange, and hierarchical ex- product line. Because increasing forward integration re-
change are based on the economic traditionof "functions quires a more complex and specific governance struc-
performed" (Williamson 1975), rather than on the or- ture, greater fixed costs are inevitable. To cover such
ganizational theory tradition of the closeness of the re- fixed costs, greater volumes are required. If channel vol-
lationship or level of coordination between firms in an ume for the product line is not extremely high but econ-
exchange (cf. Ouchi 1980). Market exchange is reflected omies of scale are still facilitated through use of an in-
by the use of merchant distributors who take title to the tegratedchannel, the firm is expected to serve the foreign
market directly from its home base.
H2: Withinthe hierarchicaloption,the use of foreignsales
3The frequency variable is not examined because we are interested subsidiaries associatedpositively with increasing
is
only in recurrent exchange. levels of channelvolume for the productline.
5. TRANSACTION
COSTANALYSIS
MODEL 199
Asset Specificity communication expenses, facilitating an adaptive, se-
Competitive foreign markets for intermediary services quential decision process, which is argued to have op-
limit the ability of independent channel members to be- timal propertiesundersuch conditions. Furthermore,high
have opportunistically because intermediaries are re- integration economizes on transactions by harmonizing
interests and permitting a wider variety of sensitive in-
placeable. When markets fail, however, behavior may
no longer be controllable at a low cost. A necessary con- centive and control processes to be activated.4
dition for market failure occurs when an exchange re- This view is in contrast to the theoretical position held
quires one party to invest in assets, whether physical or by certain organization theorists, who have argued that
looser structures(i.e., less vertically integrated)are more
intangible, that have no alternative usage outside that ex- effective under conditions of high external uncertainty
change. Such investmenthas the effect of reducinga large-
numbers bargaining situation (i.e., one in which many (cf. Lawrence and Lorsch 1967; Pfeffer and Salancik
intermediaries are available) to a small-numbers situa- 1978). A flexible organization is seen to be better able
tion. As John and Weitz (1988, p. 124) state, "Because to adapt to changing circumstances. Highly integrated
organizations are seen to be somewhat insulated from the
nonredeployable specific assets make it costly to switch environment and hence slow to react. The firm choosing
to a new relationship, the market safeguard against op-
an integrated governance structure in an uncertain en-
portunism is no longer effective." Under such conditions vironment may find it difficult to manage and then rel-
in foreign markets, firms are likely to use relatively in-
tegrated channels in which opportunism can be combat- atively difficult to dissolve.
ted through the exercise of legitimate authority, the mon- What each perspective ignores is the possibility that
external uncertainty has multiple dimensions, each with
itoring of behavior, and the offering of more varied a differential impact on organization structureand chan-
incentives than can be used with independent channel
members. The positive effect of asset specificity on nel choice. External uncertainty appears to be too broad
channel integration has received empirical support in a concept to be treated unidimensionally; different facets
studies by Anderson (1985), Anderson and Coughlan of external uncertainty may lead to either a motivation
to reduce transaction costs (the economic tradition) or a
(1987), and John and Weitz (1988).
Small-numbers bargaining may well be the more com- desire for flexibility (the organization theory tradition).
mon situation in channels decisions, particularly in an Support for opposing uncertainty effects on integration
international context. The availability of alternative in- matters is provided by Walker and Weber (1984) and
termediaries who are able and willing to handle a man- Balakrishnanand Wernerfelt (1986). Walker and Weber
ufacturer's goods may be severely restricted, especially found that volume uncertaintyhad a significant, positive
at the outset (Keegan 1984). effect on making a component part in-house, whereas
technological uncertainty was related positively (though
H3: The greaterthe transactionspecificityof assets, the not significantly) to buying the component part from the
greateris the degreeof channelintegration a for-
in market. Balakrishnan and Wernerfelt show that though
eign market.
uncertainty in general makes integration more effective,
Asset specificity is not expected to affect the firm's one particularuncertainty-the possibility of technolog-
choice of which hierarchical option to use. ical obsolescence-works the other way. Splitting the
external uncertaintyconstruct into its components allows
External Uncertainty such opposing effects to be investigated.
To date, the external uncertainty construct in trans- Two dimensions of external uncertainty are of interest
action cost analysis has been treated as unidimensional in our study-the volatility and the diversity of the en-
and viewed as another feature of market failure. High vironment in the foreign market. These two dimensions
external uncertainty, given bounded rationality, pre- correspond to the environmental volatility and diversity
cludes the writing and enforcement of contingent claims dimensions examined by Leblebici and Salancik (1981),
contracts that specify every eventuality and consequent who found them to have differential effects on decision
response (Anderson and Weitz 1986). It allows negative
information asymmetries to develop and provides the po-
tential for outside intermediaries to behave opportunist-
ically. As Williamson (1975, p. 23) states, "When, 4Inthe full theoretical statementpresented Williamson(1975),
by
bothtransaction-specific assets anduncertainty deemednecessary
are
however, transactions are conducted under conditions of for market failure.Eithereffect withoutthe othershouldnot resultin
uncertainty/complexity, in which event it is very costly, market failure(Anderson 1985). However,this rationale
maynot hold
perhapsimpossible to describe the complete decision tree, in foreignmarkets.External in
uncertainty any foreignmarket should
the bounded rationality constraint is binding and an as- be suchthatan increase assetspecificitywouldincreasetransaction
in
sessment of alternative organizational modes, in effi- costs. In addition,if externaluncertainty extremelyhigh, it should
is
influencetransaction costs independently the level of asset speci-
of
ciency respects, becomes necessary." Internalization is ficity. This reasoningaside, the interactions asset specificityand
of
seen to allow the absorption of external uncertainty environmental volatilityand diversityare examinedin the dataanal-
through specialization of decision making and savings in yses.
6. 200 OF
JOURNAL MARKETING MAY
RESEARCH, 1990
processes. They are also similar to the dimensions of a in the
rounding transaction a foreignmarket, greater
complexity and dynamism developed by Duncan (1972) is the degreeof channelintegration.
and supported by Dess and Beard (1984). H5: The greaterthe diversity of the environmentsur-
a in the
rounding transaction a foreignmarket, lesser
Volatility refers to the extent to which the environ- is the degreeof channelintegration.
ment changes rapidly and allows a firm to be caught by
surprise (cf. Leblebici and Salancik 1981). High vola- In terms of the hierarchical option, a highly volatile
tility in a foreign market is expected to lead to an in- environment is likely to make a foreign market difficult
ability to predict future outcomes, which creates prob- to serve from the firm's home base. The establishment
lems in writing contracts because these agreements will of a foreign subsidiary places the firm closer to the mar-
be incomplete in some important respects. When un- ketplace where it is in a better position to react and adapt
foreseen contingencies arise, marketcontractsare strained to unforeseen circumstances. The diversity of the envi-
in adapting to the changed circumstances because op- ronment is not expected to influence the choice between
portunistically inclined parties can try to interpret un- these two forms of hierarchy.
specified clauses to their own advantage. As John and
Weitz (1988) and Stinchcombe (1985) indicate, institu- H6: Within hierarchical
the option,the use of foreignsales
tional structuresthat permit sequential, adaptive decision is
subsidiaries associated positivelywith the level of
making are needed when such uncertainty increases. environmental volatility.
Therefore, high external uncertaintybased on high levels METHODS
of volatility should lead to relatively high transactioncosts
in marketexchanges in foreign markets, encouraginghigh Data Collection
levels of channel integrationin orderto reduce such costs. The unit of analysis for studying levels of channel in-
Diversity reflects the extent to which there are mul- tegration must be the transaction, as suggested by Wil-
tiple sources of uncertaintyin the environment(i.e., highly liamson (1975, 1985). For purposes of our study, the
heterogeneous)(cf. Aldrich 1979). A foreign marketwith transaction is the exporting of a particular product to a
high diversity would contain many customers, many fi- particularforeign market by a particularfirm.
nal users, and many competitors for the firm's product, The data necessary to test the researchhypotheses were
with high dissimilarityamong them. A firm facing a highly
diverse environment in a foreign market would have dif- gathered from a group of Canadian export firms, iden-
tified throughuse of an industrydirectory.All firms listed,
ficulty in obtaining and processing information about en- with the exception of those dealing solely in unprocessed
vironmentalentities. Moreover, because a firm must adopt
primaryproducts, were surveyed (925 firms) through use
multiple strategies to address the multiplicity of demands of a mail questionnaire.6Key informants were identified
and constraints in a diverse or heterogeneous foreign en-
vironment (cf. Keegan 1984), formulating effective stra- by name from the directory and were usually either the
owner or the general manager of the firm. A total of 510
tegic programs and responses is also very difficult. In firms responded, a 55% response rate. Data were avail-
combination, the greater amount of information needed able on each of the study constructs in 375 cases, as 33
and the greater difficulties associated with developing
questionnaires were unusable because the respondents
multiple strategies suggest that environmental diversity did not follow instructions at the beginning and 102 in-
in a foreign market will encourage the development of formants failed to provide data necessary to measure all
complex and fluid channel structures that enhance the independent variables; missing data were especially
channel's ability to cope with specialized demands (cf. common on the items relating to channel volume. All
Achrol, Reve, and Stern 1983; Dwyer and Welsh 1985). questionnaires with data on necessary items, however,
Independent channel members within the foreign market were used for measure assessment purposes. A general
will be better able to cope with such heterogeneity (Kee-
description of the sample is presented in Table 1.7
gan 1984). Therefore, external uncertaintycaused by the
diversity of the environment is expected to lead to a de-
sire for flexibility in channel structure (i.e., less inte-
6Prestudy interviews indicated that many unprocessed primary
gration). High diversity does not, in itself, suggest that products sold by Canadian export firms are commodities. As such,
high volatility also is present in the foreign market.5 these products are not likely to provide much variation on asset spec-
The preceding discussion suggests that the external ificity. Moreover, it appeared difficult to devise an instrument that
would be appropriatefor both manufactured and unprocessed primary
uncertaintyeffect can be broken down into the following
two hypotheses: products. Unprocessed products therefore were excluded.
7In examining possible nonresponse bias, we compared late re-
H4: The greaterthe volatility of the environmentsur- spondents with early respondents on several characteristics. Late re-
spondents tended to be smaller firms (p < .10), but were similar to
early respondents on all other characteristics. Moreover, the propor-
tion of respondent firms identifying the United States as the major
export market (i.e., 63%) is similar to the proportion of Canadian
SThe presence of environmental volatility may create a need for exports that go to the United States (approximately 70%). These find-
flexibility as well. However, transaction cost pressures are likely to ings and the study's high response rate suggest that nonresponse bias
dominate any such effect. is not a serious problem.
7. TRANSACTION MODEL
COSTANALYSIS 201
Table 1 but are compared subsequently with firms using the mar-
SAMPLE
DESCRIPTION ket mode.
Channelvolume. Respondentswere asked: (1) "Of your
total export sales last year, what percentage is made up
Industry classification by this particularproduct to this particularmarket?", (2)
Agriculture and primary products 1 "Whatpercentage of your total sales last year came from
Consumer products 18 exports?", and (3) "Approximately, what was the total
Paper and printing 4 value of your firm's sales last year?" The "channel vol-
Metal machinery and equipment 41
ume measure" was derived from a combination of these
Electric and electronic 16
Minerals and chemicals 20 items, which resulted in a dollar value for a firm's an-
Export experience nual exports of the particular product to the particular
Less than 5 years 21 market (measured in $10,000s).
5 to 15 years 36 Asset specificity. Asset specificity refers to the degree
More than 15 years 43
Firm size (annual sales) to which durable, transaction-specific assets were found
Less than $1 million 10 in the export channel. Both physical and human assets
$1 million to $5 million 26 were assessed. The six items used to measure this con-
$5 million to $10 million 17 struct are listed in Table 2. These items were adapted,
$10 million to $100 million 35 in part, from those used by Anderson (1985) and have
More than $100 million 12
Export intensity (export as percentage of total sales) strong face validity. The items were measured on 7-point
Less than 20% 48 scales ranging from 1, "completely disagree," to 7,
20% to 50% 23 "completely agree." They were summed and averaged
Over 50% 29 to arrive at the measure of asset specificity (a = .65).
Owner managed 58
External uncertainty. Separate measures for the two
types of uncertainty were developed. "Volatility" refers
to the extent to which the environment changes rapidly
Operational Measures and allows a firm to be caught by surprise and "diver-
Prestudy interviews with approximately 10 Canadian sity" refers to the extent to which there are multiple
exporters were instrumental in devising the operational sources of uncertaintyin the environment. The items de-
measures for the study. They were especially useful in veloped to reflect these constructs are listed in Table 2.
identifying the relevant channel integration options in in- Seven-point scales ranging from 1, "completely disa-
ternational markets and developing the measures of en- gree," to 7, "completely agree," were used. The items
vironmental volatility and diversity. were summed and averaged to obtain a measure of each
Channel integration. Respondents were informed that construct. Coefficient alpha is .70 for the volatility mea-
the focus of the study was the marketing channels used sure and .55 for the diversity measure.8
by Canadian exporters to serve foreign markets and were Assessment of Multiple-Item Measures
told to concentrate on their firm's most importantexport
product in its most importantforeign market. After writ- Factor analysis was used to assess the psychometric
ing down the product and market (i.e., country), each properties of the three constructs representing asset spec-
respondent was asked to indicate which of the following ificity, volatility, and diversity. Initial, separate analyses
descriptions best matched the firm's export arrange- indicated that five items (not presented here) should be
ments for that product in that market: (1) "We have a dropped from further analysis. These items, though ap-
wholly owned sales subsidiary"(62 firms), (2) "We serve pearing to reflect the constructs on the basis of content
the market directly from Canada, using company per- validity, did not contribute to the reliability of the scales.
sonnel" (121 firms), (3) "We are involved in a joint ven- To assess validity, Gerbing and Anderson (1988) ar-
ture with anothercompany to handle sales of this product gue for the use of all possible scales within a single anal-
in this market" (13 firms), (4) "We use commission ysis so that an assessment of internal consistency can be
agents" (74 firms), (5) "We sell to a merchant distrib- made. The results of such a factor analysis indicated that
utor who takes title to our product and contacts buyers a three-factor solution best represented the data on the
himself" (76 firms), and (6) "Other [please specify]." basis of a scree plot, as well as factorial complexity and
The descriptions provided corresponded to two hierar- interpretability. The factor loadings after a varimax ro-
chical modes (the establishment of a foreign sales sub- tation are reported in Table 2. With one exception, all
sidiary or serving the foreign market from home), an in-
termediate mode (the use of commission agents or joint
ventures), and a market mode (the use of merchant dis-
8An argument could be made that the diversity measure represents
tributors). an index rather than a scale. In a theoretical sense, simply because
By checking more than one description or checking there are many customers does not necessarily mean that there are
the "other" category, 29 firms indicated use of "dual" also many competitors. As Howell (1987) indicates, coefficient alpha
channels. These firms are held out of the hypothesis tests, is appropriate only for measurement scales, not indexes.
8. 202 JOURNAL MARKETING
OF MAY
RESEARCH, 1990
Table2 products through the channel in question. The channel
PROPERTIE MEASURES
PSYCHOMETRIC ESOF volume variable reflects only the volume of the export
firm's most important product in the foreign market. If
F'actorI Factor Factor
2 3 other products went through the same channel, the abil-
A. Asset specificity ity of the firm to reap economies of scale would in-
1. It is difficultfor an outsiderto crease; production costs would be reduced (cf. Anderson
learnour ways of doing things. -.21 .01 .53 and Coughlan 1987). A dummy variable was created to
2. To be effective, a salesperson equal one if the channel was shared and zero otherwise.
has to take a lot of time to get In 50% of the cases, the channel was in fact shared. On
to know the customers.
3. It takes a long time for a sales- the basis of prestudy interviews, we expected that export
personto learnaboutthis prod- firms would tend to use integrated channels when mul-
uct thoroughly. .18 -.01 .69 tiple product lines were distributed through the same
4. A salesperson's inside informa- channel in foreign markets.
tion on our procedures would
be very helpfulto our competi- The second control variable was specific to the Ca-
tors. .14 .04 .60 nadian context and represented whether or not the mar-
5. Specializedfacilitiesare needed ket in question was the United States. A majority of Ca-
to market product.
this .03 -.01 .67 nadian exports go to the United States and prestudy
6. A large investment equip-
in interviews indicated that Canadianexport firms have sig-
ment and facilitiesis neededto
market product.
this .32 -.01 .52 nificantly different perceptions of the U.S. in compari-
son with other foreign markets. Many Canadian export-
Eigenvalue 2.30 ers regardthe U.S. as an extension of the domestic market.
% varianceexplained 19.20
B. Volatility
Hence, we believed that, ceteris paribus, the degree of
1. We are often surprised the
by integration was likely to be greater in the U.S. market
actionsof retailersand whole- than elsewhere. A dummy variable was created to equal
salers. .08 .77 -.00 one if the U.S. market was involved (65% of the cases)
2. We are often surprised the
by and zero otherwise.9
actionsof our competitors. .07 .81 .09
3. We are often surprised cus-
by Estimation Results
tomerreaction. -.07 .78 .02
Hypotheses about the factors encouraging or discour-
Eigenvalue 1.87 aging forward integration were tested by using a multi-
% varianceexplained 15.60 nomial logit model (Malhotra 1984) that estimated the
C. Diversity
1. Thereare manyfinal users of impact of the independent variables on the probability
this productin this market. .78 .04 -.06 that each channel mode would be selected. This analyt-
2. Thereare manycompetitors for ical model is appropriatebecause the determinantsof the
this productin this market. .62 .18 .07 utilityof one mode could differ from those of other modes.
3. We have only a few immediate Estimation entailed the use of LIMDEP and the Newton-
customers this product
for in
this market(reversescaled). .64 -
-.13 -.07 Raphson method of maximizing the log-likelihood func-
tion (see Bunch and Batsell 1989).
Eigenvalue 1.56 When the multinomial logit model is estimated, one
% varianceexplained 13.00
option must be used as the base mode because once j -
1 alternative probabilities are known, the /h is deter-
mined. In testing HI, H3, H4, and H5, the base mode
variables have a loading of less than .25 on inappropriate used was that which theoretically represents the default
factors, whereas all hypothesizedloadings are greaterthan option, namely the market option (i.e., use of distribu-
.5. Even with an oblique rotation, little chiange occurred tors). The utility of the market option was assigned a
in the factor loadings and factor correlatio were min-
)ns value of zero and the utilities of the other options were
imal (.09, .04, -.005). estimated and interpretablewith reference to it. The size
A confirmatory factor analysis also was conducted on of the various coefficients indicates the extent to which
these items. The resulting X2was 142, siginificant at be- the corresponding variables contribute to the utility of
yond .001. However, the goodness-of-fit iindex was .94, choosing that option beyond their contribution the utility
to
all individual t-tests for factor loadings we: significant,
re of the market option. Results are reported in Table 3.
and only two of 55 normalized residuals wiere above .2. The chi square statistic tests the hypothesis that the es-
The discriminant validity of the measures appears to be timated coefficients, except the constant, are all zero. As
established on the basis of the results of the common and
the confirmatory factor analyses.
Control Variables
9Thecorrelations among the independent variables,includingthe
Two control variables were included in the estimation no
controlvariables,rangefrom -.08 to .22, indicating problems of
model. One was whether or not the firm also sells other multicollinearity.
10. 204 JOURNAL MARKETING
OF MAY
RESEARCH, 1990
is evident from Table 3, this hypothesis is rejected at hierarchy-subsidiaryoption, whereas environmental di-
beyond the .01 level. versity is related inversely and significantly to use of the
The coefficients of channel volume are positive and hierarchy-domestic option.
significant in each case, and especially high for the hi- In testing H2 and H6, the base mode used is the hi-
erarchical options. These findings indicate that as vol- erarchy-domestic option, which is compared with the hi-
ume of the product line increases, the probability of us- erarchy-subsidiaryoption. Results are reported in Table
ing a highly integrated channel increases in relation to 3, part C. The chi square statistic is again significant
the probability of using the market option. HI is sup- beyond the .01 level and the tau statistic is larger than
ported by the results. in the preceding two tests. As is evident, the probability
H3 predicts that asset specificity is related positively of utilizing a foreign subsidiary increases as channel vol-
to the level of channel integration. The coefficients re- ume and volatility each increase, providing support for
lating to asset specificity in Table 3 support this predic- H2 and H6. Use of a shared channel also facilitates the
tion, as they are positive and significant on each hier- foreign subsidiary option. Unexpectedly, environmental
archical option. diversity has a significant positive relationship to that
Mixed support is found for H4 and H5 on the relation- option. One interpretationis that if other factors indicate
ships between volatility, diversity, and channel integra- the necessity of a hierarchical exchange, the firm prefers
tion. As volatility increases, the probability of using the to be located within a foreign market when faced with
hierarchy-subsidiary option increases in relation to the high diversity.
probability of using the market option, as expected. Interestingly, the constant term for the hierarchy-sub-
However, the volatility coefficient is insignificant on the sidiary option is negative and significant in each part of
other hierarchy option. As predicted, diversity is related Table 3, indicating that this option is always least pre-
inversely to the use of the hierarchy-domesticoption, but ferred among the channel options. This finding is con-
it is unrelatedto the hierarchy-subsidiaryoption. At least sistent with what transactioncost analysis would predict.
some support for opposing external uncertainty effects However, it must be interpreted with caution, as it is
is offered by these results. based on the assumption of a correctly specified model.
The two control variables are related significantly to If other importantconstructs were included in the model,
channel integration as anticipated. The use of shared the size, sign, and significance of the constant could
channels and distribution to the U.S. market are both change.
related positively to the level of channel integration. On the basis of arguments and results of Anderson
To ascertain whether the model in part A of Table 3 (1985), the interaction of asset specificity and the two
fits the data well, we used it to classify the observations. external uncertainty dimensions also was examined in
A total of 43% of the observations are classified cor- separateanalyses (not reportedhere). The interactionterms
rectly. The tau statistic is .19, indicating that the 43% were included first without the main effects of asset
classification rate represents 19% fewer classification er- specificity, diversity, and volatility, and then with those
rors than would be expected by chance. However, this main effects. Each solution was clearly inferior to those
finding must be interpretedwith caution, as the data that in part A of Table 3, especially the second solution.
were classified were also used to estimate the model. Multicollinearity was a problem in each case, leading to
Much of the classificatory power of the model comes unstable coefficients.
from channel volume and the control variables. Overall, Finally, the results of the comparison between the dual
the fit of the model is weak, suggesting that measure- channel and market options are reported in Table 3, part
ment problems are present and/or other constructs with D. Only country destination and channel volume are sig-
an important impact on channel integration were ex- nificant. The probability of using dual channels in-
cluded from the model. Clearly, attempting to classify creased when the U.S. was the export country and the
correctly across four different options is difficult. channel volume of the product line was high. However,
For a more complete view of the tradeoffs among the the classificatory power of the model is extremely weak,
integration options, results of a comparison of the two as only two of the 29 dual channel cases are classified
hierarchical options with the intermediate option as the correctly.
base mode are reported in Table 3, part B; exporters us-
ing the market option were withheld from this analysis. DISCUSSION
The chi squarestatistic is significantbeyond the .01 level,
with the classificatory power of the model improving Our study is the first in which a model based on trans-
slightly as reflected by the tau statistic. The coefficients action cost analysis has been applied to explain levels of
in part B are similar to those in part A, with the excep- forward integration within distribution channels in inter-
tion that neither of the coefficients for asset specificity national markets. The results provide support for some
is significant in differentiating intermediate exchanges of the fundamental predictions of transaction cost anal-
from hierarchical exchanges. This finding suggests that ysis.
asset specificity may serve only to differentiate ex- Asset specificity is shown to differentiate significantly
changes at either extreme. Again, environmental vola- between the use of market exchanges and hierarchical
tility is related positively and significantly to use of the exchanges. When specialized knowledge and invest-
11. COSTANALYSIS
TRANSACTION MODEL 205
ments are necessary to facilitate transactions in foreign In previous research on channel integration, the im-
countries, the ability of the market to curb the oppor- pact of productioncosts has been mixed. Anderson (1985)
tunistic tendencies of outside intermediaries is limited. found no support for production cost effects, whereas
Under such conditions, hierarchical exchanges are likely John and Weitz (1988) found a weak production cost
to be preferred because opportunism can be combatted effect that was dominatedby transactioncost effects. Only
within the firm through the exercise of legitimate au- Lilien's (1979) results strongly support the importance
thority, the monitoring of behavior, and the offering of of production costs. Our results clearly suggest that the
more varied incentives. Our findings related to asset ideal channel arrangement should reflect both the vol-
specificity are consistent with those of Anderson (1985), ume of goods involved and the vulnerability of the firm
Anderson and Coughlan (1987), and John and Weitz to opportunistic behavior by outside intermediaries. In
(1988), as well as with those from a variety of other other words, both the costs of actually performing dis-
studies that have focused on this construct (see Joskow tribution functions and the costs involved in governing
1988 for a review). the channel should be considered.
Interestingly, asset specificity does not distinguish the In addition to finding support for the impact of trans-
use of intermediate exchanges from the use of either action costs and production costs on channel integration,
market exchanges or hierarchical exchanges, suggesting we attemptedto extend the transaction cost analysis per-
that the need for specialized knowledge and investments spective in terms of its treatmentof external uncertainty.
affects channel choice only at either extreme. When only The two dimensionsof externaluncertaintyexamined have
one or a few functions are entrusted to outside inter- different effects on channel integration. Environmental
mediaries in foreign markets, the firm may believe it can volatility is related positively to the hierarchy-subsidiary
maintain control of the channel and associated transac- option, and environmental diversity is related inversely
tion costs irrespective of the level of asset specificity. to use of the hierarchy-domestic option. When faced with
The external uncertainty prediction of TCA is sup- a highly diverse environment in a foreign market, the
ported by the positive relationships between environ- firm may be motivated to maintain a flexible organiza-
mental volatility and the probability that a foreign sales tion and rely on local intermediaries to cope with the
subsidiary would be established by the firm (see Table market's complexity. These results suggest that an un-
3, A and B). An adaptive, sequential decision process bundling of the external uncertaintyconstruct is essential
is needed to cope with rapid environmental change and to an understanding of the often opposing desires for
is most likely to be implemented in a highly integrated flexibility and efficiency. Such a conclusion cannot be
channel (John and Weitz 1988; Stinchcombe 1985). drawn with certainty, however, because neither environ-
Transaction costs will be reduced as a result. However, mental dimension is related significantly to both hier-
given that environmental volatility is unrelated to the hi- archical options. It is encouraging to note that empirical
erarchical option of serving the foreign market directly results of both Balakrishnan and Wernerfelt (1986) and
from home, empirical support for this effect is mixed. Walker and Weber (1984) support the need to examine
Results relating to external uncertainty and channel in- different dimensions of uncertainty separately when ap-
tegration also have been mixed in previous research. plying TCA.
The results related to asset specificity and environ- The international context of our study enabled us to
mental volatility suggest that the ability of the market to compare two different hierarchical options. Production
enforce desired or contracted behavior cannot be taken cost factors are found to be most important in differ-
at face value in noncompetitive international markets. entiating the use of either option. As channel volume
Rather, the degree of market failure present in particular increases and when shared channels are present, the
foreign markets for particularproduct lines must be rec- probability of having a wholly owned foreign subsidiary
ognized and taken into account. is found to increase. To offset the fixed costs of estab-
Strong support is found for the impact of production lishing and maintaininga foreign subsidiary, greater sales
costs on channel choice. As channel volume for the firm's volumes are necessary. Furthermore, both environmen-
product line in the foreign market increased, the inte- tal volatility and diversity are related significantly and
gration of the channel increased as well. High product positively to the subsidiary option (in comparison with
volume is likely to afford economies of scale in acquir- the hierarchy-domestic option). The volatility effect is
ing necessary resources and developing management expected, as the establishment of a subsidiary places the
skills, thereby lowering production costs. Beyond the in- firm in a better position to react and adapt to unforeseen
dividual transaction, a production cost effect also is sug- circumstances in the foreign market. However, the di-
gested by the findings associated with the use of shared versity effect is unexpected. When other factors indicate
channels. The firm is more likely to use an integrated the use of a highly integrated channel, the firm may pre-
channel when it can distribute multiple product lines fer to establish a local presence in a foreign market when
through the channel because of economies of scale. The faced with high external uncertainty, whether it comes
channel volume and shared channel constructs have from rapid change or from the presence of a heteroge-
stronger relationships with the level of channel integra- neous group of customers and competitors or both.
tion in our study than do asset specificity and environ- The preceding results shed light on the nature of chan-
mental volatility. nel integration in international markets. However, the
12. 206 JOURNAL MARKETING
OF MAY
RESEARCH, 1990
study raises more questions than it answers. Though evi- keting (cf. Graham and Gr0nhaug 1989). The measures
dence of opposing external uncertainty effects is found, of environmental volatility and environmental diversity
the results are somewhat weak. Future research must ex- especially appear to need improvement. Other ap-
amine in greater detail how environmental volatility and proaches to measuring channel volume also are needed,
diversity, as well as other dimensions of external un- as data were missing on many of the sales items used in
certainty (e.g., volume uncertainty, technological un- our study.
certainty), relate to alternative levels of channel integra- Further, the transaction cost analysis model was es-
tion in both domestic and foreign markets. timated on the basis of current practices of the export
Similar to results of Walker and Weber (1984) is our firms, whereas the theoretical foundation of TCA is es-
finding that production cost factors have a stronger im- sentially normative. In basing the hypothesis tests on
pact than transaction cost factors. However, Anderson currentpractices, we assume that inefficient practiceshave
(1985) and John and Weitz (1988) found transactioncost been selected out and that some type of equilibrium has
factors to dominate. Additional research is needed to ex- been reached. Such an assumption may not be strictly
amine furtherthe relative importance of production cost true in the case of all export firms, as government re-
factors and transaction cost factors in affecting channel strictions, contractual commitments, resource scarcity,
integration. Contingency theories perhaps should be de- lack of capable intermediaries, and other factors can lead
veloped. to a significant positive-normativegap. To the extent that
Ours is the first empirical study on channel integration there are differences between what firms desire in their
to examine intermediate exchanges. Though the ability channel arrangementsand what they are able to achieve,
of the model to differentiate these exchanges from hi- the fit of the TCA model would be lessened.
erarchical exchanges is reasonable, intermediate ex- Another possible explanation for the low classificatory
changes are weakly distinguished from market ex- power of the model is that the transaction cost analysis
changes. Futureresearch must furtherstudy intermediate model we tested is incomplete. Several researchers have
exchanges and build better models to distinguish them commented on the limited scope of TCA (cf. Harrigan
from other forms of exchange. Moreover, we identified 1983; Heide and John 1988). The literature on foreign
intermediate exchanges on the basis of channel mem- market entry suggests that factors related to political and
bers' performance of the selling function. It would be economic risks influence channel integration (Cavusgil
interesting to examine intermediate exchanges in which 1980; Keegan 1984; Root 1987). Other explanations for
other functions are involved. Differentiating market ex- vertical integration have been proposed, including the
changes from intermediate exchanges on the basis of the product characteristic (Aspinwall 1958), market char-
organization theory tradition of the "closeness of the re- acteristic (Bucklin 1966), resource dependence (Pfeffer
lationship" also warrants attention (cf. Dwyer, Schurr, and Salancik 1978), political economy (Dwyer and Welsh
and Oh 1987). 1985; Stern and Reve 1980), strategic purposes (Harri-
Country destination, whether the U.S. or elsewhere, gan 1983; Porter 1985), and entry-deterring (Stigler 1951)
has a consistently strong and positive effect on the level frameworks. Several of these frameworks are non-effi-
of integration in the channel. This finding could be due ciency-based.
primarily to the study context, as the establishment of Future research must address these limitations. An in-
integrated channels in the U.S. market by Canadian ex- tegration of alternative perspectives on forward integra-
porters is facilitated by both geographical proximity and tion would be especially welcome. The importance of
culturalsimilarity. Futureresearch must evaluate whether such an integration is suggested by our study, in which
or not this is an isolated effect. Empirical results of An- organization theory is used to explain the environmental
derson and Coughlan (1987) suggest otherwise. diversity effect on channel integration.
Finally, our model does a very poor job of explaining
the use of dual channels, though part of the problem may CONCLUSION
be the small number of dual cases examined. Previously, We developed and applied a model based on trans-
only John and Weitz (1988) had examined empirically action cost analysis, including production cost indica-
the use of dual-multiple channels. The question of when tors, to explain levels of forward integration within dis-
the use of dual-multiple channels is appropriate is ex- tribution channels in foreign markets. Four different
tremely important managerially and warrants greater at- integration choices-market exchange, intermediate ex-
tention in the future. change, and two forms of hierarchical exchange-were
examined. Based on data collected from a group of Ca-
LIMITATIONS nadian export firms, the results provide support for the
The transaction cost analysis model and the control transaction cost model.
variablesare weak in classifying the export firms in terms The most original findings of the study relate to two
of their channel choices. Imperfect measurement ac- dimensions of external uncertainty. Environmental vol-
counts for part of this lack of fit, especially given the atility and environmental diversity are shown to have
developmental stage of this research area and the fact differential effects on the level of integration in distri-
that our study pertains to an aspect of internationalmar- bution channels in foreign markets, suggesting that the
13. COSTANALYSIS
TRANSACTION MODEL 207
TCA model must deepen its treatment of the external Gatignon, Hubert and Erin Anderson (1988), "The Multina-
uncertainty construct. tional Corporation's Degree of Control Over Foreign Sub-
Though most empirical research in the channels lit- sidiaries: An Empirical Test of a Transaction Cost Expla-
eraturehas centeredon the managementof ongoing dyadic nation," Journal of Law, Economics, and Organization, 4
channel relationships, a recent stream of research has fo- (Fall), 305-36.
cused on the nature of the firm's channel integration Gerbing, David and James Anderson (1988), "An Updated
choices. In our study this focus is shifted to an inter- Paradigm for Scale Development Incorporating Unidimen-
sionality and Its Assessment," Journal of Marketing Re-
national context, which we hope will stimulate addi- search, 25 (May), 186-92.
tional research on channel integration in the future. Graham, John and Kjell Gr0nhaug (1989), "Why We Haven't
Learned Much About International Marketing in the Last
REFERENCES Twenty-Five Years," Journal of Higher Education, 60
(March/April), 152-87.
Achrol,Ravi Singh, TorgerReve, andLouisW. Stern(1983), Harrigan, Kathryn (1983), Strategies for Vertical Integration.
of
"TheEnvironment Marketing ChannelDyads:A Frame- Lexington, MA: Lexington Books.
work for Comparative Analysis," Journal of Marketing, 47 Heide, Jan and George John (1988), "The Role of Dependence
(Fall), 55-67. Balancing in Safeguarding Transaction-Specific Assets in
Aldrich, Howard (1979), Organizations and Environments. Conventional Channels," Journal of Marketing, 52 (Janu-
EnglewoodCliffs, NJ: Prentice-Hall,
Inc. ary), 20-35.
Anderson,Erin(1985), "TheSalesperson OutsideAgentor
as Howell, Roy (1987), "Covariance Structure Modeling and
Employee:A Transaction Cost Analysis," MarketingSci- Measurement Issues: A Note on Interrelations Among a
ence, 4 (Summer), 234-54. Channel Entity's Power Sources," Journal of Marketing Re-
andAnneCoughlan Market
(1987), "International En- search, 24 (February), 119-26.
try and Expansionvia Independent Integrated
or Channels John, George and Barton Weitz (1988), "ForwardIntegration
Journalof Marketing,51 (January),
of Distribution," 71- Into Distribution: Empirical Test of Transaction Cost Anal-
82. ysis," Journal of Law, Economics, and Organization,4 (Fall),
and HubertGatignon(1986), "Modesof ForeignEn- 121-39.
try:A Transaction AnalysisandPropositions,"
Cost Journal Joskow, Paul (1988), "Asset Specificity and the Structure of
of International Business Studies, 17 (Fall), 1-26. Vertical Relationships:EmpiricalEvidence," Journal of Law,
and BartonWeitz (1986), "Makeor Buy Decisions: Economics, and Organization, 4 (Spring), 95-117.
A Framework AnalyzingVerticalIntegration
for Issues in Keegan, Warren (1984), Multinational Marketing Manage-
Marketing," Sloan Management Review, 27 (Spring), 3-19. ment, 3rd ed. Englewood Cliffs, NJ: Prentice-Hall, Inc.
Aspinwall, Leo (1958), "The Characteristics of Goods and Lawrence, Paul and Jay Lorsch (1967), Organizations and En-
Parallel Systems," in Managerial Marketing, Harold Kelley vironments. Cambridge, MA: Harvard Business School.
and William Lazar, eds. Homewood, IL: Richard D. Irwin, Leblebici, Huseyin and Gerald Salancik (1981), "Effects of
Inc., 434-50. Environmental Uncertainty on Information and Decision
Balakrishnan, Srinivasan and Birger Wernerfelt (1986), Processes in Banks," Administrative Science Quarterly, 26
"Technical
Change,Competition, VerticalIntegration,"
and (December), 578-96.
Strategic Management Journal, 7 (July-August), 347-59. Lilien, Gary (1979), "Advisor 2: Modeling the Marketing Mix
Bucklin, Louis (1966), A Theory of Distribution Channel Decision for Industrial Products," Management Science, 25
Structure. Berkeley, CA: Institute of Business and Eco- (February), 191-204.
nomic Research. Malhotra, Naresh (1984), "The Use of Linear Logit Models
Bunch, David and Richard Batsell (1989), "A Mont6 Carlo in Marketing Research," Journal of Marketing Research, 21
of for
Comparison Estimators the Multinomial
LogitModel," (February), 20-31.
Journal of Marketing Research, 26 (February), 56-68. Ouchi, William (1980), "Markets, Bureaucracies, and Clans,"
Cavusgil, S. Tamer (1980), "On the InternationalizationPro- Administrative Science Quarterly, 25 (March), 129-41.
cess of Firms," European Research, 8 (November), 273- Pfeffer, Jeffrey and Gerald Salancik (1978), The External
81. Control of Organizations: A Resource-Dependence Per-
Coughlan, Anne (1985), "Competition and Cooperation in spective. New York: Harper & Row Publishers, Inc.
ChannelChoice:TheoryandApplication,"
Marketing Mar- Porter, Michael (1985), Competitive Advantage. New York:
keting Science, 4 (2), 110-29. The Free Press.
Dess, GregoryandDonaldBeard(1984), "Dimensions Or-
of Reid, Stan (1983), "FirmIntemationalization,TransactionCosts,
ganizational Task Environments," Administrative Science and Strategic Choice," International Marketing Review, 1
Quarterly, 29 (March), 52-73. (Winter), 44-56.
Duncan, Robert (1972), "Characteristicsof Organizational En- Root, Franklin (1987), Entry Strategies for International Mar-
vironments PerceivedEnvironmental
and Ad-
Uncertainty," kets. Lexington, MA: Lexington Books.
ministrative Science Quarterly, 17 (September), 313-27. Rosenbloom, Bert (1987), Marketing Channels, 3rd ed. Chi-
Dwyer, F. Robert, Paul Schurr,and Sejo Oh (1987), "De- cago: The Dryden Press.
veloping Buyer-Seller Relationships," Journal of Market- Ster, Louis and Adel El-Ansary (1988), Marketing Channels,
ing, 51 (April), 11-27. 3rd ed. Englewood Cliffs, NJ: Prentice-Hall, Inc.
and M. Ann Welsh (1985), "EnvironmentalRelation- and Torger Reve (1980), "Distribution Channels as
ships of the Internal Political Economy of Marketing Chan- Political Economies: A Framework for Comparative Anal-
nels," Journal of MarketingResearch, 22 (November), 397- ysis," Journal of Marketing, 44 (Summer), 52-64.
414. Stigler, George (1951), "The Division of Labor Is Limited by
14. 208 JOURNALOF MARKETING MAY 1990
RESEARCH,
the Extent of the Market," Journal of Political Economy, ence Quarterly, 29 (September), 373-91.
54 (June), 185-93. Williamson, Oliver (1975), Markets and Hierarchies: Analysis
Stinchcombe, Arthur (1985), "Contractsas Hierarchical Doc- and Antitrust Implications. New York: The Free Press.
uments," in Organization Theory and Project Management, (1985), The Economic Institutions of Capitalism. New
A. Stinchcombe and C. Heimer, eds. Oslo: Norwegian Uni- York: The Free Press.
versity Press.
Walker, Gordon and David Weber (1984), "A TransactionCost
Approach to Make-or-Buy Decisions," Administrative Sci- ReprintNo. JMR272106
ASSOCIATION
MARKETING
AMERICAN
REPRINT AND PERMISSION POLICIES AND PROCEDURESFOR JOURNALARTICLES
TheAmerican
Marketing is of infornation
Association eagerto serveyouin thesharing thevaluable articles.
foundin AMA'sjournal
REPRINTS
If you wishto order
REPRINTS anyarticle,
of writeto theAmerican Association,
Marketing Department, S. Wacker
Reprints 250 Drive,Chicago,
IL60606.Call(312)648-0536or Fax(312)993-7540.
Thepricesbelowapplyto eachtitleordered.
Multipletidtles be combined reach minimum
may to the order of
amount $15.00.
cover,andreproduced 20# mate paper
Eachreproduction be provided anattached
will with on stock.Ordersmustbe prepaidby checkor credit
card.NO REFUNDS EXCHANGES.
OR
REPRINT order $15.00)
PRICES(minimum is
1-14 reproductionsthesamearticle an assortment
of or $1.50each
of articles.
I 15-99 reproductionsthesamearticle
of $1.25each
100-500 reproductionsthesamearticle
of $1.00each
>500 reproductionsthesamearticle
of callforquote
I ALL PRICESINCLUDE
SHIPPING CHARGES
ANDHANDLING
fromUniversity
Also, a singlecopyreprint an order less than50 copiesmaybe obtained
or of MicrofilmsInternational, N. ZeebRoad,Armnn
300
MI are
Arbor, 48106.Articles priced prepaid $17.50plus$2.25foreachadditional of the samearticle.
at copy account
Thosewhohavea standing
withUMIpay$7.50perarticle $2.25foreachadditional of thesamearticle.
plus copy Completeissuesareobtainable $35.00each.
at
PERMISSIONS
be to
AMAwould proud grant PERMISSION THE
FOR REPRODUCTION of copyrighted To
materials. obtain
permission
to reproduce copiesfromAMApublications,
multiple contact American
the Marketing Pennissions
Association, Department
Under 'fair provision theCopyright effective
the use' of Law, January, anyone makea photocopy a copyrighted
1978, may of
AIERICAN
workfortheirownuse without seekingpermission. M4RKETING
Tosecure to
permission reproduce in
one'sownworks quantity, the
pleasecontact Permissions Department. A$OCIATION
I .