留学生国际商务管理作业写作需求由第一代写网提供。Measures of Strategic Alliance Performance: An Analysis of Construct Validity
Author(s): Africa Ariño
Source: Journal of International Business Studies, Vol. 34, No. 1 (Jan., 2003), pp. 66-79
Published by: Palgrave Macmillan Journals
Stable URL:Accessed: 26/06/2010 23:45
Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at
. 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 youmay 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
Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printedpage of such transmission.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range ofcontent in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new formsof scholarship. For more information about JSTOR, please contact Macmillan Journals is collaborating with JSTOR to digitize, preserve and extend access to Journal ofInternational Business Studies.
%1g~~~~~~.W./ journaolf InternationBaul sinesSst udie(s2 003)3 4,6 6-79
(? 2003 PalgraveM acmillanL td. Allr ightsr eserved0 047-2506 $25.00
Measures of strategic alliance performance: an analysis of construct validity Africa Arino
IESEB, usinessS choolU niversitoyf Navarra, BarcelonaS, pain
Correspondence:
Dr Africa Ariio, IESE, Business School
University of Navarra, Ave. Pearson, 21,
Online publication date: 5 December 2002
Abstract
This study evaluates the construct validity of measures of the performance ofstrategic alliances in 34 equity strategic alliances (SAs) and 45 contractual SAs.We discuss the content validity of existing measures in the light of currentconcepts of SA performance. Empirical results show that the fulfilment ofstrategic goals, and other organizational effectiveness measures of SAperformance - such as overall performance satisfaction and net spillover effects- capture different underlying factors. The latter two measures displayconvergent validity. The results also support the discriminant validity of thesetwo measures with respect to contractual changes and survival, but not withrespect to longevity. Building on our findings, we propose an expandeddefinition of SA performance that considers both outcome and processperformance.
journal of International Business Studies (2003) 34, 66-79. doi:10.1057/
palgrave.jibs.8400005
Introduction
The measurement of the performance of strategic alliances (SAs)has been an important research topic in the field of internationalmanagement (Geringer and Hebert, 1991; Yan and Zeng, 1999).
However, the validity of the underlying measures is stillquestionable, and no attempt has been made to estimate theirempirical validity (Parkhe, 1993a).The validity of the concepts that underlie measures of SAperformance needs to be assessed, for at least three reasons:
(1) To evaluate the soundness of inferences drawn from therelationships between SA performance and other theoreticalconstructs.
(2) To help future researchers in selecting measures of SAperformance.
(3) To make valid recommendations to managers on a soundbasis of research results.
The main aspects of construct validity are: content validity;reliability; and convergent, discriminant and criterion-relatedvalidity. There is still no clear description of the dimensions thatdefine SA performance. Geringer and Hebert (1991) relied oncorrelations alone to establish the reliability and comparability ofalternative measures of performance. This paper proposes an initialdefinition of SA performance that allows us to assess the contentStrateqic alliance performancevalidity of these measures. Their empirical validityis assessed using a method based on confirmatoryfactor analysis applied to two subsamples of equityand contractual SAs. This method is both broaderand more powerful than that used in previouswork: broader, because it covers both content andempirical validity; more powerful, because the useof confirmatory factor analysisallows us to isolatemeasurement error.Different types of measure correspond to differentlevels of performance (Venkatraman and
Ramanujam, 1986). Financial measures includevarious measures of profitability, growth and costposition (for a review see Geringer and Hebert,1991). Frequently used operational measures includestability measures such as longevity of the SAownership or contract stability, and survival (seereviews in Geringer and Hebert, 1991; Yan andZeng, 1999). The most commonly used organizationaleffectiveness measure is an overall assessmentof the firm's satisfaction with SA performance(Beamish, 1988; Geringer and Hebert, 1991; Parkhe,1993b; Lin and Germain, 1998). Other organizationaleffectiveness measures used include howwell the SA has fulfilled its strategic goals(Parkhe, 1993a, b), and the net spillover effects ofthe SA on the firm's other activities (Parkhe,
This paper begins by defining SA and SAperformance. After presenting the concept ofconstruct validity and its different aspects, thecontent validity of the various measures of SAperformance is assessed. I suggest that the organizational
effectiveness measures capture the conceptbetter than the operational ones. Next, I assess theempirical aspects of construct validity using twosubsamples of 34 equity SAs, and 45 contractual SAsinvolving at least one company operating in Spain.Based on the discussion of my findings, I suggest anexpanded definition of SA performance thatconsiders both out-come and process performanceaspects. Implications for research are offered in theconclusion section.Strategic alliances and their performance
In order to assess the validity of SA performance
measures, I first need to define what I mean by SAand by SA performance. I define an SA as a formalagreement between two or more business organizationsto pursue a set of private and commoninterests through the sharing of resources in
contexts involving uncertainty over outcomes(Arifio et al., 2001). An alliance is strategic when itis the means by which a firm seeks to implement, inpart or in whole, elements of management'sstrategic intent (Hamel and Prahalad, 1989).
SAs may be classified in various ways. They mayseek exploration or exploitation purposes (Koza and
Lewin, 1998). They may be cross-border or domestic,be governed as equity ventures or as contractualagreements, involve dyads of partners or multiplepartners. They may link firms in different stages of
/ value chain or seek the benefit of scale byconnecting companies in the same stage of thechain (Dussauge et al., 2000). The resources shared
may be complementary or they may constitute apool of similar resources (Nohria and Garcia-Pont,
1991). Depending on its purpose, an SA maytemporary or enduring.
These, and other, differences are likely to result in
diverse dynamics, and to have differing impacts onSA performance. For instance, equity-based SAs cantake a longer time to negotiate and organize thancontractual SAs. They provide stronger control andincentive mechanisms, but are less flexible; theymake knowledge transfer easier, but can involvehigher exit costs. The management efforts involvedin the various types of SA also vary (Gulati, 1995),and the SAs are likely to evolve in different ways(Doz, 1996), with consequent differences in performance.
There is no agreed definition of SA performancein the literature (Yan and Zeng, 1999), but goal
accomplishment underlies most interpretations(Beamish, 1988; Anderson, 1990; Beamish andDelios, 1997; Lin and Germain, 1998). However,drawing from the strategy literature, we mayrecognize three levels of performance that dependon the goals under consideration: financialperformance, operational performance, and organizationaleffectiveness (Venkatraman andRamanujam, 1986).
(1) Financial performance is relevant when thepartners in an SA have explicit financial goalsfor it.
(2) Operational performance 'focuses on those keyoperational success factors that might lead tofinancial performance' (Venkatraman andRamanujam, 1986, 804, emphasis in original).It can be measured by indicators of suchkey 代写留学生作业success factors. I depart here fromVenkatraman and Ramanujam's assertion: asfinancial performance is not always relevantto SAs (Anderson, 1990), it is sufficient to
Journal of International Business StudiesAfrica Arino
67
.PtratMqw~nir%aI. Iinnem"W Fvrn%~F~nr n -e II arI-.68say that the operational success factorsmight lead to success understood aseffectiveness.
(3) Organizational effectiveness refers to the fulfilmentof the organization's goals, taking intoaccount the interests of multiple constituencies.How should it be measured? This iswidely debated in the literature (Gawandeand Wheeler,1999).
Organizational effectiveness is the most comprehensive
of these three. If profitability is a specific
goal, then organizational effectiveness will explicitly
include financial performance. Also, if key
operational success factors lead to achievement of
organizational goals, then organizational effectiveness
will implicitly encompass operational performance.
Because there are several parties to an SA, a
multiple-constituencies approach is needed in
evaluating organizational effectiveness, including
each of the partners to the SA as an independent
organization, the SA management team, and the
community - particularly the host government in
the case of a cross-border SA. Throughout this paper
I shall assume that the goals of SA management and
of the community constrain the scope of the
partners' goals. I therefore assume that the relevant
goals in evaluating SA performance are those of the
partners, and that they reflect the goals of other
constituencies insofar as they are constrained by
them.
There are two issues to consider in relation to the
partners' goals. First, each partner will usually have
goals for the SA that are not shared by the other
partner. The shared interests are the common goals
of the SA; the goals that each firm has for the SA
and which it does not share with its partner are the
private goals (Arifio, 1995). Second, both the
common and the private goals may change over
time (Doz, 1996), yielding emergent goals that differ
from the initial ones - whether common or private.
Thus I define SA performance as the degree of
accomplishment of partners' goals, be these common
or private, initial or emergent.
Initial goals - both common and private - set the
stage for potential conflict. The level of conflict,
high or low, will influence the way the partners
interact: that is, the SA process. And the way the SA
process develops will influence both the partners'
ability to achieve their goals for the SA, and the
emergence of new goals. SA outcomes and process
are thus intertwined.
Construct validity
Construct validity represents 'the correspondence
between a construct (conceptual definition of a
variable) and the operational procedure to measure...
that construct' (Schwab, 1980, 5-6). It includes
content validity, reliability, and convergent, discriminant
and criterion-relatevda lidity of the measures of
the construct.
(1) Content validity is 'a qualitative type of validity
where the domain of a concept is made clear
and the analyst judges whether the measures
fully represent that domain' (Bollen, 1989,
185). The domain of a concept is bounded by
its theoretical definition, which should reflect
the meanings associated with the concept in
prior research and make its dimensions clear
(Bollen, 1989).
(2) Reliability is traditionally defined as the
consistency of measurement. An alternative
definition, appropriate from a structural
perspective, is that the reliability of a measure
is the magnitude of the direct relations that all
variables (except the error terms) have on that
measure (Bollen, 1989).
(3) Convergenvt alidity is 'the degree to which two
or more attempts to measure the same
concept... are in agreement' (Bagozzi and
Phillips, 1982, 468). Measures of the same
construct should display a large common
variance. Convergence is especially important
in areas in which there is a proliferation of
measures presumably assessing the same
construct (Schwab, 1980).
(4) Discriminant validity is 'the degree to which
measures of distinct concepts differ' (Bagozzi
and Phillips, 1982, 469). Measures of different
constructs should share little variance. The
analysis of discriminant validity may help to
corroborate issues of content validity when it
is suspected that some measures actually
correspond to another concept.
(5) Criterion-related validity is 'the degree of
correspondence between a measure and a
criterion variable' (Bollen, 1989, 186). To
evaluate it, a variable is needed that is a
standard against which to compare the
measure in question.
Content validity of measures of strategic
alliance performance
I turn now to evaluate the content validity
of existing operational and organizational
Journal of International Business Studies
r
Stratenir alliance Derfnrmanrce Africa Arino
Strateaic alliance performance
69
Africa Arino
effectiveness measures of SA performance in the
literature, by judging how far these measures reflect
the degree of accomplishment of the partners'
common and private, initial and emergent goals
for an SA. The analysis does not incorporate
financial measures, as financial performance is not
always relevant to SAs, and because I lacked
financial data to run the empirical analysis.
Operational measures
Operational measures of SA performance are
indicators of key success factors that lead to SA
effectiveness. In practice, they are measures of SA
stability. Although it is widely recognized that
stability and performance are different concepts
(Yan and Zeng, 1999), it is questionable whether
measures of stability are valid measures of performance.
At the conceptual level, Yan and Zeng's
(1999) discussion offers a solid ground for asserting
that an SA may be successful or unsuccessful
independently of how long it lasts. Longevity may
be a precursor to or a result of performance: and it is
difficult to decide which it is without knowing the
goals pursued with an SA. If an SA is meant to
endure, but actually has an existence shorter than
necessary to fulfil its intended goals, then it can be
said to have failed, but in general longevity does
not display content validity as a measure of SA
performance.
Similarly, ownership or contractual changes may
respond to a number of causes: an initial governance
misfit (Gomes-Casseres, 1987), an expected
step to move the SA forward (Kogut, 1991), or a
successful adaptation to new environmental conditions
(Yan, 1998). Thus ownership or contractual
changes are not conclusive about SA success. They
therefore do not display content validity as
measures of SA performance.
Turning to survival, we need to consider how an
SA terminates. If it is by dissolution, the discussion
on longevity applies: survival displays content
validity only when the SA was meant to endure
but actually lasted shorter than necessary to fulfil
the partners' goals. If the SA is acquired by one of
the partners, the discussion on ownership changes
applies: survival does not display content validity as
a measure of SA performance.
In sum, the operational measures can be related
to performance only in particular instances. These
instances require some knowledge about the
temporal nature of SA goals. It would be misleading
to use them as measures of performance without
this prior knowledge.
Organizational effectiveness measures
Organizational effectiveness measures of SA performance
assess the degree of fulfilment of several
goals from the perspective of one of the partners. A
partner's satisfaction with the overall performance of
the SA is one of the most frequently used measures
of SA performance (Beamish, 1988; Geringer and
Hebert, 1991; Parkhe, 1993b; Lin and Germain,
1998). As it is an overall assessment, this measure
evaluates the fulfilment of a partner's current
common and private goals - which may be the
initial ones or others that emerged during the
course of the SA.
Parkhe (1993b) measured SA performance as the
degree of fulfilment of strategic goals. This measure
allows respondents to rate the importance of the
particular items suggested by the researcher for
their SA. The formation of an SA responds to some
strategic goal that constitutes its initial purpose.
Thus this measure captures the degree of fulfilment
of the initial goals, be these common or private.
Finally, another measure is the net spillover effects
of the SA on other activities of the company. It is
defined as the difference between positive spillover
effects - such as the application of know-how
gained from SA activities to non-SA operations -
and negative spillover effects - such as competition
between the SA and other operations of the parent
firm (Parkhe, 1993b). By definition, spillover effects
pertain to private goals, either initial or emergent.
Thus this measure captures the degree of fulfilment
of a partner's subset of private goals for the SA.
To summarize, the various organizational effectiveness
measures of SA performance display
different degrees of content validity. A partner's
satisfaction with the SA's overall performance
evaluates the degree of fulfilment of this partner's
goals (common and private, initial and emergent),
and displays the highest content validity of the
organizational effectiveness measures. The assessment
of the accomplishment of strategic goals
evaluates the degree of fulfilment of initial goals -
common and private. Finally, the estimation of
spillover effects evaluates the degree of fulfilment
of a subset of private goals - initial and emergent.
Empirical validity of strategic alliance
performance measures
Whereas assessment of content validity relies on
conceptual arguments, evaluation of the other
aspects of construct validity has an empirical
component. For this purpose, I follow a structural
approach. This procedure requires that I specify a
Journal of International Business Studies
?
- 'IL
70
Strategic alliance performance AfricAa rino
E1
E2
E3
Figure 1 Initial measurement model.
measurement model, which should include only
those measures shown to have content validity.
From the conceptual discussion of content validity
it follows that the measurement model to be tested
has to incorporate organizational effectiveness
measures as indicators of SA performance, while
operational measures are allowed to correlate freely
with SA performance: that is, the values of these
correlations are parameters to be estimated. The
diagram in Figure 1 represents the expected
relationships. The first condition for measures to
be valid is that they be reliable. Reliability will be
assessed using Cronbach's alpha, and by means of
the reliability coefficient derived from the measurement
model. As a result of the reliability tests, the
measurement model may need to be modified. The
convergent validity and discriminant validity of
reliable organizational effectiveness measures will
be assessed on the grounds of the revised measurement
model. Assessment of criterion-related validity
is not relevant in this case, as there is no readily
available model of SA performance from which to
draw ex ante predictions with which to compare
measures of SA performance. 'For many measures
no criterion is available' (Bollen, 1989, 188).
Methods
Sample
The information needed to carry out a study of this
nature is not readily available from secondary
sources. I therefore collected information from
primary sources through a mail questionnaire
survey. The sample was drawn from Spanish firms
that appeared in Funk and Scott's Countries Index -
Europe (1986-1992) as having announced their
engagement in venturing activities. The selected
time period begins with Spain's admission to the
European Community (1986), and concludes with
the establishment of the Single European Market
(1992), a period that can be expected a priori to
display considerable venturing activity. I selected as
target industries those with a higher number of SAs
(see Table 1), and focused on dyadic SAs so as to
control for the number of partners. I detected a
total of 436 SAs.
The target informant was the person in each firm
most directly related to the SA. Sacrificing quantity
for quality, I sent out questionnaires only to those
firms in which I was able to identify this person.
This gave a total of 189 questionnaires mailed to
Table I Industries and responses
Industryd escription Questionnairesm ailed No. of responses Percentageo f responses
Energy (petroleum and electricity)
Chemicals
Machinery except electrical
Electronic equipment
Transportation equipment
Transportation
Communications
Financial services
Other services
Total
Journal of International Business Studies
19
15
7
7
5
8
2
95
31
6
14
5
4
4
6
0
37
15
189 91
6.6
15.4
5.5
4.4
4.4
6.6
0.0
40.6
16.5
100
Strateqic alliance Derformance
firms involved in 160 different SAs. Of the mailed
questionnaires, 91 (48%) were returned. I attribute
this rather high response rate to the care taken in
identifying the target respondent and in the followup
process. For this, I followed the procedure that
Dillman (1978) suggests, supplemented with phone
calls. To encourage response further, respondents
were assured of confidentiality and access to the
study's findings. More than 63% of the informants
had participated in the negotiation of their firm's
SA; on average, they had been involved with the SA
for 4.9 years, with 91% having been involved at
least since the formation of the alliance, and 4.5%
during its first year of operation. These details
indicate the competence of key informants.
For the purpose of this study, eight of the
returned questionnaires were incomplete. In four
cases. I received the answers from each side of the
SA dyad. To ensure independent data points I
dropped one of the parties to each SA, selecting it
randomly by the flip of a coin. This left a final
sample of 79 questionnaires for this study. The
sample was split in two subsamples. The first
included 34 equity SAs, and the second 45
contractual SAs (28 marketing agreements, eight
manufacturing-related agreements, three joint product
development agreements, two technology
exchanges, and one licence agreement; this information
was missing in three cases). The reason
for the split was to investigate differences in the
validity of performance measures in different types
of SA. Although a larger sample size would be
desirable, the necessary recourse to primary sources
poses an important constraint to enlarging it
significantly.
Survey-based research leaves open the possibility
of consistency artefacts and common method bias.
To address this possibility I took the caution of
arranging the questionnaire items (see Appendix)
so that the subjective items appeared prior to the
more objective ones (Salancik and Pfeffer, 1977). I
also examined whether a significant amount of
common method variance exists in the data using
Harman's (1967) single-factor test. If such a bias
exists, then a factor analysis of the items will
generate a single factor or a general factor that
accounts for most of the variance in the data (e.g.
Podsakoff and Organ, 1986). Unrotated factor
analysis using the eigenvalue-greater-than-one
criterion revealed that the first factor explained
only 8.74% of the variance in the data. Thus I
concluded that the analysis was not subject to
common method bias.
Africa Arino *^
71
Measures
Organizational effectiveness measures of SA performance
Overall performance satisfaction is a fivepoint
scale measuring the informants' assessment
of how far their firm is satisfied with the overall
performance of the SA.
Strategicg oals fulfilmenti s a composite measure of
the importancem ultiplied by the degree of fulfilment
of a partner's specific strategic goals that an SA may
be meeting (Parkhe, 1993b). Importance is a fivepoint
scale measuring the informants' assessment
of how far each of the possible goals embraced by
the firm for an SA was important to their firm.
Fulfilment is a five-point scale measuring the
informants' assessment of how far each of the
identified strategic goals for the SA was fulfilled.
The list of strategic goals was adapted from the
literature (Contractor and Lorange, 1988; Parkhe,
1993a, b). I added an open-ended item so that the
informants might consider additional goals not
listed. I also provided the category 'not applicable'
as an option.
Net spillover effects is a five-point scale measuring
the informants' assessment of the difference
between positive and negative effects of the SA on
other activities of their firm (Parkhe, 1993a).
Operationalm easures Longevitym easures the number
of years elapsed between the moment the SA
was formed and the moment it was terminated, or
the time of data collection if the SA was still
operating then. Contractual changes measures the
number of changes in the ownership structure or in
the SA contract. Survival measures whether the SA
was still operating at the time of data collection.
Estimation and results
Table 2 shows the descriptive statistics and correlations
of the variables in the study for each of the
two subsamples. Seventy per cent of the SAs were
cross-border. Possible nonresponse bias was examined
by comparing differences between early and
late respondents in all of the variables used in the
study. There were significantly (P<0.05) more
survivors among late than among early respondents.
This means that, although the number of
nonsurvivors in the sample is low relative to the
number of survivors, our sample includes a
significantly high proportion of nonsurvivors. This
bias needs to be taken into consideration when
interpreting the results.
Journal of International Business Studies
*^
72
Stratenir alliance ncrfnrmance Africa Arino
Table 2 Descriptive statistics and correlations by subsample
Variable Equity SAsa
Mean s.d.
Non-equity SAsb
1 2 3 4 5 6 Mean s.d. 1 2 3 4 5 6
(1) Overall performance
satisfaction
(2) Strategic goals fulfilment
(3) Net spillover effects
(4) Longevity
(5) Contractual changes
(6) Survival
3.56 1.16 1.00
10.30 3.99
3.65 0.81
4.38 3.20
0.65 0.85
0.79 0.41
0.52
0.76
0.20
-0.35
0.50
1.00
0.45
0.41
-0.14
0.56
1.00
0.32 1.00
-0.49 0.06
0.41 0.29
3.27 0.99 1.00
10.11 3.59 0.50 1.00
3.64 0.77 0.67 0.45
3.82 5.46 0.13 0.04
1.00 0.20 0.40 0.28 0.20
0.39 1.00 0.76 0.43 0.44 0.11
1.00
0.10 1.00
0.16 0.37 1.00
0.34 0.13 0.28 1.00
aN= 34. Correlationasb ove r= 0.34 significanat t P<0.05. bN=45. Correlationasb ove r=0.30 significanat t P<0.05. Pearsonc orrelationrse ported
except for survival (Spearman correlations reported in this case).
The first step was assessing reliability. As I shall
show shortly, the expected measurement model
needed some modifications. The revised model was
used to assess convergent validity and discriminant
validity, using the maximum likelihood estimation
method of the EQS structural equation package
(Bentler, 1992a). This method has been proved to
reject models more frequently than the generalized
least squares method with small-sized samples (Hu
et al., 1992), thus providing a more trustworthy test
for my case.
Reliability
Test 1: Cronbach's alpha. Cronbach's alpha for the
three organizational effectiveness measures
was below the recommended 0.60 limit
both for the equity SAs subsample
(a=0.51), and for the non-equity SAs
subsample ( =0.48). Removing strategic
goals fulfilment improved the coefficient
to 0.84 and 0.62, respectively. (Apparently,
this result is in contrast to that of Parkhe
(1993b) who reported an alpha of 0.79 for
his strategic goals fulfilment measure. However,
Parkhe's alpha referred to the 16 items
he used to form his composite index of
strategic goals fulfilment. The alpha I report
refers to the reliability of using strategic
goals fulfilment, overall performance satisfaction,
and net spillover effects as measures
of the same latent construct. For comparison,
I applied Parkhe's calculation to my
data, yielding a value of 0.73.)
Test 2: Reliability coefficient. The reliability coefficient
is the square of the multiple correlation
coefficient for the variable (Bollen,
1989, 221). This coefficient measures the
proportion of variance in a measure that is
explained by the variables directly affecting
it in the model. Although there is no
standard cutoff point, the higher the
coefficient, the more reliable the measure.
I calculated the reliability coefficients from
the measurement model in Figure 1. In the
equity SAs subsample, the coefficient was
0.73 for overall performance satisfaction,
meaning that 73% of the variance of this
variable is explained by the latent construct
SA performance. The coefficient was 0.34
for strategic goals fulfilment, and 0.77 for
net spillover effects. In the contractual SAs
subsample, the reliability coefficients were
0.86, 0.30 and 0.53, respectively.
The results from these tests suggest that strategic
goals fulfilment does not measure the same as
overall performance satisfaction and net spillover
effects do, and that these two are reliable measures
of the one same latent construct. As discussed more
in depth later, strategic goals fulfilment captures
the SA outcome performance, while the other two
measures capture both outcome and process
performance (Kumar and Nti, 1998). Therefore the
measurement model was changed to one that
incorporates overall performance satisfaction and
net spillover effects as indicators of a latent
construct, which is allowed to correlate freely with
strategic goals fulfilment, and with each of the
operational measures. The latter correlate with each
other, with the value of these correlations fixed and
derived from the data. Figure 2 shows the results
obtained from estimating this revised model with
both subsamples.
Convergent validity Convergent validity is examined
through three tests involving the revised
measurement model (see Table 3).
Journaol f InternationaBl usinessS tudies
73
E
2
Goal .41
fulfillment (.04)
Equity SAs sub-sample: N = 34; contractual SAs sub-sample: N = 45.
Non-parenthetical figures correspond to the model estimated with the equity SAs sub-sample; those in parenthesis, to the model
estimated with the contractual SAs sub-sample.
Figures in bold are estimated parameters; others, fixed parameters from the correlation matrices.
Significance levels obtained from non-standardized solutions:
*** significant at p< .001
** significant at p< .01
* significant at p< .05
El, E2: error terms.
Figure 2 Revised measurement model with standardized coefficients for equity and contractual SA subsamples.
Table 3 Tests of structural equation models with goodness-of-fit indices
Models X2 d.f. p X2dif d.f.a P CFI NNFI Evidence of
validity
Equity SAs subsample (N= 34)
Baseline modelb 6.54 9 0.69 1.00 1.06
(1) Convergent validity testc
Overall performance satisfaction = net
spillover effects
(2) Discriminant validity testsc
SA performance = strategic goals fulfillment
SA performance = longevity
SA performance = contractual changes
SA performance = survival
0.72 -1 <0.50 1.00 1.06 Yes
2.81 -1
0.00 -1
LDM
LDM
<0.10 1.00 1.02 No
<0.95 1.08 1.00 No
Yes
Yes
Contractual SAs subsample (N=45)
Baseline modelb
(1) Convergent validity testc
Overall performance satisfaction = net
spillover effects
(2) Discriminant validity testsc
SA performance= strategic goals fulfilment
SA performance = longevity
SA performance = contractual changes
SA performance = survival
0.56 9 0.99 - 1.00 1.29
3.06 -1 <0.10 1.00 1.20 Yes
4.95 -1
0.12 -1
LDM
LDM
<0.05 1.00 1.14 Yes
<0.95 1.00 1.29 No
Yes
Yes
aDegrees of freedom change. bEstimated coefficients shown in Figure 2. CModels tested against baseline model. CFI=
NNFI = non-normed fit index; LDM = linearly dependent model.
Test 1: Overall goodness of fit. The goodness of fit is
assessed with the x2 test, the comparative fit
index (CFI), and the non-normed fit index
(NNFI). These tests compare the hypothesized
model with the null model, where
:comparative fit index;
there is complete independence among all
the variables in the model. A model x2 that
is statistically nonsignificant rejects the null
hypothesis of complete independence, and
thus shows a good fit of the hypothesized
Journaol f InternationaBl usinessS tudies
_-l[Ikl# IIIqn% perlrman [l eC Iri a /" IiIn v
N
W Ctratonle allSanea nsrCfAn ana Afri,a Ariin
Strateqic alliance performance Africa Arino
74
relationships to the data. Magnitudes of
0.90 or greater for the CFI and the NNFI also
evidence a good fit of the model to the data
(Bentler, 1992b). The model x2 is statistically
nonsignificant in both subsamples (see
Table 3), and thus rejects the null hypothesis
of complete independence. Both the
CFI and the NNFI are above the required
level. These results show a good fit of the
measurement model to the data.
Test2 : Factorl oadingso f the observedv ariables.T here
is evidence of convergent validity if a
z-test shows all of the observed variables
measuring a construct to be significant at
least at the 0.05 level (Bentler, 1992a). The
estimated factor loadings for the indicators
of SA performance are significant below the
0.001 level in both subsamples (see Figure 2).
Test 3: Comparison of baseline model with more
restricted models. The baseline model is
compared with other models in which
factor loadings of measures of the same
construct are restricted to equal each other
(Hoskisson et al., 1993). A nonsignificant
increase in x2 provides evidence for convergent
validity (Byrne, 1994). The results
for both subsamples (see Table 3) show a
nonsignificant x2 increase in the restricted
model where the factor loadings for overall
performance satisfaction and net spillover
effects were forced to be equal.
Discriminant validity Discriminant validity is assessed
with the x2 difference test, comparing the
baseline model with a more restricted model in
which the correlation between the two constructs
under examination is constrained to equal 1.0
(Joreskog, 1971). A significantly higher x2 for the
model in which the correlation is restricted would
indicate a nonperfect correlation between the
constructs, which is evidence of discriminant
validity (Bagozzi and Phillips, 1982). This test is
done for one pair of constructs at a time.
The restricted model in which the correlation
between SA performance and strategic goals fulfilment
is set to 1 shows a nonsignificant increase in
X2 in the equity SAs subsample, but a significant
one in the contractual SAs subsample (see Table 3).
When the restriction is imposed on the correlation
between SA performance and longevity, the increase
in x2 is nonsignificant in both cases. Estimation of
the models restricting the correlation between SA
performance and contractual changes and survival,
respectively, was not feasible owing to linear
dependence problems, which showed that these
models were very poor. These results therefore
indicate that the organizational effectiveness measures
of SA performance have discriminant validity
with respect to contractual changes and survival, but
not with respect to longevity. There is evidence of
discriminant validity with respect to strategic goals
fulfilment for contractual SAs, but not for equity SAs.
One additional result in Figure 2 is worth
mentioning. The estimated covariance between
the SA performance factor and contractual changes
is negative for equity SAs, but it takes the positive
sign for contractual SAs. This is consistent with
other research (Reuer and Leiblein, 2000) that
shows that equity SAs inhibit adaptive and flexible
responses to changing events.
Discussion: towards an outcome and process
performance concept
I found that strategic goal fulfilment, on the one
hand, and overall performance satisfaction and net
spillover effects, on the other, do not measure the
one same construct. This is consistent with Parkhe's
(1993b) finding that strategic goal fulfilment was
captured by a factor different than the one
capturing other dimensions of SA performance,
which included net spillover effects and overall
performance satisfaction. I also found the two latter
to show convergent validity, meaning they are good
measures of the one same construct. One possible
explanation follows on from my earlier discussion of
the different goals captured by one or another
measure: strategic goal fulfilment captures the accomplishment
of initial goals - common and private -
whereas net spillover effect evaluates the degree of
fulfilment of a subset of private goals (whether initial
or emergent), and overall performance satisfaction
evaluates the accomplishment of common and
private, initial and emergent goals altogether.
However, new insights may be gained by shifting
the focus from the statics of outcomes to the
dynamics of process (Yan and Zeng, 1999). When
partners are asked for their assessment of overall SA
performance - and, in particular, when they are
asked to report their satisfaction with the SA - they
may evaluate not only the SA outcomes, but also
the SA process. When entering an SA, each partner
brings certain initial expectations about the outcomes
as well as about partner interactions. As the
SA evolves, outcome or process discrepancies may
affect a partner's evaluation of the SA (Kumar and
Nti, 1998), and may influence the emergence of
Journal of International Business Studies
1, Strateqic alliance performance Africa Arino
Strateqic alliance performance
new goals for the SA. A specific measure such as
strategic goals fulfilment can capture outcome
performance, but the other measures may capture
assessments of both SA outcome performance (the
degree of accomplishment of the partners' goals, be
these common or private, initial or emergent) and
process performance (how far their pattern of
interactions is acceptable to the partners). This is
in accordance with the fact that, since net spillover
effects and overall performance satisfaction are
single-item measures whereas strategic goal fulfilment
is a composite one, respondents may reflect
their general feeling about the SA in the former
measures, but not in the latter. This underscores
how important it is to develop new measures that
capture process performance and unbundle it from
outcome performance.
Additional evidence shows that strategic goal
fulfilment displays discriminant validity with
respect to the other organizational effectiveness
measures for contractual SAs, but not for equity
SAs. This suggests that outcome and process
discrepancies are easier to unbundle in contractual
SAs than in equity SAs. Also, it may indicate that
partners' disagreement regarding measures to assess
SA progress (Beamish and Delios, 1997) is larger in
contractual SAs than in equity SAs. As contractual
SAs are more flexible and adaptable than equity SAs
(Reuer and Leiblein, 2000), their performance
measures would require more frequent readjustments,
which may not actually occur.
This leads us to a second issue revisited by Yan
and Zeng (1999): the relationship between SA
performance and stability. I found that contractual
renegotiation is negatively related to equity SA
performance, and positively related to contractual
SA performance. Equity SAs tend to involve higher
asset specificity than contractual SAs (Williamson,
1991). Firms tend to selectively renegotiate an SA in
which they have made transaction-specific investments,
as it is worth the costs of adjustment (Reuer
and Arifio, 2002). Such a change would be the
response to a delay problem, especially when one
firm has a lower bargaining power than its partner.
This is true for local companies that form an equity
joint venture with a foreign company: the local
partner loses bargaining power as the foreign
partner acquires knowledge of the local market
(Inkpen and Beamish, 1997). Unless new goals
emerge that make the foreign company dependent
on the local one, the foreign company may attempt
to renegotiate the contract to its own benefit. This
type of interaction influences the relationship
negatively, leading to greater instability. By contrast,
as contractual SAs have a smaller hold up
potential than equity SAs, their renegotiation is
more likely to be a mutually agreed response to
changing circumstances, as the potential for holdup
is smaller in these SAs than in equity ones.
Undergoing a successful renegotiation strengthens
the relationship, which improves a partner's overall
assessment of SA performance. This helps us to
understand instability as a dependent rather than
an independent variable (Yan and Zeng, 1999), an
issue that deserves further attention.
The tests show that contractual changes and
survival display discriminant validity with respect
to measures of the SA performance construct.
However, this result does not hold for longevity.
One explanation for this is that, in practice, only
successful alliances may be long-lived; short-lived
SAs are either failed SAs that have been dissolved or
SAs still in their first years of operation that have
not yet had enough time to be considered as
successful. This explanation is consistent with the
fact that survival is not a good indicator of the SA
performance construct: there may be SAs in
operation not yet considered successful because
they have not passed the test of time. Evidence
suggests that most alliances fail in their first 2 years
of operation (Osborn et al., 1997), which agrees
with my interpretation. However, we cannot overlook
the bias towards nonsurvivors in the sample.
Alternatively, survival may not be a good measure
of SA performance because it is a goal of the SA
management team (Schaan and Beamish, 1988),
whereas the other performance indicators take the
partner company's perspective.
Although correlations between operational measures
and strategic goals fulfilment are not significant
in either of the subsamples, their indirect
relationship through the SA performance construct
is significant in most cases (see Figure 2). We can go
back to the distinction between outcome performance
and process performance to help us
interpret this. Although there may not be any
direct link between outcome performance and SA
longevity, contractual changes and survival, the
picture changes when process performance is
added. Outcome performance influences, and is
influenced by, process performance; in turn this
interaction affects, and is affected by, SA stability.
When the cause of a poor outcome performance is
ambiguous, a firm will attribute it either to the
partner's behaviour or to external causes, depending
on how satisfied it is with the pattern of
Journal of International Business Studies
Africa Arino
75
Stratenir allianrce DerFrmanre
interaction with the partner (Arifio and Doz, 2000).
For instance, given a poor outcome performance,
the decision to keep the SA alive will be influenced
by process performance: if partners are satisfied
with their pattern of interaction, they may decide
to maintain the relationship, waiting for better
times when external circumstances allow goal
fulfilment; by contrast, if partners are unsatisfied
with their pattern of interaction the chances are
that they will terminate the SA. These views are
consistent with research on SA evolutionary
processes (Ring and Van de Ven, 1994; Doz, 1996;
Arifio and de la Torre, 1998; Kumar and Nti, 1998;
Lin and Germain, 1998; Yan, 1998; Arifio et al.,
2001). Having both partners' views would have
allowed us to validate these points.
This discussion suggests that we need to broaden
our concept of SA performance to encompass both
outcome and process performance. I propose the
following definition:
SA performance refers to the degree of accomplishment of
the partners' goals, be these common or private, initial or
emergent (outcome performance), and the extent to which
their pattern of interactions is acceptable to the partners
(process performance).
Conclusions
The results of this study shed light on the construct
validity of existing measures of the performance of
equity SAs and contractual SAs. Identifying two
important aspects of SA performance - outcome
and process performance - broadens this concept,
and poses a challenge for researchers to develop
new measures that help distil these aspects. In fact,
one limitation of this study is that I was not able to
count on such a measure, which would have served
to test the validity of the distinction between
outcome and process performance.
This study takes the perspective of only one
partner - a limitation common to most SA research.
Also, it treats goals of the SA management team as a
constraint on those of the partners. However, as the
goals of these groups differ, they are also likely to
References
Anderson, E. (1990) 'Two firms, one frontier: on assessing
joint venture performance', Sloan Management Review 32: 19-
30.
Arino, A. (1995) 'Inter-firm collaborative ventures: performance
and cooperative behavior', Ph.D. thesis, The Anderson School,
University of California, Los Angeles.
Arino, A. and Doz, Y. (2000) 'Rescuing troubled alliances...before
it's too late', European Management Journal 18(2): 173-
182.
dissent in their assessments of SA performance: this
is an issue in SAs that result in a separate
organizational entity (Anderson, 1990). Geringer
and Hebert (1991) found significantly positive
correlations between the assessment of partners
and that of the SA management team. These
correlations were stronger in the more general
measures, and less strong in the more specific
dimensions. The further removed from SA operations
evaluators are, the more likely they are to
focus on broad performance indicators.
Replication with a different sample, ideally a
larger one, is necessary if the results are to be
generalized, as they may be bound to the specific
sample used: equity and contractual SAs with only
two partners, and one of them operating in Spain.
Future research should look at differences in the
validity of SA performance measures using other
criteria to classify SAs, such as the purpose of the
SA, or the number of partners involved. It may also
be worth the effort of developing multiple-item
scales for concepts that are currently measured by
single items. We are also in need of measures of SA
process performance.
Despite its limitations, this study provides useful
guidelines for researchers engaged in understanding
SA performance by offering suggestions regarding
the construct validity of existing measures of the
performance of equity and contractual SAs. Its results
help us to interpret past research on the subject, and
will assist future researchers in their selection of
measures of SA performance. The distinction traced
between outcome and process performance has
important implications for practice, as SA managers
may have to be evaluated on the basis not only of
outcome but also of process performance.
Acknowledgments
I thank Peter Bentler, Jose de la Torre, Dick Goodman,
Elaine Mosakowski, Paul Olk and Bill Ouchi for their
helpful comments on earlier drafts of this paper.
Funding for this research was provided by IESE
BusinessS chool,U niversitoyf Navarra(B arcelonaS, pain).
Arino, A. and de la Torre, J. (1998) 'Learning from failure:
towards an evolutionary model of collaborative ventures',
Organization Science 9(3): 306-325.
Arifo, A., de la Torre, J. and Ring, P.S. (2001) 'Relational quality:
managing trust in corporate alliances', CaliforniaM anagement
Review 44(1): 109-131.
Bagozzi, R.P. and Phillips, L.W. (1982) 'Representing and testing
organizational theories: a holistic construal', Administrative
Science Quarterly2 7: 459-489.
Journaol f InternationaBl usinessS tudies
*^
76
Africa Arino JLBMLIU%11%. 95INIMINSOU /-%I I 1%-C
?
/,%I 11
II
77
Beamish, P.W. (1988) Multinationall oint Venturesi n Developing
Countries, Routledge: New York.
Beamish, P.B. and Delios, A. (1997) 'Improving joint venture
performance through congruent measures of success', in P.W.
Beamish and J.P. Killing( eds.) CooperativeS trategies:E uropean
PerspectivesN, ew Lexington Press:S an Francisco.p p: 103-127.
Bentler, P.M. (1992a) EQS:S tructuraEl quationsP rogramM anual,
BMDP Statistical Software: Los Angeles.
Bentler, P.M. (1992b) 'On the fit of models to covariances and
methodology to the Bulletin', Psychological Bulletin 112: 400-
404.
Bollen, K.A. (1989) Structural Equations with Latent Variables,
Wiley: New York.
Byrne, B.M. (1994) Structural Equation Modeling with EQS and
EQS/WindowsS, age: Thousand Oaks, CA.
Contractor, F.J. and Lorange, P. (1988) 'Why should firms
cooperate? The strategy and economic basis for cooperative
ventures', in F.J. Contractor and P. Lorange (eds.) Cooperative
Strategies in International Business, Lexington Books:
Lexington, MA. pp: 3-30.
Dillman, D.A. (1978) Mail and Telephone Surveys: The Total
Design Method, Wiley: New York.
Doz, Y.L. (1996) 'The evolution of cooperation in strategic
alliances: initial conditions or learning processes? Strategic
Management Journal 17(Special summer issue): 55-84.
Dussauge, P., Garrette, B. and Mitchell, W. (2000) 'Learning
from competing partners: outcomes and durations of scale
and link alliances in Europe, North America and Asia', Strategic
Management Journal 21(2): 99-126.
Gawande, K. and Wheeler, T. (1999) 'Measures of effectiveness
for governmental organizations', Management Science 45(1):
42-58.
Geringer, J.M. and Hebert, L. (1991) 'Measuring performance of
international joint ventures', Journal of International Business
Studies 22(2): 249-264.
Gomes-Casseres, B. (1987) 'Joint venture instability: is it a
problem? Columbia Journal of World Business 22(2): 97-107.
Gulati, R. (1995) 'Does familiarity breed trust? The implications
of repeated ties for contractual choice in alliances', Academy of
Management Journal 38: 85-112.
Hamel, G and Prahalad, C.K. (1989) 'Strategic intent', Haward
Business Review 67(3): 63-76.
Harman, H.H. (1967) Modern Factor Analysis, University of
Chicago Press: Chicago, IL.
Hoskisson, R.E., Hitt, M.A., Johnson, R.A. and Moesel, D.D.
(1993) 'Construct validity of an objective (entropy) categorical
measure of diversification strategy', Strategic Management
Journal 14(3): 215-235.
Hu, L.-T., Bentler, P.M. and Kano, Y. (1992) 'Can test statistics in
covariance structure analysis be trusted? Psychological Bulletin
112(2): 351-362.
Inkpen, A.C. and Beamish, P.W. (1997) 'Knowledge, bargaining
power, and the instability of international joint ventures',
Academy of Management Review 22(1): 1 77-202.
Joreskog, K.G. (1971) 'Simultaneous factor analysis in several
populations', Psychometrika3 6: 409-426.
Kogut, B. (1991) 'Joint ventures and the option to expand and
acquire', Management Science 37(1): 19-33.
Koza, M.P. and Lewin, A.Y. (1998) 'The co-evolution of strategic
alliances', Organization Science 9(3): 255-264.
Kumar, R. and Nti, K.O. (1998) 'Differential learning and
interaction in alliance dynamics: a process and outcome
discrepancy model', Organization Science 9(3): 356-367.
Lin, X. and Germain, R. (1998) 'Sustaining satisfactory joint
venture relationships: the role of conflict resolution strategy',
Journal of International Business Studies 29(1): 179-198.
Nohria, N. and Garcia-Pont, C. (1991) 'Global strategic linkages
and industry structure', Strategic Management lournal 12(Summer
special issue): 105-124.
Osborn, R.N., Denekamp, J.G. and Baughn, C.C. (1997)
'Assessing the durability of US-Japanese alliances', Paper
presented at the Academy of Management Meetings: Boston,
MA.
Parkhe, A. (1993a) 'Partner nationality and the structureperformance
relationship in strategic alliances', Organization
Science 4(2): 301-324.
Parkhe, A. (1993b) 'The structuring of strategic alliances: a
game-theoretic and transaction-cost examination of interfirm
cooperation', Academy of Management Journal 36(4): 794-829.
Podsakoff, P.M. and Organ, D.W. (1986) 'Self-reports in
organizational research: problems and prospects', Journal of
Management 12: 531-544.
Reuer, j.J. and Arino, A. (2002) 'Contractual renegotiations in
strategic alliances', Journal of Management 28(1): 47-68.
Reuer, J.J. and Leiblein, M.J. (2000) 'Downside risk implications
of multinationality in international joint ventures', Academy of
Management Journal 43(2): 203-214.
Ring, P.S. and Van de Ven, A.H. (1994) 'Developmental
processes of cooperative interorganizational relationships',
Academy of Management Review 19(1): 90-118.
Salancik, G.R. and Pfeffer, J. (1977) 'An examination of needsatisfaction
models of job attitudes', Administrative Science
Quarterly2 2: 427-456.
Schaan, j.-L. and Beamish, P.W. (1988) 'Joint venture general
managers in LDCs', in F.J. Contractor and P. Lorange (eds.)
Cooperative Strategies in International Business, Lexington
Books: Lexington, MA. pp: 279-299.
Schwab, D.P. (1980) 'Construct validity in organizational
behavior', Researchi n OrganizationalB ehaviour2 : 3-43.
Venkatraman, N. and Ramanujam, V. (1986) 'Measurement of
business performance in strategy research: a comparison of
approaches', Academy of Management Review 11(4): 801-814.
Williamson, O.E. (1991) 'Comparative economic organization:
the analysis of discrete structural alternatives', Administrative
Science Quarterly3 6: 269-296.
Yan, A. (1998) 'Structural stability and reconfiguration of
international joint ventures', Journal of International Business
Studies 29(4): 773-796.
Yan, A. and Zeng, M. (1999) 'International joint venture
instability: a critique of previous research, a reconceptualization,
and directions for future research', Journal of International
Business Studies 30(2): 397-414.
Appendix: Questionnaire items
(1) Collaborative ventures can be aimed at different strategic goals. How would you describe the
importance for YOUR FIRM of each of the following strategic goals when the venture agreement
WAS SIGNED?
(a) Reducing costs/obtaining
scale economies
Minimal
1 2
Medium
3
Vital NA
4 5 6
journalo f InternationaBl usinessS tudies
Stra...t eanloi, ua,l. liancae-l-,, ~ nerr -f-n, rv manlcu er .= /A,, frI,iIrL a Li AtArIIi inIoL)
(b) Gaining access to a market
in the same industry
(c) Gaining access to a market in
another industry
(d) Developing new technologies
(e) Blocking the competition
(f) Meeting government
requirements
(g) Developing new skills
(h) Reducing risks
(i) Other (specify)
1 2
1 2
1
1
1
1
1
1
2
2
2
2
2
2
3
3
3
3
3
3
3
3
4 5 6
4 5 6
4 5 6
4
4
4
4
5
5
5
5
6
6
6
6
4 5 6
(2) Overall, to what extent do you think YOUR FIRM is satisfied with the global results of this venture?
Very unsatisfied
Unsatisfied
Somewhat satisfied
Satisfied
Very satisfied
(3) How do you think each of YOUR FIRM'S strategic goals in relation to this venture has been met?
Very poorly Normal
(a) Reducing costs/obtaining
scale economies
(b) Gaining access to a market
in the same industry
(c) Gaining access to a market
in another industry
(d) Developing new
technologies
(e) Blocking the competition
(f) Meeting government
requirements
(g) Developing new skills
(h) Reducing risks
(i) Other (specify)
1 2
1 2
1
1
1
1
2
2
2
2
1
1
1
2
2
2
3
3
3
3
3
3
3
3
3
4 5 6
4 5 6
4 5 6
4 5 6
4
4
4
4
4
5
S
5
S
5
6
6
6
6
6
(4) Many collaborative ventures result in SIDE EFFECTS for their parent firms. For example, there are
POSITIVE side effects when the skills that are being developed through the venture can be applied
profitably to other operations within the company. There are NEGATIVE side effects if the
collaboration has damaging repercussions on other activities in the company. In this venture,
the NET side effects for YOUR FIRM have been:
Strongly negative 1
Negative 2
Zero 3
Positive 4
Very positive 5
(5) When was the venture agreement signed? Year: 19
(6) Is the venture still operating?
Journal of International Business Studies
Very well NA
S.JLt9r1IaUMteILnUNicNU MILC alliance nJ eCrflfVofflrllMm ance 1A6IIfLrCi ca tAA rIi nl o
Strateqic alliance performance AfricAa rino
79
No 0
Yes 1 (go to question 8)
(7) When did the venture end? Year: 19
(8) When the agreement WAS SIGNED, did it include equity participation?
No 0 (go to question 10)
Yes 1
(9) Have there been any changes in the ownership structure?
No 0
Yes 1 How many?
(10) Was the initial contract renegotiated during the course of the venture?
No 0
Yes 1 Explain it briefly:
Accepted by Tom Brewer, outgoing Editor, received 11 September 2000. This paper has been with the author for 4 revisions.
Journal of International Business Studies