The University of Manchester Research
The use of partial least squares path modeling in
international marketing
DOI:
10.1108/S1474-7979(2009)0000020014
Link to publication record in Manchester Research Explorer
Citation for published version (APA):
Henseler, J., Ringle, C. M., & Sinkovics, R. R. (2009). The use of partial least squares path modeling in
international marketing. In Advances in International Marketing|Adv. Int. Mark. (Vol. 20, pp. 277-319). (Advances in
International Marketing). Emerald Publishing Limited. https://doi.org/10.1108/S1474-7979(2009)0000020014
Published in:
Advances in International Marketing|Adv. Int. Mark.
Citing this paper
Please note that where the full-text provided on Manchester Research Explorer is the Author Accepted Manuscript
or Proof version this may differ from the final Published version. If citing, it is advised that you check and use the
publisher's definitive version.
General rights
Copyright and moral rights for the publications made accessible in the Research Explorer are retained by the
authors and/or other copyright owners and it is a condition of accessing publications that users recognise and
abide by the legal requirements associated with these rights.
Takedown policy
If you believe that this document breaches copyright please refer to the University of Manchester’s Takedown
Procedures [http://man.ac.uk/04Y6Bo] or contact uml.scholarlycommunications@manchester.ac.uk providing
relevant details, so we can investigate your claim.
Download date:10. Aug. 2022
THE USE OF PARTIAL LEAST
SQUARES PATH MODELING IN
INTERNATIONAL MARKETING
Jo
¨
rg Henseler, Christian M. Ringle and
Rudolf R. Sinkovics
The advent of structural equation modeling (SEM) with latent variables has
changed the nature of research in international marketing and management.
Researchers acknowledge the possibilities of distinguishing between
measurement and structural models and explicitly taking measurement
error into account. As Gefen, Straub, and Boudreau (2000, p. 6) point out,
‘‘SEM has become de rigueur in validating instruments and testing linkages
between constructs.’’ They furthermore distinguish between two families of
SEM techniques: covariance-based techniques, as represented by LISREL,
and variance-based techniques, of which partial least squares (PLS) path
modeling is the most prominent representative.
PLS has been used by a growing number of researchers from various
disciplines such as strategic management (e.g., Hulland, 1999), management
information systems (e.g., Dibbern, Goles, Hirschheim, & Jayatilaka, 2004),
e-business (e.g., Pavlou & Chai, 2002), organizational behavior (e.g.,
Higgins, Duxbury, & Irving, 1992), marketing (e.g., Reinartz, Krafft, &
Hoyer, 2004), and co nsumer behavior (e.g., Fornell & Robinson, 1983).
Since 1987, for instance, more than 20 studies using PLS have been
published in five top-tier marketing journals (Eggert, 2007) – the majority in
New Challenges to International Marketing
Advances in International Marketing, Volume 20, 277–319
Copyright r 2009 by Emerald Group Publishing Limited
All rights of reproduction in any form reserved
ISSN: 1474-7979/doi:10.1108/S1474-7979(2009)0000020014
277
the last six years. PLS is the method of choice for success factor studies in
marketing (Albers, 2009) and for estimating the various national customer
satisfaction index models (e.g., Fornell, 1992). The PLS methodology has
also achieved an increasingly popular role in empirical research in
international marketing, which may represent an appreciation of distinctive
methodological features of PLS. As of March 2008, more than 30 articles on
international marketing using PLS were published in double-blind reviewed
journals. However, these publications show a rather large variability in the
way how PLS is applied and how its outcomes are reported. In many
instances, the rationale for choosing PLS among a possible set of alternative
analytical techniques is not made explicit.
Although several articles offer guidance for the use of covaria nce-based
structural eq uation modeling (CBSEM) in international marketing (e.g.,
Steenkamp & Baumgartner, 1998; Malhotra, 2001; Iacobucci, Grisaffe,
Duhachek, & Marcati, 2003), there are no similar subject-specific guidelines
for the use of PLS. While there are general recommendations for the use of
PLS (e.g., Hulland, 1999), the specific requirements and typical research
problems of international marketing have not been addressed yet. Our main
aims are to shed light on PLS path modeling as an SEM techni que, to reveal
the strengths and weakne sses of PLS in general, and to deliver guidelines for
its use in international marketing in particular.
1. APPLICATIONS OF PLS PATH MODELING
IN INTERNATIONAL MARKETING
In order to determine the status quo of PLS path modeling in international
marketing research, we conducted an exhaustive literature review. An
evaluation of double-blind reviewed journals through important academic
publishing databases (e.g., ABI/Inform , Elsevier ScienceDirect, Emerald
Insight, Google Scholar, PsycINFO, Swetswise) revealed that more than
30 academic articles in the domain of international marketing (in a broad
sense) used PLS path modeling as means of statistical analysis. We assessed
what the main motivation for the use of PLS was in respect of each article.
Moreover, we checked for applications of PLS in combination with one or
more additional methods, and whether the main reason for conducting any
additional method(s) was mentioned.
Table 1 lists all the identified academic articles that use PLS as a method
of analysis in the context of international marketing. Green and Ryans
JO
¨
RG HENSELER ET AL.278
Table 1. Studies Using PLS Path Modeling in International Marketing Research.
Study Motivation for Using PLS Path Modeling Additional Analysis Reason for
Additional Analysis
Acedo and Jones (2007, JWB) ‘‘The PLS technique is justified where theory is insufficiently grounded
and the variables or measures do not conform to a rigorously
specified measurement model, or fit a certain distribution’’ (p. 242)
t-test Group comparison
Ainuddin, Beamish, Hulland,
and Rouse (2007, JWB)
‘‘Use of PLS is especially suited to exploratory studies such as this,
where the measures [...]arenewandtherelationships [ . . . ] have not
been previously tested’’ (p. 56)
N/A N/A
Alpert, Kamins, Sakano,
Onzo, and Graham
(2001, IMR)
‘‘Formative indicators can only be analyzed using partial least squares
(PLS), and not by using the more common structural equation
technique of LISREL’’ (p. 177–178)
Multiple regression
and Chow test
Group comparison
Birkinshaw, Morrison, and
Hulland (1995, SMJ)
‘‘PLS is most appropriate when sample sizes are small, when
assumptions of multivariate normality and interval scaled data
cannot be made, and when the researcher is primarily concerned with
prediction of the dependent variable’’ (pp. 646–647)
Multiple regression Analysis of subgroups
Calantone, Graham, and
Mintu-Wimsatt
(1998, IJRM)
‘‘The PLS parameter estimates better reveal the strength and direction
(i.e., positive vs. negative) of the relationships among variables
compared to correlation coefficients’’ (p. 28), ‘‘PLS avoids
parameters estimation biases common in regression analysis’’ (p. 28)
LISREL Path significances
Festge and Schwaiger
(2007, AIM)
‘‘The researcher’s focus is placed on the explanation of an endogenous
construct’’ (p. 192)
Logistic regression Verify model for
binary variables
Gerpott and Jakopin
(2005, SBR)
Not explicitly mentioned N/A N/A
Graham, Mintu, and Rodgers
(1994, Mgmt.Sc.)
‘‘Parameters can be estimated independent of sample size’’ (p. 79),
‘‘PLS avoids parameter estimation biases inherent in regression
analysis’’ (p. 80), ‘‘PLS provides the most flexibility regarding
measurement of the constructs’’ (p. 80)
LISREL In order to test path
significances
Green and Ryans (1990, JPIM) ‘‘Given that the purpose of this study is to predict [...], PLS has thus
been chosen as the structural equation modeling approach’’ (p. 53),
‘‘[PLS] is more robust with small sample sizes’’ (p. 53), ‘‘the data do
not have to be multivariate normal because of the fixed point
estimation’’ (p. 53)
N/A N/A
Holzmu
¨
ller and Sto
¨
ttinger
(1996, JIM)
Not explicitly mentioned N/A N/A
Partial Least Squares Path Modeling in International Marketing 279
Table 1. (Continued )
Study Motivation for Using PLS Path Modeling Additional Analysis Reason for
Additional Analysis
Holzmu
¨
ller and Kasper
(1991, MIR)
‘‘Fewer restrictive assumptions,’’ ‘‘ratio between sample size and the
number of parameters to be estimated,’’ ‘‘exploratory intention’’
(p. 58)
N/A N/A
Inkpen and Birkenshaw
(1994, IBR)
‘‘All relationships are modeled simultaneously, eliminating concerns
about multicollinearity’’ (p. 208)
N/A N/A
Johansson and Yip
(1994, SMJ)
‘‘Less stringent assumptions about the randomness of the sample and
the normality of the distribution of variables’’ (p. 587), ‘‘smaller
sample sizes, as each causal subsystem sequence of paths is estimated
separately’’ (p. 587)
N/A N/A
Johnson, Herrmann, and
Gustafsson (2002, JEP)
Not explicitly mentioned N/A N/A
Julien and Ramangalahy
(2003, ETP)
‘‘PLS is known to be particularly advantageous in the initial
development and assessment phase of theory building,’’ ‘‘the PLS
method is [...] more robust since its does not require either a large
sample or normally distributed data’’ (p. 233)
N/A N/A
Lee, Yang, and Graham
(2006, JIBS)
‘‘PLS [...]is more appropriate for the exploratory nature of [a] study,’’
‘‘[PLS] allows for formative indicators [ . . . ] and dichotomous
constructs’’ (p. 632)
t-test and ANOVA Group comparison
Lee (2001, JBR) ‘‘PLS [ . . . ] can accommodate a small sample size’’ (p. 153) Confirmatory factor
analysis
Examine
unidimensionality
Lee (2000, EJM) ‘‘PLS avoids many of the restrictive assumptions imposed by other
causal models that involve latent variables such as LISREL’’, ‘‘PLS
provides measurement assessment’’, ‘‘A jack-knife procedure [ . . . ]
generates an approximate t-statistic. This overcomes the
disadvantage of the lack of formal significance tests for parameters
resulting from non-parametric methods’’, ‘‘PLS enables the explicit
estimation of the multiple item construct, which affords a
comparison of [groups] at the construct level’’ (p. 196)
t-test and ANOVA Group comparison
Mahmood, Bagchi, and Ford
(2004, IJEC)
‘‘The PLS technique imposes minimal demand on measurement scales,
sample sizes, and residual distributions. [It] is often used to test and
validate exploratory models’’ (p. 20)
AMOS Fit statistics
JO
¨
RG HENSELER ET AL.280
