The current issue and full text archive of this journal is available on Emerald Insight at:

www.emeraldinsight.com/0309-0566.htm

EJM
53,4 Rethinking some of the rethinking
of partial least squares
Joseph F. Hair
University of South Alabama, Mobile, Alabama, USA
566
Marko Sarstedt
Received 1 October 2018 Otto-von-Guericke-University Magdeburg, Magdeburg, Germany and
Revised 1 October 2018
Accepted 1 October 2018 Monash University of Malaysia, Malaysia, and
Christian M. Ringle
Hamburg University of Technology (TUHH), Hamburg,
Germany and University of Waikato, New Zealand

Abstract
Purpose – Partial least squares structural equation modeling (PLS-SEM) is an important statistical
technique in the toolbox of methods that researchers in marketing and other social sciences disciplines
frequently use in their empirical analyses. The purpose of this paper is to shed light on several misconceptions
that have emerged as a result of the proposed “new guidelines” for PLS-SEM. The authors discuss various
aspects related to current debates on when or when not to use PLS-SEM, and which model evaluation metrics
to apply. In addition, this paper summarizes several important methodological extensions of PLS-SEM
researchers can use to improve the quality of their analyses, results and findings.
Design/methodology/approach – The paper merges literature from various disciplines, including
marketing, strategic management, information systems, accounting and statistics, to present a state-of-the-art
review of PLS-SEM. Based on these findings, the paper offers a point of orientation on how to consider and
apply these latest developments when executing or assessing PLS-SEM-based research.
Findings – This paper offers guidance regarding situations that favor the use of PLS-SEM and discusses
the need to consider certain model evaluation metrics. It also summarizes how to deal with endogeneity in
PLS-SEM, and critically comments on the recent proposal to adjust PLS-SEM estimates to mimic common
factor models that are the foundation of covariance-based SEM. Finally, this paper opposes characterizing
common concepts and practices of PLS-SEM as “out-of-date” without providing well-substantiated
alternatives and solutions.
Research limitations/implications – The paper paves the way for future discussions and suggests a
way forward to reach consensus regarding situations that favor PLS-SEM use and its application.
Practical implications – This paper offers guidance on how to consider the latest methodological
developments when executing or assessing PLS-SEM-based research.
Originality/value – This paper complements recently proposed “new guidelines” with the aim of offering
a counter perspective on some strong claims made in the latest literature on PLS-SEM. It also clarifies some
misconceptions regarding the application of PLS-SEM.

Keywords Guidelines, Partial least squares, PLS-SEM, Structural equation modeling

Paper type Research paper

European Journal of Marketing

Vol. 53 No. 4, 2019
pp. 566-584
© Emerald Publishing Limited Even though this research does not use the statistical software SmartPLS (www.smartpls.com),
0309-0566
DOI 10.1108/EJM-10-2018-0665 Ringle acknowledges a financial interest in SmartPLS.
Partial least squares structural equation modeling (PLS-SEM) belongs to the toolbox of Rethinking of
methods that researchers in marketing and other disciplines frequently use in their partial least
empirical analyses (Hair et al., 2012; Rasoolimanesh and Ali, 2018; Nitzl, 2016). Even though
widely appreciated, PLS-SEM has been in the center of a lively scholarly debate for several
squares
years (Latan and Noonan, 2017; Richter et al., 2016; Rigdon et al., 2017b), which accompanied
and fostered the method’s continuous development and extension. Contributions pointing
out both strengths and limitations of the method have appeared at a rapid pace, with some
early articles incorrectly concluding “there is no use for PLS whatsoever” (Antonakis et al., 567
2010, p. 1103), while more recent works conclude that researchers should “feel the love for
PLS” (Petter, 2018, p. 12). The underlying discussions certainly have impacted
methodological research on PLS-SEM (Khan et al., 2019), and methods scholars were quick
to tackle some of the criticisms by advancing the method, while debunking other issues as
being unsubstantiated or grossly exaggerated (Henseler et al., 2014; Sarstedt et al., 2016).
Unfortunately, a recent article with alleged “new guidelines” was published (Henseler et al.,
2016) that was incorrectly positioned as the latest thinking in PLS-SEM[1].
While the concept of “new guidelines” sounds intuitively appealing, a careful
consideration of recent research shows that several of these guidelines have led to
substantial controversy, and thereby prevented unambiguous agreement on how to
deal with emerging topics such as goodness-of-fit testing, the nature of composite
models, and several more. At the same time, some users of PLS-SEM seem to be
unaware of recent developments which have clearly extended and improved the toolbox
for analyzing PLS path models (Latan and Noonan, 2017; Hair et al., 2018a, 2018b;
Matthews et al., 2018).
Taking up Henseler et al.’s (2017b) approach, this research note highlights some of these
recent developments and offers a counterpoint to emerging calls to change PLS-SEM use,
several of which we argue are misguided and add little or no value. Our overarching aim of this
research note is extending prior efforts (Henseler, 2017b) to enable researchers, reviewers, and
editors to judge recent controversies on their own (Hair et al., 2013). Most importantly, this note
offers guidance on how to consider the latest developments when executing or assessing PLS-
SEM-based research as several review studies across disciplines have called for (Hair et al.,
2012; Ringle et al., 2019; Nitzl, 2016; Kaufmann and Gaeckler, 2015; do Valle and Assaker, 2016).
Our discussion considers ten topics that have recently attracted considerable debate in
scholarly articles or have provoked discussions in forums such as http://forum.smartpls.com/.
Addressing them in the proposed order does not imply a hierarchy, but follows the steps for
systematically applying PLS-SEM (Hair et al., 2017a; Garson, 2016; Ramayah et al., 2016) – all
perspectives have the same raison d’être. Specifically, we highlight the following points, which
relate to the reasons for using PLS-SEM, model estimation specification, and evaluation (see
also Henseler, 2017b):

PLS-SEM’s small sample size capabilities are real.

Models with formatively specified constructs should be analyzed with PLS-SEM.

PLS-SEM is superior to regression analysis when assessing mediation.

Researchers should use the two-stage approach to moderator analysis,

PLSc-SEM adds very little to existing knowledge of SEM.

Researchers should use the HTMT criterion for discriminant validity testing.

Researchers do not necessarily need to assess a PLS path model’s goodness-of-fit.

PLS-SEM results should be assessed for both in-sample explanatory power and out-
of-sample predictive power.
EJM
Examining unobserved heterogeneity should be part of a PLS-SEM analysis.
53,4
Endogeneity can be assessed with PLS-SEM.

Partial least squares structural equation modeling’s small sample size

capabilities are real
568 Researchers often justify their choice of PLS-SEM on the grounds of a small sample size
(Ringle et al., 2019; Ali et al., 2018; Nitzl, 2016; Hair et al., 2012; Latan, 2018). This argument
has attracted considerable criticism, with several studies (Goodhue et al., 2012; Rönkkö and
Evermann, 2013; Marcoulides and Saunders, 2006) attempting to show that PLS-SEM’s
capabilities are indeed less favorable than implied in seminal articles on the method (Chin
and Newsted, 1999; Hair et al., 2011). As a result of this criticism, some reviewers have
begun claiming that PLS-SEM is unsuitable for handling small sample size research designs
(Lee, 2017). As it is often the case in research, the truth may lie somewhere in between. While
it is certainly correct that PLS-SEM offers no wildcard for estimating complex models with
small sample sizes (Hair et al., 2013; Henseler, 2017b; Rigdon, 2016), it is also true that PLS-
SEM has been shown to yield high levels of statistical power in composite model
populations under these conditions (Hair et al., 2017b). Similarly, when used to estimate data
from factor model populations, PLS-SEM has been shown to achieve favorable levels of
statistical power while controlling for Type I errors at the expense of parameter accuracy
(Reinartz et al., 2009; Sarstedt et al., 2016).
Apart from achieving higher levels of statistical power, PLS-SEM also has favorable
convergence behavior compared to covariance-based SEM when sample sizes are small
(Reinartz et al., 2009; Henseler et al., 2014). Compared to other SEM estimators, PLS certainly
has much better capabilities, but “it will be the nature of the population that justifies the
small sample size, and not the small sample size that justifies the choice of PLS” (Rigdon,
2016, p. 600). For example, many business-to-business populations are limited to a total of
100 or so respondents. Clearly in these situations, PLS is a suitable SEM method capable of
achieving convergence. Similarly, as Petter (2018, p. 11) notes: “With small sample sizes, the
problem is not PLS, but rather how researchers are using PLS”. We concur.

Models with formatively specified constructs should be analyzed with partial

least squares structural equation modeling
Many scholars claim that covariance-based SEM can be executed with formatively specified
measurement models (Temme et al., 2014; Diamantopoulos et al., 2008). To ensure model
identification with formative measurement models when using covariance-based SEM,
researchers must impose specific constraints on the model parameters – regardless of
whether the formative measurement model comprises causal indicators (Diamantopoulos
and Riefler, 2011) or composite indicators (Grace and Bollen, 2008). But, as Hair et al. (2012,
p. 420) note, “these constraints often contradict theoretical considerations, and the question
arises whether model design should guide theory or vice versa”.
A proposed alternative to include formative measurement models with causal indicators
in covariance-based SEM is to specify a multiple indicators and multiple causes (MIMIC)
model (Bollen, 1989; Jöreskog and Goldberger, 1975). By simultaneously including formative
and reflective indicators, the MIMIC model overcomes identification problems (Posey et al.,
2015). However, the causal indicators in a MIMIC model underrepresent the variance in the
construct, which has adverse consequences for the validity of the results. As Lee and
Cadogan (2013, p. 243) note, “researchers should not be misled into thinking that achieving
statistical identification enables one to obtain information about the variance of a formative Rethinking of
latent variable”. partial least
PLS-SEM on the other hand does not permit explicit modeling of a construct’s error term
measured with causal indicators. Instead, the method treats all indicators as composite
squares
indicators that jointly define the construct under consideration (McDonald, 1996) and
requires convergent validity (Cheah et al., 2018). The construct may thus be viewed as an
artifact which can be understood as a human-made object (Henseler, 2017a). Examples of
such artifacts in marketing include concepts such as the retail price index or the marketing 569
mix. Others argue that composite indicators can also be used to measure attitudes,
perceptions, and behavioral intentions (Sarstedt et al., 2016; Rossiter, 2011; Rossiter, 2016).
That is, when composite indicators exhibit conceptual unity, they can be used to measure
concepts in accordance with a clear theoretical definition. In specifying measurement
models, researchers have to be aware that “omitting a composite indicator is equivalent to
omitting a part of the composite” (Aguirre-Urreta et al., 2016, p. 94).
Just as in covariance-based SEM, formative measures in PLS-SEM derive their meaning
to a certain degree from their position in the model (Aguirre-Urreta et al., 2016). That is, the
indicator weight estimates depend on the context in which the construct is embedded, as
defined by its predecessor and successor constructs in the PLS path model. PLS-SEM,
however, offers a considerably higher degree of flexibility in terms of model specification.
Therefore, if construct measurement is specified formatively, the most appropriate method
is PLS-SEM, and not covariance-based SEM. Similarly, if the PLS path model includes
higher-order models, in which the lower-order components jointly form the higher-order
component (i.e. reflective-formative and formative-formative type higher-order models;
Becker et al., 2012; Wetzels et al., 2009), PLS-SEM is the preferred modeling method.

Partial least squares structural equation modeling is superior to regression

analysis when assessing mediation
Mediation has become increasingly prominent in PLS-SEM-based research. Mediation
examines a statistical model in which a construct (i.e. the mediator construct) intervenes
between two other constructs (Baron and Kenny, 1986). Analyzing a mediator model
involves examining whether a change in the exogenous construct produces a change in the
mediator construct, which in turn produces a change in the endogenous construct, where the
mediator construct facilitates the relationship between the exogenous and endogenous
constructs (Nitzl et al., 2016; Cepeda Carri
on et al., 2017; Matthews et al., 2018). Thus, a priori
theoretical support is a prerequisite for investigating mediating effects (Preacher and Hayes,
2008). A structural model may also comprise several mediators, whose singular or joint
effects can be analyzed in multiple mediation analyses (Nitzl, 2016).
When testing statistical models with latent variables, researchers frequently use
regression analysis together with Preacher and Hayes’ (2008) PROCESS method (Hayes,
2018). By applying bootstrapping to test the indirect and direct effects between exogenous
and endogenous constructs, the PROCESS method overcomes well-known limitations of
prior methods for analyzing mediation (Baron and Kenny, 1986). However, PROCESS only
allows for the sequential testing of model parts without taking the entire model structure
into account. PLS-SEM, on the other hand, considers the entire theoretical structural model
in the estimation process, making it superior to regression analysis.
Even more critical, however, is the practice of averaging several indicators that together
comprise a measurement model, and using them as input in a PROCESS regression model.
PROCESS does not account for measurement error inherent in the items and thus introduces
bias in the results, whereas PLS-SEM removes the measurement error and reduces the bias.
EJM The averaging of construct indicator values is problematic since the constructs become less
53,4 reliable, which in turn deflates structural model estimates, potentially triggering Type II
errors. Henseler et al. (2014) show that PLS-SEM substantially reduces the effects of
measurement error, thereby increasing the reliability of construct scores. Therefore,
researchers should use PLS-SEM to estimate mediating effects among latent variables
rather than regression analysis in tandem with the PROCESS macro.
570
Researchers should use the two-stage approach to moderator analysis
Following the common notion to “stay away from main effects,” researchers using PLS-SEM
often study moderation models in which the relationship between two constructs is not
constant but depends on the values of a third variable (Hair et al., 2017a). When estimating
moderating effects using PLS-SEM, researchers can choose from a variety of approaches to
model the moderator’s influence on the relationship between two constructs (Memon et al.,
2019). These include the product-indicator approach (Chin et al., 2003), the two-stage
approach (Henseler and Fassott, 2010), and the orthogonalizing approach (Lance, 1988). In
addition, depending on the approach under consideration, researchers need to decide how to
calculate the indicator variables’ product terms based on unstandardized indicator data,
mean-centered indicator data or standardized indicator data.
This plethora of options has led to confusion among applied researchers regarding which
combination of options to choose. Extending earlier works by Henseler and Fassott (2010)
and Fassott et al. (2016), Becker et al. (2018) have recently evaluated the efficacy of different
approaches for running a moderator analysis. Their results demonstrate that the
combination of approaches for generating the interaction term and data treatment options
has a pronounced effect on the parameter estimates. Becker et al.’s (2018) simulation study
shows that the two-stage approach clearly outperforms all other approaches in terms of
parameter recovery, and also suggests that it excels in terms of statistical power[2].

Consistent partial least squares structural equation modeling adds very little
to existing knowledge of structural equation modeling
The differences in parameter estimation between PLS-SEM and covariance-based SEM are
generally quite small in settings commonly encountered in applied research, as well as when
the underlying data is drawn from a common factor model population (Reinartz et al., 2009;
Sarstedt et al., 2016). Therefore, the alleged need for “corrections” of the PLS-SEM method
when estimating factor models is exaggerated. Indeed, the recently proposed alternative to
standard PLS-SEM (Dijkstra and Henseler, 2015a, 2015b), referred to as consistent PLS
(PLSc-SEM), adds very little to the body of knowledge, and by its name is deceptive,
implying that regular PLS-SEM results are not consistent. The PLSc-SEM approach follows
a composite modeling logic, but the results mimic those of covariance-based SEM, which
assumes a common factor model. The first step in applying PLSc-SEM involves estimating
the model parameters using the standard PLS-SEM algorithm. In the second step, the
procedure adjusts the initial estimates of reflectively specified measurement models for
attenuation using a variant of Dillon-Goldstein’s rho (Dijkstra and Henseler, 2015b).
PLSc-SEM relies on the questionable premise that the common factor model is “correct,”
whereas the composite model is “incorrect” (Rigdon, 2016; Sarstedt et al., 2016). In an earlier
article, Dijkstra (1983, p. 4) takes a similar position when he concluded that “PLS indicator
loadings are upwardly biased”. But that conclusion is also based on the assumption that
covariance-based SEM indicator loadings are the universal benchmark for PLS-SEM
estimates – and they clearly are not. Moreover:
 [. . .] from a philosophical standpoint, there is no need for modeling constructs as common factors Rethinking of
[. . .], and reducing SEM to common factor models is a very restrictive (unnecessarily restrictive,
we would argue) view about SEM (Henseler et al., 2014, p. 184). partial least
squares
In fact, numerous researchers have warned against reflex-like adherence to the common
factor model (Schönemann and Wang, 1972; Bandalos and Boehm-Kaufman, 2009; Rigdon,
2016), with recent research suggesting that composites may actually capture a conceptual
variable more accurately than a common factor can (Rigdon et al., 2017a; Rhemtulla et al., 571
2018).
Apart from the above, it is a mystery why PLS-SEM users would want to employ PLSc-
SEM since it produces similar results as covariance-based SEM. The covariance-based SEM
method is already widely recognized and accepted as an approach to obtain solutions to
research designed around the common factor model and also allows for the inclusion of
composites (Grace and Bollen, 2008; Rhemtulla et al., 2018). Potential reasons for using
PLSc-SEM might include that the theoretical model is underidentified, or convergence
problems arise when attempting to estimate the model using covariance-based SEM (e.g.
when structural models with six or more constructs become relatively complex from a
covariance-based SEM perspective). Otherwise, there is little value to be gained in using
PLSc-SEM.

Researchers should use the heterotrait-monotrait criterion for discriminant

validity testing
One of the most important aspects of measurement model assessment involves establishing
discriminant validity (Anderson and Gerbing, 1988). Accurate assessment of discriminant
validity is important because it ensures that each construct is empirically unique and
captures a phenomenon not represented by other constructs in a statistical model. When
analyzing relationships among constructs that lack discriminant validity, “researchers
cannot be certain whether results confirming hypothesized structural paths are real or
whether they are a result of statistical discrepancies” (Farrell, 2010, p. 324).
While numerous approaches to discriminant validity testing have been proposed (Lord,
1973; Forsyth and Feldt, 1969; McNemar, 1958), researchers using PLS-SEM typically rely
on the Fornell-Larcker criterion (Fornell and Larcker, 1981). This criterion is met if for each
of two constructs, the average variance extracted (shared variance within) is greater than
the squared correlation between constructs (shared variance between). It implies that items
share more variance with their intended underlying construct than the individual constructs
share with another construct. In the past, the Fornell-Larcker criterion has been widely
viewed as “the most stringent” test of discriminant validity (Wang and Netemyer, 2002,
p. 222). But more recently, Henseler et al. (2015, p. 129) demonstrated that the Fornell-
Larcker criterion is “largely unable to detect a lack of discriminant validity” in PLS-SEM.
Franke and Sarstedt (2019) echo this observation and also question the criterion’s general
suitability for discriminant validity assessment on conceptual grounds.
Researchers using PLS-SEM should base their discriminant validity analysis on Henseler
et al.’s (2015) heterotrait-monotrait ratio of correlations (HTMT), which contrasts the
indicator correlations between constructs with the correlations within indicators of the same
construct. For discriminant validity to be established, Henseler et al. (2015) suggest different
thresholds the HTMT statistic should not exceed, depending on whether the constructs are
conceptually similar (0.90) or not (0.85). Furthermore, researchers can formally test whether
the HTMT value is significantly lower than unity (1) using bootstrapping. Franke and
Sarstedt (2019) show that the directional cutoffs proposed by Henseler et al. (2015) have very
high power to detect discriminant validity, but at the cost of inflated Type II errors (i.e.
EJM falsely rejecting discriminant validity) when the population value matches the threshold
53,4 value. As a remedy, Franke and Sarstedt (2019) suggest that researchers should run
an inferential test to assess whether HTMT deviates significantly from a selected threshold
(e.g. 0.90). That is, researchers using PLS-SEM should test whether the bootstrap-based
confidence interval for a certain HTMT value includes the threshold value (e.g. 0.90).
The HTMT statistic is only defined for reflectively specified constructs, but can also be
572 applied to establish discriminant validity between a reflective multi-item construct and a
single-item construct. Future research should conceptually and statistically establish
approaches that allow assessing discriminant validity when either of the constructs under
consideration is specified formatively or when both constructs rely on single-item measures.

Researchers do not necessarily need to assess a partial least squares path

model’s goodness-of-fit
Testing a PLS path model’s goodness of fit is potentially useful – but only for those studies
that follow a purely confirmatory approach. One might assume that as many PLS-SEM-
based studies frame their approach in a confirmatory sense (i.e. literature review is followed
by the derivation of formal hypotheses, and finally model estimation), model fit is
imperative for PLS-SEM-based model evaluation. But many of the same studies also
formulate managerial recommendations, which are predictive statements. That is, studies in
marketing research and social sciences in general engage in what Gregor (2006, p. 626) refers
to as explanation and prediction theory, which “implies both [an] understanding of
underlying causes and prediction, as well as [a] description of theoretical constructs and the
relationships among them”.
In his seminal treatise on the philosophy of science, Conjectures and Refutations, Popper
(1962) posited that prediction is the primary criterion for evaluating falsifiability, and that
all explanatory theories must “rise and fall based on their objective predictions” (Shugan,
2009, p. 994). Similarly, Steenkamp and Baumgartner (2000, p. 197) note that:
The critical test of the marketing relevance of explanations is typically provided by the results of
actual decision making. The ability of a model to predict with reasonable accuracy the results of
one’s marketing actions plays a key role.
To summarize, only the successful combination of explanation and prediction gives rise to
the practical relevance of causal explanations.
As PLS-SEM is designed for estimating causal-predictive relations (Jöreskog and Wold,
1982; Wold, 2006), it fits the requirement of both explanation and prediction. More
specifically, PLS-SEM maximizes the amount of explained variance of the endogenous
constructs embedded in a path model (Evermann and Tate, 2016), which is grounded in well-
developed causal explanations (Sarstedt et al., 2017a). That is, PLS-SEM seeks to minimize
the combination of bias and error variance, “occasionally sacrificing theoretical accuracy for
improved empirical precision” (Shmueli, 2010, p. 293). In contrast, other methods such as
covariance-based SEM may achieve a slightly more accurate representation of the
underlying theory by focusing only on minimizing bias, but this comes at the expense of
lower predictive power (Evermann and Tate, 2016). Even more critical, however, is that the
indeterminacy of latent variable scores in covariance-based SEM renders any assessment of
the structural model’s predictive power highly problematic (Becker et al., 2013a).
It is clear, therefore, that a singular focus on achieving model fit is contrary not only to
the fundamental quest of business research and social sciences research in general but also
to the very nature of the PLS-SEM method (Sánchez, 2013; Dolce et al., 2017; Shmueli et al.,
2016). This also is evident when considering the functioning principles of (approximate)
model fit metrics that recent PLS-SEM guidelines endorse (Henseler, 2017a; Henseler et al., Rethinking of
2016). For example, the SRMR and metrics such as Euclidean distance (dULS) and geodesic partial least
distance (dG) quantify the divergence between the empirical correlation matrix and the
model-implied correlation matrix. But relying on such measures makes sense only if the
squares
method under consideration estimates parameters such that this divergence is being
minimized. This is not what PLS-SEM does. The method considers error variance in model
estimation. Hence, rejecting and modifying a model on the grounds of thresholds for these fit
metrics ignores a central component in the PLS-SEM algorithm’s objective function – as 573
long ago noted by Lohmöller (1989).
Apart from the above, thresholds for SRMR and similar metrics have not been
established yet. For example, Henseler (2017a, p. 185) notes that “SRMR values below 0.08
typically indicate that the degree of misfit is not substantial,” without substantiating this
claim. Henseler et al. (2016) also argue in favor of this threshold, referring to Henseler et al.
(2014) whose simulation study, however, does not set out to derive any such threshold and
does not do so. In fact, Henseler et al. (2014) do not explicitly mention an SRMR threshold of
0.08. Furthermore, numerous questions regarding model fit metrics remain yet to be
answered, such as whether to derive the model-implied correlation matrix on the grounds of
the estimated model or the saturated model. When grounding the model-implied matrix in
the saturated model, the question emerges what this model should look like. Gefen et al.
(2011, p. 8) note that the saturated model “includes all possible paths”, but the direction of
the paths is unclear. Consider, for example, a simple model in which two exogenous
variables (Y1 and Y2) predict an endogenous construct Y3. Would the saturated model
assume a relationship from Y1 to Y2 or Y2 to Y1?
Should we therefore abandon the search for goodness-of-fit metrics for PLS-SEM? Not
based on current knowledge, as little is known about the viability of the fit measures being
proposed for PLS-SEM in the literature. Thus, before calling for their routine use (Henseler,
2017b) we need a much better understanding of their applicability and performance in PLS-
SEM, which is not strictly confirmatory but “causal-predictive” in nature (Jöreskog and
Wold, 1982, p. 270)[3]. Indeed, this process could easily involve the development of entirely
new metrics. For example, recent research shows that information theoretic model selection
criteria such as Schwarz’s (1978) Bayesian information criterion and Geweke and Meese’s
(1981) criterion achieve a sound tradeoff between model fit and predictive power in the
estimation of PLS path models, which perfectly meets the method’s causal-predictive nature
(Sharma et al., 2019a, 2019b). But their use is restricted to model comparisons rather than the
assessment of single models. Thus, there is support for further exploring the benefits of
goodness-of-fit measures in PLS-SEM, but definitely not for application of any fit measures
yet to be fully developed.

Partial least squares structural equation modeling results should be assessed

for both in-sample explanatory power and out-of-sample predictive power
Researchers using PLS-SEM often interpret the R2 as indicative of their model’s predictive
power. The R2 statistic is appropriate for assessing a model’s in-sample explanatory power,
but is not an indication of the model’s out-of-sample predictive power (Shmueli, 2010;
Shmueli and Koppius, 2011; Dolce et al., 2017). Recognizing the differences between in-
sample explanatory power and out-of-sample predictive power has important implications
far beyond a need for more careful language. Solely focusing on the R2 value to assess the
adequacy of a theoretical model is problematic since this may cause researchers to overfit
their model to the point that it overly accommodates both the information and the
idiosyncratic noise in the data (Hair et al., 2019). This limits the prospect of the
EJM generalizability of the model results to other samples (Pitt and Myung, 2002; Sharma et al.,
53,4 2019a; Sharma et al., 2019b), which is central to empirical research.
Researchers should therefore also estimate their model’s out-of-sample predictive power
(Chin, 1995; Nitzl and Chin, 2017) by means of a training sample (i.e. a subsample of the
original data used to estimate the model parameters, also referred to as an analysis sample
(Hair et al., 2018a, 2018b) that initially excludes the cases to be predicted, and then uses the
574 model estimates to predict case values in a holdout sample (i.e. the remaining part of the
dataset not used for model estimation). To facilitate such analyses, Shmueli et al. (2016)
developed PLSpredict, a holdout sample-based procedure that generates case-level
predictions on an item or a construct level. Contrary to standard structural model evaluation
metrics like R2 and Q2, PLSpredict offers a means to effectively assess a model’s out-of-
sample predictive power. We call on researchers to routinely apply the PLSpredict
procedure.

Assessing unobserved heterogeneity should be part of a partial least squares

structural equation modeling analysis
Researchers have long noted that unobserved heterogeneity can be a substantial threat to
the validity of any SEM analysis (Jedidi et al., 1997; Becker et al., 2013b). A simple example
illustrates this issue. Consider that an analysis using the entire data yields a nonsignificant
path coefficient estimate of 0.06, suggesting that the exogenous latent variable has no effect
on the endogenous latent variable. Conclusions and recommendations based on this result
may be invalid and misleading if, for example, two equally sized segments exist in the data
with significant segment-specific coefficients of þ0.5 and 0.5. Unobserved heterogeneity
implies that differences between these two groups of data do not emerge a priori from a
specific observable characteristic, or combinations of several characteristics but stem from
latent classes underlying the data.
Techniques for dealing with unobserved heterogeneity – commonly referred to as latent
class techniques – have a long-standing tradition in covariance-based SEM (Muthén, 1989;
Jedidi et al., 1997; Masyn, 2013) and have also been introduced in the context of PLS-SEM.
Prominent examples of PLS-SEM-based latent class techniques include finite mixture
partial least squares (Hahn et al., 2002; Sarstedt et al., 2011), PLS genetic algorithm
segmentation (Ringle et al., 2014), prediction-oriented segmentation (Becker et al., 2013b) and
iterative reweighted regressions (Schlittgen et al., 2016).
Numerous researchers have long called for the routine use of latent class techniques to
substantiate whether unobserved heterogeneity has a significant impact on the model
estimates (Becker et al., 2013b). That is, the latent class analysis results may provide
evidence for the robustness of the aggregate data results or may offer a course of correction
by partitioning the data into latent classes. Nevertheless, with the notable exception of Latan
(2018), “new guidelines” for PLS analyses systematically neglect this important issue to
ensure the validity of PLS-SEM results (Henseler, 2017a; Henseler et al., 2016). In light of the
considerable biases that result from ignoring unobserved heterogeneity (Becker et al.,
2013b), we reiterate the call for making a latent class analysis an integral element of a PLS-
SEM analysis. Sarstedt et al. (2017b) recently introduced a systematic procedure for
identifying and treating unobserved heterogeneity in PLS path models. Combining several
latent-class techniques, the procedure enables researchers to check whether unobserved
heterogeneity has a critical impact on the results, decide on the number of latent segments to
retain from the data, and recover the segment-specific model estimates.
Endogeneity can be assessed with partial least squares structural equation Rethinking of
modeling partial least
Like a mantra, some critics of PLS-SEM iterate that the method is incapable of handling
endogeneity resulting from omitted variables in the PLS path model or reverse causality
squares
(Antonakis et al., 2010; McIntosh et al., 2014). Such claims are astonishing as PLS-SEM is
grounded in regression analysis for which numerous approaches for handling endogeneity
exist. Indeed, researchers have discussed and applied approaches for handling endogeneity
in a PLS-SEM context (Benitez et al., 2016; Benitez et al., 2018; Svensson et al., 2018). Hult 575
et al. (2018) integrate classic and recently proposed approaches from regression analysis in a
comprehensive framework applicable in PLS-SEM. Their framework combines a Gaussian
copula estimation (Park and Gupta, 2012) to ascertain whether model estimates are affected
by endogeneity, with control and instrumental variable approaches (Wooldridge, 2010) to
address endogeneity issues in PLS-SEM.
While Hult et al.’s (2018) study offers guidance on how to deal with endogeneity in PLS
path models, engaging in endogeneity assessments is not always required. Rather, this
decision depends on whether theory testing or prediction (e.g. to derive managerial
recommendations) is the focus of the research project (Ebbes et al., 2011). When the main
concern of the research is with prediction, controlling for endogeneity is not useful, as it may
reduce the model’s out-of-sample predictive power (Papies et al., 2016). On the contrary,
when the research has a strict confirmatory focus, controlling for endogeneity is pertinent in
order to adequately test hypotheses. Thus, researchers should confirm the central focus of
their research before attempting to assess endogeneity (Hult et al., 2018). Moreover, since
PLS-SEM analyses rarely follow purely predictive or strictly confirmatory goals, further
research must clarify which approaches to handling endogeneity best consider the interplay
between prediction and explanation in a PLS-SEM sense.

Conclusion
The PLS-SEM field has undergone a massive development during the last decade.
Extending early presentations of the method (Chin, 1998; Falk and Miller, 1992; Haenlein
and Kaplan, 2004; Tenenhaus et al., 2005) and discussions regarding its correct use (Roldán
and Sánchez-Franco, 2012; Chin, 2010; Marcoulides and Chin, 2013; Hair et al., 2013), recent
research has provided numerous methodological extensions and improvements of the
method (Hair et al., 2018a, 2018b). However, notable methodological extensions of PLS-SEM
oftentimes do not appear in general marketing or management journals, but in highly
specialized outlets such as Computational Statistics and Data Analysis (Dijkstra and
Henseler, 2015a), OR Spectrum (Ringle et al., 2014) and Decision Sciences (Sharma et al.,
2019a), making it difficult for many scholars to be aware of and comprehend the state of the
art. In addition, recent controversies surrounding the method have sparked different
understandings among methodological researchers about PLS-SEM’s strengths and
weaknesses (Rönkkö and Evermann, 2013; Henseler et al., 2014; Rigdon, 2016). These
controversies have, for example, focused on PLS-SEM’s efficacy for handling small sample
sizes (Goodhue et al., 2012; Marcoulides et al., 2012) and formatively specified constructs
(Aguirre-Urreta and Marakas, 2014; Rigdon et al., 2014). For users who just like to apply
PLS-SEM, and also for reviewers and editors who have to judge the quality of others’
research, such different understandings can be confusing.
Taking up Henseler’s (2017b) approach, this research note clarifies several
misconceptions that have emerged as a result of the alleged “new guidelines” for PLS-SEM
use (Henseler, 2017a, 2017b; Henseler et al., 2016, 2017), slow adoption of methodological
extensions among applied researchers, and general lack of knowledge about PLS-SEM by
EJM many journal reviewers and some editors. We argue that the anti-PLS-SEM attitudes in
53,4 some fields (Petter, 2018) led to a misunderstanding of the method’s small sample size
capabilities and PLS-SEM’s suitability to estimate models with formatively specified
constructs. We also emphasize that PLS-SEM is a much superior approach to the regression-
based PROCESS analysis for mediation (Hayes, 2018), as PLS-SEM, for instance, has the
ability to simultaneously consider the entire model structure in parameter estimation and to
576 remove measurement error (Henseler et al., 2014) – neither of which is possible with
regression models. With regard to the model estimation and evaluation, we provide
clarification and guidance on how to best estimate moderator models, call for the use of the
HTMT statistic for discriminant validity assessment (i.e. any model validation solely
grounded in the Fornell-Larcker criterion should be viewed critically), and note that
endogeneity testing, if necessary, can be executed and controlled for in PLS-SEM analyses.
Finally, we highlight that there is no need to use PLSc-SEM, and note that goodness-of-fit
assessment in PLS-SEM is unnecessary in general, and currently not sufficiently developed
to be considered valid. Rather, researchers should assess the PLS path model’s out-of-sample
predictive power using PLSpredict (Shmueli et al., 2016).
We invite other researchers to comment constructively on our notions expressed in this
research note. We object, however, to prematurely denoting useful common practices and
concepts as “out-of-date” without providing substantial foundation and applicable
alternatives. Such calls do not help progressing the field but trigger confusion among
authors, reviewers, and editors alike. For example, Henseler (2017a) calls for the use of the
dominant indicator approach to handle construct sign changes without, however, offering
any substantiation for its efficacy or guidance on how to implement the approach. Such
recommendations become particularly problematic and confusing when they focus only on
mathematical and statistical characteristics, which deliver the proxies of theoretical
constructs (Rigdon, 2012; Rigdon, 2014), but neglect the interplay between method,
measurement theory and the goals of the analysis (Sarstedt et al., 2016). Also, some proposed
new guidelines over-emphasize minor differences in concepts and easy to fix technical
aspects, without providing solutions or by giving only vague descriptions of potential cures.
At the same time, some of the new guidelines sell old wine in new skins, such as claiming
that model identification “has been neglected in the realm of PLS path modeling in the past”
(Henseler et al., 2016, p. 7). Lohmöller (1989, Chapter 5.6.3) devotes an entire chapter titled
“Identifiability in PLS Model” to this issue. Progressing the field needs a more careful
consideration of seminal research on the limitations of model fit metrics in PLS-SEM
(Lohmöller, 1989, Chapter 5.5), the need to routinely employ methodological extensions to
test for unobserved heterogeneity (Becker et al., 2013b), and assessing a structural model’s
predictive power (Shmueli et al., 2016). We hope this research note will spark further
discussions with the aim of establishing a unified view of PLS-SEM. Such a unified view
offers common ground for tackling long-standing issues in PLS-SEM such as common
method variance, multi-level modeling and longitudinal data analysis.

Notes
1. Henseler (2017a), Henseler et al. (2017), Henseler et al. (2018) and Müller et al. (2018) reiterate these
guidelines in different contexts in very similar form.
2. Note that some concerns expressed in some recent publications on moderator analyses in
PLS-SEM further contribute to researchers’ confusion on how to run and interpret their analyses.
For example, Fassott et al. (2016) assert that current software implementations always
standardize the interaction term’s construct scores, rendering the moderation analysis results
 inappropriate. Their statement is astonishing, as, for example, SmartPLS 3 (Ringle et al., 2015) Rethinking of
correctly implements the moderator analysis by using unstandardized construct scores of the
interaction term. partial least
squares
3. Henseler (2018) correctly indicates that PLS-SEM can also be used for other research types such
as exploratory and descriptive research. However, corresponding results are not detached from
the function principles of the PLS-SEM algorithm but have to be interpreted in the context of the
method’s causal predictive nature.
577
References
Aguirre-Urreta, M.I. and Marakas, G.M. (2014), “Partial least squares and models with formatively
specified endogenous constructs: a cautionary note”, Information Systems Research, Vol. 25
No. 4, pp. 761-778.
Aguirre-Urreta, M.I., Rönkkö, M. and Marakas, G.M. (2016), “Omission of causal indicators:
consequences and implications for measurement”, Measurement: Interdisciplinary Research and
Perspectives, Vol. 14 No. 3, pp. 75-97.
Ali, F., Rasoolimanesh, S.M., Sarstedt, M., Ringle, C.M. and Ryu, K. (2018), “An assessment of the use of
partial least squares structural equation modeling (PLS-SEM) in hospitality research”,
International Journal of Contemporary Hospitality Management, Vol. 30 No. 1, pp. 514-538.
Anderson, J.C. and Gerbing, D.W. (1988), “Structural equation modeling in practice: a review and
recommended two-step approach”, Psychological Bulletin, Vol. 103 No. 3, pp. 411-423.
Antonakis, J., Bendahan, S., Jacquart, P. and Lalive, R. (2010), “On making causal claims: a review and
recommendations”, The Leadership Quarterly, Vol. 21 No. 6, pp. 1086-1120.
Bandalos, D.L. and Boehm-Kaufman, M.R. (2009), “Four common misconceptions in exploratory factor
analysis”, in Lance, C.E. and Vandenberg, R.J. (Eds), Statistical and Methodological Myths and
Urban Legends: Doctrine, Verity and Fable in the Organizational and Social Sciences, Routledge,
New York, NY, 61-87.
Baron, R.M. and Kenny, D.A. (1986), “The moderator-mediator variable distinction in social
psychological research: conceptual, strategic and statistical considerations”, Journal of
Personality and Social Psychology, Vol. 51 No. 6, pp. 1173-1182.
Becker, J.-M., Klein, K. and Wetzels, M. (2012), “Hierarchical latent variable models in PLS-SEM:
guidelines for using reflective-formative type models”, Long Range Planning, Vol. 45 Nos 5/6,
pp. 359-394.
Becker, J.-M., Rai, A. and Rigdon, E.E. (2013a), “Predictive validity and formative measurement in
structural equation modeling: embracing practical relevance”, 2013 Proceedings of the
International Conference on Information Systems, Milan.
Becker, J.-M., Rai, A., Ringle, C.M. and Völckner, F. (2013b), “Discovering unobserved heterogeneity in
structural equation models to avert validity threats”, MIS Quarterly, Vol. 37 No. 3, pp. 665-694.
Becker, J.-M., Ringle, C.M. and Sarstedt, M. (2018), “Estimating moderating effects in PLS-SEM and
PLSc-SEM: interaction term generation*Data treatment”, Journal of Applied Structural Equation
Modeling, Vol. 2 No. 2, pp. 1-21.
Benitez, J., Henseler, J. and Roldán, J.L. (2016), “How to address endogeneity in partial least squares path
modeling”, 22nd Americas Conference on Information Systems (ACIS), San Diego, CA.
Benitez, J., Ray, G. and Henseler, J. (2018), “Impact of information technology infrastructure flexibility
on mergers and acquisitions”, MIS Quarterly, Vol. 42 No. 1, pp. 25-43.
Bollen, K.A. (1989), Structural Equations with Latent Variables, Wiley, New York, NY.
Cepeda Carri
on, G., Nitzl, C. and Roldán, J.L. (2017), “Mediation analyses in partial least squares
structural equation modeling: guidelines and empirical examples”, in Latan, H. and Noonan, R.
(Eds), Partial Least Squares Path Modeling: Basic Concepts, Methodological Issues and
Applications, Springer International Publishing, Cham, pp. 173-195.
EJM Cheah, J.-H., Sarstedt, M., Ringle, C.M., Ramayah, T. and Ting, H. (2018), “Convergent validity
assessment of formatively measured constructs in PLS-SEM: on using single-item versus multi-
53,4 item measures in redundancy analyses”, International Journal of Contemporary Hospitality
Management, Vol. 30 No. 11, pp. 3192-3210.
Chin, W.W. (1995), “Partial least squares is to LISREL as principal components analysis is to common
factor analysis”, Technology Studies, Vol. 2 No. 2, pp. 315-319.
578 Chin, W.W. (1998), “The partial least squares approach to structural equation modeling”, in
Marcoulides, G.A. (Ed.), Modern Methods for Business Research, Erlbaum, Mahwah, pp. 295-358.
Chin, W.W. (2010), “How to write up and report PLS analyses”, in Esposito Vinzi, V., Chin, W.W.,
Henseler, J. and Wang, H. (Eds), Handbook of Partial Least Squares: Concepts, Methods and
Applications (Springer Handbooks of Computational Statistics Series, vol. II), Springer,
Heidelberg, Dordrecht, London, New York, NY, pp. 655-690.
Chin, W.W., Marcolin, B.L. and Newsted, P.R. (2003), “A partial least squares latent variable modeling
approach for measuring interaction effects: results from a monte carlo simulation study and an
electronic-mail emotion/adoption study”, Information Systems Research, Vol. 14 No. 2,
pp. 189-217.
Chin, W.W. and Newsted, P.R. (1999), “Structural equation modeling analysis with small samples using
partial least squares”, in Hoyle, R.H. (Ed.), Statistical Strategies for Small Sample Research, Sage,
Thousand Oaks, pp. 307-341.
Diamantopoulos, A. and Riefler, P. (2011), “Using formative measures in international marketing
models: a cautionary tale using consumer animosity as an example”, Advances in International
Marketing, Vol. 10 No. 22, pp. 11-30.
Diamantopoulos, A., Riefler, P. and Roth, K.P. (2008), “Advancing formative measurement models”,
Journal of Business Research, Vol. 61 No. 12, pp. 1203-1218.
Dijkstra, T.K. (1983), “Some comments on maximum likelihood and partial least squares methods”,
Journal of Econometrics, Vol. 22 Nos 1/2, pp. 67-90.
Dijkstra, T.K. and Henseler, J. (2015a), “Consistent and asymptotically normal PLS estimators for linear
structural equations”, Computational Statistics and Data Analysis, Vol. 81 No. 1, pp. 10-23.
Dijkstra, T.K. and Henseler, J. (2015b), “Consistent partial least squares path modeling”, MIS Quarterly,
Vol. 39 No. 2, pp. 297-316.
do Valle, P.O. and Assaker, G. (2016), “Using partial least squares structural equation modeling in
tourism research: a review of past research and recommendations for future applications”,
Journal of Travel Research, Vol. 55 No. 6, pp. 695-708.
Dolce, P., Esposito Vinzi, V. and Lauro, C. (2017), “Predictive path modeling through PLS and other
component-based approaches: methodological issues and performance evaluation”, in Latan, H.
and Noonan, R. (Eds), Partial Least Squares Path Modeling: Basic Concepts, Methodological
Issues and Applications, Springer International Publishing, Cham, pp. 153-172.
Ebbes, P., Papies, D. and van Heerde, H.J. (2011), “The sense and non-sense of holdout sample
validation in the presence of endogeneity”, Marketing Science, Vol. 30 No. 6, pp. 1115-1122.
Evermann, J. and Tate, M. (2016), “Assessing the predictive performance of structural equation model
estimators”, Journal of Business Research, Vol. 69 No. 10, pp. 4565-4582.
Falk, R.F. and Miller, N.B. (1992), A Primer for Soft Modeling, University of Akron Press, Akron, OH.
Farrell, A.M. (2010), “Insufficient discriminant validity: a comment on Bove, Pervan, Beatty, and Shiu
(2009)”, Journal of Business Research, Vol. 63 No. 3, pp. 324-327.
Fassott, G., Henseler, J. and Coelho, P.S. (2016), “Testing moderating effects in PLS path models with
composite variables”, Industrial Management and Data Systems, Vol. 116 No. 9, pp. 1887-1900.
Fornell, C.G. and Larcker, D.F. (1981), “Evaluating structural equation models with unobservable
variables and measurement error”, Journal of Marketing Research, Vol. 18 No. 1, pp. 39-50.
Forsyth, R.A. and Feldt, L.S. (1969), “An investigation of empirical sampling distributions of correlation Rethinking of
coefficients corrected for attenuation”, Educational and Psychological Measurement, Vol. 29
No. 1, pp. 61-71.
partial least
Franke, G.R. and Sarstedt, M. (2019), “Heuristics versus statistics in discriminant validity testing: a
squares
comparison of four procedures”, Internet Research, Forthcoming.
Garson, G.D. (2016), Partial Least Squares Regression and Structural Equation Models, Statistical
Associates, Asheboro.
Gefen, D., Rigdon, E.E. and Straub, D.W. (2011), “Editor’s comment: an update and extension to
579
SEM guidelines for administrative and social science research”, MIS Quarterly, Vol. 35 No. 2,
pp. 3-14.
Geweke, J. and Meese, R. (1981), “Estimating regression models of finite but unknown order”, Journal of
Econometrics, Vol. 16 No. 1, p. 162.
Goodhue, D.L., Lewis, W. and Thompson, R. (2012), “Does PLS have advantages for small sample size
or non-normal data?”, MIS Quarterly, Vol. 36 No. 3, pp. 981-1001.
Grace, J.B. and Bollen, K.A. (2008), “Representing general theoretical concepts in structural equation
models: the role of composite variables”, Environmental and Ecological Statistics, Vol. 15 No. 2,
pp. 191-213.
Gregor, S. (2006), “The nature of theory in information systems”, MIS Quarterly, Vol. 30 No. 3,
pp. 611-642.
Haenlein, M. and Kaplan, A.M. (2004), “A beginner’s guide to partial least squares analysis”,
Understanding Statistics, Vol. 3 No. 4, pp. 283-297.
Hahn, C., Johnson, M.D., Herrmann, A. and Huber, F. (2002), “Capturing customer heterogeneity using a
finite mixture PLS approach”, Schmalenbach Business Review, Vol. 54 No. 3, pp. 243-269.
Hair, J.F., Hult, G.T.M., Ringle, C.M. and Sarstedt, M. (2017a), A Primer on Partial Least Squares
Structural Equation Modeling (PLS-SEM), Sage, Thousand Oaks, CA.
Hair, J.F., Risher, J.J., Sarstedt, M. and Ringle, C.M. (2019), “When to use and how to report the results of
PLS-SEM”, European Business Review, Forthcoming.
Hair, J.F., Hult, G.T.M., Ringle, C.M., Sarstedt, M. and Thiele, K.O. (2017b), “Mirror, mirror on the wall: a
comparative evaluation of composite-based structural equation modeling methods”, Journal of
the Academy of Marketing Science, Vol. 45 No. 5, pp. 616-632.
Hair, J.F., Ringle, C.M. and Sarstedt, M. (2011), “PLS-SEM: indeed a silver bullet”, Journal of Marketing
Theory and Practice, Vol. 19 No. 2, pp. 139-151.
Hair, J.F., Ringle, C.M. and Sarstedt, M. (2013), “Partial least squares structural equation modeling:
rigorous applications, better results and higher acceptance”, Long Range Planning, Vol. 46
Nos 1/2, pp. 1-12.
Hair, J.F., Sarstedt, M., Ringle, C.M. and Gudergan, S.P. (2018a), Advanced Issues in Partial Least
Squares Structural Equation Modeling (PLS-SEM), Sage:Thousand Oaks, CA.
Hair, J.F., Sarstedt, M., Ringle, C.M. and Mena, J.A. (2012), “An assessment of the use of partial least
squares structural equation modeling in marketing research”, Journal of the Academy of
Marketing Science, Vol. 40 No. 3, pp. 414-433.
Hair, J.F., Black, W.C., Babin, B. and Anderson, R. (2018b), Multivariate Data Analysis, 8th ed., Cengage,
London.
Hayes, A.F. (2018), Introduction to Mediation, Moderation, and Conditional Process Analysis, Guilford
Press, New York, NY.
Henseler, J. (2017a), “Bridging design and behavioral research with variance-based structural equation
modeling”, Journal of Advertising, Vol. 46 No. 1, pp. 178-192.
Henseler, J. (2017b), “Some inconvenient truths about PLS path modeling”, 9th International Conference
on PLS and Related Methods, 17-19 June, Macau.
EJM Henseler, J. (2018), “Partial least squares path modeling: quo vadis?”, Quality & Quantity , Vol. 52 No. 1,
pp. 1-8.
53,4
Henseler, J., Dijkstra, T.K., Sarstedt, M., Ringle, C.M., Diamantopoulos, A., Straub, D.W., Ketchen, D.J.,
Hair, J.F., Hult, G.T.M. and Calantone, R.J. (2014), “Common beliefs and reality about partial least
squares: comments on Rönkkö and Evermann (2013)”, Organizational Research Methods, Vol. 17
No. 2, pp. 182-209.
580 Henseler, J. and Fassott, G. (2010), “Testing moderating effects in PLS path models: an illustration of
available procedures”, in Esposito Vinzi, V., Chin, W.W., Henseler, J. and Wang, H. (Eds)
Handbook of Partial Least Squares: Concepts, Methods and Applications (Springer Handbooks of
Computational Statistics Series, vol. II), Springer, Heidelberg, Dordrecht, London, New York, NY,
pp. 713-735.
Henseler, J., Hubona, G.S. and Ray, P.A. (2016), “Using PLS path modeling in new technology research:
updated guidelines”, Industrial Management and Data Systems, Vol. 116 No. 1, pp. 1-19.
Henseler, J., Hubona, G.S. and Ray, P.A. (2017), “Partial least squares path modeling: updated
guidelines”, in Latan, H. and Noonan, R. (Eds), Partial Least Squares Structural Equation
Modeling: Basic Concepts, Methodological Issues and Applications, Springer, Heidelberg,
pp. 19-39.
Henseler, J., Müller, T. and Schuberth, F. (2018), “New guidelines for the use of PLS path modeling in
hospitality, travel and tourism research”, in Ali, F., Rasoolimanesh, S.M. and Cobanoglu, C.
(Eds), Application of Partial Least Squares: Structural Equation Modeling (PLS-SEM) in
Tourism and Hospitality Research, Emerald, Bingley, pp. 17-33.
Henseler, J., Ringle, C.M. and Sarstedt, M. (2015), “A new criterion for assessing discriminant validity in
variance-based structural equation modeling”, Journal of the Academy of Marketing Science,
Vol. 43 No. 1, pp. 115-135.
Hult, G.T.M., Hair, J.F., Proksch, D., Sarstedt, M., Pinkwart, A. and Ringle, C.M. (2018), “Addressing
endogeneity in international marketing applications of partial least squares structural equation
modeling”, Journal of International Marketing, Vol. 26 No. 3, pp. 1-21.
Jedidi, K., Jagpal, H.S. and DeSarbo, W.S. (1997), “Finite-mixture structural equation models for
response-based segmentation and unobserved heterogeneity”, Marketing Science, Vol. 16 No. 1,
pp. 39-59.
Jöreskog, K.G. and Goldberger, A.S. (1975), “Estimation of a model with multiple indicators and
multiple causes of a single latent variable”, Journal of the American Statistical Association,
Vol. 70 No. 351, pp. 631-639.
Jöreskog, K.G. and Wold, H.O.A. (1982), “The ML and PLS techniques for modeling with latent
variables: historical and comparative aspects”, in Wold, H.O.A. and Jöreskog, K.G. (Eds),
Systems under Indirect Observation, Part I, North-Holland, Amsterdam, pp. 263-270.
Kaufmann, L. and Gaeckler, J. (2015), “A structured review of partial least squares in supply chain
management research”, Journal of Purchasing and Supply Management, Vol. 21 No. 4,
pp. 259-272.
Khan, G.F., Sarstedt, M., Shiau, W.-L., Hair, J.F., Ringle, C.M. and Fritze, M.P. (2019), “Methodological
research on partial least squares structural equation modeling (PLS-SEM): an analysis based on
social network approaches”, Internet Research, Forthcoming.
Lance, C.E. (1988), “Residual centering, exploratory and confirmatory moderator analysis, and
decomposition of effects in path models containing interactions”, Applied Psychological
Measurement, Vol. 12 No. 2, pp. 163-175.
Latan, H. (2018), “PLS path modeling in hospitality and tourism research: the golden age and days of
future past”, in Ali, F., Rasoolimanesh, S.M. and Cobanoglu, C. (Eds), Application of Partial Least
Squares: Structural Equation Modeling (PLS-SEM) in Tourism and Hospitality Research,
Emerald, Bingley, pp. 53-83.
Latan, H. and Noonan, R. (2017), Partial Least Squares Structural Equation Modeling: Basic Concepts, Rethinking of
Methodological Issues and Applications, Springer, Heidelberg.
partial least
Lee, N. (2017), “Reflections on a decade of EJM and marketing scholarship: the good, the bad, and the
future”, European Journal of Marketing, Vol. 51 Nos 11/12, pp. 1774-1798.
squares
Lee, N. and Cadogan, J.W. (2013), “Problems with formative and higher-order reflective variables”,
Journal of Business Research, Vol. 66 No. 2, pp. 242-247.
Lohmöller, J.-B. (1989), Latent Variable Path Modeling with Partial Least Squares, Physica, Heidelberg.
581
Lord, F.M. (1973), “Estimation of latent ability and item parameters when there are omitted responses”,
ETS Research Bulletin Series, Vol. 1973 No. 2, pp. 1-30.
McDonald, R.P. (1996), “Path analysis with composite variables”, Multivariate Behavioral Research,
Vol. 31 No. 2, pp. 239-270.
McNemar, Q. (1958), “Attenuation and interaction”, Psychometrika, Vol. 23 No. 3, pp. 259-265.
Marcoulides, G.A. and Chin, W.W. (2013), “You write, but others read: common methodological
misunderstandings in PLS and related methods”, in Abdi, H., Chin, W.W., Esposito Vinzi, V.,
Russolillo, G. and Trinchera, L. (Eds), New Perspectives in Partial Least Squares and Related
Methods, Springer: New York, NY, pp. 31-64.
Marcoulides, G.A., Chin, W.W. and Saunders, C. (2012), “When imprecise statistical statements become
problematic: a response to Goodhue, Lewis, and Thompson”, MIS Quarterly, Vol. 36 No. 3,
pp. 717-728.
Marcoulides, G.A. and Saunders, C. (2006), “PLS: a silver bullet?”, MIS Quarterly, Vol. 30 No. 2,
pp. 3-9.
Masyn, K.E. (2013), “Latent class analysis and finite mixture modeling”, in Little, T.D. (Ed.), The Oxford
Handbook of Quantitative Methods: Statistical Analysis, Oxford University Press: New York,
NY, pp. 551-611.
Matthews, L., Hair, J.F. and Matthews, R. (2018), “PLS-SEM: the holy grail for advanced analysis”, The
Marketing Management Journal, Vol. 28 No. 1, pp. 1-13.
Memon, M., Cheah, J.-H., Ramayah, T., Ting, H., Chuah, F. and Cham, T. (2019), “Moderation analysis:
issues and guidelines”, Journal of Applied Structural Equation Modeling, Vol. 3 No. 1, pp. i-xi.
McIntosh, C.N., Edwards, J.R. and Antonakis, J. (2014), “Reflections on partial least squares path
modeling”, Organizational Research Methods, Vol. 17 No. 2, pp. 210-251.
Müller, T., Schuberth, F. and Henseler, J. (2018), “PLS path modeling: a confirmatory approach to study
tourism technology and tourist behavior”, Journal of Hospitality and Tourism Technology, Vol. 9
No. 3, pp. 249-266.
Muthén, B.O. (1989), “Latent variable modeling in heterogeneous populations”, Psychometrika, Vol. 54
No. 4, pp. 557-585.
Nitzl, C. (2016), “The use of partial least squares structural equation modelling (PLS-SEM) in
management accounting research: directions for future theory development”, Journal of
Accounting Literature, Vol. 37, pp. 19-35.
Nitzl, C. and Chin, W.W. (2017), “The case of partial least squares (PLS) path modeling in managerial
accounting”, Journal of Management Control, Vol. 28 No. 2, pp. 137-156.
Nitzl, C., Roldán, J.L. and Cepeda Carri
on, G. (2016), “Mediation analysis in partial least squares path
modeling: helping researchers discuss more sophisticated models”, Industrial Management and
Data Systems, Vol. 119 No. 9, pp. 1849-1864.
Papies, D., Ebbes, P., van Heerde, H.J., (2016), “Addressing endogeneity in marketing models”, in
Leeflang, P.S.H., Wieringa, J.E., Bijmolt, T.H.A. and Pauwels, K.H. (Eds), Advanced Methods in
Modeling Markets, Springer International Publishing, Cham, pp. 581-627.
Park, S. and Gupta, S. (2012), “Handling endogenous regressors by joint estimation using copulas”,
Marketing Science, Vol. 31 No. 4, pp. 567-586.
EJM Petter, S. (2018), “Haters gonna hate’: PLS and information systems research”, ACM SIGMIS Database:
The DATABASE for Advances in Information Systems, Vol. 49 No. 2, pp. 10-13.
53,4
Pitt, M.A. and Myung, I.J. (2002), “When a good fit can be bad”, Trends in Cognitive Sciences, Vol. 6
No. 10, pp. 421-425.
Popper, K.R. (1962), Conjectures and Refutations: The Growth of Scientific Knowledge, Basic Books,
New York, NY.
582 Posey, C., Roberts, T.L., Lowry, P.B. and Bennett, B. (2015), “Multiple indicators and multiple causes
(MIMIC) models as a mixed-modeling technique: a tutorial and an annotated example”,
Communications of the Association for Information Systems, Vol. 36 No. 11.
Preacher, K.J. and Hayes, A.F. (2008), “Asymptotic and resampling strategies for assessing and
comparing indirect effects in multiple mediator models”, Behavior Research Methods, Vol. 40
No. 3, pp. 879-891.
Ramayah, T., Cheah, J.-H., Chuah, F. and Ting, H. (2016), Partial Least Squares Structural Equation
Modeling (PLS-SEM) Using SmartPLS 3.0: An Updated and Practical Guide to Statistical
Analysis, Pearson, Singapore.
Rasoolimanesh, S.M. and Ali, F. (2018), “Editorial: partial least squares (PLS) in hospitality and tourism
research”, Journal of Hospitality and Tourism Technology, forthcoming.
Reinartz, W.J., Haenlein, M. and Henseler, J. (2009), “An empirical comparison of the efficacy of
covariance-based and variance-based SEM”, International Journal of Research in Marketing,
Vol. 26 No. 4, pp. 332-344.
Rhemtulla, M., Bork, V.R. and Borsboom, D. (2018), “Worse than measurement error: consequences of
inappropriate latent variable measurement models”, Working Paper, available at: https://osf.io/
f8g4d/ (accessed 25 February 2019).
Richter, N.F., Cepeda Carri
on, G., Roldán, J.L. and Ringle, C.M. (2016), “European management research
using partial least squares structural equation modeling (PLS-SEM): editorial”, European
Management Journal, Vol. 34 No. 6, pp. 589-597.
Rigdon, E.E. (2012), “Rethinking partial least squares path modeling: in praise of simple methods”,
Long Range Planning, Vol. 45 Nos 5/6, pp. 341-358.
Rigdon, E.E. (2014), “Rethinking partial least squares path modeling: breaking chains and forging
ahead”, Long Range Planning, Vol. 47 No. 3, pp. 161-167.
Rigdon, E.E. (2016), “Choosing PLS path modeling as analytical method in European management
research: a realist perspective”, European Management Journal, Vol. 34 No. 6, pp. 598-605.
Rigdon, E.E., Becker, J.-M., Rai, A., Ringle, C.M., Diamantopoulos, A., Karahanna, E., Straub, D.W. and
Dijkstra, T.K. (2014), “Conflating antecedents and formative indicators: a comment on Aguirre-
Urreta and Marakas”, Information Systems Research, Vol. 25 No. 4, pp. 780-784.
Rigdon, E.E., Becker, J.-M. and Sarstedt, M. (2017a), “Equating unobserved conceptual variables and
common factors in structural equation models”, Working Paper.
Rigdon, E.E., Sarstedt, M. and Ringle, C.M. (2017b), “On comparing results from CB-SEM and
PLS-SEM: five perspectives and five recommendations”, Marketing ZFP, Vol. 39 No. 3,
pp. 4-16.
Ringle, C.M., Sarstedt, M., Mitchell, R. and Gudergan, S.P. (2019), “Partial least squares structural
equation modeling in HRM research”, The International Journal of Human Resource
Management, Forthcoming.
Ringle, C.M., Sarstedt, M. and Schlittgen, R. (2014), “Genetic algorithm segmentation in partial least
squares structural equation modeling”, OR Spectrum, Vol. 36 No. 1, pp. 251-276.
Ringle, C.M., Wende, S. and Becker, J.-M. (2015), SmartPLS 3, SmartPLS, Bönningstedt.
Roldán, J.L., Sánchez-Franco, M.J., (2012), “Variance-based structural equation modeling: guidelines for
using partial least squares in information systems research”, in Mora, M., Gelman, O.,
 Steenkamp, A.L. and Raisinghani, M. (Eds) Research Methodologies, Innovations and Rethinking of
Philosophies in Software Systems Engineering and Information Systems, IGI Global: Hershey,
PA, 193-221.
partial least
Rönkkö, M. and Evermann, J. (2013), “A critical examination of common beliefs about partial least
squares
squares path modeling”, Organizational Research Methods, Vol. 16 No. 3, pp. 425-448.
Rossiter, J.R. (2011), “Marketing measurement revolution: the C-OAR-SE method and why it must
replace psychometrics”, European Journal of Marketing, Vol. 45 Nos 11/12, pp. 1561-1588.
Rossiter, J.R. (2016), “How to use C-OAR-SE to design optimal standard measures”, European Journal of
583
Marketing, Vol. 50 No. 11, pp. 1924-1941.
Sánchez, G. (2013), PLS Path Modeling with R, Trowchez Editions, Berkeley.
Sarstedt, M., Becker, J.-M., Ringle, C.M. and Schwaiger, M. (2011), “Uncovering and treating unobserved
heterogeneity with FIMIX-PLS: Which model selection criterion provides an appropriate
number of segments?”, Schmalenbach Business Review, Vol. 63 No. 1, pp. 34-62.
Sarstedt, M., Hair, J.F., Ringle, C.M., Thiele, K.O. and Gudergan, S.P. (2016), “Estimation issues with
PLS and CBSEM: Where the bias lies!”, Journal of Business Research, Vol. 69 No. 10,
pp. 3998-4010.
Sarstedt, M., Ringle, C.M. and Hair, J.F. (2017a), “Partial least squares structural equation modeling”, in
Homburg, C., Klarmann, M. and Vomberg, A. (Eds), Handbook of Market Research, Springer,
Heidelberg.
Sarstedt, M., Ringle, C.M. and Hair, J.F. (2017b), “Treating unobserved heterogeneity in PLS-SEM: A
Multi-Method approach”, in Noonan, R. and Latan, H. (Eds) Partial Least Squares Structural
Equation Modeling: Basic Concepts, Methodological Issues and Applications, Springer,
Heidelberg, pp. 197-217.
Schlittgen, R., Ringle, C.M., Sarstedt, M. and Becker, J.M. (2016), “Segmentation of PLS path models by
iterative reweighted regressions”, Journal of Business Research, Vol. 69 No. 10, pp. 4583-4592.
Schönemann, P.H. and Wang, M.-M. (1972), “Some new results on factor indeterminacy”,
Psychometrika, Vol. 37 No. 1, pp. 61-91.
Schwarz, G. (1978), “Estimating the dimensions of a model”, The Annals of Statistics, Vol. 6 No. 2,
pp. 461-464.
Sharma, P.N., Shmueli, G., Sarstedt, M., Danks, N. and Ray, S. (2019a), “Prediction-oriented model
selection in partial least squares path modeling”, Decision Sciences, in press.
Sharma, P.N., Shmueli, G., Sarstedt, M., Kim, K.H. and Thiele, K.O. (2018b), “PLS-based model selection:
the role of alternative explanations in MIS research”, Journal of the Association for Information
Systems, in press.
Shmueli, G. (2010), “To explain or to predict?”, Statistical Science, Vol. 25 No. 3, pp. 289-310.
Shmueli, G. and Koppius, O.R. (2011), “Predictive analytics in information systems research”, MIS
Quarterly, Vol. 35 No. 3, pp. 553-572.
Shmueli, G., Ray, S., Velasquez Estrada, J.M. and Chatla, S.B. (2016), “The elephant in the room:
evaluating the predictive performance of PLS models”, Journal of Business Research, Vol. 69
No. 10, pp. 4552-4564.
Shugan, S.M. (2009), “Commentary: relevancy is robust prediction, not alleged realism”, Marketing
Science, Vol. 28 No. 5, pp. 991-998.
Steenkamp, J.-B.E.M. and Baumgartner, H. (2000), “On the use of structural equation models for
marketing modeling”, International Journal of Research in Marketing, Vol. 17 Nos 2/3,
pp. 195-202.
Svensson, G., Ferro, C., Høgevold, N., Padin, C., Varela, J.C.S. and Sarstedt, M. (2018), “Framing the
triple bottom line approach: direct and mediation effects between economic, social and
environmental elements”, Journal of Cleaner Production, Vol. 197, pp. 972-991.
EJM Temme, D., Diamantopoulos, A. and Pfegfeidel, V. (2014), “Specifying formatively-measured constructs
in endogenous positions in structural equation models: caveats and guidelines for researchers”,
53,4 International Journal of Research in Marketing, Vol. 31 No. 3, pp. 309-316.
Tenenhaus, M., Esposito Vinzi, V., Chatelin, Y.-M. and Lauro, C. (2005), “PLS path modeling”,
Computational Statistics and Data Analysis, Vol. 48 No. 1, pp. 159-205.
Wang, G. and Netemyer, R.G. (2002), “The effects of job autonomy, customer demandingness, and trait
competitiveness on salesperson learning, self-efficacy, and performance”, Journal of the
584 Academy of Marketing Science, Vol. 30 No. 3, pp. 217-228.
Wetzels, M., Odekerken-Schroder, G. and van Oppen, C. (2009), “Using PLS path modeling for assessing
hierarchical construct models: guidelines and empirical illustration”, MIS Quarterly, Vol. 33
No. 1, pp. 177-195.
Wold, H.O.A. (2006), “Partial least squares”, Encyclopedia of Statistical Sciences, John Wiley and Sons,
Hoboken, NJ.
Wooldridge, J.M. (2010), Econometric Analysis of Cross Section and Panel Data, MIT Press,
Cambridge, MA.

Further reading
Nitzl, C., Roldán, J.L. and Cepeda Carri
on, G. (2016), “Mediation analysis in partial least squares path
modeling: helping researchers discuss more sophisticated models”, Industrial Management &
Data Systems, Vol. 119 No. 9, pp. 1849-1864.
Streukens, S. and Leroi-Werelds, S. (2016), “Bootstrapping and PLS-SEM: a step-by-step guide to get
more out of your bootstrap results”, European Management Journal, Vol. 34 No. 6, pp. 618-632.

Corresponding author
Joseph F. Hair can be contacted at: [email protected]

For instructions on how to order reprints of this article, please visit our website:
www.emeraldgrouppublishing.com/licensing/reprints.htm
Or contact us for further details: [email protected]