Professional Documents
Culture Documents
research-article2015
JTRXXX10.1177/0047287515569779Journal of Travel ResearchValle and Assaker
Abstract
Although the number of empirical applications of partial least–squares structural equation modeling (PLS-SEM) in tourism
has increased in the last two years, Assaker, Huang, and Hallak have conducted the only assessment on the use of PLS-SEM
in four studies and with a limited number of criteria. Thus, this study aims to critically analyze how the PLS-SEM method
has been applied in 44 articles published in 11 leading tourism journals from 2000 to 2014 in terms of four key criteria: (1)
themes explored and main motivations for using PLS-SEM; (2) characteristics of proposed models; (3) how the models were
evaluated; and (4) the use of more advanced analyses within the method. The findings revealed that although applications in
tourism have improved in recent years, problematic aspects in the application of PLS-SEM in tourism research still exist. The
article provides suggestions on how to improve the use of PLS-SEM in future tourism applications.
Keywords
structural equation modeling (SEM), partial least squares–structural equation modeling (PLS-SEM), tourism research, SSCI
economy, infrastructure, and environment factors at the des- users (researchers) in tourism can only benefit from the
tination), on an empirical basis, without any theory behind unique properties of PLS-SEM if they understand its under-
the authors’ assertions in terms of how these indicators are to lying principles, apply them properly, and report the results
be grouped in order to predict tourism demand (see, e.g., correctly. Because of the complexities involved in using
Mazanec and Ring 2011). PLS-SEM, systematic assessments of how PLS-SEM was
Another example involves using formative constructs applied in prior tourism studies can help disseminate rigor-
(i.e., relationship progresses from the indicators to the con- ous research and publication practices in the field. They can
struct, suggesting that indicators used to measure the con- also provide important guidance and, if necessary, opportu-
struct are not correlated and thus contribute differently to nities for course correction in future applications. Although
form their underlying construct) to model tourism variables/ PLS-SEM usage reviews have been carried out across many
constructs, such as in the case of loyalty (see Song et al. disciplines in business research, including marketing (e.g.,
2011). For example, three indicators are typically used to Hair, Ringle, and Sarstedt 2011), strategic management
measure loyalty: positive word of mouth, revisit intention, (e.g., Hair et al. 2012a; Hulland 1999), and MIS (e.g.,
and recommendation to acquaintances. These indicators do Ringle, Sarstedt, and Straub 2012; Urbrach and Ahlemann
not move together and are assumed to form the loyalty con- 2010), corresponding assessments of PLS-SEM in tourism
struct, given that a tourist could say positive things about his/ are still limited. Assaker, Huang, and Hallak(2012) provided
her stay in the destination. Nevertheless, this does not neces- the only assessment so far, assessing four studies in tourism.
sarily mean that he/she will return to the destination in the They showed that the PLS-SEM technique had been applied
future, as this depends on other factors as well, such as avail- with considerable variability in terms of appropriately han-
able time or money. A final example of PLS-SEM use is in dling conceptual and methodological issues. In this context,
the case of complex models where the number of indicators an update and extension of the Assaker, Huang, and Hallak
compared with the sample size is large. This is the case in (2012) case study–based assessment of PLS-SEM specific
tourism studies carried out at the macro/country level, where to the tourism discipline seems timely and warranted.
the number of observations is limited because of the number Therefore, the objective of this article is to critically ana-
of countries and lack of data available for each country lyze the use of PLS-SEM in tourism and to provide recom-
(Assaker and Hallak 2012). mendations for the use of the technique in future tourism
Despite the appropriateness of PLS-SEM to advance research (for explanations of the PLS-SEM method itself,
research in tourism studies (because of PLS-SEM’s features see, e.g., Chin 2010; Hair, Ringle, and Sarstedt, 2011;
and modeling flexibility, as previously mentioned), the Henseler, Ringle, and Sinkovics 2009; Henseler, Ringle, and
number of studies using PLS-SEM in tourism has grown in Sarstedt 2012). In particular, this study analyzes 44 empirical
recent years but is still considerably smaller than in other studies following the PLS-SEM approach that were pub-
disciplines in the business field, such as marketing (Hair, lished in indexed journals in the Social Sciences Citation
Ringle, and Sarstedt 2011), strategic management (Hair et Index (SSCI) (Web of Science) from 2000 to 2014 (Thomson
al. 2012a), and management information systems (MIS) Reuters 2014). These studies were divided into two groups:
(Ringle, Sarstedt, and Straub 2012). Researchers in the tour- pre-2012 and post-2012. Year 2012 was when the first article
ism field still seem to rely predominantly on CB-SEM in assessing PLS-SEM in four tourism articles was published
their empirical studies and thus might miss opportunities (Assaker, Huang, and Hallak 2012) and when several reviews
frequently exploited in other disciplines by using the PLS- of PLS-SEM in marketing and strategic management and
SEM technique. Potential reasons for the still-limited use of MIS were published (see Hair, Ringle, and Sarstedt 2011;
PLS-SEM in tourism, which tends to apply this method cau- Hair et al. 2012a; Hair et al. 2012b; Ringle, Sarstedt, and
tiously, might stem from researchers’—though well- Straub 2012). The studies are analyzed according to four key
equipped with a basic understanding of CB-SEM dimensions: reasons for using PLS-SEM, characteristics of
models—lack of understanding and familiarity with PLS- the proposed models, model evaluation and possible use of
SEM. Indeed, textbooks on multivariate data analysis tend more advanced applications, and reporting. Where possible,
not to discuss PLS-SEM (e.g., Churchill and Iacobucci we also indicated best practices as guidelines for future
2010) or address it only superficially (e.g., Hair et al. 2010). applications and suggested avenues for further research. The
In addition, the topic is seldom found in research methodol- current study thus helps identify how the application of PLS-
ogy course syllabi, where the focus is still on traditional SEM has changed, if at all, over time (i.e., before and after
CB-SEM because of the popularity of this method and the 2012). More importantly, it helps identify which aspects of
widespread availability of CB-SEM software (e.g., Lisrel, the method have been more carefully considered and which
AMOS, Mplus; Chin 1998). need to be improved to ensure more rigorous application in
In fact, as with other statistical methods, PLS-SEM is a the field. Such information can be used to provide sugges-
complex technique requiring several choices that, if not duly tions for researchers interested in applying PLS-SEM to
considered, can mislead researchers into improper findings, future tourism applications by highlighting current areas of
interpretations, and conclusions (Hair et al. 2012b). Thus, concerns.
Partial Least Squares Structural are endogenous latent variables. The parameters of a mea-
Equation Modeling surement model, known as the loading coefficients, are also
estimated by least squares regressions by taking into account
Partial least squares (PLS) structural equation modeling the mode used (A = reflective or B = formative).
(Wold 1982, 1985) is an alternative to traditional SEM that As previously discussed, PLS-SEM can be used to achieve
can be applied to model the relationships of causality among four major purposes. First, PLS is advantageous when the
the variables. In this partial information method, rather than researcher is trying to explore, rather than confirm, theory. It
using the model to explain covariations among indicators is useful when the phenomenon being investigated is rela-
(i.e., modeling the measurement errors, as SEM does), PLS tively new and the measurement models are at the explor-
maximizes the explained variance of all dependent variables atory stage (Wold 1985). Second, PLS can be used to examine
based on how these variables relate to their neighboring con- structural models in cases of small samples and when the
structs. It uses an iterative algorithm in which the parameters multivariate normality of the data cannot be supported (Chin
are calculated with a series of least squares regressions after and Newsted 1999). Third, PLS modeling allows the unre-
explicitly creating construct scores by weighting the sums of stricted computation of models composed of “reflective” and
items underlying each construct (Chin 1998 ). The term par- “formative” measurement models (Diamantopoulos and
tial thus stems from the idea that the iterative procedure Winklhofer 2001). Finally, PLS can examine large, complex
involves separating the parameters rather than estimating models comprising several latent and manifest variables as
them simultaneously (Hulland 1999). well as hierarchical models with first-order and second-order
PLS follows a two-step process that starts with an itera- latent constructs (Wold 1985). Thus, PLS can overcome
tive estimation of latent variables scores. The method uses identification issues, nonconvergence, limitations, and
the PLS algorithm to estimate outer and inner weights based assumptions associated with CB-SEM (see Vinzi, Trinchera,
on how the structural and measurement models are specified. and Amato 2010).
This requires an iterative procedure in which two kinds of
approximations for the latent variables are estimated until
the weight estimates converge. The two weight calcula- PLS-SEM Studies in Tourism Research:
tions—inside and outside—relate to the inner relations and Journals and themes
outer relations. The algorithm starts with an arbitrary initial
weight used to calculate an outside approximation of the Our review focused on the use of PLS-SEM articles in 11
latent variables. The inner relations among latent variables tourism journals: Journal of Travel Research, Annals of
are then considered to calculate the inside approximations. Tourism Research, Tourism Management, Journal of Travel
To perform this approximation, the researcher can choose and Tourism Marketing, Journal of Sustainable Tourism,
among three possible scenarios, called weighting schemes: Asia Pacific Journal of Tourism Research, International
(1) centroid, (2) factor, or (3) path scheme. After the inside Journal of Tourism Research, Current Issues in Tourism,
approximations are obtained, the algorithm turns again to the Tourism Geographies, Tourism Economics, and Tourism
outer relations, and new weights are calculated considering Analysis. These journals (except Tourism Analysis) were
how the indicators are related to their constructs: by Mode A selected because they were included in the Social Sciences
(reflective) or Mode B (formative). Mode A implies simple Citation Index (SSCI) (Web of Science) (Thomson Reuters
linear regressions between the construct and its reflecting 2014) as leading journals in the field of tourism. Tourism
indicators as the construct is assumed to affect each indicator Analysis was included as it is another high-impact journal in
separately. Mode B implies multiple linear regressions the field that publishes significant quantitative research in
between the construct and the set of indicators as the indica- tourism, making it important to consider this journal when
tors are assumed to affect the construct on a collective basis. reviewing the use of an advanced statistical technique in
The simple or multiple regression coefficients are then used tourism (e.g., PLS-SEM) (Nunkoo et al. 2013 ).
as new weights for the outside approximation. This process In addition to being the leading publication outlets for
continues iteratively until the weights converge—that is, tourism research, these journals have different foci, aims,
until the change in the outer weights between two iterations and objectives, thereby ensuring that SEM-based articles
drops below a predefined limit. This limit is usually preset with different tourism orientations (e.g., marketing, plan-
within the PLS-SEM software (Henseler, Ringle, and ning, psychology, and economics) were included in this
Sinkovics 2009). review study. All issues between 2000 (the first year an
After the weights converge and latent variables are esti- application was found) and 2014 (March 2014, the time
mated, the second step of the process calculates the parame- when this article was written) were searched for empirical-
ters of the structural and the measurement models. The based applications of PLS-SEM. We conducted a full-text
structural coefficients, also known as path coefficients, are search in the Thomson Reuters Web of Knowledge and
calculated using ordinary least squares regression between EBSCO Business Source Premier databases using the key-
latent variables (LVs). There are as many regressions as there words partial least squares structural equation modeling,
Number of
Journals Publications Studies
Annals of Tourism Research 1 Song et al. (2012)
Asia Pacific Journal of Tourism Research 5 Huang et al. (2009); Kim et al. (2011); Kim, Ahn, and Chung (2013) ; Guan
and Jones (2014); Chang, Shu, and King (2013)
Current Issues in Tourism 4 López-Bonilla and López-Bonilla (2012a, 2012b); Blazquez-Resino, Molina,
and Esteban-Talaya (2013); Tan and Kuo (2013)
International Journal of Tourism Research 3 Song et al. (2011); Gómez and Molina (2012); Ku (2014)
Journal of Travel and Tourism Marketing 10 Loureiro and González (2008); Mistilis and D’ambra (2008); Conze et al.
(2010); Loureiro (2010); Zach and Racherla (2011); Battour, Battor, and
Ismail (2012); Kim, Chung, Lee, and Kim (2012) ; Regan, Carlson, and
Rosenberger III (2012); Agapito, Valle, and Mendes (2013); Cheng and
Loi (2014)
Journal of Travel Research 8 Hernández-Maestro, Muñoz-Gallego, and Santos-Requejo (2009); Pike,
Murdy, and Lings (2011); Ayeh, Au, and Law (2013a); Gardiner, King,
and Grace (2013); Kim, Lee, Chung, and Kim (2013); Stienmetz,
Levy, and Boo (2013); Assaker et al. (2014); Hernández-Maestro and
González-Benito (2014)
Tourism Economics 1 Mazanec and Ring (2011)
Tourism Management 11 Murphy, Pritchard, and Smith (2000); Lai and Youcheng (2009);
Camarero, Garrido, and Vicente (2010); Alexander et al. (2012); Ayeh,
Au, and Law (2013b); Casanueva, Gallego, and Sancho (2013); García,
Gómez, and Molina (2012); Prayag, Nunkoo, and Alders (2013); Roxas
and Chadee (2013) ; Taheri, Jafari, and O’Gorman (2014); Chiu, Lee,
and Chen (2014)
Journal of Sustainable Tourism 0
Tourism Analysis 1 Assaker and Hallak (2012)
Tourism Geographies 0
partial least squares path modeling, PLS-SEM, and PLSPM. The identified studies explored several themes, which can
We also looked into the online versions of the journals in be grouped into 10 categories. As shown in Table 2, “loyalty
order to double-check that we had captured all PLS-SEM and/or satisfaction regarding a destination or tourism organi-
articles in the targeted tourism journals. Ultimately, the zation” is the main theme studied using this method and is
search resulted in a total of 44 articles from the 11 journals, clearly more prevalent than the others, being analyzed in 15
after removing conceptual articles on PLS-SEM and research (34.1%) of the publications identified. The focus on this
articles from the review (see Table 1). theme is stronger in studies before 2012 compared with later
In particular, Tourism Management, Journal of Travel and studies (57.1% vs. 23.3%). “Perceived quality of organiza-
Tourism Marketing, and Journal of Travel Research pub- tions, tourist products, or services” occupies the next posi-
lished the largest number of PLS-SEM studies among the tion in the set of themes and is analyzed in 7 articles (15.9%).
reviewed journals (all 3 altogether published 29 articles, “Tourism organizations management” ranks third, being
accounting for 65.9% of the total publications). In contrast, explored in 5 (11.4%) publications. Finally, it is interesting
the Journal of Sustainable Tourism and Tourism Geographies to note that articles published between 2012 and 2014 cover
did not include a single PLS-SEM study in the period of a broader range of themes (all 10 themes) compared to those
time reviewed. Figure 1 shows the (cumulative) number of published prior to 2012 (which only cover 5 of the 10
studies between 2000 (the first year an application was themes), suggesting more variety in the recent applications
found) and 2014. It is apparent that the use of PLS-SEM has of PLS-SEM in tourism and the prevalent use of PLS-SEM
substantially increased over time. Approximately 31.8% (n to investigate different phenomena within tourism as com-
= 14) of articles were published before 2012, compared to pared to early applications in tourism (prior to 2012).
68.2% (n = 30) between 2012 and 2014 (see Table 1).
Regressing the number of studies on the linear effects of Reasons for Using PLS-SEM in Tourism
time yields a significant model (F = 23.71; p < 0.000) in
Research
which the time effect is significant (t = 4.86; p < 0.000), sug-
gesting that the use of PLS-SEM in tourism has accelerated Given that PLS-SEM is an alternative approach to traditional
significantly over time. CB-SEM when the latter reaches its limitations and cannot
50
45
40
35
30
10
Themes n % n % n %
•• Loyalty and/or satisfaction regarding a destination or tourism organization 8 57.1 7 23.3 15 34.1
•• Tourist organizations’ management 2 14.3 3 10.0 5 11.4
•• Brand management 1 7.1 2 6.7 3 6.8
•• Perceived quality of organizations, tourist products, or services 2 14.3 5 16.7 7 15.9
•• Travel intention 0 0 3 10.0 3 6.8
•• Intention to use media in travel planning 0 0 3 10.0 3 6.8
•• Destinations’ competitiveness 1 7.1 2 6.7 3 6.8
•• Residents’ attitudes toward tourism 0 0 1 3.3 1 2.3
•• Tourists’ environmental orientation 0 0 2 6.7 2 4.6
•• Competencies in higher education in tourism 0 0 2 6.7 2 4.6
be used, studies applying PLS-SEM in tourism need to pres- data (21 studies, 47.7%), and formative nature of the model
ent the rationale behind the selection of this method rather (15 studies, 34.1%). All of these characteristics have been
than CB-SEM when this is the case. Of the 44 studies identi- extensively discussed in the methodological literature on
fied, only 2 (4.5%) did not provide such justification. Both of PLS-SEM (see Vinzi, Trinchera, and Amato 2010; Lohmöller
these articles were published in the period before 2012, sug- 1989). The sample size argument in particular has been the
gesting that users/researchers in recent years (2012 and subject of much debate (e.g., Marcoulides and Saunders
onward) are more aware of the features that make the PLS- 2006). A rule of thumb for robust PLS path modeling estima-
SEM technique suitable for the model/study at hand; they tions suggests that the sample size should be equal to the
find it easier now to report their reason for using PLS-SEM, larger of the following: (1) 10 times the number of indicators
which seems to not always have been the case in earlier years of the scale with the largest number of formative indicators,
(prior to 2012) when the technique was adopted. or (2) 10 times the largest number of structural paths directed
Table 3 displays the main reasons for using the PLS-SEM at a particular construct in the inner path model (Chin and
approach: the predictive focus of the method (31 studies, Newsted 1999). This represents the largest number of regres-
70.5%), small sample size (21 studies, 47.7%), nonnormal sions performed during the PLS-SEM iterative process; as
Table 3. Motivations for Using PLS-SEM in Tourism Research. (p ≤ 0.01) smaller samples (mean = 451), in terms of sample
mean, than later studies (mean = 562), suggesting that
Number and Percentage of Studies
Reporting researchers have become more aware of the sample size limi-
tations; as such, they are using larger sample sizes to avoid
Before 2012 and compromising the sample representativeness and to accu-
2012 After Total rately achieve statistical power.
Moreover, a comparison of studies published before 2012
(n = 14) (n = 30) (n = 44) with those published in 2012 and onward shows a fairly con-
sistent pattern, with a focus on prediction, small sample size,
Motivations n % n % n %
nonnormal data, and formative measures being the most
•• Nature of the method 11 78.6 20 66.7 31 70.5 prevalent reasons in recent years. This observation is consis-
•• Sample size 7 50.0 14 46.7 21 47.7 tent with patterns observed in other business disciplines,
•• Data characteristics 6 42.9 15 50.0 21 47.7 although with a different order in relevance, including mar-
•• Type of outer model 4 28.6 11 36.7 15 34.1 keting research (Hair, Ringle, and Sarstedt 2011), strategic
•• Model complexity 1 7.1 4 13.3 5 11.4 management (Hair et al. 2012a), and management informa-
•• Other 2 14.3 4 13.3 6 13.6 tion systems (Ringle, Sarstedt, and Straub 2012), suggesting
a similar understanding and applicability of the PLS-SEM
method by tourism researchers compared with researchers
such, it would be the logical threshold for sample size to applying PLS-SEM in other disciplines.
ensure the accuracy and statistical power of the model.
Nonetheless, several authors (e.g., Hair et al. 2012b) have
argued that very small samples can be a problem when per- Model Characteristics in Tourism
forming PLS-SEM for two reasons. First, they rarely capture
more heterogeneous populations (i.e., they fail to capture the
Research
heterogeneity in the population from which the sample is Table 4 shows the main features of models using PLS-SEM
drawn), thereby resulting in a larger sampling error. Second, in tourism research. The first aspects to be highlighted con-
if the sample is too small and the data are asymmetrical, the cern the number of latent variables in the inner model and the
bootstrap standard errors will be too large, thereby reducing number of indicators used to measure these variables. On
the method’s statistical power. Thus, to be on the safe side in average, the number of latent variables per model is 6.02,
terms of sample size, one might recommend 100 cases with and the average number of indicators per model is 24.93.
the objective of improving accuracy (see, e.g., Assaker, Although both figures are less than the average number of
Huang, and Hallak 2012). Indeed, few studies have system- latent variables and indicators in path models reported in
atically evaluated PLS-SEM’s performance when the sample other business disciplines, including marketing (7.9 and
size is small (e.g., Chin and Newsted 1999). More recently, 29.55, respectively), strategic management (7.5 and 27,
Reinartz, Haenlein, and Henseler (2009) showed that PLS- respectively), and management information systems (8.12
SEM achieves high levels of statistical power—compared to and 27.42), they still suggest a relatively high level of model
its covariance-based counterpart—even if the sample size is complexity compared with studies in the CB-SEM context
relatively small (i.e., 100 observations). Similarly, Boomsma (e.g., Shah and Goldstein review of CB-SEM reported an
and Hoogland’s (2001) study underscored the CB-SEM’s average number of latent variables of 4.70, with an average
need for relatively large sample sizes to achieve robust number of indicators of 16.30).
parameter estimates. PLS-SEM is therefore generally more Table 4 also shows that the average number of latent vari-
favorable with smaller sample sizes and more complex mod- ables per model is significantly (p ≤ 0.05) larger in articles
els. However, as noted by Marcoulides and Saunders (2006) published before 2012 than in 2012 onward (6.9 vs. 5.6,
as well as Sosik, Kahai, and Piovoso (2009), PLS-SEM is not respectively); however, the average number of indicators
a silver bullet for use with samples of any size, nor is it a remained almost the same in articles published before and
panacea for dealing with empirical research challenges. after 2012 (25.3 vs. 23.9, respectively). Taking these results
Researchers need to consider the effects of size, reliability, jointly, we can argue that model complexity did not decrease
total number of indicators, data characteristics, and other over time, given that the average number of indicators used
issues likely to affect the statistical power of the PLS-SEM to measure each construct significantly increased (p ≤ 0.05)
method (Hair et al. 2012b). In the present article, 42 of the 44 between articles published before 2012 and articles pub-
analyzed studies fulfill these conditions of using a sample of lished in 2012 onward (3.7 vs. 4.3 indicators on average per
100 cases or more, and only two use samples smaller than construct for the period after 2012). The same can be said
100 (Pike et al. 2011; Zach and Racherla 2011). The average about the number of inner model relationships being ana-
and median sample sizes are 487 and 321 observations, lyzed, which remained more or less the same over time (7.23
respectively, and studies prior to 2012 use significantly vs. 8.21 for the periods before and after 2012, respectively),
Table 4. Model Characteristics in Tourism Research. observations suggest that, overall, researchers in tourism are
still not capitalizing on the full features of PLS-SEM in terms
Number and Percentage of Studies
Reporting of allowing the assessment of reflective and formative mod-
els, although this trend seems to have started changing in
2012 and recent years, with researchers significantly applying more
Before 2012 After Total formative measurement schemes in their models.
Another aspect to note is that 6 studies (13.6%; see Table
(n = 14) (n = 30) (n = 44) 4) proposed models with at least one construct measured
using a single indicator, with no significant difference (p ≥
Aspect to Report n % n % n %
0.05) in the number of articles that have used a single indica-
Number of latent variables tor across the two time periods (3 prior to 2012 and 3 in 2012
Mean 6.93 5.60 6.02 onward). While single-indicator constructs in CB-SEM gen-
Median 6 6 6 erally lead to unidentified models, this problem does not
Range 3, 17 1, 13 1, 17 occur when PLS-SEM is used. However, Reinartz, Haenlein,
Number of structural paths and Henseler (2009) showed that PLS-SEM estimates are
Mean 8.21 7.23 7.55 only “consistent at large” in smaller samples if multiple indi-
Median 8.50 7 7 cators per construct are available. Otherwise, the tendency is
Range 3, 16 0, 15 0, 16 to underestimate the relationships in the inner model and
Type of outer model overestimate the relationships in the outer model (Chin
Only reflective 9 64.3 18 60.0 27 61.4
1998). Thus, authors such as Hair et al. (2012a) have advised
Only formative 0 0 1 3.3 1 2.3
caution in using this type of construct in PLS-SEM, and
Reflective and 5 35.7 11 36.7 16 36.4
formative
researchers in tourism should be aware of this limitation in
Total number of indicators future works, as it appears that the use of single-indicator
Mean 25.29 23.93 24.93 constructs is still a common pitfall within tourism studies.
Median 23 22.5 23
Range 14, 55 8, 65 8, 65
Number of models 3.00 21.4 3.00 10.0 6.00 13.6 Evaluating PLS-SEM in Tourism
with single-item Research
constructs
PLS-SEM does not offer a universal goodness-of-fit index.
In this regard, Chin (1998) proposed assessing the model by
suggesting again that fewer constructs have been used in applying various criteria at two moments. The first moment
PLS-SEM studies in recent years. These studies still incorpo- is the evaluation of the outer model—that is, the part of the
rated large numbers of inner model relationships and higher model describing the relationships between the latent vari-
numbers of indicators, thereby adding to the complexity of ables and their indicators. The second moment is the evalua-
the model being analyzed. tion of the inner model—that is, the part of the model
Moreover, an important feature of PLS-SEM is its ability describing the relationships across the latent variables that
to incorporate both reflective (the relationship progresses make up the model. As pointed out by Henseler, Ringle, and
from the construct to the indicators, suggesting that indica- Sinkovics (2009, p. 298), “It only makes sense to evaluate
tors are correlated or move together in the same direction) the inner path model estimates when the calculated latent
and formative measures (relationship progresses from the variable scores show evidence of sufficient reliability and
indicators to the construct, suggesting that indicators used to validity.” Underlying this statement is the assumption that it
measure the construct are not correlated and thus contribute is only worth analyzing the relationships between the latent
differently to forming their underlying construct) of the variables if the researcher feels confident that the indicators
latent variables. In this perspective, Table 4 also shows that properly represent these variables. We proceed by analyzing
27 (61.4%) studies still analyze the causal relationships the forms of outer model (reflective or formative) and inner
between constructs with the reflective measurement model model validation in the identified empirical studies in
only, while 16 (36.4%) articles used a combination of both tourism.
reflective and formative measurement models, and only one
article used solely a formatively measured latent construct.
Reflective Outer Model Evaluation
Moreover, of the 16 articles involving formative and reflec-
tive measurement models, a significantly (p ≤ 0.05) higher The evaluation of a reflective outer model should observe the
number (i.e., 11) were published after 2012. The only articles aspects listed in Table 5, that is, examine the reliability (indi-
involving solely a formative measurement model were pub- vidual and composite) and the validity (convergent and dis-
lished in this time period as well (2012 onward). These criminant) of the constructs.
Table 5. Reflective Outer Model Evaluation in Tourism measuring different constructs). In tourism research, dis-
Research. criminant validity has been evaluated by applying the Fornell
Number and Percentage of Studies and Larcker (1981) criterion, whereby a latent variable
Reporting should share more variance with its indicators than with indi-
cators of other latent variables. This aspect can be observed
2012 and if the AVE of each latent variable is greater than the highest
Before 2012 After Total squared correlation between this and the remaining latent
variable. This form of discriminant validity assessment was
(n = 14) (n = 30) (n = 44) observed in 41 (93.2%) of the identified studies. Again, the
Aspect to Report n % n % n %
use of this approach is equally prevalent in both periods
under analysis. Only rarely does discriminant validity analy-
Indicator reliability sis consider the observation of the cross-loadings—namely,
Indicator loadings 5 35.7 30 100 35 79.6 the loadings of each indicator in the other latent variables.
(Bootstrap) Table 5 shows that only 9 (20.5%) of the studies (all from
Composite reliability 2012 onward) provided results of the cross-loadings in their
Composite reliability 13 92.9 30 100 43 97.7
analyses, indicative of a significant (p ≤ 0.01) increase in the
index (Goldstein rho)
use of a cross-loading indicator in recent years. In summary,
Convergent validity
Indicators’ significance 14 100 30 100 44 100.0
all of the results presented thus far suggest that tourism
AVE 13 92.9 29 96.7 42 95.5 researchers demonstrate an in-depth understanding of the
Discriminant validity reliability and validity measures of the reflective constructs
AVEs compared with 13 92.86 28 93.33 41 93.18 in PLS-SEM.
squared correlations
between constructs
Cross-loadings 0 0 9 30.00 9 20.45
Formative Outer Model Evaluation
In a formative outer model, the indicators represent indepen-
Note: AVE = average variance extracted.
dent causes of a theoretical concept (a latent variable) and do
not need to be correlated. Thus, in evaluating a formative
The individual reliability analysis, by observing loadings’ outer model, there is no need, nor does it make sense, to eval-
magnitude (i.e., the correlation coefficients between the indi- uate the reliability and validity, as in reflective outer models.
cators and the corresponding latent variable), is reported in Rather, in evaluating these models, the first aspect that mat-
35 (79.6%) of the 44 studies identified. All the studies con- ters is its theoretical rationality and the expert’s opinion
ducted in 2012 onward used this analysis, while it was per- (Diamantopoulos and Winklhofer 2001). Second, it is neces-
formed in only 5 (35.7%) of the articles prior to this date, sary to observe some statistical criteria, as shown in Table 6.
suggesting significant (p ≤ 0.05) improvement in the applica- Of the 44 studies analyzed, 17 (38.6%) include at least
tion of individual reliability measures in recent years. one formative construct. As can be seen, all the studies are
Overall, regarding the composite reliability, the Goldstein concerned with analyzing the weights and/or the loadings of
rho index (as proposed by Werts, Linn, and Jöreskog 1974) the indicators, while 14 (82.35%) of those 17 studies reported
should be observed. This index corresponds to a measure of as well loadings/weights statistical significance using the
overall correlation between a construct and its indicators, bootstrap procedure. Moreover, the issue of multicollinearity
and it was found in all 30 articles (100%) published in 2012 was reported in only 11 of the published articles (64.7%),
onward and in almost all articles (13 of 14 or 92.9% of the with significant prevalence (p ≤ 0.01) during 2012 (9 studies
total) published prior to 2012, suggesting the nonsignificant reporting multicollinearity). Both statistical significance and
and consistent use of this composite reliability measure multicollinearity measures are important when assessing for-
across time. mative models because, if the outer loading or weights are
Moreover, all the analyzed studies assess convergent insignificant, it means there is no empirical support for the
validity (i.e., the degree of connection between the indicators indicator’s relevance regarding providing content to the for-
and their construct, or whether they represent one and the mative index; in this case, it should be considered a strong
same latent concept), by observing indicators’ statistical sig- candidate for removal. The analysis of multicollinearity in a
nificance, and 42 of the articles (95.5% of the total) analyze set of indicators linked to a formative construct is also impor-
the average variance extracted (AVE), that is, the variance tant because it can produce a lack of statistical significance
shared between the indicators and the construct. The use of in the weight estimate of a formative indicator (Grewal,
this approach is quite similar in both periods under analysis. Cote, and Baumgartner 2004).
Regarding discriminant validity, the aim is to assess the Thus, the above observations suggest that, overall, a forma-
extent to which different theoretical concepts are sufficiently tive outer model assessment in tourism discipline leaves much
measured by distinct indicators (therefore, they are indeed to be desired. Researchers neglect fundamental principles of
Table 6. Formative Outer Model Evaluation in Tourism Table 7. Structural Model Evaluation in Tourism Research.
Research.
Number and Percentage of Studies
Number and % of Studies Reporting
Reporting
Before 2012 and
Before 2012 and 2012 After Total
2012 After Total
(n = 14) (n = 30) (n = 44)
(n = 5) (n = 12) (n = 17)
Aspect to Report n % n % n %
Aspect to Report n % n % n %
Explained variance of each endogenous latent variable
Indicator’s contribution to the latent variable (indicator validity) R 2 for endogenous latent 12 85.7 29 96.7 41 93.2
Indicators’ weight and loading 5 100 12 100 17 100 variables
ts (bootstrap) 5 100 9 75 14 82.35 Effect size ( f 2 ) 0 0 5 16.7 5 11.4
Construct validity Path coefficient estimates
R2 for endogenous latent 0 0 0 0 0 0 Sign, magnitude, and t 12 92.9 30 100 42 95.5
variables bootstrap of each path
Multicollinearity coefficient
Indicator’s variance inflation 2 40 9 75 11 64.71 Predictive relevance for each endogenous latent variable
2
method or tolerance factor Stone–Geisser índex ( Q ) 3 21.4 11 36.7 14 31.8
2
Effect size ( q ) 0 0 0 0 0 0
of advanced criteria that could further help support the valid- Table 8. Advanced Analyses Using PLS-SEM in Tourism
ity of the model. Research.
Second, the review also demonstrated consistent under- of the PLS-SEM method, ranging from evaluation tech-
standing of the reasons behind using PLS-SEM, with focus niques such as blindfolding (e.g., Chin 1998) and confirma-
on prediction, small sample size, non-normal data, and for- tory tetrad analysis (e.g., Gudergan et al. 2008) to interaction
mative measures being the most prevalent reasons over time; effects and PLS-SEM multigroup analyses (e.g., Sarstedt,
suggesting that tourism researchers show full awareness of Henseler, and Ringle 2011) and response-based segmenta-
the basic features of the PLS-SEM technique, and, in recent tion approaches, such as finite mixture-PLS (e.g., Rigdon,
years, they have even become aware of more advanced fea- Ringle, and Sarstedt 2010). Results from this study showed
tures/types of analyses that are available within PLS-SEM that the majority of research articles in the discipline still do
software such as multigroup analysis, interaction effects, and not fully exploit such potentials and thus miss opportunities
response-based segmentation using FIXMIX that helps cope to further substantiate the appropriateness of findings and
with data heterogeneity. improve their analyses. Although results show that the use of
Third, a review of the model characteristics suggests that such advanced applications (10 of the 14 advanced analyses
researchers in tourism should pay closer attention to model- articles were published in 2012–2014) has significantly
specification issues, particularly when using formatively grown in recent years, it is important for tourism researchers
measured constructs in PLS-SEM. In formative constructs, to stay updated in terms of the latest developments in PLS-
indicators are assumed to capture the entire construct domain SEM by consulting statistical journals (e.g., Structural
(or at least major parts of it). This issue needs to be further Equation Modeling: A Multidisciplinary Journal;
stressed in future applications, as it was not identified in any Multivariate Behavioral Research) that frequently publish
of the 17 articles that used either formative only or a combi- innovations on this technique. Tourism scholars should also
nation of formative and measurement models in our review. refer to recently available textbooks on PLS-SEM (see, e.g.,
Similarly, researchers in tourism should be careful when A Primer on Partial Least Squares Structural Equation
using single-item measures, which likely lag behind multi- Modeling [PLS-SEM] by Hair et al. 2014), which contain
item scales in terms of predictive validity. Their use should further comprehensive and detailed explanation of the differ-
be avoided in PLS-SEM in most situations, especially in ent types of analyses available within PLS-SEM. The present
light of the technique’s predictive focus. Again this issue was review and identified recommendations on the use of PLS-
not brought up in any of the six articles that used single-item SEM in tourism should be particularly useful for tourism
latent variables. More careful thought should also be given to researchers adopting PLS-SEM in their studies in the future.
data characteristics. Although PLS-SEM performs well with Hopefully, these recommendations will help enhance the
small samples and nonnormal data, researchers should not be quality of empirical research in tourism and advance theory
careless in implementing these advantages. Small sample and practices in the future.
sizes and skewed data easily increase sampling error, yield-
ing inflated bootstrap standard errors. When this occurs, the
technique’s statistical power is reduced, offsetting one of Limitations and Future Research
PLS-SEM’s major advantages. However, as shown in our As with all research, the present study is not without limita-
review, recent articles have relied on larger sample sizes with tions; for example, it assessed applying PLS-SEM based on
a mean sample size of 562 for articles published in 2012 information reported in published articles. In some cases, the
onward compared with a mean of 451 for articles published authors of the reviewed articles may have made appropriate
prior to 2012, thereby showing tourism researchers’ greater decisions and discussed decisions with referees during the
awareness with respect to this issue. review process but did not include such material in their arti-
Fourth, researchers should make greater use of model- cles, especially with regard to the suitability of PLS-SEM as
evaluation criteria, especially when assessing the quality of it related to their work and model evaluation. This informa-
formatively measured constructs and the quality of inner tion, however, cannot be known just by reviewing the article.
models. Our review showed that current practice still indi- Another limitation is the small number of articles (44) exam-
cate gaps in this regard, casting doubt on the validity of some ined as compared with the number of articles examined in
of the measures. Similarly, researchers should make use of other disciplines (e.g., 204 in marketing; 109 in management
the full range of criteria available to assess the model’s pre- information systems). As the use of PLS-SEM becomes
dictive capabilities, such as the cross-loading measures, more common in tourism studies, a more comprehensive
effect size redundancy index (f2) or the Stone–Geisser index review can be prepared, and the rationale for using PLS-
(Q2). Tourism researchers need to understand that these mea- SEM and assessing its use in the context of tourism research
sures are by definition not indicative of model fit in a covari- will be more effective.
ance-based sense, and any effort to interpret them as such
should clearly be rejected.
Finally, besides basic PLS-SEM analyses, our review Declaration of Conflicting Interests
demonstrated that researchers still have to take further The author(s) declared no potential conflicts of interest with respect
advantage of a much larger set of methodological extensions to the research, authorship, and/or publication of this article.
Chinese Tourists’ Heterogeneous Preferences.” Asia Pacific Ubiquitous Tour Information Services: A Service Quality
Journal of Tourism Research. DOI:10.1080/10941665.2014. Perspective.” Asia Pacific Journal of Tourism Research, 18
889727. (6): 598-617.
Gudergan, S. P., C. M. Ringle, S. Wende, and A. Will. (2008). Kim, M., N. Chung, C. Lee, and J. Kim. (2012). “Do Loyalty
“Confirmatory Tetrad Analysis in PLS Path Modeling.” Groups Differ in the Role of Trust in Online Tourism
Journal of Business Research, 61 (12): 1238-49. Shopping? A Process Perspective.” Journal of Travel and
Hair, J. F., B. Black, B. Babin, R. E. Anderson, and R. L. Tatham. Tourism Marketing, 29 (4): 352-68.
(2010). Multivariate Data Analysis, 6th edition. Upper Saddle Kim, M., C. Lee, N. Chung, and W. Kim. (2013). “Factors Affecting
River, NJ: Prentice Hall. Online Tourism Group Buying and the Moderating Role of
Hair, J. F., G. T. M. Hult, C. M. Ringle, and M. Sarstedt. (2014). A Loyalty.” Journal of Travel Research, 53 (3): 380-94.
Primer on Partial Least Squares Structural Equation Modeling Kim, M., M. Lee, C. Lee, and H. Song. (2011). “Does Gender
(PLS-SEM). Thousand Oaks, CA: Sage. Affect Korean Tourists’ Overseas Travel? Applying the Model
Hair, J. F., C. Ringle, and M. Sarstedt. (2011). “PLS-SEM: Indeed of Goal-Directed Behavior.” Asia Pacific Journal of Tourism
a Silver Bullet.” Journal of Marketing Theory and Practice, 19 Research, 14 (1): 59-75.
(2): 139-51. Kock, N. (2011). “Using WarpPLS in e-Collaboration Studies:
Hair, J. F., M. Sarstedt, T. M. Pieper, and C. R. Ringle. (2012a). Mediating Effects, Control and Second Order Variables, and
“The Use of Partial Least Squares Structural Equation Algorithm Choices.” International Journal of e-Collaboration,
Modeling in Strategic Management: A Review of Past Practices 7 (3): 1-13.
End Recommendations for Future Applications.” Long Range Ku, E. (2014). “Distributed Fascinating Knowledge Over an
Planning, 45 (5-6): 320-40. Online Travel Community.” International Journal of Tourism
Hair, J. F., M. Sarstedt, C. Ringle, and J. A. Mena. (2012b). “An Research, 16 (1): 33-43.
Assessment of the Use of Partial Least Squares Structural Lai, H., and F. Youcheng. (2009). “Understanding the Relationships
Equation Modeling in Marketing Research.” Journal of the of Quality, Value, Equity, Satisfaction, and Behavioral
Academy of Marketing Science, 40:414-33. Intentions among Golf Travelers.” Tourism Management, 30
Henseler, J., C. Ringle, and R. R. Sinkovics. (2009). “The Use (2): 298-308.
of Partial Least Squares Path Modeling in International Lohmöller, J. B. (1989). Latent Variable Path Modeling with
Marketing.” Advances in International Marketing, 20:277-319. Partial Least Squares. Heidelberg: Physica.
Henseler, J., C. M. Ringle, and M. Sarstedt. (2012). “Using Partial López-Bonilla, J. M., and L. M. López-Bonilla. (2012a).
Least Squares Path Modeling in International Advertising “Environmental Orientation in Tourism: The RTEO Scale.”
Research: Basic Concepts and Recent Issues.” In Handbook of Current Issues in Tourism, 15 (6): 591-96.
Research in International Advertising, edited by S. Okazaki. López-Bonilla, J., and L. M. López-Bonilla. (2012b). “Holistic
Cheltenham: Edward Elgar, pp. 252-76. Competence Approach in Tourism Higher Education: An
Hernández-Maestro, R., and O. González-Benito. (2014). “Rural Exploratory Study in Spain.” Current Issues in Tourism, 17
Lodging Establishments as Drivers of Rural Development.” (4): 312-26.
Journal of Travel Research, 53 (1): 83-95. Loureiro, S. M. (2010). “Satisfying and Delighting the Rural
Hernández-Maestro, R. M., P. A. Muñoz-Gallego, and L. Santos- Tourists.” Journal of Travel and Tourism Marketing, 27 (4):
Requejo. (2009). “Small-Business Owners’ Knowledge and 396-408.
Rural Tourism Establishment Performance in Spain.” Journal Loureiro, S. M., and F. J. González. (2008). “The Importance of
of Travel Research, 48 (1): 58-77. Quality, Satisfaction, Trust, and Image in Relation to Rural
Huang, Y., F. Huang, M. Hsu, and F. Chang. (2009). “Determinants Tourist Loyalty.” Journal of Travel and Tourism Marketing,
and Outcomes of Relationship Quality: An Empirical 25 (2): 117-36.
Investigation on the Chinese Travel Industry.” Asia Pacific Marcoulides, G. A., and C. Saunders. (2006). “PLS: A Silver
Journal of Tourism Research, 14 (1): 59-75. Bullet?” Management Information Systems Quarterly, 30 (2):
Hulland, J. (1999). “Use of Partial Least Squares (PLS) in Strategic iii-ix.
Management Research: A Review of Recent Studies.” Strategic Mazanec, J. A., and A. Ring. (2011). “Tourism Destination
Management Journal, 20 (2): 195-204. Competitiveness: Second Thoughts on the World Economic
Jöreskog, K. G. (1973). “A General Method for Estimating a Linear Forum Reports.” Tourism Economics, 17 (4): 725-51.
Structural Equation System.” In Structural Equation Models Mistilis, N., and J. D’ambra. (2008). “The Visitor Experience
in the Social Sciences, edited by A. S. Goldberger and O. D. and Perception of Information. Quality at the Sydney Visitor
Duncan. New York: Academic Press, pp. 85-112. Information Centre.” Journal of Travel and Tourism Marketing,
Jöreskog, K. G. (1978). “Structural Analysis of Covariance and 24 (1): 35-47.
Correlation Matrices.” Psychometrika, 43 (4): 443-77. Murphy, P., M. P. Pritchard, and B. Smith. (2000). “The Destination
Jöreskog, K. G., and H. Wold. (1982). “The ML and PLS Product and Its Impact on Traveller Perceptions.” Tourism
Techniques for Modeling with Latent Variables: Historical and Management, 21 (1): 43-52.
Comparative Aspects.” In Systems under Indirect Observation: Nunkoo, R., and H. Ramkissoon. (2011). “Structural Equation
Part I, edited by K. G. Jöreskog and H. Wold. Amsterdam: Modelling and Regression Analysis in Tourism Research.”
North-Holland, pp. 263-70. Current Issues in Tourism, 15 (8): 1-26.
Kim, I., K. Ahn, and N. Chung. (2013). “Examining the Factors Nunkoo, R., H. Ramkissoon, and D. Gursoy. (2013). “Use of
Affecting Perceived Enjoyment and Usage Intention of Structural Equation Modeling in Tourism Research: Past,
Present, and Future.” Journal of Travel Research, 52 (6): 759- Taheri, B., A. Jafari, and K. O’Gorman. (2014). “Keeping Your
71. Audience: Presenting a Visitor Engagement Scale.” Tourism
Pike, S., S. Hosany, S. Murdy, and I. Lings. (2011). “Visitor Management, 42:321-29.
Relationship Orientation of Destination Marketing Tan, W., and C. Kuo. (2013) “The Effect of Aesthetic-Image
Organizations.” Journal of Travel Research, 50 (4): 443-53. Value of Travel Expert Blogs on the Intention to Travel: An
Prayag, G., R. Nunkoo, and T. Alders. (2013). “London Residents’ Exploratory Study.” Current Issues in Tourism. DOI:10.1080/
Support for the 2012 Olympic Games: The Mediating Effect 13683500.2013. 840270.
of Overall Attitude.” Tourism Management, 36 (June): Thomson Reuters. (2014). Web of Knowledge. thomsonreuters.
629-40. com/journal-citation-reports/. (accessed June 20, 2014).
Regan, N., J. Carlson, and P. J. Rosenberger, III. (2012). “Factors Urbrach, N., and F. Ahlemann. (2010). “Structural Equation
Affecting Group-Oriented Travel Intention to Major Events.” Modeling in Information Systems Research Using Partial
Journal of Travel and Tourism Marketing, 29 (2): 185-204. Least Squares.” Journal of Information Technology Theory
Reinartz, W. J., M. Haenlein, and J. Henseler. (2009). “An Empirical and Application, 11 (2): 5-40.
Comparison of the Efficacy of Covariance-Based and Variance Vinzi, V. E., L. Trinchera, and S. Amato. (2010). “PLS Path Modeling:
Based SEM.” International Journal of Market Research, 26 From Foundations to Recent Developments and Open Issues for
(4): 332-44. Model Assessment and Improvement.” In Handbook of Partial
Rigdon, E. E., C. M. Ringle, and M. Sarstedt. (2010). “Structural Least Squares, edited by V. E. Vinzi, W. W. Chin, J. Henseler,
Modeling of Heterogeneous Data with Partial Least Squares.” and H. Wang. Berlin: Springer-Verlag, pp. 47-81.
In Review of Marketing Research, edited by N. K. Malhotra. Werts, C. E., R. L. Linn, and K. G. Joreskog. (1974). “Interclass
Armonk, Armonk, NY: ME Sharpe, pp. 255-96. Reliability Estimates: Testing Structural Assumptions.”
Ringle, C. M., M. Sarstedt, and W. W. Straub. (2012). “A Critical Education and Psychological Measurement, 34 (1): 25-33.
Look at the Use of PLS-SEM in MIS Quarterly.” MIS Wold, H. O. (1982). “Soft Modeling: The Basic Design and Some
Quarterly, 36 (1): 1-19. Extensions.” In Systems under Indirect Observations, Part II,
Ringle, C. M., S. Wende, and A. Will. (2005). SmartPLS 2.0 (Beta), edited by K. G. Jöreskog and H. O. Wold. Amsterdam: North-
SmartPLS. Hamburg (www.smartpls.de). Holland, pp. 1-54.
Roxas, B., and D. Chadee. (2013). “Effects of Formal Institutions Wold, H. O. (1985). “Partial Least Squares.” In Encyclopedia
on the Performance of the Tourism Sector in the Philippines: of Statistical Sciences, Vol. 6, edited by S. Kotz and N. L.
The Mediating Role of Entrepreneurial Orientation.” Tourism Johnson. New York: John Wiley, pp. 581-891.
Management, 36:629-40. Zach, F., and P. Racherla. (2011). “Assessing the Value of
Sarstedt, M., J. Henseler, and C. Ringle. (2011). “Multi-Group Collaborations in Tourism Networks: A Case study of Elkhart
Analysis in Partial Least Squares (PLS) Path Modeling: County, Indiana.” Journal of Travel and Tourism Marketing,
Alternative Methods and Empirical Results.” Advances in 28 (1): 97-110.
International Marketing, 22:195-218.
Song, H., G. Li, R. Van der Veen, and J. L. Chen. (2011). “Assessing Author Biographies
Mainland Chinese Tourists’ Satisfaction with Hong Kong
Patrícia Oom do Valle, PhD, is an Associate Professor of
Using Tourist Satisfaction Index.” International Journal of
Quantitative Methods Applied to Economics and Business at the
Tourism Research, 13 (1): 82-96.
Faculty of Economics, University of the Algarve. Her research
Song, H., R. Van der Veen, G. Li, and J. L. Chen. (2012). “The
focuses on the use of advanced multivariate statistical methods in
Hong Kong Tourist Satisfaction Index.” Annals of Tourism
modelling consumer behavior, especially in the fields of tourism
Research, 39 (1): 459-79.
and environment. Tel: + 351 289 800 915 ext. 7277; fax: +351 289
Sosik, J. J., S. S. Kahai, and M. J. Piovoso. (2009). “Silver Bullet
800 064; email: pvalle@ualg.pt.
or Voodoo Statistics? A Primer for Using Partial Least Squares
Data Analytic Technique in Group and Organization Research.” Guy Assaker, PhD, is an Assistant Professor of Business and
Group and Organization Management, 34 (1): 5-36. Tourism at the Lebanese American University (LAU). His research
Stienmetz, J., S. Levy, and S. Boo. (2013). “Factors Influencing focuses on the use of advanced multivariate statistical methods to
the Usability of Mobile Destination Management Organization model various business and tourism subjects. Tel: + 961 54 72 54
Websites.” Journal of Travel Research, 52 (4): 453-64. ext. 2443; fax: +961 9 54 72 56; email: guy.assaker@lau.edu.lb.