WEB SITE BRAND ATTRIBUTES AND E-SHOPPER LOYALTY
A Comparative Study of Spain and Scotland
Sandra Loureiro and Silvina Santana
Department of Economy, Management, and Industrial Engineering
Aveiro University, Campus of Santiago, Aveiro, Portugal
Keywords: Web site brand, Online shopping, B2C, Brand relationship, Brand image, Brand personality, Brand
association, Loyalty, PLS.
Abstract: This study examines the impact of web site brand personality, web site brand association, web site brand
image, and web site brand relationship on e-shopper loyalty to the web site. The model was estimated on
data from consumers of online products in Spain and Scotland using PLS technique. The findings suggest
that web site brand association and web site brand personality are good predictors of web site brand image.
However, web site brand image does not explain the intention of Spanish students to recommend a web site
and to use it to by again.
1 INTRODUCTION
Brands are important sources of competitive
advantage. Therefore, knowing how actual and
potential clients perceive a brand is fundamental
information for its management. In brand theory, a
brand is said to have attributes such as brand
personality, brand association, and brand image to
which brand knowledge is always linked (e.g.,
(Aaker, 1991; Keller, 1993, 1998)). Some authors
defend that consumer-brand relationship depends
largely on the successful establishment of brand
knowledge (Keller, 2003).
Brand knowledge can be formed directly from a
consumer’s experience. Therefore, brand attributes
might be crucial mediators between brand
experience and consumer-brand relationship. If such
a relation proves, understanding the way these
concepts interrelate with each other might be
valuable to inform marketing strategy formulation,
namely, in what concerns brand management.
The main objective of this work is to study the
direct impact of web site brand relationship and web
site brand image on loyalty. In addiction, we also
study the direct effect of web site brand association
and web site brand personality on web site brand
image. The model was estimated on data from 195
consumers of online products from two countries,
Scotland and Spain, using PLS technique.
To the best of our knowledge, this is the first
time that web site brand knowledge, mediated by
attributes like web site brand association,
personality, image, and relationship, is addressed in
such a way and the study differs from previous work
which has related brand knowledge of goods and
services (Bart, Shankar, Sultan, & Urban, 2005;
Chang & Chieng, 2006), sold through virtual stores
(web site) or physical stores. Secondly, this study
focuses on consumers’ experiences in two European
countries with very different levels of Internet use
for shopping.
Given the paucity of cross-country studies in this
area, using PLS (Partial Least Squares) might prove
to be valuable to considerably advance existing
knowledge and enhance current practices of web use
for retailing.
2 LITERATURE REVIEW
2.1 Constructs Definition
Brand image is defined here as perceptions about a
brand as reflected by the brand associations held in
consumer memory (Keller, 1993).
Brand personality is defined as the set of human
characteristics associated with a brand (Aaker,
1997). It is a comprehensive concept, which includes
257
Loureiro S. and Santana S.
WEB SITE BRAND ATTRIBUTES AND E-SHOPPER LOYALTY - A Comparative Study of Spain and Scotland.
DOI: 10.5220/0002783302570262
In Proceedings of the 6th International Conference on Web Information Systems and Technology (WEBIST 2010), page
ISBN: 978-989-674-025-2
Copyright
c
2010 by SCITEPRESS Science and Technology Publications, Lda. All rights reserved
all the tangible and intangible traits of a brand, such
as beliefs, values, prejudices, features, interests, and
heritage. A brand personality makes it unique. Brand
personality is seen as a valuable factor in increasing
brand engagement and brand attachment, in much
the same way as people relate and bind to other
people. Researchers have proposed that brand
personality is an aspect of brand image (Keller,
1993, 1998; Plummer, 2000) and results from
empirical studies indicate that brand personality
have a statistically significant positive influence on
brand image (O'Cass & Lim, 2001).
According to previous studies (Chang & Chieng,
2006; Keller, 1998), brand association is defined as
the information linked to the node in memory. This
information reflects an association between a range
of aspects and the brand in the mind of the
consumer. Brand associations have been presented
as critical components in developing a brand image
(Keller, 1993) and empirical studies have shown that
brands associations lead to the formation of a
distinct brand image in the minds of consumers
(Hsieh, 2002).
In this study brand relationship is defined as the tie
between a person and a brand that is voluntary or is
enforced interdependently between the person and
the brand (Chang & Chieng, 2006). A relationship
between the brand and the consumer results from the
accumulation of consumption experience.
Finally, loyalty is the intention to recommend a
product to other people and to buy it again
(Zeithaml, Berry, & Parasuraman, 1996).
In this work, the above concepts are transposed
to the context of web site brand. We postulate that
web site brand personality, web site brand
association, web site brand image, and web site
brand relationship all hold different information that
link to web site brand, as happens with other
products (Aaker, 1991). Furthermore, we defend that
brand personality, brand association, brand image,
and brand relationship are antecedents of loyalty to a
web site brand (Chang & Chieng, 2006; O'Cass &
Lim, 2001).
2.2 Structural Equations Explained
A structural equation model approach using Partial
Least Squares (PLS) (Ringle, Wende, & Will, 2005)
is used to test the hypotheses of this study. PLS is
based on an iterative combination of principal
components analysis and regression, and it aims to
explain the variance of the constructs in the model.
In terms of advantages, PLS simultaneously
estimates all path coefficients and individual item
loadings in the context of a specified model, and as a
result, it enables researchers to avoid biased and
inconsistent parameter estimates. Based on recent
developments (Chin, Marcolin, & Newsted, 2003),
PLS has been found to be an effective analytical tool
to test interactions by reducing type II error. By
creating a latent construct which represents the
interaction term, a PLS approach significantly
reduces this problem by accounting for the error
related to the measures. In fact, PLS models are
based on prediction-oriented measures, not
covariance fit like covariance structure models
developed by Karl Jöreskog, or LISREL program
developed by Jöreskog and Sörborn.
LISREL estimates causal model parameters
aiming at minimizing the discrepancies between the
initial empirical covariance data matrix and the
covariance matrix deduced from the model structure
and the parameter estimates (Barclay, Higgins, &
Thompson, 1995). PLS seeks to maximize variance
explained in constructs and/or variables, depending
on model specification. In addition, LISREL offers a
number of measures of overall model “fit” such as
the χ
2
goodness-of-fit, which are related to the
ability of the model to account for the sample
covariance. PLS does not possess these kind of
overall fit measures, relying instead on variance
explained (i.e., R
2
) as an indicator of how well the
technique has met its objective (Barclay et al.,
1995). In spite of that, there are several fit indices
available on PLS software (Ringle et al., 2005) such
as communality and redundancy measures and
Stone-Geisser’s Q
2
measure, which can be used to
evaluate the predictive power of the model.
As a substitute to parametric global goodness of
fit measures that are used in LISREL technique, the
geometric mean of the average communality (outer
model) and the average R
2
(inner model) (going
from 0 to 1) has been proposed (Tenenhaus, Vinzi,
Chatelin, & Lauro, 2005) as overall goodness of fit
(GoF) measures for PLS (Cross validated PLS GoF),
according to equation (1).
2
.GoF communality R=
(1)
2.3 Hypothesis Proposed
Five hypotheses are formulated in this study and
tested with PLS:
H1: Web site brand personality significantly and
positively influences web site brand image
H2: Web site brand association significantly and
positively influences web site brand image
WEBIST 2010 - 6th International Conference on Web Information Systems and Technologies
258
H3: Web site brand image significantly and
positively influences web site brand relationship
H4: Web site brand image significantly and
positively influences consumer loyalty
H5: Web site brand relationship significantly and
positively influences consumer loyalty
3 METHODS
3.1 Participants and Procedure
The surveys were conducted in June 2008 through
face-to-face interviews in universities in Spain and
Scotland. The same two interviewers, specially
trained, were used in the two countries. We choice
Spain and Scotland to consider different cultural
contexts.We collected 95 completed questionnaires
from students in Spain and 100 from students in
Scotland. Each sub-sample has the same average age
of 24 years. The respondents split almost equally in
terms of gender for both countries.
In this study, the web site brands involved
belong to different industry branch, such as: clothes,
books, music, and airlines.
3.2 Measures
Web site brand association was measured using two
dimensions (product and organization) (Barclay et al.,
1995; Carmines & Zeller, 1979)
. Web site brand
personality was operationalized using 5 dimensions
(sincerity, excitement, competence, sophistication,
and ruggedness) (Aaker, 1997), web site brand
image with 3 dimensions (function, experience, and
symbolic) (Chang & Chieng, 2006; Keller, 1993),
web site brand relationship with 6 dimensions
(functional, love, commitment, attachment, self-
connection, and partner quality) (Chang & Chieng,
2006), and loyalty with 2 dimensions (recommend
and by again) (Zeithaml et al., 1996). Each statement
of the questionnaire was recorded on a 5-point
Likert scale (1=strongly disagree, 5=strongly agree).
The instrument was elaborated in English and
translated to Spanish using a dual focus method
(Erkut, Alarcón, Coll, Tropp, & Garcia, March
1999).
3.3 Data Analysis
The PLS model is analyzed and interpreted in two
stages. First, the adequacy of the measures is
assessed by evaluating the reliability of the
individual measures and the discriminant validity of
the constructs (Hulland, 1999). Then, the structural
model is appraised.
Composite reliability is used to analyze the
reliability of the constructs since this has been
considered a more exact measurement than the
Cronbach’s alpha (Fornell & Larcker, 1981). To
determinate convergent validity, we compute the
average variance of manifest variables extracted by
constructs (AVE) that should be at least 0.5,
indicating that more variance is explained than
unexplained in the variables associated with a given
construct. To assess discriminant validity we follow
the rule that the square root of AVE should be
greater than the correlation between the construct
and other constructs in the model (Fornell &
Larcker, 1981).
Bootstrap (a nonparametric approach) is used to
estimate the precision of the PLS estimates and
support the hypotheses. Accordingly, 500 samples
sets were created in order to obtain 500 estimates for
each parameter in the PLS model. Each sample was
obtained by sampling with replacement to the
original data set (Chin, 1998; Fornell & Larcker,
1981).
Finally, the differences between the Scottish and
the Spanish samples are compared using a t-test of
m+n+2 degrees of freedom (where m=Spain sample
size and n=Scotland sample size). This test uses the
path coefficients and the standard errors of the two
structural paths calculated by PLS with the samples
of both countries, according to equation (2).
(2)
4 RESULTS
All the loadings of reflective constructs approach or
exceed 0.707 (Table 1), which indicates that more
than 50% of the variance in the manifest variable is
explained by the construct (Carmines & Zeller,
1979), except for the construct brand personality and
brand relationship. Ruggedness, functional,
attachment, self connection and partner quality were
eliminated from the Scottish sample. Results in
Table 1 shows that all constructs are reliable since
the composite reliability values exceed the threshold
of 0.7 and even the strictest one of 0.8 (Nunnally,
1978).
WEB SITE BRAND ATTRIBUTES AND E-SHOPPER LOYALTY - A Comparative Study of Spain and Scotland
259
The measures also demonstrate convergent
validity and discriminant validity (Table 2),
according to the criteria defined in Methods.
The structural results for Spain are presented in
Figure 1. All the path coefficients are found to be
significant at the 0.001 level and all the coefficients’
signs are in the expected direction, excepting for the
causal order between brand image and loyalty.
Multiplication of the Pearson correlation value
for the path coefficient value of each of the two
constructs reveals that 49.4% of the brand image
variability is explained by brand association, 34.1%
of the brand relationship variability is explained by
brand image, and 18.6% of the loyalty variability is
explained by brand relationship.
Brand
Association
Brand
Relationship
R
2
= 34.1 %
Q
2
= 0.190
Brand
Image
R
2
= 68.4 %
Q
2
= 0.474
Brand
Personality
Loyalty
R
2
= 18.3 %
Q
2
= 0.153
0.584***
34.1 %
0.631***
49.4 %
0.308***
19.0 %
-0.011ns
-0.3 %
0.434***
18.6 %
GoF
Spain
= 0.5374
*** p < 0.001 ; ns : n o significant
Figure 1: Structural results for Spain.
Brand
Association
Brand
Relationship
R
2
= 36.0 %
Q
2
= 0.292
Brand
Image
R
2
= 60.6 %
Q
2
= 0.394
Brand
Personality
Loyalty
R
2
= 40.6 %
Q
2
= 0.325
0.600***
36.0 %
0.438***
31.7 %
0.404***
28.9 %
0.430***
25.7 %
0.278**
14.9 %
GoF
Scotland
= 0.5932
*** p < 0.001 ; ** p < 0.01
Figure 2: Structural results for Scotland.
Table 1: Measurement Results.
Variable
LV
Index
Values
Item
Loading
Composite
reliability
AVE
Spain
Brand association
3.5
0.843
0.729
AS1:Product 0.909
AS2:Organization 0.795
Brand personality
3.4
0.901
0.646
PS1:Excitement 0.822
PS2: Sophistication 0.763
PS3: Ruggedness 0.802
RS4:Sincerity 0.793
RS5: Competence 0.836
Brand Image 3.3 0.874
0.698
IS1: Function 0.777
IS2: Experience 0.889
IS3:Symbolic 0.837
Brand Relationship 2.8 0.903 0.609
RS1:Functional 0.710
RS2:Love 0.867
RS3:Commitment 0.833
RS4:Attachment 0.744
RS5:Self Connection 0.727
RS6:Partner quality 0.789
Loyalty 3.8 0.949 0.903
LS1:Recommendation 0.967
LS2:By again 0.933
Scotland
Brand association
3.6
0.954
0.912
ASc1:Product 0.954
ASc2:Organization 0.956
Brand personality
3.3
0.886
0.682
PSc1:Excitement 0.880
PSc2: Sophistication 0.730
PSc3: Sincerity 0.837
PSc4: Competence 0.799
Brand Image 3.2 0.867
0.686
ISc1: Function 0.781
ISc2: Experience 0.837
ISc3:Symbolic 0.864
Brand Relationship 2.8 0.902 0.821
RSc1:Love 0.891
RSc2:Commitment 0.927
Loyalty 3.6 0.868 0.768
LSc1:Recommendation 0.921
LSc2:By again 0.830
The structural results for Scotland are presented in
Figure 2. All the path coefficients are significant at
the 0.001 level and all the coefficients’ signs are also
in the expected direction, excepting for the causal
order between brand relationship and loyalty which
is significant at the 0.01 level. As in the Spanish
sample, the Bootstrap approach with n = 500 was
used and all the hypothesized relations were
supported. Multiplication of the Pearson correlation
value for the path coefficient value of each of the
two constructs reveals that 31.7% of the brand image
variability is explained by brand association, 36.0%
of the brand relationship variability is explained by
WEBIST 2010 - 6th International Conference on Web Information Systems and Technologies
260
Table 2: Discriminant validity: square root of AVE and
correlations of constructs.
Correlations of constructs
Construct
Brand
associati
on
Brand
image
Brand
perso
nality
Brand
relations
hip
Loyalty
Spain
AVE
1/2
0.96 0.83 0.81 0.91 0.88
Brand
association 1.00 0.72 0.71 0.47 0.50
Brand image
0.72 1.00 0.71 0.60 0.60
Brand
personality 0.71 0.71 1.00 0.55 0.54
Brand
relationship 0.47 0.60 0.55 1.00 0.54
Loyalty 0.50 0.60 0.54 0.54 1.00
Scotland
AVE
1/2
0.85 0.84 0.80 0.78 0.95
Brand
association 1.00 0.78 0.49 0.43 0.31
Brand image
0.78 1.00 0.62 0.58 0.24
Brand
personality 0.49 0.62 1.00 0.60 0.46
Brand
relationship 0.43 0.58 0.60 1.00 0.43
Loyalty 0.31 0.24 0.46 0.43 1.00
brand image, and 25.7% of the loyalty variability is
explained by brand image.
The results of t-test (Table 3) show that there are
not statistically significant differences between the
two countries in any of the two structural paths (at
critical t-value=|1.960|), excepting for the causal
order between brand image and loyalty.
5 CONCLUSIONS
This study represents the first attempt to considerer
the web site brand in a structural model using the
PLS approach, which analyzes simultaneously the
causal orders among web site brand association, web
site brand image, web site brand personality, web
site brand relationship, and loyalty.
The results show that web site brand association
and web site brand personality are good predictors
of web site brand image and that the hypotheses H1
and H2 are confirmed for the Scottish and the
Spanish samples. Hypotheses H3 and H5 are also
supported, but the hypothesis H4 is not supported by
the Spanish sample. Thus, web site brand
relationship seems to be more important than brand
image in explaining the intention to recommend the
web site and to buy again. The Scottish students give
more importance to web site brand image than the
Spanish students. However, web site brand
Table 3: Results of multi-group analysis: Spain and
Scotland.
Structural
paths
Standard
error
Spain
Standard
error
Scotland
Sp
1
β
Spain
-
β
Scotland
t-test
Brand
association
Brand
image
0.098 0.095 0.950 0.192 1.414
Brand
personality
Brand
image
0.090 0.097 0.921 -0.097
-
0.733
Brand
image
Brand
relationship
0.091 0.060 0.749 -0.016
-
0.147
Brand
image
Loyalty
0.132 0.098 1.137 -0.441
-
2.709
Brand
relationship
Loyalty
0.115 0.099 1.048 0.157 1.044
1 Unbiased estimator of average error standard variance
association exercises a stronger effect on web site
brand image than the web site brand personality, for
the two groups of students in the different countries.
Traditionally, brand image and brand personality
are different constructs. However, the PLS technique
seems to give evidence of some correlation between
the competence (eliminated in this analyze) of brand
personality and the symbolic part of brand image.
Further directions for future work have been
indicated by this first study of web site brand
knowledge. The model is being redesigned to
include other constructs and we are planning to
extend our research to other countries, such as
Brazil, USA, Germany, Portugal and Poland. With a
cross-country approach we will be able to analyze
the impact of culture on consumers’ perception and
test the effect of globalization, advancing existing
knowledge and generating valuable information for
decision makers, marketers and web designers.
REFERENCES
Aaker, D. (1991). Managing brand equity: Capitalizing on
the value of a brand name. New York: The Free Press.
Aaker, J. (1997). Dimensions of brand personality.
Journal of Marketing Research, 34, 347-356.
Barclay, D., Higgins, C., & Thompson. (1995). The partial
least squares (PLS) approach to causal modeling,
personal computer adoption and use as an illustration.
Technology Studies, 2, 285-309.
WEB SITE BRAND ATTRIBUTES AND E-SHOPPER LOYALTY - A Comparative Study of Spain and Scotland
261
Bart, Y., Shankar, V., Sultan, F., & Urban, G. (2005). Are
the drivers and the role of online trust the same for all
web sites and consumers? A large-scale exploratory
empirical study. Journal of Marketing, 69, 133-152.
Carmines, E., & Zeller, R. (1979). Reliability and Validity
Assessment. London: Ed. Sage Publications, Inc.
Chang, P.-L., & Chieng, M.-H. (2006). Building
consumer-brand relationship: A cross-cultural
experential view. Psychology and Marketing, 23(11),
927-959.
Chin, W. (1998). The Partial Least Squares approach to
structural equation modeling. In G. A. Marcoulides
(Ed.), Modern Methods for Business Research (pp.
295-336). NJ: Mahwah, Lawrence Erlbaum Associates
Publisher.
Chin, W., Marcolin, B., & Newsted, B. (2003). A partial
least squares latent variable modelling approach for
measuring interaction effects: results from a Monte
Carlo simulation study and an electronic mail emotion/
adoption study. Information Systems Research, 14(2),
189-217.
Erkut, S., Alarcón, O., Coll, C., Tropp, L., & Garcia, H. A.
V. (March 1999). The dual-focus approach to creating
bilingual measures. Journal of Cross-Cultural
Psychology, 30(2), 206-218.
Fornell, & Larcker, D. (1981). Evaluating structural
models with unobservables variables and measurement
error. Journal of Marketing Research, 28, 39-50.
Hsieh, M. (2002). Identifying brand image dimensionality
and measuring the degrees of brand globalization: A
cross-nation study. Journal of International
Marketing, 10, 46-67.
Hulland, J. (1999). Use of partial least squares (PLS) in
strategic management research: a review of four recent
studies. Strategic Management Journal, 20(2), 195-
204.
Keller, K. (1993). Conceptualizing, measuring,, and
managing customer-based brand equity. Journal of
Marketing, 57, 1-22.
Keller, K. (1998). Strategic brand management: Building,
measuring and managing brand equity. New York:
Prentice Hall.
Keller, K. (2003). Brand synthesis: the
multidimensionality of brand knowledge. Journal of
Consumer Research, 29(4), 595-600.
Nunnally, J. (1978). Psychometric Theory (2nd ed.). New
York: McGraw-Hill.
O'Cass, A., & Lim, K. (2001). The influence of brand
association on brand preference and purchase
intention: As Asia perspective on brand association.
Journal of International Consumer Marketing, 14, 41-
71.
Plummer, J. (2000). How personality makes a difference.
Journal of Advertising Research, 40(6), 79-83.
Ringle, C., Wende, S., & Will, A. (2005). SmartPLS 2.0
(beta). from www.smartpls.de
Tenenhaus, M., Vinzi, E., Chatelin, Y.-M., & Lauro, C.
(2005). PLS path modeling. Computational Statistics
& Data Analysis, 48, 159-205.
Zeithaml, V., Berry, L., & Parasuraman, A. (1996). The
behavioral consequences of service quality. Journal of
Marketing, 60(2), 31-46.
WEBIST 2010 - 6th International Conference on Web Information Systems and Technologies
262