Content uploaded by Jose Mauro C. Hernandez
Author content
All content in this area was uploaded by Jose Mauro C. Hernandez
Content may be subject to copyright.
Adoption of internet banking:
proposition and implementation
of an integrated methodology
approach
Jose
´Mauro C. Hernandez
Centro Universita
´rio Nove de Julho (UNINOVE), Sa
˜o Paulo, Brazil, and
Jose
´Afonso Mazzon
Universidade de Sa
˜o Paulo, Sa
˜o Paulo, Brazil
Abstract
Purpose – This study proposes a new method to investigate adoption of new technologies and tests
this method by looking into the determinants of internet banking adoption in Brazil.
Design/methodology/approach – A total of 600 respondents living in one of the biggest cities in
Brazil were sampled for interviewing: 300 were internet bank users, 150 were internet but not internet
bank users, and 150 were neither internet nor internet bank users.
Findings – The adjusted coefficient of determination in the multiple linear regression equation
(dependent variable: intention to use/continue to use IB) was 60 percent and the main effects of the
eight variables proved significant (relative advantage of control, compatibility with lifestyle, image,
subjective norm, self-efficacy, relative advantage of security and privacy, results demonstrability, and
trialability). The McFaden pseudo coefficient of determination in the multinomial logistic regression
equation (dependent variable: a dummy variable for each of the three groups analyzed) ranged from
45 percent to 69 percent. The findings show that the variables that influence the intention to
use/continue to use IB are not exactly the same as those that influence actual adoption. Specifically, the
results seem to suggest that intention to use IB is influenced solely by people’s beliefs about IB, while
its actual adoption is influenced also by individual characteristics.
Originality/value – The findings herein suggest that the proposed integrated model offers superior
ability to explain adoption of internet banking to that of the models elected by previous studies.
Furthermore, the model looks not only into the intentions but also into actual adoption.
Keywords Consumer behavior, Internet, Electronic commerce
Paper type Research paper
Introduction
The 1980s witnessed a marked shift in the distribution channels of banking services
towards self-servicing alternatives. Pressured by rising costs, ever more demanding
customers, and the need to preserve profitability while standing out from the
competition, banks found themselves forced to invest in new customer service channels
such as internet banking (IB).
In Brazil, the IB growth rate over the past years has exceeded that of the internet
itself. According to Febraban (2004), the number of IB registered current accounts has
jumped from 8.3 million in 2002 to 18.1 million in 2004, of which 16.2 million belong to
individual account holders. The 2 billion or so transactions processed via IB in 2004
accounted for 7 percent of the year’s transaction total, surpassing the number of check
The current issue and full text archive of this journal is available at
www.emeraldinsight.com/0265-2323.htm
IJBM
25,2
72
Received July 2006
Revised October 2006
Accepted November 2006
International Journal of Bank
Marketing
Vol. 25 No. 2, 2007
pp. 72-88
qEmerald Group Publishing Limited
0265-2323
DOI 10.1108/02652320710728410
transactions. Financial sites are among the most visited by internet users, with 50
percent of the 13.5 million home users accessing them regularly (IBOPE, 2005).
Additionally, taking into consideration that in 2004 alone, Brazilian banks invested
2.1 billion dollars in technology (Febraban, 2004), including the development of remote
access technology such as IB, a relevant question is what prevents people from using
the internet to control their bank accounts. In the marketing literature, this topic is
usually referred to as adoption of new technologies (Rogers, 1983).
Recent literature on marketing in the banking industry shows that several scholars
have investigated the adoption of internet banking. However, the utilization of several
theoretical models combined with a broad variety of data collection and analysis tools
makes a comparison difficult between distinct studies and consolidation of their
respective results.
Therefore, this study has two key objectives – to propose a method that lends itself
to studying adoption of new technologies and test it to find out the factors that
influence adoption of internet banking in Brazil. The proposed method is based on
three pillars: a model that reconciles the key constructs in the best known theoretical
models of new technology adoption; a sampling process that includes both current and
prospective IB users; and the use of multiple linear regression and multinomial logistic
regression techniques as data analysis tools.
Below, in the first section, we review the better known models of new technology
adoption and the literature on IB adoption. In the following section, we set out the main
principles of method proposed to look into new banking technologies. We then describe
how the method was brought into operation and show the results of the empirical
study. Finally, we discuss the results obtained and share the conclusions of the study.
Adoption of new technologies
Several competing theoretical approaches have been used to investigate the
determinants of acceptance and use of new information technology (Venkatesh et al.,
2003). One of the most important lines of study in this area focuses on the determinants
of individual acceptance of new technologies by using behavioral intention (intention
to adopt a new technology) or behavior itself (actual adoption of a new technology) as
dependent variables (Davis, 1989; Taylor and Todd, 1995). These models are based on
the Theory of Reasoned Action (TRA) (Fishbein and Ajzen, 1975) and on the Theory of
Planned Behavior (TPB) (Ajzen, 1985) and the adoption determinants are based on
beliefs, attitudes, subjective norm, and perceptions of behavioral control.
A second line of research has considered adoption of new information technologies
from the perspective of the Innovation Diffusion Theory (Rogers, 1983; Tornatzky and
Klein, 1982). This line also uses behavioral intention or behavior itself as dependent
variables but the determinants are usually established according to the characteristics
of the new technology.
Below we look into the main features of the theoretical models most frequently used
in the studies of innovation adoption, and more specifically into studies in the area of
information technologies such as internet banking.
Innovation Diffusion Theory (IDT)
The Innovation Diffusion Theory (IDT) has been used since the 1960s to explain the
process of innovation adoption. Following an extensive review of the literature, Rogers
Adoption of
internet banking
73
(1983) found five attributes that consistently proved to be determinants of the diffusion
rate of an innovation:
(1) Relative advantage: refers to the extent to which the innovation is perceived as
superior to all other options.
(2) Compatibility: extent to which the innovation is perceived as being in line with
the values, needs and experiences of prospective adopters.
(3) Complexity: the extent to which the innovation is perceived as difficult to
understand or use.
(4) Observability: extent to which the benefits or attributes of the innovation can be
observed, pictured or described to prospective adopters.
(5) Trialability: the extent to which the innovation can be experienced before its
actual adoption.
The relationship between each of these characteristics and the intention to adopt an
innovation is positive, with the exception of the complexity construct, which bears a
negative relationship to the intention to adopt.
With the purpose of developing an instrument to measure the initial adoption and
possible diffusion of information technology innovations within organizations, Moore
and Benbasat (1991) built on the five basic characteristics originally proposed by
Rogers (1983) by adding two other constructs. The first, image, refers to “the degree to
which use of an innovation is perceived to improve image or status in a given social
system” (p. 195). According to the authors, the reasoning for including this construct is
that, although Rogers (1983) sees image as one of the aspects in relative advantage,
Tornatzky and Klein (1982) suggested that image must be considered separately from
relative advantage owing to the importance of its effect in previous studies. Therefore,
by hypothesis, the more clearly people realize that adopting an innovation will improve
their status within their group, the stronger their intention to adopt it.
The second construct added to Moore and Benbasat’s measuring instrument was
the voluntariness of use, defined as “the degree to which use of an innovation is
perceived as being voluntary or an act of free will.” (p. 195). The reasoning for
including this construct is that, when considering an innovation, the fact whether
individuals are free to either adopt or reject it must be taken into consideration. By
hypothesis, the greater is their freedom to adopt it, the higher the odds that the new
technology will be adopted. In situations where adoption is free, this construct will
prove irrelevant.
During fine-tuning of the scale, Moore and Benbasat broke down the observability
construct into two other: result demonstrability and visibility. The first, result
demonstrability, refers particularly to the extent to which an innovation can be
observed before it is adopted and the second construct, Visibility, focuses on the extent
to which the benefits of an innovation are visible to prospective adopters. The relation
between these two characteristics and the intention to adopt is also positive.
The instrument developed by Moore and Benbasat tries to capture not the primary
characteristics of an innovation per se but how these characteristics are perceived. The
reasoning behind this decision is that the behavior of prospective adopters can be
explained by the extent to which they are able to perceive the characteristics of an
IJBM
25,2
74
innovation. Since it is true that different people may have different perceptions of the
primary characteristics of an innovation, their behavior may also differ as a result.
Theory of Reasoned Action (TRA)
The Theory of Reasoned Action (TRA), developed by Fishbein and Ajzen (1975), is
probably one of the most influential theories used to explain human behavior
(Venkatesh et al., 2003). Simply put, according to this theory, the behavioral intention
can be explained by the attitude towards behavior and subjective norm. The attitude
towards behavior is defined as “an individual’s positive or negative feelings
(evaluative effect) about performing the target behavior” (Fishbein and Ajzen, 1975, p.
216). Subjective norm refers to perception that most people who really matter to the
individual think that he either should or should not perform the behavior in question”
(Fishbein and Ajzen, 1975, p.302). Attitude towards behavior, in turn, can be explained
by the salient beliefs in the behavior.
Technology Acceptance Model (TAM)
The Technology Acceptance Model (TAM) was proposed by Davis (1989) to predict the
acceptance and use of new information technology (software and information systems)
within organizations. The model derives from the TRA and its final version is rather
parsimonious. In the model, behavioral intention can be explained by the attitude
towards use of the system and its perceived usefulness. Attitude towards use of the
system, in turn, can be explained both by its perceived usefulness and its perceived
ease of use. perceived usefulness was defined by Davis (1989) as “the degree to which
individuals believe that using a particular system would enhance their job
performance” (p. 320) whereas perceived ease of use relates to “the degree to which
individuals believe that using a particular system would require no effort” (p. 320).
By hypothesis, the greater the perceived usefulness and the perceived ease of use,
the better are people’s reactions towards the innovation and the higher their intention
to adopt it. Later, based on the theory of reasoned action, Venkatesh and Davis (2000)
added to TAM the subjective norm construct, and this new model became known as
TAM2.
Theory of Planned Behavior (TPB)
The Theory of Planned Behavior (TPB) was proposed by Ajzen (1985) as an extension
of TRA (Fishbein and Ajzen, 1975) for situations where people do not have complete
control over their behavior. Basically, TPB adds a determinant to the behavioral
intention and the attitude towards behavior constructs which is the perceived
behavioral control. This construct reflects how people perceive the internal and
external limitations to their behavior. On more formal terms, it refers to how easy or
difficult people believe it would be to perform certain behaviors (Ajzen, 1985).
In TPB, behavior itself is a function of both the behavioral intention and the
perceived behavioral control. Behavioral intention, in turn, is influenced by the attitude
towards behavior, the subjective norm and the perceived behavioral control. The
determinants of intention (attitude, subjective norm, and perceived behavioral control)
are established by the structure of the underlying (attitudinal, normative and control)
beliefs.
Adoption of
internet banking
75
Decomposed Theory of Planned Behavior (DTPB)
In the context of information technology, Taylor and Todd (1995) proposed a model
known as the Decomposed Theory of Planned Behavior (DTPB), bringing together
concepts from two distinct lines of research: Innovation Diffusion Theory and Theory
of Planned Behavior.
According to Taylor and Todd (1995), Decomposed Theory of Planned Behavior
(DTPB) offers a number of advantages: it renders more transparent and easier to grasp
the relations among beliefs, attitudes and intentions, it enables application of the model
to a variety of situations and in managerial terms it is more relevant because it helps to
determine specific factors that lead to adoption and use of new technology.
In DTPB, attitudinal beliefs are broken down into three constructs extracted from
the literature on the perceived characteristics of innovations (Rogers, 1983):
(1) perceived usefulness;
(2) ease of use; and
(3) compatibility.
Normative beliefs are related to disagreement among the opinions of key reference
groups in an organizational environment (peers, superiors, and subordinates). Control
beliefs also break down into two groups: self-efficacy and facilitating conditions.
Self-efficacy is related to the perceived ability of using a new technology and
facilitating conditions refers to the available physical (time and money) and
technological resources for adoption.
The hypothesis here is that the clearer the perceptions of both self-efficacy in the use
of a new technology and the existence of facilitating conditions, the stronger the
intention to adopt the innovation.
Adoption of internet banking
Literature on banking technology has developed particularly as of the late 1980s and
the early 1990s, with the emergence of new technologies that simplified remote access
to banks. Innovations such as telephone banking, ATMs, the growing use of debit
cards and the forerunner of internet banking – access via proprietary software
installed on PCs - have aroused the interest of both scholars and practitioners. This
literature focused particularly on four areas: new retail bank services, the distribution
channels chosen for these services, banks’ and bankers’ perception of new banking
technologies and the clients’ perception about adopting them (Akinci et al., 2004).
In this latter line, the first studies focused on analyzing consumers’ perception of
specific technologies such as direct banking, telephone banking and home banking (for
instance, Howcroft et al., 2002). In the late 1990s, all eyes turned to internet banking
and its adoption. Revisiting the articles published on the topic over the past 10 years,
studies on IB adoption can be seen to fit two categories: descriptive and relational.
Descriptive studies focus on identifying the characteristics of IB adopters, their
reactions, attitudes, and the barriers to adoption they face, or the attributes that make
IB more attractive to prospective adopters (for example, Lee et al., 2005; Akinci et al.,
2004).
Relational studies aim exclusively at detecting the variables influencing adoption of
IB, in general using one of the new technology adoption models mentioned above (IDT,
TRA, TAM, TAM2, TPB, and DTPB).
IJBM
25,2
76
Analysis of this group of studies seems to suggest, at first sight, that the literature
on IB adoption is already mature and makes up a consistent theoretical body. However,
a more thorough analysis will conclude that the use of a diversified set of supporting
theoretical models and different tools of data collection and analysis add to the
difficulty of integrating the results from these studies. Therefore, despite the vast
number of existing studies, very little is known about the variables that truly
determine the adoption of IB.
Studies on banking innovation – proposed methodology
The current study aims to propose a methodology that can be used as a reference for
future studies on the adoption of new technologies such as internet banking. The
characteristics that set this methodology apart from others used in previous studies
rest on three pillars, discussed in more details in the sections below.
Proposition of an Integrated Theoretical Model
By and large, the previous studies on the adoption of IB took variables from one or
more of the best known models of technology adoption and added other variables
expected to enhance their explanatory ability (for instance, Eriksson et al., 2005;
Pikkarainen et al., 2004; Suh and Han, 2002). However, addition of new variables has
contributed very little to increasing the explanatory ability of these models, probably
owing to the effect of multi-collinearity between the new variables and those already
present in the original models. This study proposes a model that includes only those
variables from traditional models that are actually separate and theoretically sound
constructs.
This new theoretical model encompasses four sets of independent variables. From
IDT and TAM, the innovation characteristics included in the model are proven
determinants of adoption of new technologies: relative advantage, visibility, results
demonstrability, compatibility, complexity, trialability and image (Moore and
Benbasat, 1991). Since adoption of IB is voluntary by nature, that is, people do not
depend on an order from a superior to accept or reject internet banking, Moore &
Benbasat’s voluntariness construct was not included in the model. From DTPB, the
model has incorporated the idea of breaking down attitudes to enhance its explanatory
ability. From TAM2 came the subjective norm construct. From DTPB, the perceived
behavioral control and facilitating conditions constructs. Following Tan and Teo
(2000), Facilitating Conditions was broken down into technological support and
government support.
IDT’s complexity construct is similar to TAM’s perceived ease of use whereas IDT’s
relative advantages is in fact the same concept as TAM’s perceived usefulness.
Therefore, by including one, we are also including the other.
A few individual characteristics (Has a home PC, Age, Income, Education and
Gender) have been included in the model as they proved relevant in previous studies on
the adoption of IB (Lee et al., 2005; Kolodinsky et al., 2004; Chang, 2003). Therefore,
based on previous evidence, we can argue that younger male PC owners, with a college
degree and higher income are more likely to adopt IB.
As an independent variable, previous studies included either the intention to use IB
(for example, Tan and Teo, 2000) or the level of IB utilization (for example, Pikkarainen
et al., 2004), whereas a handful of other studies included both the intention and the
Adoption of
internet banking
77
actual use of IB (for example, Lee et al., 2005; Suh and Han, 2002). The model herein
includes both the Intention to Use IB and the Actual Adoption of IB (adopted/didn’t
adopt). The advantages of this decision are discussed in the next section. Finally, like
in many studies on technology adoption that use TAM as their theoretical base (for
example, Hong et al., 2001; Adams et al., 1992) and other previous studies on the
adoption of IB (Lee et al., 2005; Chang, 2003; Tan and Teo, 2000), the attitude towards
behavior construct has been removed. The model proposes that attitudinal, normative,
and control beliefs have a direct impact on both the intention to behave and the
behavior itself, rendering it simpler and parsimonious. Figure 1 shows a representation
of the model proposed (the signs in brackets after each variable indicate the type of
influence – positive or negative – on the intention to adopt IB and the actual adoption
behavior).
Joint Sampling Process: users and non-users
By including both the intention to use IB and the actual adoption of IB in the new
technology adoption model, this present method assumes that the sample will cover
both current and prospective IB users. By investigating current users alone, it is
possible to detect what drove them to adopt the innovation. However, the perceptions
of this group reveal a decision made long ago in the past and they can be masked by
the effects of time and the experience acquired through use of the innovation.
On the other hand, by investigating prospective users only, all we could find out
was which variables influence their intention. Fishbein et al. (2003) argues, however,
Figure 1.
Theoretical model
proposed for IB adoption
IJBM
25,2
78
that a lack of skills and/or environmental constraints may prevent intention from
turning into action, therefore, intention might not be a good enough predictor of
behavior under some circumstances. In order to improve the ability to predict behavior
born of intention, researchers have tested alternative scales (for example, Wansink and
Ray, 2000) and developed models in which intention scores are based on analysis of the
actual purchase behavior of specific consumers whose purchase intention had been
previously measured. However, Chandon et al. (2005) draw attention to the limitations
of these models since the very measurement of intention might feed back into the
association between intention and behavior, improving the chance that individuals will
actually act upon their intention after being interviewed.
By including in the sampling process both current and prospective adopters, it is
possible to circumvent some of the shortcomings related to the prediction of adoption
based solely on the measurement of intentions. Therefore, the differentiating factor in
this method is the incorporation of a sampling process that includes three populations
with distinct and defined characteristics:
(1) internet /IB users;
(2) internet/non-IB users; and
(3) non-internet /non-IB users.
As they have greater internet experience and ease of access, the hypothesis is that
internet users are more likely to adopt IB than non-internet users (Chau and Lai, 2003).
Data analysis: Multinomial Logistic Regression and Multiple Linear Regression
The data analysis technique that has predominated in studies on IB adoption (for
example, Chan & Lu, 2004) is the structural equation modeling. This is due to the fact
that most studies have used the structure proposed initially by TRA, that is, beliefs –
attitudes – intention – behavior. However, structural equation modeling has certain
disadvantages, such as the need for larger samples and the complexity associated with
incorporating non latent and ordinal variables such as age, income, and gender, as well
as dichotomy variables (adopted/didn’t adopt). Since the proposed model has done
away with the attitude construct, it can be appropriately tested through the use of
multiple linear regression and logistic regression.
Method and procedures
Implementation of the method proposed in the previous section to determine the factors
that influence IB adoption in Brazil is described below. The respondents engaged in
this study had at least one current bank account at the time of the interview. They were
sampled by convenience and approached close to bank branches at several locations in
a Brazilian major capital. Clients from several banks were interviewed roughly in the
same proportion as the market share that each of the banks then held on the Brazilian
market.
Three types of questionnaire were prepared to cater for the three groups elected for
the study: internet/IB users (IIB), internet/non-IB users (INIB) and non-internet/non-IB
users (NINIB). Interviewers picked the appropriate questionnaire according to the
answers given to the screening questions.
As suggested by Moore and Benbasat (1991), the items of latent constructs were
formulated in terms of perception. The questionnaires explored similar topics which
Adoption of
internet banking
79
enabled an investigation into the same topic across distinct populations with a
comparison of the results. In this case, constructs were expressed in different verb
tenses according to IB users’ status. Therefore, the items in the IB-user questionnaire
were either formulated in the present or in the past tense (for instance, “I can tell my
friends about the advantages and disadvantages of using internet banking”, “My
friends influenced my decision to use internet banking”) whereas when the
respondents were non-IB users, the same items were in a conditional tense (for
instance, “I would be able to tell my friends about the advantages and disadvantages of
using internet banking”, “My friends would influence my decision to use internet
banking”). Some items were put in the same verb tense in both questionnaires (for
instance, “The benefits of internet banking are clear to me”).
The same type of adjustment had to be applied to the Intention to use IB construct.
For IB users, the items tried to capture their intention to continue to use IB (for
example, “I will continue using IB to carry out my banking transactions”), while for
non-IB users the phrases tried to capture their intention to start using IB (for example,
“I intend to use IB to carry out my banking transactions.”). With the exception of the
verb tenses, the items in the constructs had the same wording in all tree questionnaires.
In order to maintain comparability with most previous studies, the items of all latent
constructs were measured by 6-point Likert type scales (1 ¼totally disagree;
6¼totally agree). The items for the visibility, results demonstrability, ease of use,
trialability and image constructs were adapted from Moore and Benbasat (1991). The
Compatibility scale of Moore and Benbasat (1991) was developed to look into the
adoption of new technologies in organizational environments. Therefore, it captures
only the compatibility between the technology and the type of work that the
prospective adopter performs, rendering it inappropriate for this study. Instead, we
chose to use an adaptation of the scale employed by Tan and Teo (2000).
Since the relative advantage construct was broken down, the items therein had to be
created for this study. Definition of the items in the relative advantage of convenience
construct were based on the convenience that the internet offers (24/7 availability all
year round, transactions can be carried out quickly from anywhere). The items in the
relative advantage of security and privacy construct tried to capture users’ common
worries related to use of the internet, such as the risk of information theft, identity theft
or sale of private information (Liebermann and Stashevsky, 2002). The items in the
relative advantage of economic benefits construct try to capture perceptions about the
cost of internet transactions (it’s cheaper to perform transactions through IB, IB fees
are lower) while the items related to the relative advantage of control construct refer to
the possibility of controlling transactions (IB affords greater control over personal
finances, IB permits scheduling of account payments, and IB permits retrieval of old
transactions).
According to the definition of subjective norm, the items in this construct were
defined according to the influence that friends, family and workmates may have on the
adoption of internet banking (for example, “My friends influenced my decision to use
IB.”).
The items in the self-efficacy construct were adapted from Compeau and Higgins
(1995). The items in the government support construct were adapted from Tan and Teo
(2000), and the technology support items were specifically created for this study. The
IJBM
25,2
80
items in the intention to use/continue to use IB construct were formulated along the
same lines as in other studies on IB adoption.
The questionnaire went through several testing steps, including evaluation of the
items by a professional translator and marketing scholar, and preliminary tests with
graduate students and prospective respondents.
Results
Sample characteristics
A total 300 IB users, 150 internet/non-IB users, and 150 non-internet/non-IB users were
interviewed. The resulting sample was demographically well balanced: 55 percent of
interviewees were female, 57 percent were between 21 and 40 years old, 57 percent had
a college degree, and 46 percent had monthly income higher than US$750.00. When
asked about the bank in which they performed most of their transactions, 94 percent of
the interviewees mentioned the top seven Brazilian banking groups.
Construct validity, dimensionality and reliability
Content validity (Devellis, 2003, p. 49) was established during preparation of the
questionnaire by using scales already validated in the literature, carefully analyzing
the items during translation, and judiciously developing new items.
The uni-dimensionality of constructs (Devellis, 2003, p. 94) was investigated by
means of exploratory factorial analysis (Hair et al., 1998). The principal components
extraction method was adopted and the results were submitted to the Varimax rotation
method. The number of factors was restricted to the exact number (13) of constructs
analyzed.
In the first step of the analysis, the main concern involved those items that didn’t
have well defined factorial loads in their respective constructs (Hair et al., 1998). Based
on these criteria, six items of different constructs had to be dropped from the
subsequent analysis. A second exploratory factorial analysis was then conducted and,
this time, 13 factors were found to be responsible for 84 percent of the variance. The
communalities extracted from all items were higher than 0.7 (Hair et al., 1998 suggest a
lower limit of 0.5) and there were no cross-loadings superior to 0.3. By the same token,
the factorial loads of every item in their respective constructs were higher than 0.7 (a
lot higher than the lower limit of 0.3 suggested by Hair et al., 1998).
Bartlett’s sphericity test was significant at 1 percent (
x
2
599 df ¼16,662, p,0.01)
and the Kaiser-Meyer-Olkin test, an assessment of the partial correlations between
variables, was 0.92, above the 0.7 lower limit suggested by Hair et al. (1998).
Construct reliability was evaluated on internal consistency, by calculating
Cronbach’s Alpha coefficient (Devellis, 2003). The resulting coefficients were higher
than 0.7 (Nunnally, 1978) for all constructs, except for Government Support (0.66).
Since the items in this scale had been adapted from a past study (Tan and Teo, 2000) in
which they proved reliable, we chose to keep the construct in the analysis. All the latent
constructs were then obtained from an average of their respective items.
Unfortunately, the relative advantage of economic benefits construct had to be
dropped from subsequent analyses due to the high number of missing values.
Adoption of
internet banking
81
Model testing – intention to use/continue to use IB as a dependent variable
The method of multiple linear regression was used to test the model against the
intention to use/continue to use IB dependent variable. The independent variables
included in the analysis were all from the latent constructs of the proposed model;
individual characteristics were coded as dummy variables. The results from this
analysis can be seen in Table I.
The model is significant (F
18.572
¼49.906, p,0.01) and the adjusted coefficient of
determination for the model was 60 percent. Residuals were analyzed in search for
departures from the multiple linear regression assumptions (linearity,
homoscedasticity and normality). The Kolmogorov-Smirnov test did not allow us to
reject the hypothesis that the distribution of residuals was normal (z¼1.044,
p¼0.226).
Analysis of the results in Table I leads to the conclusion that the main effects of 8
variables were significant. At a confidence level of 1 percent, the significant coefficients
corresponded to the following constructs: relative advantage of control, compatibility
with lifestyle, image, subjective norm and self-efficacy; at a confidence level of 5
percent, the significant coefficients corresponded to the following constructs: relative
advantage of security and privacy, results demonstrability, and trialability. It is
important to notice that the sign of the trialability coefficient is opposite to that in the
model hypothesis, that is, according to results, the higher the trialability of IB, the
lower the intention to use/continue to use IB.
In addition to the main effects model, an alternative model was analyzed adding the
effect of interaction between the dummy variables for two groups (IIB and INIB) and
each of the independent variables. If the intention to use/continue to use IB
determinants were different for each of the three groups, the coefficients for this new
Non-standard
coefficients Standard coefficients
BStandard error Beta TSig.
(Constant) 20.24 0.30 20.80 0.42
Relative advantage of convenience 20.07 0.06 20.04 21.13 0.26
Relative advantage of security and privacy 0.09 0.04 0.07 2.17 0.03
Relative advantage of control 0.24 0.05 0.18 4.61 0.00
Visibility 20.02 0.04 20.02 20.61 0.54
Result demonstrability 0.10 0.05 0.10 2.03 0.04
Compatibility with lifestyle 0.29 0.05 0.28 5.93 0.00
Ease of use 20.06 0.04 20.05 21.55 0.12
Trialability 20.09 0.04 20.08 22.55 0.01
Image 0.14 0.04 0.11 3.55 0.00
Subjective norm 0.11 0.04 0.08 2.66 0.01
Self-efficacy 0.30 0.05 0.27 6.42 0.00
Technological support 0.05 0.04 0.04 1.08 0.28
Government support 0.06 0.04 0.04 1.52 0.13
Has a home PC 0.02 0.10 0.00 0.15 0.88
College degree 20.05 0.01 20.02 20.48 0.63
Age between 21 and 40 years 0.02 0.09 0.00 0.22 0.82
Male 20.09 0.09 20.03 20.97 0.34
Income higher than 750 dollars 20.08 0.01 20.02 20.82 0.41
Table I.
Results of the multiple
linear regression
(dependent variable:
intention to use/continue
to use IB)
IJBM
25,2
82
set of variables would be significant. However, since at 5 percent confidence, none of
these variable coefficients worked out significant, the model was not taken into account
and the hypothesis that the determinants of intention to adopt are different in each of
the groups was rejected.
Model testing – IB/internet user status as dependent variable
In order to test the determinants of actual adoption, the multinomial logistic regression
technique was used; the dependent variable included in the analysis was a dummy
variable for the IB/internet user status (IIB, INIB and NINIB) and the independent
variables included were all from the proposed model. Along the lines of the multiple
linear regression analysis, individual characteristics were coded as dummy variables.
The reference chosen for the analysis was the IB users group (IIB). The estimated
coefficients for the main effects only model are shown in Table II.
The model fit test based on the difference between twice the log of likelihood (-2LL)
for the model with no independent variables and the full model is significant (
x
2
,36
df ¼549, p,0.01) and the pseudo coefficients of determination for the model are
relatively high (Cox & Snell ¼0.61; Nagelkerke ¼0.69; McFadden ¼0.45). The model
classification table shows that 90 percent of the IIB group, 75 percent of the NINIB
group, and only 47 percent of the INIB group were correctly classified. As a result, 75
percent of all observations were correctly classified. In the latter group, some 29
percent was misplaced in the IIB group and 24 percent, in the NINIB group.
Taking into consideration the characteristics of this analysis, a positive (negative)
sign before the coefficient of a specific variable means that a higher value of that
variable for a certain individual increases (reduces) the odds that he/she belongs to the
group analyzed relative to the IB users group (IIB group). Therefore, for instance, a
positive sign before the Visibility coefficient of the INIB group (0.36) means that
respondents in this group have a higher perception of the IB visibility than IB users
themselves. By the same token, a negative sign before the College degree coefficient of
the NINIB group (21.38), means that respondents in this group are less likely to have a
college degree than IB users.
Analysis of the coefficients in Table II leads to the conclusion that the results show
the opposite of what had been predicted for three hypothesis on the INIB group
(Visibility, Technological and Government Support) and for four hypothesis on the
NINIB group (Visibility, Technological Support, Government Support and Image). The
hypotheses that cannot be rejected refer to the constructs of Results Demonstrability,
Compatibility with Lifestyle, Self-efficacy and Has a home PC in both groups, and the
constructs of age between 21 and 40 years and College degree in the INIB group.
Discussion
With a few exceptions (in general, studies that incorporate the Attitude Toward IB
construct and analyzed data by means of Structural Equation Model), results from the
proposed model suggest that its ability to explain Adoption of internet banking
exceeds that of previous studies. The adjusted coefficient of determination for the
model using the Intention to use/Continue to use IB as a dependent variable was 60
percent and the pseudo-coefficients of determination for the model using the Actual
Adoption of IB as a dependent variable ranged from 45 percent to 69 percent.
Adoption of
internet banking
83
According to Venkatesh et al. (2003), conventional models of innovation adoption
“usually explain about 40 percent of the variance in the intention to use a specific
technology” (p. 40) while the various models tested by these authors managed to
explain between 17 percent and 53 percent of the intention to use the innovation
proposed.
Earlier research into IB adoption employed a broad variety of data analysis tools,
making a comparison difficult between previous studies and the present method in
terms of explanatory ability. However, we can say that the performance of the intention
B Standard error Est. Wald df Sig. Exp(B)
Internet/non-IB users (INIB)
(Constant) 3.91 1.01 14.81 1 0.000
Relative advantage of convenience 0.02 0.20 0.01 1 0.92 1.02
Relative advantage security and privacy 0.01 0.13 0.56 1 0.45 1.10
Relative advantage of control 20.16 0.17 0.82 1 0.37 0.85
Visibility 0.36 0.12 8.75 1 0.00 1.44
Result demonstrability 20.68 0.16 18.45 1 0.00 0.51
Compatibility with lifestyle 20.72 0.16 18.98 1 0.00 0.49
Ease of use 20.15 0.13 1.40 1 0.24 0.86
Trialability 0.02 0.11 0.02 1 0.89 1.02
Image 20.06 0.13 0.22 1 0.64 0.94
Subjective norm 0.12 0.13 0.82 1 0.37 1.12
Self-efficacy 20.55 0.15 13.07 1 0.00 0.58
Technological support 0.32 0.15 4.52 1 0.03 1.38
Government support 0.23 0.13 3.00 1 0.08 1.26
Has a home PC 20.69 0.34 4.13 1 0.04 1.99
College degree 2.32 .29 1.16 1 .28 1.37
Age between 21 and 40 years 20.13 0.28 0.20 1 0.65 1.13
Income higher than 750 dollars 0.03 0.29 0.01 1 0.92 0.97
Male 0.22 0.28 0.61 1 0.44 0.80
Non-internet users/non-IB users (NINIB)
(Constante) 3.08 1.20 6.58 1 0.01
Relative advantage of convenience 20.09 0.23 0.15 1 0.70 0.91
Relative advantage of security and privacy 20.23 0.18 1.68 1 0.19 0.80
Relative advantage of control 0.03 0.21 0.02 1 0.90 1.03
Visibility 0.41 0.16 6.38 1 0.01 1.50
Result demonstrability 21.09 0.20 28.67 1 0.00 0.34
Compatibility with lifestyle 20.87 0.20 19.70 1 0.00 0.42
Ease of use 20.10 0.15 0.48 1 0.49 0.90
Trialability 20.13 0.15 0.78 1 0.38 0.88
Image 0.39 0.17 5.30 1 0.02 1.47
Subjective norm 0.23 0.17 1.93 1 0.17 1.26
Self-efficacy 20.54 0.19 7.75 1 0.01 0.58
Technological support 0.09 0.17 0.28 1 0.60 1.01
Government support 0.30 0.17 3.04 1 0.08 1.34
Has a home PC 22.40 0.40 35.88 1 0.00 10.81
College degree 21.38 0.39 12.86 1 0.00 3.99
Age between 21 and 40 years 20.91 0.36 6.42 1 0.01 2.49
Income higher than 750 dollars 20.18 0.39 0.22 1 0.64 1.20
Male 20.30 0.37 0.68 1 0.41 1.36
Table II.
Results of the
multinomial logistic
regression (internet bank
users (IIB) was taken as a
reference group)
IJBM
25,2
84
to use/continue to use IB model herein is at least equivalent to previous ones (for
instance, Wang et al., 2003) while the performance of the Actual IB Adoption model is
superior to that reported for previous studies (for instance, Pikkarainen et al., 2004).
In the first model using the intention to use/continue to use IB as a dependent
variable, most coefficients of belief-related variables had a significant main effect,
which means that the multi-collinearity effect among them was relatively low.
Additionally, all significant coefficients, with a single exception (Trialability), had
signs in the expected direction.
On the other hand, in the multinomial logistic regression model, fewer belief-related
coefficients proved significant, although the coefficients of some individual variables
were significant. Some belief-related coefficients had signs in the opposite direction to
what had been anticipated.
Maybe the most important point to be observed when comparing the two models is
that the variables that influence the intention to use/continue to use IB are not exactly
the same that influence actual adoption. While the first model produced no significant
individual characteristics, the second model had three:
(1) has a home PC;
(2) age; and
(3) college degree.
These results seem to suggest that, while actual adoption is also influenced by some
individual characteristics such as possession of a home PC, age and a college degree,
these factors do not influence the intention to use internet banking.
Conclusions
This study has two key objectives – to propose a new method to study innovation
adoption and to use this method to investigate the factors that influence adoption of
internet banking in Brazil. The results obtained from the proposed integrated model
proved to be superior to those from most general studies on innovation adoption and at
least equivalent to those on the adoption of internet banking. This corroborates the
belief that the set of explanatory variables chosen is consistent and the decision to
include in the model only constructs from the best known innovation adoption models
was sensible. The main advantages of the proposed method are that this model is more
comprehensive that those preceding it and therefore more suitable for a multitude of
situations; in addition, the variables that influence adoption, are determined through
analyzing not only intention measures but also actual adoption.
Study results have also led to the identification of a broad set of variables that
proved significant in influencing both the intention to use/continue to use IB and actual
IB adoption in Brazil. Based on these results, Brazilian banks could develop marketing
programs to encourage their clients to adopt IB.
As an example, take the compatibility with lifestyle construct to illustrate how the
results of this study can be put to managerial use. Compatibility with lifestyle refers to
the extent to which people perceive that an innovation is compatible with the way they
think, act and lead their lives. Aware of how influential this variable is on IB adoption,
Brazilian banks could develop communications reinforcing the message that IB is in
line with different lifestyles. Another example is the result demonstrability construct.
In order to encourage adoption of IB, Brazilian banks could set up IB terminals at bank
Adoption of
internet banking
85
branches, shopping malls, airports and other busy venues so that their clients could
personally check on the advertised benefits.
Few differences were found between the two groups of non-adopters (internet users
and non-users). One of the key conclusions is that experience on the internet doesn’t
seem to have a strong moderating effect on IB beliefs, that is, those who use the
internet but don’t use IB share with non-internet users a similar perception of IB and
are not more inclined to adopt IB. This finding certainly deserves further investigation.
By comparing results from the two models, we have concluded that while actual
adoption is influenced by some individual characteristics (Has a home PC for the INIB
group and has a home PC, college degree and age between 21 and 40 for the NINIB
group), these variables do not play an important role in determining the intention to
use/continue to use IB. From the managerial point of view, Brazilian banks should first
focus on those clients who already have a home PC, are more educated and younger
since they are the most likely to adopt internet banking.
Despite the remarkable results, the study has some limitations that should be
addressed in the future. The first limitation refers to the fact that the model was tested
against a single type of innovation, internet banking. In order to be considered better
than previous models, a model has to undergo testing in multiple and different
categories. Once consumers have got to telephone banking, internet banking can be
easily sold by staff, which could differentiate internet banking from other innovative
technologies. In order to prove its explanatory ability, the present model has yet to be
tested on different technologies in the banking industry (for instance, mobile banking)
or technological innovations in general (for instance, video on demand on mobile
phones).
The second limitation comes from the selection criteria of respondents that do not
accommodate generalization of results. Finally, it should be noted that measurement of
certain items in the questionnaire did not work out as expected. It is suggested that
future studies should focus on refining the proposed scales to include larger,
probabilistically selected samples and investigate different types of technology
innovations.
References
Adams, D.A., Nelson, R.R. and Todd, P.A. (1992), “Perceived usefulness, ease of use, and usage of
information technology: a replication”, MIS Quarterly, Vol. 16 No. 2, pp. 227-47.
Ajzen, I. (1985), “From intentions to actions: a theory of planned behavior”, in Kuhl, J. and
Beckmann, J. (Eds), Action Control: From Cognition to Behavior, Springer, Heilderberg,
pp. 11-39.
Akinci, S., Aksoy, S. and Atilgan, E. (2004), “Adoption of internet banking among sophisticated
consumer segments in an advanced developing country”, The International Journal of
Bank Marketing, Vol. 22 No. 3, pp. 212-32.
Chandon, P., Morwitz, V.G. and Reinartz, W.J. (2005), “Do intentions really predict behavior?
Self-generated validity effects in survey research”, Journal of Marketing, Vol. 69 No. 2,
pp. 1-14.
Chang, Y.T. (2003), “Dynamics of banking technology adoption: an application to internet
banking”, Royal Economic Society Annual Conference, No. 41, available at: http://ideas.
repec.org/p/ecj/ac2003/41.html (accessed April 15, 2006).
IJBM
25,2
86
Chau, P.Y.K. and Lai, V.S.K. (2003), “An empirical investigation of the determinants of user
acceptance of internet banking”, Journal of Organizational Computing & Electronic
Commerce, Vol. 13 No. 2, pp. 123-45.
Compeau, D.R. and Higgins, C.A. (1995), “Computer self-efficacy: development of a measure and
initial test”, MIS Quarterly, Vol. 19 No. 2, pp. 189-211.
Davis, F.D. (1989), “Perceived usefulness, perceived ease of use, and user acceptance of computer
technology”, MIS Quarterly, Vol. 13 No. 3, pp. 319-40.
Devellis, R.F. (2003), Scale Development: Theory and Applications, Sage Publications, Thousand
Oaks, CA.
Eriksson, K., Kerem, K. and Nilsson, D. (2005), “Customer acceptance of internet banking in
Estonia”, The International Journal of Bank Marketing, Vol. 23 No. 3, pp. 200-16.
Febraban (2004), available at: www.febraban.com.br/Arquivo/Servicos/Dadosdosetor/
tecnologia_2005_dadossetor.asp (accessed April 15, 2006).
Fishbein, M. and Ajzen, I. (1975), Belief, Attitude, Intention, and Behavior: An Introduction to
Theory and Research, Addison-Wesley, Reading, MA.
Fishbein, M., Hennessy, M., Yzer, M. and Douglas, J. (2003), “Can we explain why some people do
and some people do not act on their intentions?”, Psychology, Health & Medicine, Vol. 8
No. 1, pp. 3-18.
Hair, J.F. Jr, Anderson, R.E., Tatham, R.L. and Black, W.C. (1998), Multivariate Data Analysis,
5th ed., Prentice-Hall, Upper Saddle River, NJ.
Hong, W., Thong, J.Y.L., Wong, W.M. and Tam, K.Y. (2001), “Determinants of user acceptance of
digital libraries: an empirical examination of individual differences and system
characteristics”, Journal of Management Information Systems, Vol. 18 No. 3, pp. 97-124.
Howcroft, B., Hamilton, R. and Hewer, P. (2002), “Consumer attitude and the usage and adoption
of home-based banking in the United Kingdom”, The International Journal of Bank
Marketing, Vol. 20 No. 2, pp. 111-21.
IBOPE (2005), “Bank sites are among the most frequently hit by Brazilian internet users”,
available at: www.ibope.com.br (accessed April 15, 2006).
Kolodinsky, J.M., Hogarth, J.M. and Hilgert, M.A. (2004), “The adoption of electronic banking
technologies by US consumers”, The International Journal of Bank Marketing, Vol. 22
No. 4, pp. 238-45.
Lee, E.J., Kwon, K.N. and Schumann, D.W. (2005), “Segmenting the non-adopter category in the
diffusion of internet banking”, The International Journal of Bank Marketing, Vol. 23 No. 4,
pp. 414-37.
Liebermann, Y. and Stashevsky, S. (2002), “Perceived risks as barriers to internet and
E-commerce usage”, Qualitative Market Research, Vol. 5 No. 4, pp. 291-300.
Moore, G.C. and Benbasat, I. (1991), “Development of an instrument to measure the perceptions of
adopting an information technology”, Information Systems Research, Vol. 2 No. 3,
pp. 173-91.
Nunnally, J.C. (1978), Psychometric Theory, 2nd ed., McGraw-Hill, New York, NY.
Pikkarainen, T., Pikkarainen, K., Karjaluoto, H. and Pahnila, S. (2004), “Consumer acceptance of
online banking: an extension of the technology acceptance model”, Internet Research,
Vol. 14 No. 3, pp. 224-35.
Rogers, E.M. (1983), Diffusion of Innovations, The Free Press, New York, NY.
Suh, B. and Han, I. (2002), “Effect of trust on customer acceptance of internet banking”, Electronic
Commerce Research & Applications, Vol. 1 No. 3, pp. 247-63.
Adoption of
internet banking
87
Tan, M. and Teo, T.S.H. (2000), “Factors influencing the adoption of internet banking”, Journal of
the Association for Information Systems, Vol. 1.
Taylor, S. and Todd, P.A. (1995), “Understanding information technology usage: a test of
competing models”, Information Systems Research, Vol. 6 No. 2, pp. 144-76.
Tornatzky, L.G. and Klein, K.J. (1982), “Innovation characteristics and innovation
adoption-implementation: a meta-analysis of findings”, IEEE Transactions on
Engineering Management, Vol. 29 No. 1, pp. 28-45.
Venkatesh, V. and Davis, F.D. (2000), “A theoretical extension of the technology acceptance
model: four longitudinal field studies”, Management Science, Vol. 45 No. 2, pp. 186-204.
Venkatesh, V., Morris, M.G., Davis, G.B. and Davis, F.D. (2003), “User acceptance of information
technology: toward a unified view”, MIS Quarterly, Vol. 27 No. 2, pp. 425-78.
Wang, Y., Wang, Y., Lin, H. and Tang, T. (2003), “Determinants of user acceptance of internet
banking: an empirical study”, International Journal of Service Industry Management,
Vol. 14 No. 5, pp. 501-19.
Wansink, B. and Ray, M.L. (2000), “Estimating an advertisement’s impact on one’s consumption
of a brand”, Journal of Advertising Research, Vol. 40 No. 6, pp. 106-13.
Further reading
Limayem, M., Hirt, S.G. and Chin, W.W. (2001), “Intention does not always matter: the contingent
role of habit on IT usage behavior”, The 9th European Conference on Information
Systems, Bled, June, pp. 274-86.
Corresponding author
Jose
´Mauro C. Hernandez can be contacted at: jmhernandez@uninove.br
IJBM
25,2
88
To purchase reprints of this article please e-mail: reprints@emeraldinsight.com
Or visit our web site for further details: www.emeraldinsight.com/reprints
