Kybernetes

A comparison of the different versions of popular technology acceptance models:

A non-linear perspective
Francisco Javier Rondan-Cataluña, Jorge Arenas-Gaitán, Patricio Esteban Ramírez-Correa,
Article information:
To cite this document:
Francisco Javier Rondan-Cataluña, Jorge Arenas-Gaitán, Patricio Esteban Ramírez-Correa, (2015)
"A comparison of the different versions of popular technology acceptance models: A non-linear
perspective", Kybernetes, Vol. 44 Issue: 5, pp.788-805, https://doi.org/10.1108/K-09-2014-0184
Permanent link to this document:
Downloaded by Göteborgs Universitet At 13:50 23 November 2017 (PT)

https://doi.org/10.1108/K-09-2014-0184
Downloaded on: 23 November 2017, At: 13:50 (PT)
References: this document contains references to 38 other documents.
To copy this document: permissions@emeraldinsight.com
The fulltext of this document has been downloaded 1573 times since 2015*
Users who downloaded this article also downloaded:
(2015),"The unified theory of acceptance and use of technology (UTAUT): a literature review", Journal
of Enterprise Information Management, Vol. 28 Iss 3 pp. 443-488 <a href="https://doi.org/10.1108/
JEIM-09-2014-0088">https://doi.org/10.1108/JEIM-09-2014-0088</a>
(2015),"What catalyses mobile apps usage intention: an empirical analysis", Industrial
Management &amp; Data Systems, Vol. 115 Iss 7 pp. 1269-1291 <a href="https://doi.org/10.1108/
IMDS-01-2015-0028">https://doi.org/10.1108/IMDS-01-2015-0028</a>

Access to this document was granted through an Emerald subscription provided by emerald-
srm:387340 []
For Authors
If you would like to write for this, or any other Emerald publication, then please use our Emerald
for Authors service information about how to choose which publication to write for and submission
guidelines are available for all. Please visit www.emeraldinsight.com/authors for more information.
About Emerald www.emeraldinsight.com
Emerald is a global publisher linking research and practice to the benefit of society. The company
manages a portfolio of more than 290 journals and over 2,350 books and book series volumes, as
well as providing an extensive range of online products and additional customer resources and
services.
Emerald is both COUNTER 4 and TRANSFER compliant. The organization is a partner of the
Committee on Publication Ethics (COPE) and also works with Portico and the LOCKSS initiative for
digital archive preservation.

*Related content and download information correct at time of download.

 The current issue and full text archive of this journal is available on Emerald Insight at:
www.emeraldinsight.com/0368-492X.htm

K
44,5
A comparison of the different
versions of popular technology
acceptance models
788
A non-linear perspective
Francisco Javier Rondan-Cataluña and Jorge Arenas-Gaitán
Department Business Management and Marketing, University of Seville,
Seville, Spain, and
Patricio Esteban Ramírez-Correa
Downloaded by Göteborgs Universitet At 13:50 23 November 2017 (PT)

School of Business Studies, Catholic University of the North, Coquimbo, Chile

Abstract
Purpose – The purpose of this paper is to provide a complete and chronological view of the evolution
of the main acceptance and use of technology models, from the 1970s to the present day.
Design/methodology/approach – A comparison of partial least squares (linear model) and WarpPLS
(non-linear model) has been run for each acceptation of technology model: TRA, TAM0, TAM1, TAM2,
TAM3, UTAUT, UTAUT2. The data set collects the information of mobile internet users.
Findings – The authors have concluded that UTAUT2 model obtains a better explanation power than
the rest of technology acceptance models (TAMs) in the sample of mobile internet users. Furthermore,
all models have a better explanation power using non-linear relationships than the traditional linear
approach.
Originality/value – The vast majority of research published to date with regard to the Theory
of Reasoned Action (TRA), the Technology Acceptance Model (TAM), and the Unified Theory of
Acceptance and Use of Technology (UTAUT) are based on structural equation models assuming linear
relationships between variables. The originality of this study is that it incorporates non-linear relationships
and compares the same models using both approaches.
Keywords Internet, Information systems, Information technology, Non-linear systems, Modelling
Paper type Research paper

1. Introduction
The acceptance, adoption, and use of technology at an individual level are ripe topics in
information systems (IS) literature. Although there are earlier works, the first technology
acceptance models (TAMs) of acceptance of technology appear in the 1970s. These models
try to understand how users accept and use a technology. When users are introduced to a
new technology, many variables affect their choice about how and when they will use it
(Fishbein and Ajzen, 1975). Currently, however, this is still a research line in clear growth.
An examination of the technology acceptance literature discloses that some academics are
confident that the acceptance of technology has scarcely been reached (Bauer and Kenton,
2005; Franklin and Molebash, 2007; Hew and Brush, 2007), while others propose that it has
been more effective in some cases than others (Drucker, 2006; Hughes and Ooms, 2004).
Affirmations on the acceptance of technology are frequently founded on models made
available by the technology acceptance literature (Ifenthaler and Schweinbenz, 2013).
Kybernetes Its success is due to the fact that technology has become the heart of economic growth.
Vol. 44 No. 5, 2015
pp. 788-805 At first these studies were mainly related to the entrance of the first computers in
© Emerald Group Publishing Limited
0368-492X
DOI 10.1108/K-09-2014-0184 Research project – Proyectos de Excelencia de la Junta de Andalucía – reference code no. P11-SEJ-7042.
 workplaces. Over time, technologies, and especially computers, have been applied to Different
almost all human activity. Not only have we found applications in offices, but also in versions of
industry, agriculture, medicine, and leisure. Nowadays, the most recent models have
focussed on the adoption of new technologies by consumers. The fact that these models
popular
explain the success or failure of new technologies makes them a tool of utmost importance. TAMs
In this work, the technology discussed is internet on mobile phones, in Chile, an
emerging country where this technology has been recently adopted, but is experiencing a 789
tremendous boom. Specifically, international statistics show that the proportion of people
using the internet has been increasing in Chile. These data exceed the regional average
and are double the mean scores of developing countries. Currently, this figure is very
near to the value of some European countries (International Telecommunications Union
(ITU), 2013).
In recent decades, beginning with the Theory of Reasoned Action (TRA) a multitude
Downloaded by Göteborgs Universitet At 13:50 23 November 2017 (PT)

of models trying to explain the acceptance of technologies have been appearing in the
academic literature. But do these models really explain a better reality than the first ones?
Have there been significant advances with the different versions and improvements
of models? What this paper actually offers is a comparison of the different versions of
popular TAMs: TRA, TAM, and UTAUT. Some articles have reviewed different studies
involving the TAM (Yousafzai et al., 2007a, b), trying to evolve toward an integrated vision
of TAM, or looking for moderating effects in TAM constructs (Schepers and Wetzels,
2007). Similarly, the Unified Theory of Acceptance and Use of Technology (UTAUT) has
been revised through statistical meta-analysis to investigate its performance (Nüttgens
et al., 2011). However, as far as we know, the comparison and assessment between TAM,
TRA, and UTAUT developments have not been published. Furthermore, one of the major
drawbacks of most models covered in these themes is that relations between the variables
and constructs of the models are assumed as linear. New structural equation models (SEM)
techniques which can use non-linear relationships to estimate these models can vary in the
results obtained with linear relations (Schumacker and Marcoulides, 1998).
The main objective of this work is to analyze the evolution of TAMs using non-
linear relationships between variables. This objective has been decomposed into other
operational objectives:
(1) to quantify the improvement in the explanation of use and intention to use
through the different models of technology acceptance based on its explanation
levels; and
(2) to compare if assuming non-linear relationships in the models improve the
goodness of Fit (GoF) and quality of the models.
This work makes various contributions to the literature on the acceptance of technology
within the IS scope. First, it provides a complete and chronological view of the evolution
of the most popular models of acceptance and use of technology, from the 1970s to the
present day. From a theoretical point of view this study discusses the justifications on
the variations and improvements in models, and from an empirical point of view it
debates improvements in the explanation of use of new technologies in statistical
indicators provided by SEM.
Second, the vast majority of research published to date is based on SEM assuming
linear relationships between variables, based both on variance models, such as partial
least squares (PLS), or covariance models, such as EQS, LISREL, and AMOS. New
software for the application of SEM based on PLS has been recently released which
 K allows using non-linear relationships between the constructs included in models. This
44,5 improvement can be particularly relevant in social sciences such as marketing,
management, or IS where the behavior of individuals and the relationships between
constructs do not have to be necessarily linear (Moosbrugger et al., 2009; Schumacker
and Marcoulides, 1998), it is reasonable to consider that there will be improvements on
existing knowledge using this new technique of non-linear relationships. Nevertheless,
790 there is currently a lack of works comparing the results of SEM linear and non-linear
models (Cariou et al., 2014). Numerous interactions, including relationships concerning
social variables, are non-linear and their shape is known as a U-curve or inverted U-
curve. In this configuration one variable influences another in a way that points to a
maximum or minimum value, where the effect is either maximized or minimized,
respectively. An S-curve is also very usual in socio-economic relationships (Kock, 2013).
Therefore, trying to estimate real non-linear coefficients with estimated-linear
Downloaded by Göteborgs Universitet At 13:50 23 November 2017 (PT)

coefficients leads to biased and inconsistent estimates of the models.

Third, this work provides an application with a sample of users of a Latin American
country: Chile. The vast majority of currently published works have been carried out
with Asian and Anglo-Saxon cultures, mainly from the USA and the UK. We believe
that validating the theoretical models in other cultural contexts is an important
contribution.
To achieve the goals the work is structured in the following way. First, we carry out
a review of the literature on models of acceptance and use of technology, focussing on
TRA, TAM, and UTAUT models with their respective expansions and modifications.
Second, we deal with the problems of non-linear SEM from a theoretical point of view.
Third, the empirical work is shown, describing the sample, variables, and statistical
tools employed. Fourth, the statistical results of the tests performed are collected and
discussed. Finally, the main conclusions of the work are provided and its main
limitations are listed as well as future research lines.

2. Conceptual background
To understand why people accept or reject a technology has become one the main lines
of research in the IS scope. This is partly because of the development of computers in
virtually all facets of life, in the past four decades. This is evidenced by the multitude of
theoretical models that have attempted to address this objective: TRA proposed by
Fishbein and Ajzen (1975), TAM suggested by Davis (1986), TAM2 proposed by
Venkatesh and Davis (2000), TAM3 recommended by Venkatesh and Bala (2008),
UTAUT proposed by Venkatesh et al. (2003), and UTAUT2 planned by Venkatesh et al.
(2012). These and other authors have developed a relevant research stream trying to
improve the explanatory models. This research line was born with TRA, it was
developed with the various models of TAM, and it was recently updated with UTAUT.
Next, we make a brief approximation to these models. After that, a review of non-linear
models applied to SEM is offered (Figure 1).

2.1 TRA
TRA was one of the first models that studied acceptance of technology. From social
psychology, TRA analyzes the determinants of conscious behavior (Ajzen and Fishbein,
1980; Fishbein and Ajzen, 1975). According to this theory, the specific behavior of a
person is determined by his/her intention to carry out this behavior; this is called
behavioral intention (BI). At the same time, this BI is determined by his/her attitude (A)
and the subjective norms (SN) relating to the conduct in question (Figure 2).
 Other theories about technology
acceptation: Different
• Theory of Planned Behavior (1985)
• Model of PC Utilization (1977)
versions of
• Innovation Diffusion Theory (1995) popular
• Social Cognitive Theory (1986)
TAMs
UTAUT UTAUT2
Venkatesh et al. (2003) Venkatesh et al. (2012)
791
Fishbein and Ajzen (1975)
TRA
Downloaded by Göteborgs Universitet At 13:50 23 November 2017 (PT)

Figure 1.
TAM TAM2 TAM3 Evolution of theories
Davis (1986) Venkatesh and Davis (2000) Venkatesh and Bala (2008) about technology
acceptation

Belief and
Attitude
Evaluations

Behavioral
Actual Behavior
Intention

Normative Beliefs
and Motivation to Subjective Norm
Figure 2.
comply Theory of
reasoned action
Source: Fishbein and Ajzen (1975)

TRA is a general model, not designed for a specific behavior or technology, which has
allowed it to be applied to countless fields. A particularly notable feature of TRA from
the point of view of BI is that any other factors that influence behavior do so only
indirectly by influencing A, SN, or their relative weights. This implies that TRA mediates
the impact of uncontrollable environmental variables and controllable intentions on user
behavior.

2.2 TAM
TAM model was developed by Davis (1986). It is an adaptation of TRA specifically
tailored for modeling user acceptance of IS. The goal of TAM is to provide an
explanation of the determinants of computer acceptance that is general, capable of
explaining user behavior across a broad range of end-user computing technologies and
user populations. At the same time it is both parsimonious and theoretically justified.
TAM was formulated to identify a small number of fundamental variables suggested
by previous research, dealing with the cognitive and affective determinants of
computer acceptance. TAM uses TRA as a theoretical backdrop for modeling the
relationships between these variables. Specifically, TAM is based on two particular
beliefs, perceived usefulness (PU) and perceived ease of use (PEOU), as the main
antecedents of computer acceptance. Like TRA, TAM maintains that the use of
 K computers is determined by BI, although it differs from TRA in that BI is determined
44,5 by PU, as well as A, toward using the system. In the same way, TAM does not include
the construct SN used by TRA because of its uncertain theoretical and psychometric
status (Figure 3).
Later, Davis (1989) found that PU and PEOU exert a strong impact on BI, and the
effect of A decreases with time. With this argument they decided to remove the latter
792 construct from the TAM model. When Venkatesh and Davis (1996) analyzed the
antecedents of PEOU, they no longer included A in the model (Figure 4).
Over time, the TAM model has been implemented in a variety of contexts, beyond the
mere acceptance of computers in the workplace. Therefore, TAM has become well-
established as a robust, powerful, and parsimonious model for predicting user acceptance.
The first of the extensions of TAM, the so-called TAM2 (Venkatesh and Davis, 2000), is
based on the expansion of the antecedents of PU. Across the many empirical tests of
Downloaded by Göteborgs Universitet At 13:50 23 November 2017 (PT)

TAM, PU has consistently been a strong determinant of BI. Using TAM as the starting
point, TAM2 incorporates additional theoretical constructs spanning social influence
processes (SN, voluntariness, and image) and cognitive instrumental processes (job
relevance, output quality, result demonstrability, and PEOU). It should be emphasized
that the inclusion of SN affects both BI directly and through PU (Figure 5).
Later on, and having the same intention as in TAM2, to complete the model
incorporating the antecedents of the original TAM, Venkatesh and Bala (2008) developed
TAM3. More specifically, if TAM2 added the antecedents of PU, TAM3 was enlarged by
the constructs that precede PEOU and which were already set forth in Venkatesh and
Davis (1996) and Venkatesh (2000). In particular, building on the anchoring (computer
self-efficacy, computer anxiety, computer playfulness, and perceptions of external control)

Perceived
Usefulness

Behavioral Actual
External
Attitude Intention to System
Variables
Use Use

Figure 3. Perceived
Technology Ease of Use
acceptance model
Source: Davis (1986)

Perceived
Usefulness

Behavioral Actual
External
Intention to System
Variables
Use Use

Figure 4. Perceived
Technology Ease of Use
acceptance model 1
Source: Venkatesh and Davis (1996)
 Experience Voluntariness
Different
versions of
popular
Subjective
Norm
TAMs

793
Image
Perceived
Usefulness

Job
Relevance Intention Use
to Use Behavior
Downloaded by Göteborgs Universitet At 13:50 23 November 2017 (PT)

Output
Quality
Perceived
Ease of Use

Result Technology Acceptance Model Figure 5.

Demonstrability Technology
acceptance model 2
Source: Venkatesh and Davis (2000)

and adjustment framing (perceived enjoyment and objective usability) of human

decision making, Venkatesh and Bala (2008) developed a model of the determinants of
PEOU (Figure 6).
Finally, TAM models in the last decades have been widely used, extending their
application to a multitude of technologies, especially to web site applications. TAM
models have found a lot of support within the literature. Proof of this are more than
4,100 citations inside the Social Science Citation Index database in November 2013, and
more than 17,600 identified by Google Scholar for the article of Davis (Davis, 1989).

2.3 UTAUT
As mentioned above, the explanation of the use and acceptance of a new technology
has become one of the main lines of research within the literature of IS. As a result, in
addition to TRA and TAM many models have appeared. This development has
contributed to the fact that many researchers publish ad hoc models, mixing concepts
of various theories, or using only those most favorable to their objectives without
considering the contributions of other alternatives. Consequently, Venkatesh et al.
(2003) affirmed that there is a need for a review and synthesis in order to progress
toward a unified view of user acceptance. For this reason, they reviewed the acceptance
literature and discussed eight prominent models: TRA, TAM, the motivational model,
the theory of planned behavior (TPB), a model combining TAM and TPB, the model of
PC utilization, the innovation diffusion theory, and the social cognitive theory. They
empirically compared the eight models and their extensions. Based on them, they
formulated a unified model that integrates elements across the eight models. Finally,
they empirically validated the new model to give greater consistency to their
contribution. By encompassing the combined exploratory power of the individual
models and key moderating influences, UTAUT advances cumulative theory while
retaining a parsimonious structure (Figure 7).
 K Experience Voluntariness
44,5 Subjective
Norm

Image

Job Perceived
794 Relevance Usefulness
Output
Quality
Intention Use
Result to Use Behavior
Demonstrability

Computer Self-
Eficacy
Downloaded by Göteborgs Universitet At 13:50 23 November 2017 (PT)

Perceived
Perceptions of
Ease of Use
Anchor

External Control
Technology Acceptance Model
Computer Anxiety

Computer
Playfulness
Adjustment

Perceived
Enjoyment

Figure 6. Objective
Technology Usability

acceptance model 3
Source: Venkatesh and Bala (2008)

Performance
Expectancy

Effort
Expectancy
Behavioral Use
Intention Behavior
Social
Influence

Figure 7.
Facilitating
Unified theory of
Conditions
acceptance and use
of technology
Source: Venkatesh et al. (2003)

UTAUT pointed out four constructs that play a significant role as direct determinants
of user acceptance and usage behavior: performance expectancy, effort expectancy,
social influence, and facilitating conditions. These four constructs directly affect BI. As
well as TRA and TAM, BI directly affects the use of the technology (U). On the other
hand, unlike TRA and TAM, BI is not the only direct antecedent of U. Furthermore, the
facilitating conditions directly determine U (Figure 8).
 Performance Different
Expectancy
versions of
Effort
popular
Expectancy TAMs
Behavioral Use
Social Intention Behavior
Influence 795
Facilitating
Conditions

Hedonic
Motivations
Downloaded by Göteborgs Universitet At 13:50 23 November 2017 (PT)

Price Value

Figure 8.
Habit
Unified theory of
acceptance and use
of technology 2
Source: Venkatesh et al. (2012)

UTAUT has distilled the critical factors and contingencies related to the prediction of
BI to use a technology and to technology used primarily in organizational contexts.
Over time, UTAUT has served as a baseline model and has been applied to the study of
a variety of technologies in both organizational and non-organizational settings.
However, like TAM and TRA, UTAUT is designed from an internal perspective of the
organization. That is, developed from the point of view of the implementation of new
technologies within organizations. For this reason the constructs which form it have a
distinctly utilitarian character. While the various studies contribute to understanding
the utility of UTAUT in different contexts, there is still the need for a systematic
investigation and theorizing of the salient factors that would apply to a consumer
technology use context.
On the basis of UTAUT, a new model is built designed to be applied in the context of
consumer technologies. This is the so-called UTAUT2 proposed by Venkatesh et al.
(2012). Three new determinants of BI are added to the constructs already employed by
UTAUT: hedonic motivation, price value, and habit. In addition, the habit construct is
also related to U.

2.4 Non-linear SEM in IS models

Non-linear SEM is gaining more and more interest in the context of IS models and other
closed disciplines (Schumacker and Marcoulides, 1998). Non-linear SEM offers many
benefits compared to linear SEM, but it is more difficult to conduct and is interfered by
methodological problems. Some estimation procedures have been developed recently.
These procedures intend to provide unbiased and efficient parameter estimates for the
non-linear effects (Moosbrugger et al., 2009).
One of the most critical complications of non-linear SEM is the multivariate
non-normality of the non-linear terms. This circumstance has two potential
consequences. First, acceptable estimation approaches should take the multivariate
non-normality unequivocally into account. Second, if an estimation technique is
 K used under the supposition of normally distributed indicator variables that does
44,5 not consider non-normality, the robustness of the models should be examined
meticulously (Moosbrugger et al., 2009).
The other big problem of non-linear SEM is multicollinearity. In non-linear SEM the
correlation between latent predictor variables is usually greater than the correlation
between manifest indicator variables because of the decrease caused by the unreliability
796 of the indicators. As a consequence, as the multicollinearity grows, the estimation is more
problematic, estimates are biased, and standard errors can be estimated incorrectly
(Kelava et al., 2008; Moosbrugger et al., 2009).
The literature classifies SEM methods into two classes: covariance and variance
based (PLS) (Gefen et al., 2000; Haenlein and Kaplan, 2004). PLS is currently one of the
most used SEM tools in social sciences, especially in areas such as management,
marketing, and IS (Hair et al., 2012, 2014). This is due to PLS having some benefits over
Downloaded by Göteborgs Universitet At 13:50 23 November 2017 (PT)

covariance models: it always produces a solution, variables are not required to fit
multivariate normality and large samples, and it permits the estimation of parameters
in models with formative latent variables (LV). However, one of the biggest problems of
both approaches is that they suppose linear relationships between the variables.
Conversely, most relationships between variables in social and economic sciences are
non-linear (Cariou et al., 2014; Weidlich, 1991). One of the few types of software that
allows non-linear relationships (concretely, U-shape and S-shape) between LV in the
SEM approach is WarpPLS 3.0 (based on PLS regression) (Kock, 2010).

3. Methodology
A comparison of PLS (linear model) and WarpPLS (non-linear model) has been run for
each acceptation of technology model: TRA, TAM0, TAM1, TAM2, TAM3, UTAUT,
UTAUT2. The data set collects the information of mobile internet users. WarpPLS 3.0
is the software used to run all the analyses. Some of the advantages of WarpPLS: it
estimates model fit indices, it provides scatter plots of each of the relationships between
LV, and it offers variance inflation factor coefficients (Kock, 2010).
The survey was carried out with mobile internet users from Chile in the month of
November 2012. The survey was presented in Spanish to interviewees. Previously the
questionnaire was translated from English to Spanish and then back to English to
safeguard correspondence in translation. Subsequently a pilot test was conducted
to test the survey. We used a non-random sampling method: quota sampling method,
on the basis of age range, and gender of the Chilean internet users. In total, 501 valid
questionnaires were collected, 46.9 percent females and 53.1 percent males.
The scales applied to measure the LV of the model have been tested previously
(Davis, 1989; Venkatesh et al., 2003, 2012) and they have been adapted to the case of
mobile internet. All of them were measured through seven-point Likert-type scales.
The scale ranged from “never” to “many times per day” for use behavior, and from
“strongly disagree” to “ strongly agree” for the other items. All of them are reflective
LV, and only U is a formative LV (all items of the scales are presented in Appendix 1).

4. Results
The main objective of this study is not to confirm the relationships between the
constructs of each model, but to compare the GoF, through some indicators, between
each acceptation of the technology model and between PLS and WarpPLS. In addition,
all the constructs have acceptable levels of validity and reliability in the models
 analyzed. If the aim is to observe whether one model has an improved fit than another, Different
then the model fit and quality indices are a suitable set of measures linked to model versions of
superiority.
The results of data analysis are presented in Tables I and II. We present the results
popular
of different indicators of model fit: average path coefficient (APC), average R2 (ARS), TAMs
and average variance inflation factors (AVIF). In addition, the R2 of U and BI and GoF
are presented. 797
Adding new LV into a model will usually raise the ARS. Nonetheless, that will
normally lead to a diminution of APC because the path coefficients associated with the
new LV will be smaller. Thus, the APC and ARS will compensate each other, and both
of them will only rise if the LVs added to the model improve the total predictive and
explanatory quality of the model. The AVIF index will grow if other LV are added to
the model enlarging its multicollinearity. In addition, the GoF index is the square root
Downloaded by Göteborgs Universitet At 13:50 23 November 2017 (PT)

of the product between the average communality index and the ARS. If GoF values are
equal to or greater than 0.36 the model has a high-explanatory power (Kock, 2013).
In Table I the R2 of U is the same in the TRA and all the TAM models, but increases
in UTAUT and UTAUT2. However, the R2 of BI is highest in UTAUT2 and TAM0.
The best GoF is for UTAUT2 followed by the TRA models. It is suggested that AVIF
be lower than 3.3, mainly in models where most of the variables are measured through
two or more indicators. Although ARS and AVIF indices are higher in the UTAUT2
model than in the TRA or TAM models, these values are within the recommended
thresholds. This is because of the greater number of LV included in the UTAUT2
model. However, the most important idea is that the R2 of U (0.247) in the
UTAUT2 model is the highest of all of them (26 percent better than the TAM
models and 10.6 percent better than the UTAUT model), and taking into account
that explaining use of the new technology is the main aim of all these models,
UTAUT2 is the best model, and has a very good model fit and quality indices.

Model R2 U R2 BI GoF APC ARS AVIF

TRA 0.196 0.614 0.589 0.436 0.405 1.141

TAM0 0.196 0.639 0.571 0.434 0.398 1.578
TAM1 0.196 0.402 0.517 0.440 0.331 1.651
TAM2 0.196 0.441 0.517 0.260 0.318 1.438 Table I.
TAM3 0.196 0.441 0.549 0.245 0.375 1.401 Comparison of
UTAUT 0.232 0.441 0.524 0.276 0.337 1.470 models using linear
UTAUT2 0.247 0.640 0.606 0.161 0.443 1.975 relationships

Model R2 U R2 BI GoF APC ARS AVIF

TRA 0.223 0.616 0.599 0.446 0.420 1.178

TAM0 0.223 0.642 0.581 0.444 0.412 1.557
TAM1 0.223 0.421 0.534 0.458 0.353 1.557 Table II.
TAM2 0.223 0.460 0.531 0.264 0.335 1.411 Comparison of
TAM3 0.223 0.460 0.564 0.250 0.397 1.418 models using
UTAUT 0.266 0.460 0.544 0.288 0.363 1.403 non-linear
UTAUT2 0.278 0.643 0.618 0.165 0.461 1.918 relationships
 K The results offered in Table II are in line with those of Table I. The best R2 of mobile
44,5 internet use is achieved by UTAUT2 (0.278, 24.66 percent better than the TAM models
and 4.5 percent better than the UTAUT model), also using non-linear relationships
between the LVs of each model. Furthermore, UTAUT2 has a very good model fit and
quality indices, as have the rest of models. But, if we compare each model of Table I
to the same model of Table II, we can find that the R2 of U are higher in models from
798 Table II. This means that considering non-linear relationships between the LVs of all
the models improves the explanation level of each model (13.77 percent of improvement
in the TAM models, 14.66 percent in UTAUT, and 12.55 percent in UTAUT2).
An example of an individual improving the relationship between two constructs can be
seen in Appendix 2.

5. Discussion
Downloaded by Göteborgs Universitet At 13:50 23 November 2017 (PT)

Both of the two objectives of this study have been achieved. First, we have shown that
the UTAUT2 model obtains a better explanation power (26 percent better) than the rest
of the TAMs in our sample of mobile internet users. This is true for linear PLS and for
non-linear PLS. It is a remarkable the fact that the sample is made up of end-users of
mobile internet, and this is a consumer use context. They are not employees of firms who
have to compulsorily adopt a new technology. TAM and TRA models were specified for
other contexts. This study validates the better performance of the UTAUT2 model in a
consumer use context. This fact is very relevant because many studies are based on the
consumer use of social network sites, search engines, web sites, etc. The problem is that
these consumer-focussed technologies have been tested and modeled through models
created for other contexts and this fact could lead to biased conclusions. Price value,
habit, and hedonic motivations are variables that should be taken into account in
technologies oriented toward consumers (Venkatesh et al., 2012).
According to the results of this study, TAM developments do not improve previous
versions of TAM, and even worsen the R2 of BI of use. When attitude is removed from
TAM1 model, the R2 of BI of use drops more than 34 percent, and the R2 of use remains
constant. Therefore, the new versions of TAM do not seem to offer a better explanation
for the acceptance of new technologies than TAM0. If the aim of these model
developments was to improve the explanation of use of new technologies, it seems that
they do not achieve this objective. With regards to attitude, it is likely that when TAM
models are not applied to consumers but to employees or students who are obliged to
use a technology, attitude does not play an important role. But when TAM models are
applied to final consumers, voluntary intentions toward using the technology could be
a key element for explaining BI. When the use of technology is voluntary, attitude
gains weight in the relationship with BI of use (Yousafzai et al., 2007a). Despite the
expectations of improvement of TAM models a decade ago (Lee et al., 2003), these
results show that there have been no significant improvements, at least applying the
methods with voluntary users.
The second objective was to compare the performance of models using linear
relationships between each construct vs using non-linear relationships between them, for
each one of the models. Results show that all models have a better explanation power
using non-linear relationships than with the traditional approach. The improvements of
the R2 of use ranged between 12 and 15 percent in all models. In addition, the fit and
quality indices were still very good in all cases. This is an interesting result because the
use of non-linear PLS is novel and not many applications using WarpPLS have been
published. Some recent studies (Schmiedel et al., 2014) indicate the advantages of this
 statistical tool, such as permitting non-linear relationships between constructs Different
and providing a set of model goodness of fit statistics. To sum up, these results versions of
indicate that using non-linear relationships in TAMs could be an improvement in
this research topic. This study offers some interesting comparisons that can provide
popular
theoretical controversies in the years to come about the advantages and shortcomings of TAMs
UTAUT2 and non-linear PLS.
The TAMs revised in this study rely both on the TRA (Fishbein and Ajzen, 1975) 799
and on TPB (Ajzen, 1991). These theories indicated that the best predictor of people’s
behavior is their intention to perform the behavior. Specifically, in a technology
acceptance context, Turner et al. (2010) performed a systematic literature review with a
total of 73 articles, and found that BI fruitfully predicted actual usage. Nevertheless, for
some authors the relationship between intention and behavior is insufficiently
validated by empirical research (Nistor, 2014). Due to this, caution should be taken
Downloaded by Göteborgs Universitet At 13:50 23 November 2017 (PT)

using these models in other contexts outside management IS in a business setting

(Turner et al., 2010).
In this study, the data were collected from users of mobile internet services: a
context outside of a business setting where the voluntariness must be a variable to be
considered. Since the use of these services is not mandatory, the correlation between
the BIs of an individual and his/her subsequent behavior is explicitly supported by the
literature in the general global context (Sheppard et al., 1988), and also in the technology
acceptance context (Venkatesh et al., 2003). In particular, the process of acceptance of
these services is an ongoing phenomenon. The users consume these services at present
because of their intentions to use them in the past and their current intentions are a
predictor of future use. While today they have adopted these services, it is possible that in
the future they will stop doing so. Consequently, the BI variable is capturing their
intention to continue using mobile internet services.

6. Implications and future research

The penetration rate of mobile internet in Chile is over 55 percent of Chile’s population.
Data released by the sub-secretary of telecommunications in 2014 suggest the market is
at this time facing modernization – that is, current mobile phone users are switching to
advanced mobile devices such as smartphones. These are statistical data that reflect
human behavior. But what causes these changes? TAM and UTAUT models have been
widely used in many contexts and countries, trying to explain why people adopt new
technologies. The question is to confirm if these kinds of models are good explainers of
this adoption. This paper tries to answer this query in relation to a technology that is
rapidly increasing the internet adoption of millions of people around the world: internet
mobile. We cannot imagine the radical changes that humans are experiencing because
of being connected to the internet everywhere and all the time. All attempts to
understand this process better are relevant and necessary.
Furthermore, this study has other implications. First, it offers a comprehensive
review of the evolution of the most popular models of acceptance and use of technology.
From a theoretical point of view, this study debates the explanations about the
variations and improvements of these models. From a practical standpoint it debates
the developments in the explanation of use of new technologies using GoF indices
provided by the statistical tool. Second, the bulk of the research published to date
is based on SEM approaches assuming linear relationships between LV. New software
based on PLS allows us to use non-linear relationships between the constructs contained
within the models. This advance can be particularly pertinent in social sciences where the
 K actions of people and the relationships concerning constructs do not have to be inevitably
44,5 linear (Moosbrugger et al., 2009; Schumacker and Marcoulides, 1998). Third, this
work offers an application with a sample of users from a Latin American country: Chile.
The majority of articles published about this topic have been carried out with European,
Asian, and Anglo-Saxon cultures. We consider that confirming these models in other
cultural situations is an essential contribution.
800 One of the limitations of this study is related to the low levels of the R2 of use in all
the models, although the R2 of BI to use mobile internet is close to the typical values
(Venkatesh et al., 2003, 2012). The measures of use of this study are based on Venkatesh
et al. (2012). However, the explanatory power is considerably lower in our research.
Perhaps the explanation is that Chilean mobile users simply use more traditional calls
instead of internet services on their mobile devices than other cultures where this
technology is more usual and established. Low levels of the R2 of use may be due to the
Downloaded by Göteborgs Universitet At 13:50 23 November 2017 (PT)

fact that with mobile internet, aspects such as the social influence, hedonic motivations,
perceived utility, or PEOU have a direct influence on use, while the tested models do not
estimate direct influences on use.

References
Ajzen, I. (1991), “The theory of planned behaviour”, Organizational Behavior and Human Decision
Processes, Vol. 50 No. 2, pp. 179-211.
Ajzen, I. and Fishbein, M. (1980), Understanding Attitudes and Predicting Social Behavior,
Prentice-Hall, Englewood Cliffs, NJ.
Bauer, J. and Kenton, J. (2005), “Toward technology integration in schools: why it isn’t
happening”, Journal of Technology and Teacher Education, Vol. 13 No. 4, pp. 519-546.
Cariou, V., Verdun, S. and Qannari, E.M. (2014), “Quadratic PLS regression applied to external
preference mapping”, Food Quality and Preference, Vol. 32, March, pp. 28-34.
Davis, F.D. (1986), A Technology Acceptance Model for Empirically Testing New End-User Information
Systems: Theory and Results, Massachusetts Institute of Technology, Cambridge, MA.
Davis, F.D. (1989), “Perceived usefulness, perceived ease of use, and user acceptance of
information technology”, MIS Quarterly, Vol. 13 No. 3, pp. 319-340.
Drucker, M.J. (2006), “Commentary: crossing the digital divide: how race, class, and culture
matter”, Contemporary Issues in Technology and Teacher Education, Vol. 6 No. 1, pp. 43-45.
Fishbein, M. and Ajzen, I. (1975), Belief, Attitude, Intention, and Behavior: An Introduction to
Theory and Research, Addison-Wesley, Reading, MA.
Franklin, C.A. and Molebash, P.E. (2007), “Technology in the elementary social studies classroom:
teacher preparation does matter”, Theory and Research in Social Education, Vol. 35 No. 2,
pp. 153-173.
Gefen, D., Straub, D.W. and Boudreau, M.C. (2000), “Structural equation modeling and regression:
guide-lines for research practice”, Communications of the AIS, Vol. 4 No. 7, pp. 1-77.
Haenlein, M. and Kaplan, A.M. (2004), “A beginner’s guide to partial least squares analysis”,
Understanding Statistics, Vol. 3 No. 4, pp. 283-297.
Hair, F.J. Jr, Sarstedt, M., Hopkins, L. and Kuppelwieser, V. (2014), “Partial least squares
structural equation modeling (PLS-SEM) an emerging tool in business research”, European
Business Review, Vol. 26 No. 2, pp. 106-121.
Hair, J.F., Sarstedt, M., Ringle, C.M. and Mena, J.A. (2012), “An assessment of the use of partial
least squares structural equation modeling in marketing research”, Journal of the Academy
of Marketing Science, Vol. 40 No. 3, pp. 414-433.
 Hew, K.F. and Brush, T. (2007), “Integrating technology into K-12 teaching and learning: current Different
knowledge gaps and recommendations for future research”, Educational Technology
Research and Development, Vol. 55 No. 3, pp. 223-252.
versions of
popular
Hughes, J.E. and Ooms, A. (2004), “Content-focused technology inquiry groups: preparing urban
teachers to integrate technology to transform student learning”, Journal of Research on TAMs
Technology in Education, Vol. 36 No. 4, pp. 397-411.
Ifenthaler, D. and Schweinbenz, V. (2013), “The acceptance of tablet-PCs in classroom instruction: 801
the teachers’ perspectives”, Computers in Human Behavior, Vol. 29 No. 3, pp. 525-534.
International Telecommunications Union (ITU) (2013), World Telecommunication/ICT Indicators
Database, ITU, Geneva, available at: www.itu.int/ITU-D/ict/statistics/ (accessed January
25, 2015).
Kelava, A., Moosbrugger, H., Dimitruk, P. and Schermelleh-Engel, K. (2008), “Multicollinearity
and missing constraints: a comparison of three approaches for the analysis of latent
Downloaded by Göteborgs Universitet At 13:50 23 November 2017 (PT)

nonlinear effects”, Methodology: European Journal of Research Methods for the Behavioral
and Social Sciences, Vol. 4 No. 2, pp. 51-66.
Kock, N. (2010), “Using WarpPLS in e-collaboration studies: an overview of five main analysis
steps”, International Journal of e-Collaboration, Vol. 6 No. 4, pp. 1-11.
Kock, N. (2013), WarpPLS 4.0 User Manual, ScriptWarp System, Laredo, TX.
Lee, Y., Kozar, K.A. and Larsen, K.R.T. (2003), “The technology acceptance model: past, present,
and future”, Communications of the Association for Information Systems, Vol. 12 No. 1,
pp. 50.
Moosbrugger, H., Schermelleh-Engel, K., Kelava, A. and Klein, A.G. (2009), “Testing multiple
nonlinear effects in structural equation modeling: a comparison of alternative estimation
approaches”, in Teo, T. and Khine, M.S. (Eds), Structural Equation Modeling in Educational
Research: Concepts and Applications, Teo, T.Khine, M.S., Amsterdam, pp. 103-136.
Nistor, N. (2014), “When technology acceptance models won’t work: non-significant intention-behavior
effects”, Computers in Human Behavior, Vol. 34, May, pp. 299-300.
Nüttgens, M., Gadatsch, A., Kautz, K., Schirmer, I., Blinn, N., Dwivedi, Y., Rana, N., Chen, H. and
Williams, M. (2011), Governance and Sustainability in Information Systems. Managing
the Transfer and Diffusion of IT: IFIP WG 8.6 International Working Conference,
September 22-24, Vol. 366, Springer Science & Business Media, Hamburg, pp. 155-170.
Schepers, J. and Wetzels, M. (2007), “A meta-analysis of the technology acceptance model:
investigating subjective norm and moderation effects”, Vol. 44 No. 1, pp. 90-103.
Schmiedel, T., vom Brocke, J. and Recker, J. (2014), “Development and validation of an instrument
to measure organizational cultures’ support of business process management”,
Information & Management, Vol. 51 No. 1, pp. 43-56.
Schumacker, R.E. and Marcoulides, G.A. (1998), Interaction and Nonlinear Effects in Structural
Equation Modeling, Lawrence Erlbaum Associates, Mahwah, NJ.
Sheppard, B.H., Hartwick, J. and Warshaw, P.R. (1988), “The theory of reasoned action: a meta-
analysis of past research with recommendations for modifications and future research”,
Journal of Consumer Research, Vol. 15 No. 3, pp. 325-343.
Turner, M., Kitchenham, B., Brereton, P., Charters, S. and Budgen, D. (2010), “Does the technology
acceptance model predict actual use? A systematic literature review”, Information and
Software Technology, Vol. 52 No. 5, pp. 463-479.
Venkatesh, V. (2000), “Determinants of perceived ease of use: integrating control, intrinsic
motivation, and emotion into the technology acceptance model”, Information Systems
Research, Vol. 11 No. 4, pp. 342-365.
 K Venkatesh, V. and Bala, H. (2008), “Technology acceptance model 3 and a research agenda on
interventions”, Decision Sciences, Vol. 39 No. 2, pp. 273-315.
44,5
Venkatesh, V. and Davis, F.D. (1996), “A model of the antecedents of perceived ease of use:
development and test”, Decision Sciences, Vol. 27 No. 3, pp. 451-481.
Venkatesh, V. and Davis, F.D. (2000), “A theoretical extension of the technology acceptance
model: four longitudinal field studies”, Management Science, Vol. 46 No. 2, pp. 186-204.
802 Venkatesh, V., Thong, J.Y.L. and Xu, X. (2012), “Consumer acceptance and use of information
technology: extending the unified theory of acceptance and use of technology”, MIS
Quarterly, Vol. 36 No. 1, pp. 157-178.
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. 3, pp. 425-478.
Weidlich, W. (1991), “Physics and social science – the approach of synergetics”, Physics Reports,
Downloaded by Göteborgs Universitet At 13:50 23 November 2017 (PT)

Vol. 204 No. 1, pp. 1-163.

Yousafzai, S.Y., Foxall, G.R. and Pallister, J.G. (2007a), “Technology acceptance: a meta-analysis
of the TAM: part 1”, Journal of Modelling in Management, Vol. 2 No. 3, pp. 251-280.
Yousafzai, S.Y., Foxall, G.R. and Pallister, J.G. (2007b), “Technology acceptance: a meta-analysis
of the TAM: part 2”, Journal of Modelling in Management, Vol. 2 No. 3, pp. 281-304.

Corresponding author
Dr Francisco Javier Rondan-Cataluña can be contacted at: rondan@us.es

(The Appendix follows overleaf.)

 Appendix 1 Different
versions of
popular
Itemsa TAMs
Attitude toward using (ATT)
ATT1 Using mobile internet services is a good idea 803
ATT2 Using mobile internet services is a wise idea
ATT3 I like the idea of using mobile internet services
Behavioral intention (BI)
BI1 I intend to continue using mobile internet services in the future
BI2 I will always try to use mobile internet services in my daily life
BI3 I plan to continue to use mobile internet services frequently
Downloaded by Göteborgs Universitet At 13:50 23 November 2017 (PT)

Computer anxiety (CANX)

CANX1 Mobile internet services do not scare me at all
CANX2 Working with mobile internet services makes me nervous
CANX3 Mobile internet services make me feel uncomfortable
CANX4 Mobile internet services make me feel uneasy
Computer playfulness (CPLAY)
CPLAY1 When I use mobile internet services, I feel spontaneous
CPLAY2 When I use mobile internet services, I feel creative
CPLAY3 When I use mobile internet services, I feel playful
CPLAY4 When I use mobile internet services, I feel imaginative
Computer self-efficacy (CSE)
CSE1 I could complete the job using mobile internet services if there was no one around to tell me
what to do as I go
CSE2 I could complete the job using mobile internet services if I had just the built-in help facility
for assistance
CSE3 I could complete the job using mobile internet services if someone showed me how to do it
first
CSE4 I could complete the job using mobile internet services if I had used similar technologies
before this one to do the same job
Effort expectancy (EE)
EE1 Learning how to use mobile internet services is easy for me
EE2 My interaction with mobile internet services is clear and understandable
EE3 I find mobile internet services easy to use
EE4 It is easy for me to become skillful at using mobile internet services
Facilitating conditions (FC)
FC1 Mobile internet services is compatible with other technologies I use
FC2 I feel comfortable using mobile internet services
Hedonic motivation (HM)
HM1 Using mobile internet services is fun
HM2 Using mobile internet services is enjoyable
HM3 Using mobile internet services is very entertaining
Habit (HT)
HT1 The use of mobile internet services has become a habit for me
HT2 I am addicted to using mobile internet services
HT3 I must use mobile internet services
Table AI.
Measurement
(continued ) instrument
 K Itemsa
44,5
Image (IMG)
IMG1 People who use mobile internet services have more prestige than those who do not
IMG2 People who use mobile internet services have a high profile
IMG3 Having mobile internet services is a status symbol
804 Performance expectancy (PE)
PE1 I find mobile internet services useful in my daily life
PE2 Using mobile internet services helps me accomplish things more quickly
PE3 Using mobile internet services increases my productivity
Price value (PV)
PV1 Mobile internet services are reasonably priced
PV2 Mobile internet services are good value for money
Downloaded by Göteborgs Universitet At 13:50 23 November 2017 (PT)

PV3 Overall, the use of mobile internet services delivers good value
Job relevance (REL)
REL1 In my job, usage of mobile internet services is important
REL2 In my job, usage of mobile internet services is relevant
REL3 The use of mobile internet services is pertinent to my various job-related tasks
Result demonstrability (RES)
RES1 I have no difficulty telling others about the results of using mobile internet services
RES2 The results of using mobile internet services are apparent to me
Social influence (SI)
SI1 People who are important to me think that I should use mobile internet services
SI2 People who influence my behavior think that I should use mobile internet services
SI3 People whose opinions I value prefer that I use mobile internet services
Use behavior (UB)
USE1 Usage frequency for short message service
USE2 Usage frequency for games apps
USE3 Usage frequency for browsing on web sites
USE4 Usage frequency for mobile electronic mail
Notes: aThe items were measured on a seven-point Likert scale. The scale ranged from “never”
to “many times per day” for use behavior, and from “ strongly disagree” to “ strongly agree” for the
Table AI. other items
 Appendix 2. Example of improvement by considering non-linear relationships Different
The relationship between computer self-efficacy and perception of ease of use in the TAM3 model
is an example of improvement associated with considering non-linear relationships in technology
versions of
acceptance models: if we consider this relationship as linear, then its β is −0.073 and its effect size popular
is 0.017, however, if we consider this as a non-linear relationship, then the magnitude of its β TAMs
increases to −0.124, and its effect size consistently increases to 0.040. Figure A1 shows the linear
relationship and Figure A2 shows the non-linear relationship.
805
Linear relationship

0.89

–0.32
Downloaded by Göteborgs Universitet At 13:50 23 November 2017 (PT)

Perception of ease of use

–1.53

–2.74

–3.95
Figure A1.
Linear relationship
between computer
–5.16 self-efficacy and
–0.91 –0.22 0.48 1.18 1.87 2.57 perception of ease of
use in TAM3 model
Computer self-efficacy

Nonlinear relationship

0.89

–0.32
Perception of ease of use

–1.53

–2.74

–3.95
Figure A2.
Non-linear
relationship between
computer
–5.16 self-efficacy and
–0.91 –0.22 0.48 1.18 1.87 2.57
perception of ease of
use in TAM3 model
Computer self-efficacy
 This article has been cited by:

1. Bonnie J Wakefield, Carolyn L Turvey, Kim M Nazi, John E Holman, Timothy P Hogan,
Stephanie L Shimada, Diana R Kennedy. 2017. Psychometric Properties of Patient-Facing eHealth
Evaluation Measures: Systematic Review and Analysis. Journal of Medical Internet Research 19:10,
e346. [Crossref]
Downloaded by Göteborgs Universitet At 13:50 23 November 2017 (PT)