Taiwo Downe 2013 PDF

Download as pdf or txt
Download as pdf or txt
You are on page 1of 12

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/279653519

The theory of user acceptance and use of technology (UTAUT): A meta-analytic


review of empirical findings

Article  in  Journal of Theoretical and Applied Information Technology · January 2013

CITATIONS READS

126 1,765

2 authors:

Ayankunle A Taiwo Alan G. Downe


Dallas County Community College District Multimedia University
7 PUBLICATIONS   206 CITATIONS    34 PUBLICATIONS   428 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Cultural UX/Usability View project

All content following this page was uploaded by Ayankunle A Taiwo on 23 September 2015.

The user has requested enhancement of the downloaded file.


Journal of Theoretical and Applied Information Technology
10th March 2013. Vol. 49 No.1
© 2005 - 2013 JATIT & LLS. All rights reserved.

ISSN: 1992-8645 www.jatit.org E-ISSN: 1817-3195

THE THEORY OF USER ACCEPTANCE AND USE OF


TECHNOLOGY (UTAUT): A META-ANALYTIC REVIEW OF
EMPIRICAL FINDINGS
1
AYANKUNLE ADEGBITE TAIWO, 2ALAN G. DOWNE
1
Department of Computer and Information science, Universiti Technologi PETRONAS,Malaysia.
2
Assoc. Prof., Department of Marketing and Management, Curtin University Sarawak, Malaysia
E-mail: 1kunletaiwong@gmail.com, 2alan.downe@curtin.edu.my

ABSTRACT

The unified theory of acceptance and use of technology (UTAUT) proposes that performance expectancy,
effort expectancy, and social influence predict behavioral intention towards the acceptance of information
technology. The theory further proposes that facilitating conditions and behavioural intention predicts use
behavior in the acceptance of information technology. Ever since its inception, the theory has been assessed
using different applications, and it has become a dè factor model of measuring user acceptance.
Nonetheless, in terms of statistical significant magnitude and direction, reports on the model are diverse.
Therefore, in this study, based on 37 selected empirical studies, a meta-analysis was conducted in order to
harmonize the empirical evidence.
The outcome of the study suggests that only the relationship between performance expectancy and
behavioural intention is strong, while the relationships between effort expectation, social influence and
behavioural intention are weak. Similarly, the relationship between facilitating condition, behavioural
intention and use behaviour is also weak. Furthermore, the significance of the relationship between
facilitating condition and use behaviour does not pass the fail safe test while the significance of the
relationship between behavioural intention and use behaviour does not pass the fail safe test satisfactorily.
Implications for further studies are also discussed.
Keywords: Information Systems (IS), Adoption, UTAUT, Meta-Analytic Review.

1. INTRODUCTION models have been proposed to give explanations to


end users acceptance behaviour. The newest
Information technology pervades the amongst them is the Unified theory of adoption and
international community from programmable home use of technology (UTAUT) by Venkatesh et al.
appliances to organization applications. Increase in (2003), which has been applied and empirically
technological innovation and application with tested in different domains. Since its inception
awesome advantages brought changes to human life many empirical studies have been conducted using
and work endeavours. As people, organizations and UTAUT. The model is believed to be more robust
governments moved towards the use of Information than other Technology acceptance model in
Technology. Such move of change has increased evaluating and predicting technology acceptance
the human computer interaction, which is the sole (Venkatesh et al., 2003). Although, the model has
aim of performing a task (Card, Moran and Newell been widely used, tested and validated, the outcome
,1983). Interaction between humans and computers of empirical studies has been inconclusive in
is affected by quite a number of human factors and respect to the magnitude, direction and significance
its characteristics (Whitley, 1997), to which studies of the relationships amongst the model. In social
have come up with theories and models to sciences the issue of variety in statistical
investigate factors that influences humans to use significance is common because of complexity in
computers and its applications. human behaviour. Therefore, mixed outcomes in
different studies are not uncommon, but it does
The design, development and acceptance of
undermine the accuracy of the models, UTAUT
information technologies have received substantial
inclusive. Consequently, identifying users’ history
attention in the past few decades. Many theoretical
towards technology acceptance is difficult and

48
Journal of Theoretical and Applied Information Technology
10th March 2013. Vol. 49 No.1
© 2005 - 2013 JATIT & LLS. All rights reserved.

ISSN: 1992-8645 www.jatit.org E-ISSN: 1817-3195

complicated for the academia and information acceptance model and the theory of planned
technologist. behavior (C-TAM-TPB) ,The model of PC
Utilization (MPCU),The innovation diffusion
Information Systems (IS) researchers,
theory (ID) and Socio Cognitive Theory (SCT).
Information Technology (IT) managers and e-
commerce decision makers can benefit from the The unified model outperformed the eight
importance of meta-analysis on UTAUT as a individual models (adjusted variance (R2) of 70
knowledge cumulating tool (Hwang and percent). The UTAUT model uses four core
Schmidt,2011) by having better understanding of determinants of usage and intention (performance
concrete pre-cursors to users acceptance towards a expectancy, effort expectancy, social influence, and
technology and its applications. Armed with this facilitating conditions) alongside with four
knowledge IT managers and other decision makers moderators (gender, age, experience and
can take more successful steps in attaining increase voluntariness of use) of key relationships. Previous
in technological patronage and usage. Studies have works that have used UTAUT are briefly discussed
shown that to achieve a top level IT management in the following paragraphs.
success, accurate IT prescription is of paramount
AlAwadhi and Morris (2008) investigated the
importance (Benbasat & Zmud, 1999).
adoption of e-government services using UTAUT,
The objective of this study is to investigate the the survey was carried out on 880 students revealed
validity of UTAUT and reveal how much this that performance expectancy, effort expectancy and
validity is substantiated in present literature. In peer influence determine students’ behavioural
order to achieve this we harmonized existing results intention. Similarly facilitating conditions and
on UTAUT through a meta-analysis. Integrating behavioural intentions determine students’ use of e-
empirical results of the theory can assist in government services. Also, Biemans, Swaak,
understanding the application of UTAUT to variety Hettinga & Schuurman (2005) used the UTAUT
of technology in general. Meta-analysis also fosters model to examine nurses behavioural intentions
examination of relationship between the dimensions towards the use of Medical Teleconferencing
of a model as a whole. Thus, analyzing Application, the study revealed that performance
relationships between the constructs of UTAUT expectancy and effort expectation are high
with a larger sample of subjects becomes feasible predictors of behavioural intention but social
than any individual study. influence prediction power is low. In a cross
cultural study of IT adoption, Oshlyansky,Cairns
The outline of this study is as follows, a revision
and Thimbleby (2007) found that performance
on UTAUT with discrepancies and consistent
expectancy, effort expectancy and social influence
results in the existing literature, methods of study
predicts use intention. Furthermore, Šumak,
selection and coding of empirical findings based on
Polančič and Heričko (2010) found that social
37 carefully selected studies. We concluded by
influence have a significant impact on students
discussing the outcome of the study, limitation of
behavioural intention to use moodle and students’
the study and implications for future studies. We
behavioural intentions is a powerful predictor of the
anticipate that the outcome of this study can be
use of the e-learning system. Cheng, Liu, Song and
relatively used as a point of reference while testing
Qian (2008) investigated the validity of UTAUT
UTAUT in the nearest future.
using 313 intended users of Internet banking in
2. LITERATURE REVIEW China, the result suggest that performance
expectancy and social influence are strong
The UTAUT is a unified model that was predictors of behavioural intention. In a similar
developed by Vankatesh et al (2003) based on study, Cheng, Liu, and Qian (2008) found
social cognitive theory with a combination of eight performance expectancy and social influence of the
prominent information technology (IT) acceptance UTAUT constructs as predictors of users
research models. The authors examined the behavioural intention towards internet banking.
predictive validity of eight models in determining
In addition, an empirical study by Fang, Li, and Liu
the behavioural intention and usage to allow fair
(2008) suggests that performance expectancy, effort
comparison of the models. The eight models are;
expectancy and social influence significantly
The Theory of reasoned action (TRA), The theory
predicts managers intention to engage in knowledge
of Planned behavior (TPB), The technology
sharing using web2.0. Maldonado, Khan, Moon and
acceptance model (TAM), The motivational Model
Rho (2009) examined the acceptance of an e-
(MM), A model combining the technology
learning technology in secondary school in Peru,

49
Journal of Theoretical and Applied Information Technology
10th March 2013. Vol. 49 No.1
© 2005 - 2013 JATIT & LLS. All rights reserved.

ISSN: 1992-8645 www.jatit.org E-ISSN: 1817-3195

240 Students took part in the survey. Result from the authors also found performance expectancy,
their study suggests that social influence effort expectation, social influence and facilitating
significantly predicts behavioural intention. In the conditions as predictors of use behaviour. He and
same study, Maldonado et al. (2009) found Lu (2007) further suggest that performance
behavioural intention to significantly predict use expectancy and social influence are predictors of
behaviour. Carlsson, Carlsson and Hyvönen (2006) behavioural intention towards consumer’s
examined the acceptance of mobile telephone and acceptances of mobile advertising. The authors also
found that performance expectancy, effort found that facilitating condition and behavioural
expectancy and social influence are predictors of intention predicts use behaviour. Cheng, Liu, Qian
behavioural intention. & Song (2008) examined the acceptance of internet
banking, results suggest that performance
Also, Wu, Tao and Yang (2007) investigated the
expectancy and social influence predicts intention.
acceptance of 3G services in Taiwan and found
performance expectancy and social influence as
predictors of behavioural intention. Interestingly,

Figure1. Unified Theory Of Acceptance And Use Of Technology (UTAUT) (Source: Venkatesh Et Al., 2003).

As much as some studies have supported that the Similarly, Šumak, Polančič and Heričko (2010)
four predictive factors of UTAUT predicts intention suggested that performance expectancy and effort
and use behavior, results from some other studies expectancy are non-significant predictor of
suggest otherwise. Li & Kishore (2006) studied the behavioural intention. In a related study, Cheng,
Use of Online Community Weblog Systems, the Liu, Song and Qian (2008) discovered that effort
results indicated that scales for the four constructs expectancy does not significantly predict
in UTAUT including performance expectancy, behavioural intention. Similar studies have also
effort expectancy, social influence, and facilitating found effort expectancy to be non-significant in
conditions have invariant true scores across most predicting behavioural intention (See Cheng, Liu
but not all subgroups. The authors expressed need and Qian 2008; He and Lu 2007;Wu, Tao and Yang
for caution when interpreting UTAUT. In a 2007). In a study to investigate the role played by
structured PLS-Graph Conceptual Model, motivation in e-learning technology adoption,
Tibenderana and Ogao (2008) found performance Maldonado, Khan, Moon and Rho (2009) found
expectancy and social influence to be non- facilitating condition to be non- significant in
significant in predicting behavioural intention to predicting use behaviour. Cheng, Liu, Qian, Song
use electronic Library services in Ugandan (2008) also examined the acceptance of internet
Universities. Performance expectancy, effort banking and found that effort expectancy does not
expectancy and social influence were found to be predict customers intention to use internet banking.
non-significant in predicting intention in a study In the context of eGovernment, Schaupp, Carter
investigating the acceptance of an interface robot and Hobbs (2009) investigated the acceptance of E-
and a screen agent by elderly users Filing by the American tax payers. Results from the
(Heerink,Kröse,Wielinga and Evers, 2009). study suggest that performance expectancy and

50
Journal of Theoretical and Applied Information Technology
10th March 2013. Vol. 49 No.1
© 2005 - 2013 JATIT & LLS. All rights reserved.

ISSN: 1992-8645 www.jatit.org E-ISSN: 1817-3195

social influence predicts behavioural intention. reported in articles are larger than unpublished
Interestingly, the study revealed that effort studies (Ma and Liu, 2004). Therefore, reporting
expectancy is not a predictor of behavioural only the published papers will result into file
intention. The inconsistency in the outcomes of the drawers problem since an important method in meta-
studies on UTAUT leaves the output of the analysis is the calculation of the average effect size
relationships in the model inconclusive. of individual studies. There is every likelihood that
meta-analytic results may be inflated due to the file-
3. METHODOLOGY
drawers problem. In order to avoid this, we included
academic conference papers from the Association of
Glass (1976) defined meta-analysis as statistical computer machinery, IEEE digital library and
analysis of a large collection of analysis results for Proquest Academic database for online thesis,
the purpose of integrating the findings. Shaughnessy besides the papers from journals.
et al. (2006) further defined meta-analysis as the Furthermore, we took caution in assuming that
approach in which data is summarized and reported. individual study is independent (independent effect
Lastly, DeCoster (2004) described meta-analysis as sizes), which is a general assumption in meta-
a method used to provide information to support a analysis. Because some authors violate this by
particular theoretical statement concerning strength reporting two or more correlation based on a single
or consistency of a specific relationship. Following sample. Thus, we cross checked to see if the
the approach of Lipsey & Wilson, (2001) and Ma correlations were not based on the same study before
and Liu (2004) we employed their approach for data final selection. In this study, a total of 96 empirical
collection, selection criteria, articles usage studies were discovered, seventy one (71) studies
permission, coding and subsequent data analysis. were without correlation coefficient or other
statistical metrics that can be used. We corresponded
3.1 Selection of Studies with 20 authors with contact details in their paper
A detailed literature search was performed using and 12 of the authors provided us with inter-item
Science Direct, Emerald, ProQuest, EBSCO, correlations between the constructs in their study.
PsycArticles, and Dissertation Abstracts Online, Amongst the 37 selected information systems (IS)
IEEE, GoogleScholar and ACM. A compendium of studies, there are 24 Journal Papers, 6 conference
studies on technology acceptance that used UTAUT papers, 7 PhD thesis papers. A total of 153
was gathered using search words such as "UTAUT, correlation coefficients were obtained from the
User Acceptance, and Technology Adoption". studies.
Furthermore, we identified the use of the UTAUT
model using references in other articles (Lipsey & 3.2 Variables Recorded
Wilson, 2001). In order to be included in the meta- The six (6) variables examined in each study were
analysis, the paper has to meet the following criteria coded as performance expectation (PE), effort
adapted from Ma and Liu (2004): expectancy (EE), social influence (SI), facilitating
a. Be a behavioural study. condition (FC) behavioural intention (BI) and Use
b. Involve technology investigation. Behaviour (UB). The calculation of effect sizes and
c. Involves empirical testing of UTAUT directly or other derivatives of meta-analysis were based on
indirectly. these important variables of the UTAUT model.
d. Reported correlation co-efficient between the Other variables investigated by Venkatesh et al.
constructs of UTAUT or other values that can be (2003) but without significant effect were not
converted to correlations. included in this study.
e. Reports a sample size.
f. It must be published or dated after 2003 when 3.3 Computation of Effect Sizes
UTAUT was first published to 2011. Correlation co-efficient (r) was the effect size metric
In addition, effect sizes (r) from various samples chosen for this study because of its wide
vary, therefore, studies that fail to report enough accessibility and availability in technology accepted
statistical data to calculate effect sizes of the study literature, also because of its ease of interpretation
were not included in the meta-analysis. A general and availability of formulae to convert other test
concern while conducting a meta-analysis is statistics to correlation coefficient. In many studies,
publication bias. Owing to the fact that only studies it is not uncommon to have researchers reporting
with statistical significant results are published in different statistical value such as correlation
academic journals, the sizes of the effect sizes coefficient, F,t-value, p-value and chi square.

51
Journal of Theoretical and Applied Information Technology
10th March 2013. Vol. 49 No.1
© 2005 - 2013 JATIT & LLS. All rights reserved.

ISSN: 1992-8645 www.jatit.org E-ISSN: 1817-3195

Nonetheless, achieving insight into the depth of the PE-BI correlations were reported from 36 studies,
strength of the relationships between factors in a (36) SI-BI correlations were reported from 31
study becomes cumbersome without a uniform studies, 16 FC-BI correlations were reported from
metric. Therefore, converting metrics into a uniform 13 studies and 16 BI-UB correlations were reported
format becomes pertinent before a meta-analysis can from 13 studies. The number of studies for PE-BI
be conducted. In this study, we adopted the and EE-BI were approximately the same, the
Pearson’s correlation co-efficient as effect size index number of studies for SI-BI is a little lower than the
representing the empirical strength of the first two. While the number of studies for FC-BI and
relationship between each pair of the UTAUT BI-UB were low and reporting the same low 16
construct. We followed the approach described by correlations from 13 studies respectively.
Lipsey & Wilson (2001) and Ma & Liu (2004), for Concerning the strength of individual correlation
each of the pair of the UTAUT construct: coefficient, Table1 shows that the range of the effect
performance expectancy (PE), effort expectancy sizes moves from insignificant to strongly
(EE), social influence (SI), faciliatating condition significant. Although, most studies reported
(FC), behavioural intention (BI) and use behaviour significant results, it could be noted that some of the
(UB) the effect size was computed such that it is studies reported insignificant results. All the five
simply a correlation coefficient (r) if reported, relationships reported high positive significance but
otherwise a conversion is made using equation (1) if PE-BI reports the highest positive significant
other metrics such as t-value was reported. This correlations. Moreover, FC-BI and BI-UB have
procedure by Rosenthal (1984) and has been widely equal but highest negative non-significant
adopted by several studies (See Szymansky& correlation. Furthermore, the sample size varies
Henard, 2001; Ma and Liu 2004). The effect sizes of from one study to another, as the sample size was as
variables in each study were computed to access low as 41 in a study, while it is as high as 1607 in a
prediction effect towards behavioural intention and related study. The average sample size shows that
use behaviour. Effect sizes reported by authors were number of subjects in the PE-BI, EE-BI and SI-BI
not recalculated but were used directly. The are very close while the sample size for both FC-BI
computed outcome of effect values were computed and BI-UB are the same and lower than the first
into excel spreadsheet. Generally, about 96% of the three relationships.
effect sizes were calculated using the means and the 4.2 Direct Effect Analysis
mean and standard deviation spreadsheet. 4% of the
effect sizes were calculated using the F or t test
Using the simple means, sample size adjusted mean
spreadsheet.
and the Fisher r to Z transformation method (Fisher
1932; Lipsey and Wilson (2001). The average of all
the individual effect sizes is regarded as a simple
…..equation (1) mean. A sample size-weighted average of the
individual effect size is considered as sample size
adjusted mean (equation 2). Correlations were
4. DATA ANALYSIS
transformed to Fisher's Z (using equation 3) for
analysis and later back-transformed in the
The analysis of the data computed was reported in correlation metrics for result presentation.
two phases. The first phase described the range,
direction, statistical significance and the sample size
of the correlation. The multiplicities of the existing
findings were described. The univariate analysis of
the correlation was investigated in the second phase. ..equation (2)
Therefore, statistical significance and essential
tendencies of the findings were deducted.

4.1 The descriptive Statistics


Using the 153 correlation coefficients obtained from ..equation (3)
the studies, Table 1 shows that some studies did not
report all the five correlation or their equivalents.
Out of the 153 correlation coefficients, (43) PE-BI
correlations were obtained from (37) studies, (42)

52
Journal of Theoretical and Applied Information Technology
10th March 2013. Vol. 49 No.1
© 2005 - 2013 JATIT & LLS. All rights reserved.

ISSN: 1992-8645 www.jatit.org E-ISSN: 1817-3195


Table1:Summary Of Selected Correlation
Link No of No of Range of Positive Negative Range of Sample Cumulative
Studies Correlation correlation significant significant size sample size
Coefficients correlation correlation
from To # % # % from To Ave
PE-BI 37 43 0.10 0.70 42 97.7 1 2.3 41 1607 257 11057
EE-BI 36 42 0.07 0.70 40 95.2 2 4.8 41 1607 262 10995
SI-BI 31 36 0.12 0.89 35 94.6 2 5.4 41 1607 252 9304
FC-UB 13 16 0.11 0.79 13 81.3 3 18.7 55 722 191 3048
BI-UB 13 16 0.12 0.84 13 81.3 3 18.7 55 722 191 3048

Table2: Summary Of Means, Variances, Correlations


Link Sample Simple Standard Correlation Standard Sample Fail safe Confidence
Adjusted r mean r Deviation from Zr Error Variance Interval
@99%
PE-BI 0.4982 0.4919 0.1486 0.5361 0.1581 0.0221 167.03 (.43,.55)
EE-BI 0.4224 0.4131 0.1509 0.4356 0.1601 0.0228 76.67 (.35,.47)
SI-BI 0.4235 0.4019 0.1658 0.4236 0.1715 0.0275 47.15 (.33,.47)
FC-UB 0.3566 0.3556 0.2047 0.3769 0.2774 0.0419 -1.05 (.22,.50)
BI-UB 0.4104 0.4125 0.0221 0.4356 0.2774 0.0490 5.57 (.40,.42)
In meta-analysis, some argue that there is no much computed for each mean estimate to depict its
difference between the simple mean and the Fisher’s statistical significance. With the existence of
r to Z transformation (See Szymanski & Henard variances and error, the confidence provides a range
(2001); Mo et al 2004), but we decided to engage of effects that may exist in a true population. None
both methods is our study (equation 2 and 3) for of the intervals in Table2 is 0, thus we can conclude
clarity. Reliability adjusted mean was not computed that all the mean effect is significantly different from
in this study but sample size adjusted mean. zero. The significance of the effect sizes were
The Fisher’s r to Z transformation results is quite further tested by computing the failsafe test. Using
larger than the simple mean and the sample adjusted equation 3, the failsafe test (N) shows the amount in
mean except for some of the values SI-BI are almost numbers of additional studies that is needed to
the same. All in all the three values from the three confirm the null hypothesis (r=0) required to annul
methods are approximately the same. Therefore we the conclusion that there exist a significant
can interpreted the result from these study based on relationship amongst the pair of variables.
both the Fisher’s r to Z transformation method and Consequently, Table 2 shows that the mean effect
sample size adjusted method. The magnitude of an sizes of PE-BI, EE-BI and SI-BI are significantly
effect size is regarded as small when it is close to different from zero to the extent that 47-167 null
0.10, medium when it is close to 0.30 and large effect size is needed to revert mean effect sizes to a
when it is close to 0.50 (Cohen 1977;Mo et al 2004). level considered as non – significant statistically, BI-
Following this rule, our meta-analysis implies a UB requires 6 of null effect to make the relationship
medium size effect for FC-BI and large size effect non-significance(Many study suggest significant
for all other relationship. Furthermore, our study relationship with behavioural intention).
suggests that the PE-BI relationship is the strongest Nevertheless, the mean effect size of FC-UB did not
while FC-UB is the weakest of the relationships. pass the fail safe test with the negative (Nfs.05).
Since individual level analysis (use of individual
Also, EE-BI relationship is stronger than SI-BI correlation) was employed in this study against the
relationship (See Table 2). For more confidence that study level analysis (use of average correlation). We
intervals include means (µ) a wider confidence computed the sampling error variance and the
interval is recommended (Cumming and Finch standard deviation for each of the relationship to
2005). Therefore, a 99% confidence interval was avoid under estimation of the sampling error
sampling and generalization (Mo et al., 2004). Our
result shows that the variances of sampling error are

53
Journal of Theoretical and Applied Information Technology
10th March 2013. Vol. 49 No.1
© 2005 - 2013 JATIT & LLS. All rights reserved.

ISSN: 1992-8645 www.jatit.org E-ISSN: 1817-3195

close to each other. Therefore, individual level theorized parameter and the observed estimate in
analysis is satisfactory for this meta-analytic study. this study is not statistically significant at the 1%
level using a 99% confidence Interval.
5. DISCUSSIONS
5.2 Effort Expectancy and behavioural intention
The summary of the results computed in the meta- (EE-BI)
analysis is depicted by Table 3. Using Cohen (1992) Table 3 shows that the effect size (Zr) of effort
criteria of categorizing mean effect sizes as non- expectancy is small with a weight of 0.4356.
significant, small, medium or large. As depicted by However this value is consistent with previous
the table, most of the effect sizes are small. This literature (Venkatesh et al 2003; Dijk et al 2008).
result however is consistent with the result of Users of Information Systems are concerned with
Venkatesh et al.(2003), except the results of BI-USE the ease that is associated with the use of the
which is lower. This could be as a result of inability information system. A complex system or a web
of authors to measure actual usage of systems being interface that is difficult to navigate can make users
investigated. uninterested in adopting the system or website
Table3: Summary Of Effect Sizes By Dependent Variables (Byun & Finnie, 2011). The issue regarding the level
of computer literacy amongst the population can
Link Correlation Size/weight alter the perception of respondents to the ease
from Zr associated with using an information system,
PE-BI 0.5361 Medium because computer savvy users may be indifferent.
EE-BI 0.4356 Small The fail safe test shows that 76 null effect sizes are
expected to make the computed effect size non-
SI-BI 0.4236 Small significant. However this is hard to realize
FC-UB 0.3769 Small considering that fourty two (42) coefficients were
BI-UB 0.4356 Small used to compute the effort expectancy effect size.
Furthermore, the derived effect size in this study is
Note: Effect sizes classification is based on Cohen’s further tested with it statistical significance and we
1992 s = small (.2 <d <.5); M = Medium (.5 <d <.8); found that the out-put is statistically different from
L = Large (d >.8); zero with a confidence interval of (.35,.47). Thus the
difference between the theorized parameter and the
5.1 Performance Expectancy and behavioural observed estimate in this study is not statistically
intention (PE-BI) significant at 1% level using a 99% confidence
The effect size of performance expectancy can be interval.
classified as medium with a Zr of 0.5361. This is the
largest effect size in the study, and it is consistent 5.3 Social Influence and behavioural intention
with previous literature showing that amongst the (SI-BI)
four major construct of UTAUT, performance The effect size of social influence can be classified
expectancy has the highest co-efficient path weight as small (See Table 3) with a Zr of 0.4236. This
(Venkatesh et al 2003, Wnag & Shih 2009; Dijk et result is consistent with previous literature showing
al 2008). Users of Information Systems gives high the effect of social influence on intention to adopt a
regard to the level at which the system is technology (Venkatesh et al 3003; Wang &Shih
advantageous to them in their daily routine. The 2009).
ability of the system to assist users to achieve task Besides an effective and easy to use information
quickly will motivate users to adopt the system. system, end-users might not be obliged to use the
The fail safe test shows that 167 null effect sizes are system until they are motivated by important others
expected to make the computed effect size non- (people) that can influence their attitude and
significant. This however is not possible considering behaviour. With the way people's life are molded
that fourty three (43) coefficients were used to round role models, public figures, sportsmen and
compute the effect size. celebrities, an encouragement by such important
However, the derived effect size of PE-BI in this figures to use the system can motivate users to adopt
study is further tested with it statistical significance the use of an information system (Taiwo et al.,
and we found that the out-put is statistically different 2012).
from zero with a confidence interval of (.43,55) and The fail safe test shows that 47 null effect sizes are
thus we can say that the difference between the expected to make the computed effect size non-

54
Journal of Theoretical and Applied Information Technology
10th March 2013. Vol. 49 No.1
© 2005 - 2013 JATIT & LLS. All rights reserved.

ISSN: 1992-8645 www.jatit.org E-ISSN: 1817-3195

significant. This however is hard to obtain could be as a result of inability of many studies that
considering that thirty six (36) coefficients were employed UTAUT in their investigation of
used to compute the effect size. acceptance and adoption of technologies to measure
However, the derived effect size in this study is the actual use behaviour of the new information
further tested with it statistical significance and we systems being investigated. Therefore, few studies
found that the out-put is statistically different from actually investigate the effect of intention on use
zero with a confidence interval of (.33, .47) and thus behaviour rather many authors relied on the premise
we can say that the difference between the theorized that there exist a strong relationship between
parameter and the observed estimate in this study is intention and usage which Venkatesh et al had
not statistically significant at the 1% level using a hypothesized and found significant.
99% confidence Interval. The derived effect size in this study is further tested
with it’s statistical significance and we found that
5.4 Facilitating condition and behavioural the out-put is statistically different from zero with a
intention (FC-BI) confidence interval of (.40, .42). Thus, the difference
between the theorized parameter (Venkatesh
Table 3 depicts a small effect size (Zr) of facilitating
outcome) and the observed estimate in this study is
condition with a weight of 0.3769.The FC-BI
not statistically significant at the 1% level using a
relationship accounts for the lowest effect size in the
99% confidence Interval. Users that show positive
study. This could be as a result of inability of most
intention towards a technology actually exhibit that
studies to measure the actual usage of the
specific behaviour at a later time. However, our
information systems being invested. Therefore, few
study found the BI-UB relationship had a less than
studies actually reports the outcome of the effect of
desired fail safe test value of 5.56, thus about 6 null
facilitating conditions on use behaviour, rather many
effect sizes are expected to make the computed
authors reports the effect of facilitating condition on
effect size non-significant.
intention which Venkatesh et al had hypothesized
and found non-significant. In some studies,
facilitating condition has been found to be 6. LIMITATIONS AND FUTURE-WORKS
significant in predicting intention (Foon and Fah,
2011; Venkatesh et al 2011a; Venkatesh et al Besides the selection criteria surrounding studies
2011b). Thus, the outcome on the relationship involved in this study, one of the limitations of this
between the facilitation condition and use behaviour study is that the UTAUT theory is merely cited in
can be said to be inconclusive. Although many articles but not actually used. This led to the
empirically, there might be inconclusive argument relative small sample size of studies that were
on the effect of FC on UB or BI, it is important to employed in this study. The inability of a
note that qualitative research have shown that the standardized and generally accepted effect size
contribution of provision of organizational and statistics would enhance meta-analytic outcomes
technical infrastructure for users towards the (Lipsey and Wiley 2000;Ma and Liu,2004).
acceptance of a technology cannot be Secondly, meta-analysis has the ability of
overemphasized (Alawadhi and Morris 2009). indentifying whether the variation of correlation is
Furthermore, the derived effect size of the FC-UB due to chance, dimensions or methods employed.
relationship is further tested with it statistical We were unable to achieve this because of the small
significance and we found that the out-put is size of the selected studies, although we earlier
statistically different from zero with a confidence planned to examine the effects of the moderators and
interval of (.22,.50). Thus the difference between the methods used by running multiple regressions.
theorized parameter and the observed estimate in Thirdly, inability of the studies to measure use
this study is not statistically significant at the 1% behaviour and the integration of some dimensions of
level using a 99% confidence Interval. Nevertheless, UTAUT led to uneven number of correlations
the effect of facilitating condition on intention do between the constructs. Therefore, some caution is
not pass the fail safe test showing a negative value, advised when interpreting the results in the study.
thus the consistency in the outcome of FC-BI Future works could to attempts to resolve the above
relationship is questionable. limitations. Besides the use of coefficient paths
5.5 Behavioural intention and use behaviour (BI- magnitude, reports on future studies of UTAUT
UB) should include correlations, T-test and other
The effect size of BI-USE can be classified as small statistical measures that can be used to compute a
with a Zr of 0.4356(See table 3). This however meta-analysis. This shall enhance measuring the

55
Journal of Theoretical and Applied Information Technology
10th March 2013. Vol. 49 No.1
© 2005 - 2013 JATIT & LLS. All rights reserved.

ISSN: 1992-8645 www.jatit.org E-ISSN: 1817-3195

consistency of findings and the parsimonious nature [3] Benbasat,I and Zmud,R. (1999). “Empirical
of UTAUT. research in information systems”. The practices
of relevance.MIS Quarterly,23(1),3-16.
7. CONCLUSIONS [4] Biemans,M, Swaak,J, Hettinga,M and
Schuurman,J.G (2005). “Involvement Matters:
Through the meta-analysis, this study has succeeded The Proper Involvement of Users and
in combining and investigating existing empirical Behavioural Theories in the Design of a
literatures on Unified theory of acceptance and use Medical Teleconferencing Application”.
of technology (UTAUT). The relationship between Proceeds of GROUP’05, November 6–9, 2005,
UTAUT was examined using a larger sample size of Sanibel Island, Florida, USA.
over 11,000 which could have been difficult to [5] Byun,D-H and Finnie,G.2011. “Evaluating
achieve in a single study. On the basis of meta- usability, user satisfaction and intention to
analysis reported in this article, generally, our revisit for successful e-government websites”.
findings confirms Venkatesh et al. initial findings Electronic Government, An International
amongst the five constructs of UTAUT, only the Journal, Vol. 8, No. 1, 2011,pp.1–19.
relationship between PE and BI is strong while
[6] Card, S K, Moran, T P and Newell, A (1983).
others are slightly weak but significant.
“The psychology of human-computer
Also the relationship between BI and UB is also interaction”. Hillsdale, NJ: Erlbaum.
reliable while the relationship between FC and UB is
[7] Cheng,D,Liu,G.,Song,Y-F and Qian,C. “(2008)
found to be fairly less than desired. Furthermore, all
Adoption of Internet Banking: An Integrated
the five mean effects are statistically positive at
Model”. Proceeds of the 4th International
α=0.01 to attest to their statistical significance. The
Conference on Wireless Communications,
fail safe test further asserts the significance of the
Networking and Mobile Computing,12-14 Oct.
relationships. We discovered that 47-167 null effect
2008,Dalian.
size is needed to be hidden in file drawer for the
mean correlation between the trio of PE,EE,SI and [8] Cheng,D, Liu,G and Qian,C. “(2008) On
BI to be non significant, this seems unlikely. Determinants of User Acceptance of Internet
However the mean effect size of BI-UB has a weak Banking: A Theoretical Framework and
fail safe value, suggesting that six (6) reports with Empirical Study”. Advanced Management of
null effects can make the effect non-significant. Information for Globalized Enterprises, 2008.
While the mean effect size of FC-UB failed the fail AMIGE 2008.
safe test, suggesting that addition of just one report [9] Carlsson,C, Carlsson,J., Hyvönen,K.,
with null effect can make the effect non-significant. Puhakainen,J and Walden. (2006). “Adoption of
In conclusion, we discovered that majority of Mobile Devices/Services – Searching for
researchers cited UTAUT in their articles in order to Answers with the UTAUT”. Proceedings of the
support an argument rather than using it. Others that 39th Hawaii International Conference on
reported the use of UTAUT actually used it partially System Sciences.
while only a few have reported the use of the [10] Cohen, J (1992). “A power Primer”.
actually theory. This paper contributes to the area of Psychological Bulletin, 112,155-159
IS/IT adoption and diffusion research by showing [11] Cumming G and Finch S. (2005) “Inference by
the inadequacy and inconsistency in the use and Eye Confidence Intervals and How to Read
output of a theory. Pictures of Data”. American Psychological
Association. Vol. 60, No. 2, 170–180 DOI:
REFRENCES: 10.1037/0003-066X.60.2.170
[13] DeCoster, J. (2004). “Meta-analysis Notes".
[1] Alawadhi, A and Morris,A (2008). “The use of Retrieved on 20th May, 2011, from
the UTAUT model in the adoption of e- http://www.stat-help.com/notes.html.
government services in Kuwait”. Proceedings of [14] Dijk,J.A.G.M.V, Peters,O and Ebbers,W.
the 41st Hawaii International Conference on (2008). “Explaining the acceptance and use
System Sciences, Hawaii, 2008. government Internet services: A multivariate
[2] Alawadhi,A and Morris,A (2009). “Factors analysis of 2006 survey data in Netherlands”.
Influencing the Adoption of E-government Government information Quarterly. 25,375-
Services”. Journal of Software, Vol. 4, NO. 6, 399.
AUGUST 2009

56
Journal of Theoretical and Applied Information Technology
10th March 2013. Vol. 49 No.1
© 2005 - 2013 JATIT & LLS. All rights reserved.

ISSN: 1992-8645 www.jatit.org E-ISSN: 1817-3195

[15] Fang,W.C., Li,M.W and Liu,C.W (2008). [25] Oshlyansky,L.,Cairns,P. and Thimbleby,H
Measurement of the Knowledge-Sharing (2007). “Validating the Unified Theory of
Efficacy of Web2.0 Site Constructed on the Acceptance and Use of Technology (UTAUT)
Basis of Knowledge-based Systems by Tool Cross-Culturally”. Proceedings of the 21st
Applying the Model of UTAUT: Evidence of BCS HCI Group Conference.3-7 September
the early adopters”. Proceeds of the 3rd 2007, Lancaster University, UK.
Intetnational Conference on Innovative [26] Rosenthal, R., and Dimatteo, M. R. (2001)
Computing Information and Control Meta-analysis: Recent developments in
(ICICIC'08). quantitative methods for literature reviews.
[16] Foon, Y.S and Fah, B.C.Y. (2011). “Internet Annual Review of Psychology, 52, 59- 82.
banking adoption in Kuala Lumpur: An [27] Schaupp, LC, Carter, L and Hobbs, J (2009). “
application of UTAUT model. International E-File Adoption: A Study of U.S. Taxpayers
Journal of Business and Management. Intentions”. In proceedings of the 42nd Hawaii
Vol.6,No.4, pp161-167. International Conference on System Sciences –
[17] Glass, G V (1976). “Primary, secondary, and 2009.
meta-analysis of research”. Educational [28] Shaughnessy, J. J., Zechmeister, E. B and
Researcher, 5, 3-8. Zechmeister, J. S (2006). “Research methods in
[18] He,D and Lu,Y (2007). “Consumers psychology”. Seventh edition. New York:
Perceptions and Acceptances towards Mobile McGraw-Hill.
Advertising: An Empirical Study in China”. In [29] Šumak,B, Polančič,G, and Heričko,M. (2010)
proceedings of International Conference “An Empirical Study Of Virtual Learning
Wireless Communications, Networking and Environment Adoption Using UTAUT”. In
Mobile Computing. Proceedings of 2010 Second International
[19] Heerink,M., Kröse,B, Wielinga,B and Evers,V Conference on Mobile, Hybrid, and On-Line
(2009). “Measuring the influence of social Learning.
abilities on acceptance of an interface robot and [30] Szymantki, D.M and Henard, DH (2001)
a screen agent by elderly users”. People and Customer Satisfaction: A Meta-Analysis of
Computers XXIII(HCI) – Celebrating people empirical evidence. Journal of the academy of
and technology. Marketing Science,29(1),16-35.
[20] Hwang M I and Schmidt F L (2011). [31] Taiwo, A.A., Downe,A.G And Mahmood, A.K
“Assessing moderating effect in meta-analysis: (2012). “User Acceptance of Egovernment:
a re-analysis of top management support studies Integrating Risk and Trust Dimensions with
and suggestions for researchers”. European UTAUT Model”. In proceedings at the
Journal of Information Systems online International Conference on Computer and
publication. doi:10.1057/ejis.2011.12 Information sciences. Kuala Lumpur,
[21] Li,J.P and Kishore,R (2006). “How Robust is [23] Tibenderana,P.K and Ogao,P.J (2008)
the UTAUT Instrument? A Multigroup “Acceptance and Use of Electronic Library
Invariance Analysis in the Context of Services in Ugandan Universities”. In
Acceptance and Use of Online Community Proceedings of JCDL 2008, Pittsburgh,
Weblog Systems”. Proceeds of SIGMIS- Pennsylvania, USA.
CPR’06, April 13–15, 2006, Claremont, [33] Venkatesh,V, Morris, M. G, Davis, G. B. And
California, USA. Davis,F. D (2003) “User acceptance of
[22] Lipsey, M. And Wilson, D. (2001). “Practical information technology: toward a unified view”.
meta-analysis”. CA: Sage Publications. MIS Quarterly, vol. 27, no. 3, pp. 425-478,
[23] Ma,Q and Liu,L (2004). “The technology 2003.
Acceptance Model:A meta-Analysis of [34] Venkatesh V, Zhang X and Sykes T A (2011a)
Empirical Findings”. Journal of Organizational “Doctors do too little technology : a
and End User Computing.16(1),59-72. longitudinal field study of an electronic
[24] Maldonado,U.P.T., Khan,G.F., Moon,J And healthcare”. Journal Information Systems
Rho,J.J. (2009). “E-learning motivation, Research. Volume: 22 Issue: 3 pp.523-546
Students’ Acceptance/Use of Educational Portal [35] Venkatesh,V, Sykes T A And Zhang X (2011b)
in Developing Countries”. Proceeds of the 4th “Just What the Doctor Ordered’: A Revised
International Conference on Computer Sciences UTAUT for EMR System Adoption and Use by
and Convergence Information Technology. Doctors”. Proceedings of the 44th Hawaii

57
Journal of Theoretical and Applied Information Technology
10th March 2013. Vol. 49 No.1
© 2005 - 2013 JATIT & LLS. All rights reserved.

ISSN: 1992-8645 www.jatit.org E-ISSN: 1817-3195

International Conference on System Sciences - [46] Duyck, P, Pynoo B, Devolder P, Voet T, Adang L
2011 and Vercruysse J (2008a). “User acceptance of a
[36] Wang, Y-S and Shih,Y-W (2009). “Why do Picture Archiving and Communication System -
people use information Kiosks?A validation of Applying the unified theory of acceptance and use
the Unified theory of Acceptance and use of of technology in a radiological setting”. Methods
of Information in Medicine 47(2): 149-156.
Technology”. Government Information
Quaterly,26,158-165. [47] Duyck, P, Pynoo B, Devolder P, Voet T, Adang L
and Vercruysse J (2008b). “Do hospital physicians
[37] Whitley,B.E (1997). “Gender Differences in really want to go digital? Acceptance of a picture
Computer-Related Attitudes and Behavior: A archiving and communication system in a
Meta-Analysis”. Computers in Human university hospital”. Röfo-Fortschritte Auf dem
Behavior, Vol. 13, No. l, pp. 1-22. Gebiet der Röntgenstrahlen und der Bildgebenden
[38] Wu, Y, Tao, Y and Yang, P (2007). “Using Verfahren 180(7): 631-638.
UTAUT to explore the behavior of 3G mobile [48] Duyck P, B Pynoo, Devolder P, Voet,T,Adang.,
communication users”. In Proceedings of The Ovaere, D and Vercruysse, J (2010). “Monitoring
International Conference on Industrial the PACS implementation process in a large
Engineering and Engineering Management University Hospital – Discepancies between
(IEEM). radiologist and physicians”. Journal of Digital
References included in the Meta-Analysis Imaging.Vol.23,No1,pp73-80.
[39] Al-Gahtani,S.S,Hubona,G.S and Wang,J. (2007). [49] Foon YS and Fah, B.C.Y (2011). “Internet
“Information technology (IT) in Saudi banking adoption in Kuala Lumpur: An
Arabia:Culture and the acceptance and use of IT”. application of UTAUT model”. International
Journal of Information and management. Journal of Business and Management. Vol.6,No.4,
(44),pp.140-154 pp161-167.
[40] Basaglia,S, Caporarello,L, Magni,M and [50] Goh, S and Yoon, T. (2011) “If you build it will
Pennarola, F (2010) Individual adoption of they come?An empirical investigation of
convergent mobile phone in Italy. Review of facilitators and inhibitors of hedonic Virtual world
Managerial Science.3,pp1-18 acceptance”. Proceedings of the 44th Hawaii
International conference on system sciences.
[41] Chea, S (2006). “Understanding the roles of
emotion in technology acceptance. University of [51] Gupta B, Dasgupta,S and Gupta,A (2008)
Hawaii at Manoa)”. ProQuest Dissertations and “Adoption of ICT in a government organization in
Theses, 146 p. ( UMI No 3216051).Retrieved from a developing country: An empirical study”.
http://search.proquest.com/docview/305319501?ac Strategic Information systems,17,pp140-154.
countid=47520 on Retrieved June 20, 2011 [52] He D and Lu Y (2007). “Consumers perception
[42] Chiu,C-M and Wang, E.T.G. (2008) and acceptances towards Mobile advertising: An
“Understanding web-based learning continuance empirical study in China”. International
intention: The role of subjective task value”. Conference on Wireless Communications
Information & Management,45,pp194-201. Networking and Mobile Computing Vols 115.
[43] Chiu,Y-T,H.,Fang, S-C and Tseng,C-C (2010) [53] Johnston A.C.(2006). “An empirical investigation
“Early versus potential adopters Exploring the of the influence of fear appeals on attitudes and
antecedents of use information in the context of behavioral intentions associated with
retail service innovations”. International journal recommended individual computer security
of retail and distribution management. actions”. Mississippi State University). ProQuest
Vol.38,No.6, pp443-459. Dissertations and Theses, , 188 p. UMI No
3213966 Retrieved from
[44] Clarke, D. C (2001) “Differences in student
http://search.proquest.com/docview/305307755?ac
dropout rate, attendance rate, and grade point
countid=47520 on June 20, 2011.
average after the implementation of the North
Carolina Dropout Prevention/Drivers License
Law”. Dissertation Abstract International, 62(3),
853A. (UMI No. AAT 3010522) Retrieved August
20, 2004, from Dissertations and Theses database
[45] Dijk,J.A.G.M.V., Peters,O and Ebbers,W (2008).
“Explaining the acceptance and use government
Internet services: A multivariate analysis of 2006
survey data in Netherlands”. Government
information Quarterly. 25,375-399.

58

View publication stats

You might also like