Sie sind auf Seite 1von 18

Information & Management 43 (2006) 109126

www.elsevier.com/locate/dsw

Acceptance of electronic tax filing:


A study of taxpayer intentions
Jen-Ruei Fu a,*, Cheng-Kiang Farn a,1, Wen-Pin Chao b,2
a
Department of Information Management, National Kaohsiung University of Applied Sciences,
415 Chien Kung Road, Kaohsiung 807, Taiwan, ROC
b
National Tax Administration of Northern Taiwan Province, Ministry of Finance,
No. 186 Fushing Road, Tau-Yuan City, Tao-Yuan County, Taiwan 330, ROC
Received 2 April 2004; received in revised form 6 February 2005; accepted 1 April 2005
Available online 29 August 2005

Abstract

Government today has benefited from information technology in many ways. The importance of understanding and
influencing citizens acceptance of e-government services is critical, given the investment in technology and the potential for
cost saving. Using the filing of personal income tax in Taiwan as an example of e-government services, this study integrated two
important theories (TAM and TPB), and discussed the factors affecting the taxpayers intention to adopt a particular tax-filing
method (from manual, two-dimensional barcode, or Internet) based on empirical data gathered from a large-scale nationwide
survey. The demographic characteristics and perceptions of the taxpayers were also explored in order to identify potential
determinants. Results indicated that taxpayers tend to concentrate on the usefulness of a tax-filing method and may be fairly
pragmatic in developing general attitudes towards using the method. Interestingly, the effects of perceived ease of use, subjective
norms, and self-efficacy on behavioral intention were different for manual and electronic tax-filers. Understanding these factors
can extend our knowledge of taxpayers decision making and lead to better planning and implementation of e-government
services.
# 2005 Elsevier B.V. All rights reserved.

Keywords: Electronic tax filing; e-Government acceptance; Technology acceptance mode; Theory of planned behavior

1. Introduction

* Corresponding author. Tel.: +886 3 4515811x793; According to the annual Global e-Government
fax: +886 3 462 1348. Study of Brown University in the United States,
E-mail addresses: fred@msa.vnu.edu.tw (J.-R. Fu), Taiwans ranking rose from the worlds number 2
ckfarn@mgt.ncu.edu.tw (C.-K. Farn),
samchao@im.ntx.gov.tw (W.-P. Chao).
position in 2001 to first in 2002, among the 198
1
Tel.: +886 3 4267250. countries surveyed [10]; e-government in Taiwan
2
Tel.: +886 3 3396789x1115. provides its citizens and organizations with convenient

0378-7206/$ see front matter # 2005 Elsevier B.V. All rights reserved.
doi:10.1016/j.im.2005.04.001
110 J.-R. Fu et al. / Information & Management 43 (2006) 109126

access to government information and services. agency. Error checking on paper is fault-intensive and
Electronic Filing of Personal Income Tax (eTax) is was traditionally done by the tax agency staff.
an important application that automates tax related The explosive growth in computing and use of the
processes in an attempt to improve efficiency in Internet provided a way for the tax agency to improve
assessing and collecting tax information. It has the service quality while reducing cost. After a 2-year
potential to improve tax-filing service while at the small-scale pilot study, in 1998 the agency introduced
same time reducing costs to both taxpayers and tax Internet-based and 2D barcode tax-filing methods in
collecting agencies. However, despite the readily an attempt to reduce the costs of tax collection. Both of
available tax preparation software and its promotion these electronic tax-filing methods use tax preparation
by the government, only forty percent of all taxpayers software provided by the agency; the forms can be
switched to eTax and most taxpayers were unwilling to downloaded through the Internet or obtained from
give up their paper forms. Thus, the problem of under- widely distributed compact disks (CD-ROM) free of
utilization still remained and plagued governments. charge. Calculation, error checking and suggestion for
Understanding why people accept or reject an IS has best return option for the taxpayer are carried out
proven to be a challenging issues. The growing interest automatically by the software. Taxpayers can decide
in e-government thus raises the question of how to either directly transmit the data to the tax agency
governments can increase citizens adoption of eTax. through the Internet, or print a two- to three-page
User acceptance was defined by Swanson [30] to be paper income tax return with a 2D barcode in the lower
a potential users predisposition toward personally left-hand corner of the return and deliver it to the tax
using a specific system. A review of previous agency through the postal service. For the former, a
research suggests that many studies were anchored in taxpayer who possesses public key certification issued
behavioral intention and reported a strong causal link by a government agency can download his or her
between behavioral intention and targeted behavior personal income tax withholding records via the
[8]. Agarwal and Prasad [2] also argued that, in a Internet. Generally, a taxpayer only has to confirm the
survey-based research design, analysis of intention is validity of these records without bothering to enter
more appropriate than actual usage. them. Thus, the Internet filing method may simplify
Building on a study of technology acceptance, we and expedite the tax-filing process.
attempted to develop further understanding of tax- The primary purpose of a 2D barcode filing was to
payers behavioral intention in the Taiwanese context. facilitate a smooth transition from pen-and-paper to
Specifically, a nation-wide survey of taxpayers sup- Internet tax filing. Since most taxpayers had not
ported by the National Tax Administration (NTA) of obtained a government issued public key certificate,
Northern Taiwan Province was conducted to examine the 2D barcode provided taxpayers who had a
the actions that participants were likely to take. computer a sound alternative to filing electronically.
2D barcode tax filing was similar to 1040 filing offered
by the IRS in the U.S. A taxpayer used the tax
2. Personal tax filing in Taiwan preparation software to key in data and print a 2D
barcode which encoded all the needed information in
Starting in 1998, citizens in Taiwan were able to the return. The 2D barcode worked like a paper disk
choose among three tax-filing methods: manual, (see Fig. 1 for an example) which contained machine-
Internet-based, and two-dimensional (2D) barcode. readable information to help the agency quickly feed
Manual filing is the traditional method either by hand
or typewriter. Taxpayers usually perform complex
calculations using mental arithmetic or calculator, and
then the return is delivered to the tax agency through
the postal service or in person. After receiving a
return, the agency uses a data entry service to input the
data. This process is cumbersome, time-consuming,
and paper-intensive for both taxpayers and the tax Fig. 1. An example of 2D barcode for tax returns.
J.-R. Fu et al. / Information & Management 43 (2006) 109126 111

While it is desirable to increase the percentage of


taxpayers filing electronically, any further increase is a
challenge. It was therefore important to understand the
differences among the groups so that effective
promotional schemes could be designed.

3. Theoretical background and research model

Technology acceptance is important. We define this


Fig. 2. Number of returns filed electronically in fiscal years 1998 as an individuals psychological state with regard to
2003. his or her voluntary, intended use of a technology [16].
A fundamental intention-based theory is the Theory of
tax information forms into the agencys IS through a Reasoned Action (TRA) [13]. According to this,
scanner. The advantages to the government were external stimuli influence a persons attitude toward a
significant. Elimination of manual entry errors by behavior indirectly by influencing his or her salient
taxpayers provided the tax agency with a cleaner beliefs about the consequences of performing the
stream of income tax returns, requiring less error behavior. Adapted from this, both the technology
correction computations and adjustment notices. acceptance model (TAM) and the theory of planned
The tax agency saw electronic tax filing as an behavior (TPB) are well established in the IT arena and
opportunity to reengineer the process, making it appear to be widely accepted [23,26,31].
simpler, faster, and error free. Experience showed that
the error rate for electronically filed income tax returns 3.1. The technology acceptance model (TAM)
was less than 1% versus 20% for paper returns [14].
Electronic tax filing was fully launched in 1998 and TAM theorizes that an individuals intention
has grown steadily since then. As shown in Figs. 2 and towards using a system is jointly determined by
3, the number of electronic returns increased from perceived usefulness (PU), the users subjective
about eighteen thousand in 1998, when electronic tax probability that using a specific application system
filing became available nationwide, to over 2.1 million will increase his or her job performance [11], and
representing 43.5% of all taxpayers in 2003. For perceived ease of use (PEOU), the degree to which
electronic tax filing, it was noted that the number of the . . . user expects the target system to be free of
2D barcode returns was twice as many as that of effort. The effects of external variables (e.g., system
Internet returns in 2003, Nevertheless, Internet tax design characteristics) on behavioral intention (BI) are
filing has experienced higher growth rates than did the mediated by these beliefs. According the Davis et al.
2D barcode method in the past 2 years (Fig. 3). [12], PEOU also has a direct effect on PU [35].
In predicting usage, TAM models might be useful
within and across organizations for evaluating
applications or technologies, or to make comparisons
between user groups or applications, as suggested by
Adams et al. [1] and Subramanian [29]. However,
TAM has limitations in being applied beyond the
workplace because its fundamental constructs do not
fully reflect the variety of user task environment and
constraints. Paul and John [25] suggested that TAM is
a useful model but has to be integrated into a broader
one which would include variables related to both
Fig. 3. Number of returns filed (in thousands) for three methods in human and social factors. The theory of planned
fiscal years 19982003. behavior (TPB) takes these factors into account.
112 J.-R. Fu et al. / Information & Management 43 (2006) 109126

3.2. Theory of planned behavior (TPB) 3.3. Research model

TPB incorporates subjective norms (SN) and Neither TAM nor TPB have been found to provide
perceived behaviors control (PBC) as direct determi- consistently superior explanations or predictions of
nants of BI. Together with TAM, TPB was selected to behavior. A growing body of research has been
provide a necessary theoretical premise for the focused on integrating them to predict usage intention
research model examined in our study. from different viewpoints: SN has been included as an
TPB asserts that BI is jointly determined by ones additional predictor of intention for mandatory
attitude, which reflects positive feelings towards settings [36], while Venkatesh and Davis proposed a
performing a behavior, SN, which reflects percep- theoretical framework that integrated TPB as the
tions that other people desire the individual to determinant of PEOU [33,34] and PU; also Rie-
perform a particular way; PBC, which reflects menschneider et al. developed a series of models
internal and external constraints on the performing which involved various loose to tight combinations
of the action. In our context, TPB suggests that a and employed a combine and conquer strategy to
taxpayer is more willing to file electronically if she examine their overall fit and relative explanatory
or he has positive attitude towards using eTax, wants power. Finally, the decomposed theory of planned
to comply with other important peoples opinions on behavior, which decomposes attitude using core
the use of eTax, and has the requisite resources, constructs of TAM, was proposed by Mathieson and
skills, or opportunities. has been further validated by Taylor and Todd and

Fig. 4. Research model.


J.-R. Fu et al. / Information & Management 43 (2006) 109126 113

Chau et al. The decomposed model allows researchers As the literature in consumer behavior suggests,
to pay attention to key antecedents to intention and has when engaging in an online transaction process,
provided a richer understanding of how each of these concerns of inherent risk have surfaced as salient to
antecedents explains variance across users. Given the many potential adopters [18]. However, actual risk is
focus on behavioral intention, we used the decom- difficult to measure objectively, literature has there-
posed model in our research model. fore primarily addressed the notion of perceived risk
Using TPB as a basic structure, attitude is (PR), defined here as the taxpayers perception of the
decomposed by incorporating PU and PEOU. The uncertainty and adverse consequence of a desired
approach provides several advantages. First, it seems outcome. Perceived risk of online shopping has been
unlikely that belief structures will be consistently shown to influence attitude towards online purchasing
related to intention; the relationships should become [17]. Taxpayers may hesitate to file their tax return
clearer and more readily understood. Second, by electronically if they perceive a lack of security in the
focusing on specific beliefs, the model becomes more process.
managerially relevant, pointing to specific factors that Also included in our model is compatibility (CMP),
may influence adoption and usage. The model is which serves as an antecedent of both PU and PEOU.
shown in Fig. 4. Compatibility, defined as the degree to which the
Several studies have examined the role of attitude technology fits the potential adopters previous
on IT usage behavior. The attitude variable was experience, work practice, and needs, has been
considered as a mediating variable that influences the identified as an essential factor for innovation
BI and is affected by PU and PEOU. However, weak adoption [27]. Considerable prior research has
support of the relationship was found between the reported significant effect of CMP on the user
attitudes toward a specific IS and BI. Kay [21] technology acceptance decision. More recently, Chau
suggested that it would be better to be as specific as et al. examined the importance of CMP in physicians
possible about the content of the attitude object with technology acceptance and found that CMP is a
respect to measuring attitudes toward IS usage. The significant determinant of PU but not PEOU.
omission of attitude allows researchers to investigate
the influence of PEOU and PU on BI. Thus, the
attitudes factor is removed to simplify the model. 4. Research method and results
Additionally, a link between PEOU and BI is included
because empirical studies have found a significant 4.1. Data collection
relationship between them [22].
The PBC construct refers to an individuals Data to test the model were drawn from a cross-
perceptions of the presence or absence of resources sectional field study of Taiwanese taxpayers in 2003.
or opportunities necessary for performing an action. It A nation-wide questionnaire-based survey, supported
is decomposed into both the internal notion of by the tax agency, was conducted. Since taxpayers
individual self-efficacy (SE) [4] and to external were able to adopt any one of the three filing methods
resource constraints, similar to Triandiss [32] facil- (manual, 2D barcode, or Internet tax filing) at will,
itating conditions. SE is related to perceived ability. subjects were classified into one of three groups
With respect to eTax usage, we anticipated that higher according to the method they adopted. Three
levels of SE would lead to higher levels of BI toward variations of a questionnaire were developed and
eTax. The facilitating conditions consist of two sub- different data collection methods were employed to
dimensions: resource factors (RFC), such as computer target the different groups.
equipment, and technology issues (TFC), that may Questionnaires were distributed via an online
restrict usage. The absence of facilitating resources website, offline CD-ROM, and traditional mail for
results in barriers to usage which may inhibit usage. the convenience of respondents. First, the online
Thus, a taxpayer might be unwilling to accept eTax if survey was set up on the website of the tax agency
little or no computer equipment is available, and if (http://www.ntx.gov.tw). After completing the filing
technical support is low. of an income tax return, the tax preparation program
114 J.-R. Fu et al. / Information & Management 43 (2006) 109126

invited the taxpayer to participate in the survey. The relevant to the target populations were needed. For
respondent could choose to participate or not with a example, a sample item to assess Internet taxpayers
mouse click. The completed response was stored behavioral intention: I intend to use Internet filing
directly into a database. In order to access offline method for my income tax return next year was
users, the survey program was bundled with the tax slightly modified for manual taxpayers to: I intend to
preparation software on the CD-ROM. To solicit use manual filing method for my income tax return
responses, subjects were invited to fill out the next year. Since it is meaningless to assess manual
questionnaire soon after completing their filing. The taxpayers SE, TFC, and RFC for pen-and-paper
completed response was encoded and saved together technology, items for these constructs were consistent
with the income tax return in the 2D barcode for across the three groups in assessing taxpayers PBC
researchers. It should be noted that data collected for electronic tax-filing methods. Appendix A lists a
using electronic means (e.g., WWW, offline program) final Internet-based version of questionnaire items
have similar psychometric qualities to data collected used to measure each construct and its source.
through physical procedures [28], thus the methods A pretest was conducted to ensure that the
used here can be justified. Finally, it was not easy for instrument possessed acceptable reliability and valid-
researchers to approach pen-and-paper taxpayers who ity. The subjects in the pretest sample were working
did not use or have computer equipment, thus 1000 students who had tax-filing experience in prior years.
random respondents were selected from manual tax- One hundred and seventy-eight subjects voluntarily
filing taxpayers and questionnaires were sent by completed the survey. All of the scales had reliabilities
traditional mail to them with a stamped return of 0.8 or more. The final questionnaire contained 31
envelope addressed to the tax agency. questions to measure the constructs of interest, as well
The period of the data collection lasted from May as some demographic and other related questions.
to July of 2003 filing season. To stimulate responses, In analyzing the collected data, our study followed
reminder letters were sent out after two weeks by mail the two-step procedure recommended by Anderson
or email and respondents were told that respondent and Gerbing [3]. We estimated and re-specified the
names would be entered into a lottery for a gift. All measurement model prior to incorporating the
respondents were guaranteed confidentiality. Com- structural restrictions. Convergent and discriminant
pleted questionnaires were received from 59,166 validity of the remaining items and scales were tested
taxpayers (out of a population of 755,541 for Internet with confirmatory factory analysis (CFA) using the
and 1,427,046 for 2D-barcode taxpayers) with 26,989 EQS program [6]. Each item was modeled as a
using Internet tax filing (a 3.51% response rate), reflective indicator of its hypothesized latent con-
31,596 using 2D barcode tax filing (2.21%), and 582 struct. The nine constructs were allowed to covary in
using manual tax-filing method (58.2%). Table 1 the CFA model. The measurement models were
summarizes the demographic characteristics of the evaluated using maximum likelihood estimation.
respondents. James et al. [20] recommended having three or four
indicators for each latent construct to avoid model
4.2. Measurement of constructs identification problems. Three constructs failed to
meet this criterion. To avoid the problems associated
Constructs were measured using multiple-item with having only two indicators of three constructs
scales, drawn from pre-validated measures in previous (SN, TFC and RFC), we constrained the measurement
studies. Items for PU, PEOU, and BI were adapted errors of two items to be equal as suggested by James
from Davis and Chau et al.; items for SN, SE, PBC, et al. Appendix B presents the results of the CFA
and CMP were from Taylor and Todd; PR was analysis.
modified from Cases [7]. All items were measured Since the x2 test is sensitive to sample size and the
using a seven-point Likert-type scale with anchors on probability of rejecting any model increases as the
strongly agree and strongly disagree. Since sample size increases, even when the model is
taxpayers are divided into three groups, appropriate minimally false [5], it is suggested that other model
modifications to make questionnaire items specifically fit indices should be used, and scores of 0.9 or higher
J.-R. Fu et al. / Information & Management 43 (2006) 109126 115

Table 1
Sample demographics
Manual 2D barcode Internet
Count % Count % Count %
Computer experience (years)
None 89 15.3 174 0.6 98 0.4
13 years 132 22.7 2309 7.3 1736 6.4
46 years 114 19.6 6646 21 4926 18.2
79 years 84 14.4 7459 23.6 6218 23
10 years or above 163 28.0 15009 47.5 14011 51.9
Years on Internet
None 123 21.1 426 1.3 223 0.8
13 years 197 33.8 6473 20.5 4921 18.2
46 years 138 23.7 13006 41.2 10335 38.3
79 years 68 11.7 7489 23.7 7250 26.9
10 years or above 56 9.6 4202 13.3 4260 15.8
Gender
Female 323 55.5 14875 47.1 9721 36
Male 259 44.5 16720 52.9 17265 64
Education
Elementary school or under 30 5.2 124 0.4 78 0.3
Junior high school 24 4.1 209 0.7 154 0.6
High school 161 27.7 3960 12.5 2727 10.1
College/university 320 55 22226 70.3 17850 66.1
Graduate school 47 8.1 5072 16.1 6177 22.9
Age
Under 19 years of age 7 1.2 6 0 6 0
2024 years 30 5.2 938 3 224 0.8
2529 years 105 18 7200 22.8 852 3.2
3034 years 103 17.7 8574 27.1 6342 23.5
3539 years 111 19.1 6278 19.9 7187 26.6
4044 years 66 11.3 3969 12.6 5086 18.8
4549 years 54 9.3 2384 7.5 3341 12.4
5054 years 42 7.2 1544 4.9 2200 8.1
5559 years 19 3.3 464 1.5 1330 4.9
60 or over 45 7.7 236 0.7 420 1.6
Computer and network facilities at home
Have no computer 85 14.6 1366 4.3 690 2.6
Have computer(s) but cannot connect to Internet 64 11 1876 5.9 860 3.2
Dial up 125 21.5 8550 27.1 5893 21.8
Broadband (ADSL, cable modem) 308 52.9 19805 62.7 19545 72.4
Computer and network facilities at work
Have no computer 84 14.4 486 1.5 497 1.8
Have computer(s) but cannot connect to Internet 55 9.5 1211 3.8 1016 3.8
Dial up 31 5.3 592 1.9 536 2
LAN 100 17.2 10331 32.7 10055 37.2
Broadband (ADSL, cable modem) 312 53.6 18976 60.1 14882 55.1
Frequency of Internet use
Never 92 15.8 206 0.7 76 0.3
Less than one time per month 33 5.7 276 0.9 118 0.4
Once a month 22 3.8 212 0.7 100 0.4
Once a week 48 8.2 853 2.7 513 1.9
116 J.-R. Fu et al. / Information & Management 43 (2006) 109126

Table 1 (Continued )
Manual 2D barcode Internet
Count % Count % Count %
Once per 45 days 33 5.7 918 2.9 603 2.2
Once per 23 days 57 9.8 2951 9.3 2212 8.2
Everyday 297 51 26178 82.8 23367 86.6

Table 2
Fit statistics of the confirmatory factor analysis
x2 d.f. p-Value NFI NNFI CFI RMSEA
Manual 1056 314 <0.01 0.93 0.94 0.95 0.066
2D barcode 37102 314 <0.01 0.96 0.95 0.96 0.061
Internet 28431 314 <0.01 0.96 0.95 0.96 0.057

on Bentler Bonett Normed Fit Index (NFI), Non- (1) All indicator factor loadings (l) should be
Normed Fit Index (NNFI) and Comparative Fit Index significant and exceed 0.7.
(CFI) should be considered evidence of good fit. For (2) Construct reliabilities should exceed 0.8.
the current CFA models, NFI, NNFI, and CFI were (3) Average variance extracted (AVE) by each
over 0.9 (see Table 2), suggesting adequate model fit. construct should exceed the variance due to
Convergent validity was assessed based on the measurement error for that construct (i.e., AVE
criteria that the indicators estimated pattern coeffi- should exceed 0.5).
cient was significant on its posited underlying
construct factor. We evaluated for the measurement All l values in the CFA model exceeded 0.7 and
scales using the three criteria suggested by Fornell and were significant at p = .001 (see the t-values in
Larcker [15]: Appendix B). Composite reliabilities (rc) of con-

Fig. 5. SEM analysis for manual filers (N = 582). x2 = 1150 (d.f. = 332, p < .001), NFI = 0.93, NNFI = 0.94, CFI = 0.95, RMSEA = 0.067,
**
p-value < .001.
J.-R. Fu et al. / Information & Management 43 (2006) 109126 117

Fig. 6. SEM analysis for 2D barcode filers (N = 32,845). x2 = 45400 (d.f. = 332, p < .001), NFI = 0.95, NNFI = =0.94, CFI = 0.95,
RMSEA = 0.066, **p-value < .001.

structs ranged between 0.80 and 0.97 (see correlation parameters between factors to 1.0. Each
Appendix D). AVE, ranging from 0.67 to 0.92, was time only one correlation parameter was fixed. A
greater than the variance due to measurement error. significant x2 difference between the constrained and
Hence, all three conditions for convergent validity unconstrained CFA model indicated discriminant
were met. validity between the constrained pair of constructs.
Finally, discriminant validity was assessed using a As shown in Appendix C, such x2 differences were all
series of x2 difference tests by constraining the above 4, demonstrating adequate discriminant validity

Fig. 7. SEM analysis for Internet filers (N = 27,208). x2 = 34800 (d.f. = 332, p < .001), NFI = 0.95, NNFI = =0.95, CFI = 0.95,
RMSEA = 0.062, **p-value < .001.
118 J.-R. Fu et al. / Information & Management 43 (2006) 109126

for all scales. Fornell and Larcker recommend a Table 3


Effect on behavior intention
stronger test of discriminant validity, where the AVE
for each construct should exceed the squared Manual 2D barcode Internet
correlation between that and any other construct. Direct effect
The factor correlation matrix in Appendix D indicates PU 0.63 0.76 0.81
PEOU 0.11 0.00 0.04
that the largest squared correlation between any pair of
PR 0.05 0.03 0.04
constructs was 0.64 while the smallest AVE was 0.67. SN 0.16 0.02 0.01
Hence, the latter test of discriminant validity was also SE 0.12 0.07 0.08
met. RFC 0.04 0.02 0.04
TFC 0.00 0.03 0.06
4.3. Data analysis and results Indirect effect
PEOU 0.25 0.33 0.33
The research model presented earlier was tested CMP 0.52 0.56 0.60
using the structural equation modeling (SEM) Total effect
approach. We estimated the model for all three PU 0.63 0.76 0.81
PEOU 0.36 0.32 0.37
groups. Overall, the goodness-of-fit of the structural
PR 0.05 0.03 0.04
model was comparable to that of the previous CFA SN 0.16 0.02 0.01
models and provided evidence of adequate fit in the SE 0.12 0.07 0.08
three groups. RFC 0.04 0.02 0.04
Figs. 57 illustrate the path coefficients for the TFC 0.00 0.03 0.06
CMP 0.52 0.56 0.60
three groups, together with their significance. Two
observations are worth noting. First, the path
significance is not consistent across the three groups.
This suggests that usage intentions of different tax- found to be significant, but to have a negative impact,
filing methods are associated with different ante- suggesting that taxpayers with higher level of
cedents. Second, several coefficient of significant computer self-efficacy will tend to give up manual
paths are relatively small, especially for electronic tax-filing method. Finally, PR, TFC, and RFC have
taxpayers. The possible explanation for this is the little or no effect on BI.
statistical power of the test; the greater the power, the
more often the path will be found to be significant.
With a larger sample population of the 2D barcode and 5. Exploratory study
Internet groups, non-significant will be found less
often. Thus, the paths reported statistically significant In our exploratory study we tried to identify
may not necessarily have real meaning. potential differences among the three kinds of
Overall, PU was the strongest determinant and taxpayers. Both uni- and multivariate analysis of
explained most of the variance of BI in all three variance were used to assess the difference among the
groups, as shown in Table 3. For the TAM model, three groups. With these steps we were trying to
PEOU explained a significant amount of the variance provide a picture of the taxpayers in each group and
in PU. This is consistent with most previous research. give insight into the reasons for adoption of tax-filing
CMP was found to be a significant determinant of PU methods.
and PEOU.
The path coefficients of the other independent 5.1. Dependent and independent variables
variables in TPB model (SN, SE, TFC, and RFC) were
less than 0.1 for electronic taxpayers, which suggested The independent variable was derived by assigning
that the four variables were not important predictors of a value of 1 to Internet taxpayers, 2 to 2D barcode
BI for them. For manual taxpayers, the effects of taxpayers, and 3 to manual taxpayers. For the
PEOU, SN, and SE on BI were significant and stronger dependent variables, respondents were asked to give
than the other two types of taxpayer. SE was also their demographics (age, gender, education, prior
J.-R. Fu et al. / Information & Management 43 (2006) 109126 119

experience with computers and Internet, and computer across six demographic variables and five perceptual
and Internet resource conditions). These attributes variables; significant differences between electronic
were selected from a review of the literature [19], as and manual taxpayers were found in all variables in
well as discussion with staff of the tax agency. Table 1 the study.
shows the eight variables. Since the continuous and
interval levels for dependent variables are required for
MANOVA, gender was not included in the analysis. 6. Discussion
Finally, the perceptual variables were included to
make all possible comparisons. Through the confirmatory and exploratory ana-
lyses, this study can help practitioners and researchers
5.2. Sample selection and analysis better understand taxpayers acceptance decision of
eTax; this may shed light on interesting and subtle
MANOVA and ANOVA were performed in an difference between different kinds of taxpayers. The
attempt to understand how the three kinds of taxpayers study has the merits of conducting a technology
varied. MANOVA was employed to test if the acceptance investigation in a real-world tax-filing
centroids or the mean vectors of the three groups setting involving individual taxpayers and various tax-
were significantly different, then a pairwise, uni- filing methods. Such a setting should increase the
variate analysis was performed to determine which generalizability of the results.
variables contributed to the difference between them. While the exploratory study did not advance
Since the sample sizes of three groups varied specific hypotheses, several interesting relationships
substantially, we randomly sampled from the larger were observed. First, Table 4 suggests that manual
groups, thereby reducing their size to a level taxpayers, in comparison with electronic ones, had a
comparable to the smallest group and thus worked lower education and were older, had less computer and
with 582 samples in each group. Internet experience, had less IS resources, and were
online less frequently. Also, all the responses of
5.3. Results of exploratory study perceptions for manual taxpayers were significantly
lower than those of their counterparts. Finally,
The assumption of homoscedasticity was tested inspection of the mean and standard deviation
using Boxs M. Results indicated that M is significant revealed the fact that the electronic taxpayers intended
(M = 3570; F = 10.30; sig. = .000) which suggested to continue using the method. However, manual
that the assumption was violated. Research has shown taxpayers usage intentions were relatively diverse.
that for equal cell sizes, the significance level is not The confirmatory study represented a conceptual
appreciably affected by unequal covariance matrices replication by re-examining theoretical models in tax-
[24]. The three groups in this study are of the same filing settings. Overall, we saw major differences in
size, thus a violation of this assumption has minimal the path coefficients between manual and electronic
impact. The F test and Wilks l show that the centroid taxpayers, and little differences between electronic
of means of the dependent variables were not the same taxpayers. The same observation can also be found in
for the three groups which established that there were the exploratory study.
significant differences among them. Across the three kinds of taxpayers it is important
Along with the overall results, a post hoc, uni- to recognize that BI was largely driven by PU, which,
variate test of group differences was conducted to on its own explained, almost 40% of the variance in BI
determine which group means differed significantly for manual taxpayers and 64% for the other two kinds
from others. This helped specify the exact nature of of taxpayer. Furthermore, CMP represented the
the overall effect determined by the F test. Pairwise second largest determinant of behavioral intention.
multiple comparison tests were employed to test each Thus, it seems that taxpayers tend to concentrate on
pair of groups to identify similarities and differences. the usefulness of a tax-filing method and may be fairly
Results, as shown in Table 4, indicated that 2D pragmatic in developing general attitudes towards
barcode and Internet taxpayers were quite similar using it, and the compatibility of a tax-filing method
120 J.-R. Fu et al. / Information & Management 43 (2006) 109126

Table 4
Multi- and uni-variate analysis of taxpayers demographic and perceptual variables
Manual 2D barcode Internet F Multiple comparison
Uni-variate analysis
Computer experience
M 3.19 4.13 4.22 135.0** Int. = 2D. > Man.
S.D. 1.47 1.00 1.01
Internet experience
M 2.56 3.37 3.45 121.0** Int. = 2D. > Man.
S.D. 1.24 .97 1.02
Frequency of Internet use
M 5.16 6.66 6.76 204.0** Int. = 2D. > Man.
S.D. 2.32 .99 .71
Computer and network facilitation at home
M 3.15 3.49 3.66 48.10** Int. = 2D. > Man.
S.D. 1.12 .85 .69
Computer and network facilitation at work
M 3.97 4.61 4.49 45.30** Int. = 2D. > Man.
S.D. 1.61 .92 .99
Age
M 5.29 4.68 4.79 17.90** Man. > Int. = 2D.
S.D. 2.25 1.60 1.66
Education
M 3.57 3.98 4.11 92.60** Int. > 2D. > Man.
S.D. .89 .59 .60
Behavioral intention
M 4.15 6.19 6.43 662.0** Int. > 2D. > Man.
S.D. 1.61 0.92 0.84
Perceived usefulness
M 4.09 6.10 6.22 616.0** Int. = 2D. > Man.
S.D. 1.55 0.91 0.92
Perceived risk
M 3.47 4.48 5.28 252.0** Int. > 2D. > Man.
S.D. 1.21 1.21 1.08
Perceived ease of use
M 4.43 5.93 5.48 350.0** Int. = 2D. > Man.
S.D. 1.36 0.92 0.91
Compatibility
M 4.32 5.98 6.19 447.0** Int. > 2D. > Man.
S.D. 1.56 0.96 0.87
Subjective norms
M 3.90 5.11 4.92 101.0** Int. = 2D. > Man.
S.D. 1.53 1.55 1.61
Technology facilitating conditions
M 3.89 5.27 5.31 167.0** Int. = 2D. > Man.
S.D. 1.73 1.40 1.38
Self-efficacy
M 4.28 6.05 6.01 403.0** Int. = 2D. > Man.
S.D. 1.64 0.91 0.95
Resource facilitating conditions
M 4.33 5.40 5.53 83.20** Int. > 2D. > Man.
S.D. 1.58 1.47 1.42
Multivariate analysis Wilks l = 0.34**, F = 69.70**, h2 = .126, power = 1.0
Man.: manual, 2D.: 2D barcode, Int.: Internet.
**
p-value < .001.
J.-R. Fu et al. / Information & Management 43 (2006) 109126 121

with taxpayers practice is a crucial antecedent to their that directly influences taxpayers choice of tax-filing
behavioral intention. Given the dominant effect of PU method.
and compatibility, our findings suggested that it is
important for the agency to incorporate useful and
valuable functions into the eTax software for the 7. Conclusion
targeted taxpayers practices and service needs and
also to demonstrate the value of eTax to taxpayers. The study provided some preliminary evidence
It is interesting to note that the effects of PEOU, about the criteria that taxpayers used to evaluate their
SN, and SE on BI were inconsistent between manual adoption of a taxation IS. This could be valuable for
and electronic taxpayers. Examining the minor role of governments in designing and implementing their e-
PEOU for the electronic tax-filing systems shows that government services. However, this study has a
our results contradicted the findings reported by some number of limitations. First, the effects of PBC may
prior research. Furthermore, insignificant effect of SN depend on the taxpayers past usage experience. In our
was also found for electronic taxpayers but not for study, instruments for PBC were consistent across
manual taxpayers. The exploratory study might taxpayers but electronic taxpayers had experience
provide an explanation in that manual taxpayers with eTax while manual might not have had.
may be less adaptable to new technologies, and have Venkatesh et al. suggested that perceptions become
less mental and cognitive capacity. As a consequence, more specific after direct experience. Manual tax-
electronic taxpayers may not consider PEOU or SN an payers may base their PBC perception on their general
issue of particular importance. Moreover, they might beliefs about computers and computer use while
be more confident about their own point of view and electronic taxpayers may adjust their system-specific
place less weight on other persons opinions [9]. perceptions to reflect their interaction with eTax.
Conversely, a manual taxpayers decision to adopt tax- Second, this study used a cross-sectional design and
filing methods could be more likely to be influenced correlational data which did not provide irrefutable
by PEOU and social pressures. evidence of causation. Also, all data are self-reported,
Also interesting was the role of SE, which had giving rise to problems associated with common-
significant but negative effect on BI for manual method variance. Third, our model accounted for
taxpayers. In contrast, the effect of SE on electronic about 67.1%, 63.6% and 67.1% of the variance across
taxpayers BI was negligible. Table 4 indicates that three groups indicating that variables other than those
self-efficacy perception of electronic taxpayers was examined in this study are needed to explain
significantly higher than that for manual taxpayers. It additional variance.
seems likely that individuals with higher self-efficacy The findings might have policy implications for
in IT have more options available and feel free to the implementation of electronic tax-filing and e-
choose whatever they want. government services. Understanding adoption fac-
Another two sub-constructs of PBC (TFC and tors can extend their knowledge of taxpayers
RFC) do not seem to contribute to taxpayers future decision making and lead to better strategies. Our
choice. However, manual taxpayers perceived much research provides a starting point for government
less technology and resource support than the other seeking ways to improve citizens acceptance of
two groups, and thus facilitating conditions did eTax services. Additional research is also needed to
constrain manual taxpayers behavior. It seems likely determine whether the results of this study can be
that TFC and RFC play the role of threshold for replicated in other populations and e-government
taxpayers real behavior in choosing filing methods. services.
This suggests that the electronic tax-filing method
may not reach the entire population, and a digital
divide may exist. Acknowledgements
Finally, in spite of electronic taxpayers having
perceived higher risk than manual taxpayers, contrary We gratefully acknowledge support from the
to our prediction, PR may not be an important factor National Tax Administration of Northern Taiwan
122 J.-R. Fu et al. / Information & Management 43 (2006) 109126

Province, Ministry of Finance in the 2003 research 3. Using the Internet tax-filing method will fit well
program for time and data collection efforts. We also with the way I live.
thank the editors and anonymous reviewers for their  Subjective norms (Source: Taylor and Todd [31])
valuable suggestions on this article. 1. People who influence my behavior would think
that I should use the Internet tax-filing method.
2. People who are important to me would think that
Appendix A. Questionnaire items I should use the Internet tax-filing method.
 Perceived risk (Source: Cases [7])
 Behavioral intention (Source: Davis [11]) 1. Use of Internet tax-filing method may cause my
1. I intend to use Internet filing method for my personal income tax information to be stolen.
income tax return next year. 2. I will feel uneasy psychologically if I use
2. In choose filing methods for my income tax Internet tax-filing methods.
return, Internet tax-filing method is my first 3. I do not think it is safe to use Internet tax-filing
priority. methods because of the privacy and security
3. I would like to recommend Internet tax-filing concerns.
method to my relatives and friends.  Self-efficacy (Source: Taylor and Todd [31])
 Perceived usefulness (Source: Davis [11]) 1. I would feel comfortable using the Internet/2D
1. Internet tax filing will be of no benefit to me (R). barcode tax-filing method on my own.
2. Using Internet tax filing will speed the tax-filing 2. If I wanted to, I could easily operate any of the
process. equipment to use the Internet/2D barcode tax-
3. The advantages of Internet tax filing will filing method on my own.
outweigh the disadvantages. 3. I would be able to use the Internet/2D barcode
4. Overall, using Internet tax filing will be tax-filing method even if there was no one
advantageous. around to show me how to use it.
 Perceived ease of use (Source: Davis[11]; Chau and  Resource facilitating conditions (Source: Taylor
Hu [8]) and Todd [31])
1. Learning to use Internet tax-filing method would 1. There will not be enough computers and network
be easy for me. equipment for me to use the Internet/2D barcode
2. I find Internet tax-filing method easy to use. tax-filing method.
3. It is not easy for me to be skillful in using 2. Using the Internet/2D barcode tax-filing method
Internet tax-filing method (R). will be too expensive for me.
4. It is easy for me to input and modify data when I 3. I cannot find appropriate computer equipment
use Internet tax-filing method. when I want to use Internet/2D barcode filing
5. Instructions for using Internet tax-filing method method for my tax return.
will be easy to follow.  Technology facilitating conditions (Source: Taylor
6. My interaction with Internet tax filing is clear and Todd [31])
and understandable. 1. It is easy for me to get support if I need help
 Compatibility (Source: Taylor and Todd [31]) when I have problems using computers or
1. Using the Internet tax-filing method will fit well Internet at work.
with the way I work. 2. It is easy for me to get support if I need help
2. Using the Internet tax-filing method will fit into when I have problems using computers or
my work style. Internet at home.
J.-R. Fu et al. / Information & Management 43 (2006) 109126 123

Appendix B. Confirmatory factor analysis (CFA)


Construct Item Manual 2D Barcode Internet
Mean S.D. Loading t-Statistic Mean S.D. Loading t-Statistic Mean S.D. Loading t-Statistic
BI BI1 4.70 1.60 0.89 6.19 0.93 0.97 6.47 0.86 0.96
BI2 3.99 1.59 0.82 25.5 6.13 0.96 0.89 295 6.29 0.97 0.86 238
BI3 4.29 1.75 0.92 31.1 6.18 0.96 0.95 406 6.47 0.87 0.95 328
PU USE1 4.17 1.57 0.89 6.03 0.95 0.88 6.16 1.00 0.86
USE2 4.26 1.63 0.89 31.6 6.10 0.95 0.93 260 6.29 0.98 0.87 194
USE3 4.28 1.56 0.93 34.9 6.05 0.98 0.95 277 6.18 1.02 0.93 222
USE4 4.23 1.57 0.93 34.9 6.08 0.95 0.95 275 6.21 0.98 0.93 222
PR PR1 3.56 1.55 0.80 4.25 1.65 0.90 5.29 1.50 0.90
PR2 3.44 1.53 0.95 25.8 3.94 1.57 0.95 294 4.92 1.51 0.93 250
PR3 3.63 1.57 0.91 25.2 4.11 1.62 0.96 303 5.16 1.50 0.95 264
PEOU PEOU1 4.63 1.55 0.84 6.13 0.88 0.80 6.02 1.08 0.79
PEOU2 4.58 1.52 0.88 27.3 5.99 1.02 0.83 170 6.01 1.07 0.85 157
PEOU3 4.45 1.61 0.83 24.6 5.89 1.11 0.80 161 5.95 1.10 0.83 152
PEOU4 4.45 1.47 0.83 24.6 5.60 1.21 0.78 157 5.49 1.32 0.78 141
PEOU5 4.69 1.31 0.87 26.4 5.71 1.11 0.85 175 5.65 1.22 0.75 134
PEOU6 4.67 1.33 0.85 25.3 5.74 1.06 0.84 174 5.60 1.30 0.70 123
CMP CMP1 4.58 1.52 0.97 5.94 0.97 0.97 6.19 0.91 0.97
CMP2 4.56 1.53 0.95 53.0 5.93 0.98 0.97 465 6.19 0.91 0.95 357
CMP3 4.32 1.54 0.92 45.3 5.84 1.06 0.91 336 6.05 1.04 0.87 250
SN SN1 3.98 1.51 0.96 5.07 1.52 0.96 5.02 1.57 0.96
SN2 3.99 1.51 0.96 36.9 5.00 1.55 0.95 172 4.96 1.58 0.96 161
TFC TFC1 3.72 1.83 0.87 4.97 1.54 0.82 5.10 1.55 0.88
TFC2 4.08 1.82 0.92 20.7 5.38 1.35 0.90 116 5.42 1.41 0.88 140

SE SE1 4.16 1.78 0.84 6.12 0.88 0.89 6.07 1.01 0.88
SE2 4.26 1.73 0.90 26.7 5.90 1.03 0.86 194 5.84 1.16 0.88 189
SE3 4.31 1.76 0.91 26.9 5.97 1.06 0.77 163 6.00 1.10 0.84 176
RFC RFC1 4.32 1.66 0.74 5.31 1.55 0.88 5.45 1.56 0.87
RFC2 4.35 1.83 0.89 9.20 5.49 1.54 0.83 92.8 5.61 1.55 0.84 75.6

Appendix C. Chi-square (x2) Tests of discriminant validity


Variables constrained Degrees of freedom Manual 2D barcode Internet
x2 Dx2 x2 Dx2 x2 Dx2
None 314 1058 37102 28431
BI + PU 315 1096 38 39612 2510 31035 2604
BI + PR 315 1138 80 46849 9747 38367 9936
BI + PEOU 315 1066 8 43542 6440 33412 4981
BI + CMP 315 1070 12 40326 3224 33027 4596
BI + SN 315 1063 5 40455 3353 32392 3961
BI + TFC 315 1190 132 40873 3771 31780 3349
BI + SE 315 1202 144 43957 6855 34140 5709
BI + RFC 315 1082 24 50130 13028 41411 12980
PU + PR 315 1156 98 48581 11479 38562 10131
124 J.-R. Fu et al. / Information & Management 43 (2006) 109126

Appendix C (Continued )
Variables constrained Degrees of freedom Manual 2D barcode Internet
2 2 2 2
x Dx x Dx x2 Dx2
PU + PEOU 315 1080 22 42354 5252 31273 2842
PU + CMP 315 1078 20 39773 2671 31428 2997
PU + SN 315 1062 4 40206 3104 31001 2570
PU + TFC 315 1190 132 40324 3222 30266 1835
PU + SE 315 1175 117 43433 6331 32510 4079
PU + RFC 315 1087 29 51669 14567 40744 12313
PR + PEOU 315 1175 117 49736 12634 37636 9205
PR + CMP 315 1173 115 47461 10359 38067 9636
PR + SN 315 1139 81 40626 3524 33001 4570
PR + TFC 315 1140 82 43483 6381 34443 6012
PR + SE 315 1138 80 50779 13677 38254 9823
PR + RFC 315 1114 56 37406 304 29829 1398
PEOU + CMP 315 1083 25 41033 3931 31475 3044
PEOU + SN 315 1063 5 41232 4130 30730 2299
PEOU + TFC 315 1154 96 40603 3501 29362 931
PEOU + SE 315 1141 83 44544 7442 31472 3041
PEOU + RFC 315 1091 33 52387 15285 39049 10618
CMP + SN 315 1064 6 39200 2098 31326 2895
CMP + TFC 315 1174 116 39051 1949 29918 1487
CMP + SE 315 1176 118 41786 4684 31536 3105
CMP + RFC 315 1072 14 50444 13342 41576 13145
SN + TFC 315 1137 79 38567 1465 29192 761
SN + SE 315 1155 97 42436 5334 32008 3577
SN + RFC 315 1074 16 40210 3108 30908 2477
TFC + SE 315 1075 17 39319 2217 28874 443
TFC + RFC 315 1170 112 44424 7322 34563 6132
SE + RFC 315 1173 115 54250 17148 41173 12742

Appendix D. Scale properties and correlations

Tables D1D3 show the scale properties and correlations.

Table D1
Scale properties and correlations for manual tax filing
AVE Composite reliability Factor correlations
BI PU Risk PEOU CMP SN TFC SE RFC
BI 0.77 0.91 1.00
PU 0.83 0.95 0.81 1.00
PR 0.79 0.92 0.05 0.00 1.00
PEOU 0.72 0.94 0.68 0.79 0.03 1.00
CMP 0.89 0.96 0.64 0.70 0.06 0.76 1.00
SN 0.92 0.97 0.59 0.59 0.05 0.57 0.57 1.00
TFC 0.80 0.89 0.19 0.17 0.01 0.06 0.14 0.05 1.00
SE 0.78 0.91 0.19 0.11 0.03 0.00 0.12 0.07 0.63 1.00
RFC 0.67 0.80 0.26 0.23 0.13 0.23 0.31 0.30 0.20 0.20 1.00
P 2 P 2 P
Reliability computed as: rc = ( l) /[( l) + var(e)]; AVE is average variance extracted (i.e., proportion of variance in construct that is not
P P P
due to measurement error) = ( l2)/[( l2) + var(e)].
J.-R. Fu et al. / Information & Management 43 (2006) 109126 125

Table D2
Scale properties and correlations for 2D barcode tax filing
AVE Composite Factor correlations
reliability BI PU Risk PEOU CMP SN TFC SE RFC
BI 0.88 0.96 1.00
PU 0.86 0.96 0.80 1.00
PR 0.88 0.96 0.05 0.09 1.00
PEOU 0.67 0.92 0.61 0.75 0.08 1.00
CMP 0.90 0.97 0.64 0.74 0.09 0.77 1.00
SN 0.91 0.97 0.30 0.35 0.05 0.37 0.37 1.00
TFC 0.74 0.85 0.32 0.40 0.09 0.49 0.45 0.24 1.00
SE 0.71 0.88 0.50 0.59 0.16 0.65 0.62 0.23 0.55 1.00
RFC 0.73 0.84 0.17 0.20 0.37 0.17 0.20 0.09 0.15 0.29 1.00

Table D3
Scale properties and correlations for Internet tax filing
AVE Composite Factor correlations
reliability BI PU Risk PEOU CMP SN TFC SE RFC
BI 0.85 0.95 1.00
PU 0.81 0.94 0.81 1.00
PR 0.86 0.95 0.03 0.08 1.00
PEOU 0.72 0.91 0.56 0.70 0.05 1.00
CMP 0.86 0.95 0.62 0.70 0.05 0.68 1.00
SN 0.92 0.97 0.25 0.33 0.03 0.34 0.29 1.00
TFC 0.77 0.87 0.34 0.45 0.09 0.56 0.48 0.27 1.00
SE 0.75 0.90 0.49 0.56 0.09 0.65 0.66 0.22 0.63 1.00
RFC 0.73 0.84 0.21 0.21 0.23 0.16 0.24 0.09 0.14 0.27 1.00

References [8] Y.K. Chau, J.H. Hu, Information technology acceptance by


individual professionals: a model comparison approach, Deci-
[1] D.A. Adams, R.R. Nelson, P.A. Todd, Perceived usefulness, sion Sciences 32(4), 2001, pp. 699718.
ease of use and usage of information technology: a replication, [9] Y.K. Chau, J.H. Hu, Investigating healthcare professionals
MIS Quarterly 16(2), 1992, pp. 227247. decisions to accept telemedicine technology: an empirical test
[2] R. Agarwal, J. Prasad, Are individual differences germane to of competing theories, Information and Management 39, 2002,
the acceptance of new information technologies? Decision pp. 297311.
Science 30(2), 1999, pp. 361391. [10] M.W. Darrell, Global e-Government, 2002. Available:http://
[3] J.C. Anderson, D.W. Gerbing, Structural equation modeling in www.insidepolitics.org/egovt02int.htm (accessed April 01,
practice: a review and recommended two-step approach, Psy- 2003).
chological Bulletin 49, 1988, pp. 411423. [11] F.D. Davis, Perceived usefulness, perceived ease of use, and
[4] A. Bandura, Self-efficacy: toward a unifying theory of beha- user acceptance of information technology, MIS Quarterly
vioral change, Psychological Review 84(2), 1977, pp. 191 13(3), 1989, pp. 319339.
215. [12] F.D. Davis, R.P. Bagozzi, P.R. Warshaw, User acceptance of
[5] P.M. Bentler, Comparative fit indexes in structural models, computer technology: a comparison of two theoretical models,
Psychological Bulletin 107, 1990, pp. 238246. Management Science 35(8), 1989, pp. 9821003.
[6] P.M. Bentler, C.P. Chou, Practical issues in structural model- [13] M. Fishbein, I. Ajzen, Belief, Attitude, Intention and Behavior:
ing, Sociological Methods and Research 16, 1987, pp. 78117. An Introduction to Theory and Research, Addision-Wesley,
[7] A.S. Cases, Perceived risk and risk-reduction strategies Reading, MA, 1975.
in Internet shopping, The International Review of Retail, [14] P.D. Fletcher, E-file: Electronic Tax Administration in the
Distribution and Consumer Research 12(4), 2002, pp. 375 United States (Projet CEFRIO Nouve aux mode`les Mode`les
394. de collaborationCollaboration, realise par Center for Tech-
126 J.-R. Fu et al. / Information & Management 43 (2006) 109126

nology in Government, 2002. Available: http://www.cefrio. [31] S. Taylor, P.A. Todd, Understanding information technology
qc.ca/../../projets/Documents/NMC_ETA.pdf (accessed Febru- usage: a test of competing models, Information Systems
ary 23, 2005). Research 6(2), 1995, pp. 144176.
[15] C.R. Fornell, D.F. Larcker, Structural equation models with [32] H.C. Triandis, Interpersonal Behavior, Brooks/Cole, Mon-
unobservable variables and measurement error, Journal of terey, CA, 1977.
Marketing Research 18, 1981, pp. 3951. [33] V. Venkatesh, Determinants of perceived ease of use: integrat-
[16] U.E. Gattiker, Technology Management in Organization, ing control, intrinsic motivation, and emotion into the technol-
Sage, Newbury Partk, CA, 1990. ogy acceptance model, Information Systems Research 11(4),
[17] H. van der Heijden, T. Verhagen, M. Creemers, Predicting 2000, pp. 342365.
online purchase behavior: replications and test of competing [34] V. Venkatesh, F.D. Davis, A model of the antecedents of
models, in: Proceedings of the 34th Hawaii International perceived ease of use: development and test, Decision Sciences
Conference System Science, Maui, Hi, 2001. 27(3), 1996, pp. 451481.
[18] D. Hoffman, T. Novak, M. Peralta, Building consumer trust [35] V. Venkatesh, F.D. Davis, A theoretical extension of the
online, Communications of the ACM 14(2), 1999, pp. 8085. technology acceptance model: four longitudinal field studies,
[19] C.S. Hwang, A comparative study of tax-filing methods: Management Science 46(2), 2000, pp. 186204.
manual, Internet, and two-dimensional bar code, Journal of [36] V. Venkatesh, M.G. Morris, G.B. David, User acceptance of
Government Information 27, 2000, pp. 113127. information technology: toward a unified view, MIS Quarterly
[20] L.R. James, S.S. Mulaik, J.M. Brett, Causal analysis: Assump- 27(3), 2003, pp. 425478.
tions, Models, and Data, Sage, Beverly Hills, 1982.
[21] R.H. Kay, An exploration of theoretical and practical founda- Jen-Ruei Fu is a assistant professor in the
tions for assessing tttitudes toward computers: the computer Department of Information Management
attitude measure (CAM), Computers in Human Behavior 9(4), at the National Kaohsiung University of
1993, pp. 371386. Applied Sciences in Taiwan, ROC. He
[22] H.P. Lu, D.H. Gustafson, An empirical study of perceived received his MBA and PhD degrees in
usefulness and perceived ease of use on computerized support information management from National
system use over time, International Journal of Information Central University in Taiwan. His current
Management 14(5), 1994, pp. 317329. research interests includes government
[23] K. Mathieson, Predicting user intentions: comparing the tech- information systems and knowledge man-
nology acceptance model with the theory of planned behavior, agement.
Information Systems Research 2(3), 1991, pp. 173191.
Cheng-Kiang Farn is a professor in the
[24] C.L. Olson, Comparative robustness of six tests in multivariate
Department of Information Management
analysis of variance, Journal of the American Statistical Asso-
at the National Central University in Tai-
ciation 69(348), 1974, pp. 894907.
wan, ROC. His teaching and research inter-
[25] L. Paul, I. John, Why do people use information technology? A
ests focus on electronic business, SCM and
critical review of the technology acceptance model Informa-
knowledge management. His major pub-
tion and Management 40(3), 2003, pp. 191204.
lications have been in the areas of electro-
[26] C.K. Riemenschneider, D.A. Harrison, P.P. Mykytyn, Under-
nic business and management of IS.
standing it adoption decisions in small business: integrating
Cheng-Kiang is also consultant to many
current theories, Information and Management 40, 2003, pp.
government agencies and enterprises.
269285.
[27] E.M. Rogers, The Diffusion of Innovation, 4th ed., Free Press, Wen-Pin Chao is an IS analyst at the
New York, 1985. National Tax Administration of Northern
[28] J.M. Stanton, An empirical assessment of data collection using Taiwan Province, Ministry of Finance in
the Internet, Personnel Psychology 51, 1998, pp. 709725. Taiwan, ROC. He received his MBA
[29] G.H. Subramanian, A replication of perceived usefulness and degree in information management from
perceived ease-of-use measurement, Decision Science 25(5/6), the National Central University. His cur-
1994, pp. 863874. rent job focuses on tax information sys-
[30] E.B. Swanson, Information System Implementation: Bridging tem, including management and
the Gap Between Design and Utilization, Homewood, Irwin, development. He is one of the electronic
IL, 1988. tax-filing project initiators in Taiwan.

Das könnte Ihnen auch gefallen