Professional Documents
Culture Documents
net/publication/43980359
CITATIONS READS
249 2,310
3 authors, including:
All content following this page was uploaded by Tanya Jane McGill on 02 January 2014.
Table of Contents
i Editorial Preface—What the Next IT Revolution Should Be
David Paper, Utah State University, USA
The associate editor discusses ways to position MIS for the future.
62
BOOK REVIEW
Web Work: Information Seeking and Knowledge Work on the World Wide Web
Review by Mohamed Taher, Library and Information Consultant, Canada
ABSTRACT
DeLone and McLean’s (1992) model of information systems success has received much attention
amongst researchers. This study provides the first empirical test of an adaptation of DeLone
and McLean’s model in the user-developed application domain. The model tested was only
partially supported by the data. Of the nine hypothesized relationships tested, four were found
to be significant and the remainder not significant. The model provided strong support for the
relationships between perceived system quality and user satisfaction, perceived information
quality and user satisfaction, user satisfaction and intended use, and user satisfaction and
perceived individual impact. This study indicates that user perceptions of information systems
success play a significant role in the user-developed application domain. There was, however,
no relationship between user developers’ perceptions of system quality and independent experts’
evaluations, and user ratings of individual impact were not associated with organizational
impact measured as company performance in a business simulation. Further research is required
to understand the relationship between user perceptions of IS success and objective measures
of success, and to provide a model of IS success appropriate to end user development.
INTRODUCTION
Copyright © 2003, Idea Group Publishing. Copying without written permission of Idea Group Publishing is prohibited.
Information Resources Management Journal, 16(1), 24-45, Jan-Mar 2003 25
Individual Organizational
Impact Impact
Information User
Quality Satisfaction
Copyright © 2003, Idea Group Publishing. Copying or distributing in print or electronic forms without written permission of Idea
Group Publishing is prohibited.
26 Information Resources Management Journal, 16(1), 24-45, Jan-Mar 2003
Table 1: Summary of research that is consistent with the relationships depicted in DeLone
and McLean’s model
Relationship Study
System quality Æ user satisfaction Seddon and Kiew (1996)
Roldán and Millán (2000)
Rivard, Poirier, Raymond and Bergeron (1997)a
a
Involved UDAs
gated both prior to and subsequent to the cision support system use by Snitkin and
publication of the model), also provide em- King (1986) are consistent with the pro-
pirical support for a number of the rela- posed relationship between use and indi-
tionships. The key research that is consis- vidual impact. However, neither Gelderman
tent with DeLone and McLean’s model is (1998) nor Roldán and Millán (2000) found
summarized in Table 1. any evidence of this relationship. The rela-
Seddon and Kiew (1996) tested the tionship between user satisfaction and in-
‘upstream’ portion of the model and their dividual impact received support in Gatian’s
results provided substantial support for the (1994) study, in which significant positive
proposed relationships among system qual- relationships were found between user sat-
ity, information quality, and user satisfac- isfaction and both objective and subjective
tion. Roldán and Millán (2000) also found measures of individual impact. Gelderman’s
support for these relationships. In addition, (1998) survey of 1,024 Dutch managers
their study also considered the relationships also confirmed the relationship between sat-
between system quality and use, and infor- isfaction and both subjective and objective
mation quality and use, but failed to find a individual impact measures. Etezadi-Amoli
relationship. Baroudi, Olson, and Ives (1986) and Farhoomand (1996) and Roldán and
showed that, although user satisfaction in- Millán (2000) used only perceptual mea-
fluences use, use does not significantly in- sures of individual impact, but their results
fluence user satisfaction. Igbaria and Tan were consistent with the previously men-
(1997) and Fraser and Salter (1995) also tioned studies of this relationship. Igbaria
found support for the influence of user sat- and Tan (1997) found that user satisfac-
isfaction on system usage. tion has the strongest direct effect on indi-
The results of an earlier study of de- vidual impact, but identified a significant role
Copyright © 2003, Idea Group Publishing. Copying without written permission of Idea Group Publishing is prohibited.
Information Resources Management Journal, 16(1), 24-45, Jan-Mar 2003 27
for system usage in mediating the relation- more satisfied with the application than they
ship between user satisfaction and indi- would be if it were developed by someone
vidual impact. Empirical support for the else. This may have implications for the
relationship between individual impact and role of user satisfaction in the model.
organizational impact has been provided by Edberg and Bowman (1996) pointed out
Millman and Hartwick (1987) in their study that users may not only lack the skills to
of middle managers’ perceptions of the develop quality applications, but may also
impact of systems, and by Roldán and lack the knowledge to make realistic de-
Millán (2000). terminations about the quality of applica-
Despite the number of studies that tions that they develop. Therefore, the pos-
provide a degree of support for DeLone ited relationships between system quality
and McLean’s model of IS success, it is and user satisfaction, and system quality
difficult to compare and interpret their re- and use may also be of concern.
sults due to differences in measurement The study described in this paper was
approaches. designed to investigate the applicability of
DeLone and McLean’s (1992) model of
Concerns About the Model’s IS success to UDAs. It sought to measure
Applicability in the UDA Domain all the IS success factors included in the
model, and to demonstrate how they might
Little is known about the applicability be related in the UDA domain. In order to
of DeLone and McLean’s model in the enable testing, it was, however, necessary
UDA domain. Most support for elements to make several modifications to the model.
of the model has come from research in These are described below.
the organizational domain. Only two of the
relationships proposed in the model appear Model to be Tested
to have been specifically investigated for
UDAs (these are identified by a superscript Two modifications were made to
in Table 1). The proposed relationship be- DeLone and McLean’s model to recognize
tween system quality and satisfaction is earlier research results. DeLone and
supported by Rivard et al. (1997) who McLean had included both objective and
found a significant positive correlation be- subjective measures of system quality in
tween perceived system quality and end their single system quality category. How-
user computing satisfaction for UDAs. ever, because of concerns about the ability
Kasper and Cerveny’s (1985) study pro- of end user developers to make judgments
vided evidence for the link between indi- about system quality (Edberg & Bowman,
vidual impact and organizational impact, 1996), perceived system quality and sys-
with the improved performance of the end tem quality were specified as separate con-
user developers flowing through to their structs in the model to be tested here. In
firm’s stock price, market share, and re- addition, because prior research suggests
turn on assets. that user satisfaction causes system usage
However, the results of a study by rather than vice versa (Baroudi et al., 1986)
McGill, Hobbs, Chan, and Khoo (1998) sug- the causal path between satisfaction and
gest that the process of developing an ap- use was specified in this direction.
plication to facilitate an organizational task In the UDA domain, time spent using
predisposes an end user developer to be a system may be confounded with time
Copyright © 2003, Idea Group Publishing. Copying or distributing in print or electronic forms without written permission of Idea
Group Publishing is prohibited.
28 Information Resources Management Journal, 16(1), 24-45, Jan-Mar 2003
spent on iterative enhancement of the sys- H2: User developers are more satisfied
tem, as evolutionary change has been with systems of higher perceived in-
shown to occur in nearly all UDAs (Cragg formation quality.
& King, 1993; Klepper & Sumner, 1990). H3: User developers are more satisfied
Because of concerns that perceptions of with systems of higher perceived sys-
current UDA use might include time spent tem quality.
iteratively developing the systems, intended H4: User developers intend to use sys-
use was considered more appropriate for tems of higher perceived information
this study. Intended use has been shown to quality more often.
be a satisfactory surrogate for actual use H5: User developers intend to use sys-
in studies of organizational systems (Ajzen, tems of higher perceived system
1988; Klobas, 1995). quality more often.
A final modification to the model re- H6: Higher levels of user satisfaction re-
flects the difficulty in obtaining objective sult in higher levels of intended use.
measures of information quality, since the H7: The impact of a UDA on an
quality of information in an IS is usually individual’s work performance in-
measured by the perceptions of those who creases as intended use increases.
use the information. The measures in H8: The impact of a UDA on an
DeLone and McLean’s information qual- individual’s work performance in-
ity category were mostly of this kind. In creases as user satisfaction in-
this study, the information quality category creases.
is acknowledged to be perceived informa- H9: The organizational impact of a UDA
tion quality. The model tested in the study increases as the impact on an
is therefore the model presented in Figure individual’s work performance in-
2. creases.
Hypotheses METHOD
The hypotheses that follow directly This study was conducted in an envi-
from this model are: ronment where UDAs were used to sup-
H1: User developers’ perceptions of sys- port business decision-making. The UDAs
tem quality reflect actual system qual- studied were spreadsheet applications and
ity.
Figure 2: A Modified and Testable Representation of the DeLone and McLean (1992) Model
of IS Success Factors Showing the Hypothesized Relationships
Perceived
System H1 H5 Intended
System
Quality Use
Quality
H7
H4
H6 Individual H9 Organizational
H3 Impact Impact
H8
Information User
Quality H2 Satisfaction
Copyright © 2003, Idea Group Publishing. Copying without written permission of Idea Group Publishing is prohibited.
Information Resources Management Journal, 16(1), 24-45, Jan-Mar 2003 29
the decision-making took place in a simu- (DeLone & McLean, 1992) and in particu-
lated business environment. The partici- lar UDAs (Shayo et al., 1999), so this situ-
pants were postgraduate business students ation provided a unique opportunity to ex-
with substantial previous work experience plore the full series of relationships repre-
who were participating in a course on stra- sented in DeLone and McLean’s (1992)
tegic management. They developed and model of IS success. The opportunity to
used spreadsheet applications to support undertake a study in a partially controlled
decision-making in a business policy simu- environment, where the possible impact of
lation ‘game.’ This research environment UDAs on organizational outcomes could
was chosen for the study because it pro- be investigated with minimum confounding
vided an opportunity to explore the nature by extraneous variables, was considered
of end user development of applications, worth trading off against the greater
the impact of UDAs on organizational out- generalizability that could have been ob-
comes, and the ability of end user develop- tained from a study of end user develop-
ers to make judgments about the quality ment in actual organizations. Thus whilst
and success of the applications they de- the artificial nature of the organizational
velop, in a controlled setting. impact measures is an undeniable disad-
The major advantages of the approach vantage, the strong internal validity of the
chosen were firstly that, within the simu- approach should provide a strong founda-
lated business, participants acted as real tion for future studies with a wider range
end user developers, developing applica- of end user developers.
tions to support their ‘work.’ While con- A further reason for the choice of
ducted as part of an academic course of research environment was the fact that
study, this situation was less artificial than spreadsheets were the tool recommended
an experiment because development of for participants to develop their applications.
spreadsheets was not a requirement of the Spreadsheets are the most commonly used
business game. Whilst all participants were tool for end user development of applica-
involved in application development for the tions (Taylor, Moynihan, & Wood-Harper,
simulated business, they developed spread- 1998) and by studying their use, maximum
sheets because they recognized the poten- generalizability of results would be possible.
tial value of a UDA for decision support
rather than because of any compulsion re- The Game
sulting from the research study.
The second advantage was that be- The Business Policy Game (BPG)
cause the participants were involved in a (Cotter & Fritzche, 1995) simulates the
business simulation, it was possible to ob- operations of a number of manufacturing
tain organizational performance measures companies. Teams compete with one an-
that should have been directly linked to the other as members of the management of
performance of the individuals involved. these companies, producing and selling a
Goodhue and Thompson (1995) stressed consumer durable good. Individual partici-
the need to go beyond perceived perfor- pants assume the roles of managers, and
mance impacts and make objective mea- make decisions in the areas of marketing,
surements of performance. However, it has production, financing, and strategic plan-
proved to be difficult to measure the orga- ning. Typical decisions to be made include
nizational impact of individual applications product pricing, production scheduling, and
Copyright © 2003, Idea Group Publishing. Copying or distributing in print or electronic forms without written permission of Idea
Group Publishing is prohibited.
30 Information Resources Management Journal, 16(1), 24-45, Jan-Mar 2003
obtaining finance. As the simulation model BPG, but ultimately to have an impact on
is interactive, decisions made by one com- their performance in their unit of academic
pany influence the performance of other study. They were all Master’s of Business
companies as well as their own. Administration (MBA) students who had
In this study, the decisions required at least two years of previous professional
for the operation of each company were employment experience, as this was a con-
made by teams with four or five members. dition of entry to the MBA. Most were
Each team was free to determine its man- studying part time while working in busi-
agement structure, but in general, the ness. Their ages ranged from 21 to 49 with
groups adopted a functional structure, with an average age of 31.8; 78.5% were male
each member responsible for a different and 21.5% female. They had an average
area of decision-making. Formal group de- of 9.5 years experience using computers
cision-making sessions of about one hour (with a range from 2 to 24 years) and re-
were held before each set of decisions was ported an average of 5.9 years experience
recorded, and these were preceded by sub- using spreadsheets (with a range from 0 to
stantial preparation. Decisions were re- 15 years).
corded twice a week and the simulation The applicability of research findings
run immediately afterwards so that results derived from student samples has been
were available for teams to begin work on raised as an issue of concern (Cunningham,
the decisions for the next period. Anderson, & Murphy, 1974). However,
The simulation was run over 13 Briggs, Balthazard, & Dennis (1996) found
weeks as part of a capstone course in stra- MBA students to be good surrogates for
tegic management. It simulated five years executives in studies relating to the use and
of business performance with each bi- evaluation of technology, suggesting that the
weekly decision period equating to one fi- participants in this study can be considered
nancial quarter. Participants drew upon as typical of professionals who would be
both their previous business knowledge, and involved in user development of applica-
that acquired during their program of study. tions in organizations.
Successful decision-making required appli-
cations of equivalent complexity to those The User-Developed Applications
used in ‘real’ businesses (Cotter & Fritzche,
1995). The simulation accounted for 50% The teams developed their own deci-
of the participants’ overall course grade, sion support systems using spreadsheets to
so successful performance was very im- help in their decision-making. These deci-
portant to them. Half of these marks were sion support systems could consist of ei-
based directly on the company’s perfor- ther a workbook containing a number of
mance. linked worksheets, or a number of stand-
alone workbooks, or a combination of stand-
Participants alone and integrated worksheets and work-
books. Where several members of a team
The 79 participants in this study were worked on one workbook each was re-
end user developers, developing applica- sponsible for one worksheet, that relat-
tions to support decision-making as part of ing to their area of responsibility. Figure 3
their ‘work,’ in this case for a fictitious provides an example of the possible deci-
manufacturing company as part of the sion support configurations for the teams.
Copyright © 2003, Idea Group Publishing. Copying without written permission of Idea Group Publishing is prohibited.
Information Resources Management Journal, 16(1), 24-45, Jan-Mar 2003 31
Marketing
in industry where the ability to develop
Marketing manager
small applications is a necessary part of
Production manager Worksheet
Production
many jobs (Jawahar & Elango, 2001), yet
few spreadsheet developers have spread-
sheet development in their job descriptions
Workbook 1 (Panko, 2000). Because the successful
Workbook 2
Standalone performance of their ‘company’ had direct
Worksheet Workbook 3
Marketing manager
1
Marketing
and significant implications for their grade
Production manager
Worksheet
in the course, the allocation of grades pro-
Finance manager
Production
Worksheet vided external motivation for performance
Finance in the game. Because participants volun-
tarily developed spreadsheets as a tool to
support their performance in the game, and
Workbook 1 not as a contrived task which was in itself
Partially Workbook 2
Worksheet
evaluated, motivation to perform in this
integrated Marketing study is more similar to motivation to per-
Marketing manager
Worksheet
Worksheet
form in a business environment than to past
Production manager
Production
Finance studies that have been criticized for using
Finance manager
student participants and contrived tasks
(Cunningham et al., 1974).
In each case, a single individual was re- Procedure for Collection of Data
sponsible for the development of an identi-
fiable application: either a whole workbook Each participant was asked to com-
or one or more worksheets within a team plete a written questionnaire and provide a
workbook. Hence, the unit of the analysis copy of their spreadsheet on disk after eight
in the study was an individual’s application. ‘quarterly’ decisions had been made (four
If they wished, the participants were weeks after the start of the simulation). This
able to use simple templates available with point was chosen to allow sufficient time
the simulation as a starting point for their for the development and testing of the ap-
applications, but they were not constrained plications. Ninety-one questionnaires were
with respect to what they developed, how distributed and 79 useable responses were
they developed it, or the hardware and soft- received giving a response rate of 86.8%.
ware tools they used. The majority of ap-
plications were developed in Microsoft The Instrument
Excel© but some participants also used
Lotus 1-2-3 © and Claris Works ©. The The development of the research in-
spreadsheets themselves were not part of strument for this study involved a review
the course assessment and participants of many existing survey instruments. To
were reassured of this, so there were no ensure the reliability and validity of the
Copyright © 2003, Idea Group Publishing. Copying or distributing in print or electronic forms without written permission of Idea
Group Publishing is prohibited.
32 Information Resources Management Journal, 16(1), 24-45, Jan-Mar 2003
measures used, previously validated mea- ‘never’ and (7) is labeled ‘always.’ All
surement scales were adopted wherever items in this established scale can be inter-
possible. preted in relation to UDAs. A typical item
on this scale is ‘Does the system provide
System Quality and Perceived System the precise information you need?’
Quality
User Satisfaction
The items used to measure system Given the confounding of user satis-
quality and perceived system quality were faction with information quality and sys-
obtained from the instrument developed by tem quality in some previous studies
Rivard et al. to assess the quality of UDAs (Seddon & Kiew, 1996), items measuring
(Rivard et al., 1997). This instrument was only user satisfaction were sought. Seddon
designed to be suitable for end user devel- and Yip’s (1992) 4-item, 7-point semantic
opers to complete, yet to be sufficiently deep differential that attempts to measure user
to capture their perceptions of components satisfaction directly was used in this study.
of quality. For this study, items which were A typical item on this scale is ‘How effec-
not appropriate for the applications under tive is the system?’, measured from (1)
consideration (e.g., specific to database ‘effective’ to (7) ‘ineffective.’
applications) or which were not amenable
to independent assessment (e.g., required Intended Use
access to the hardware configurations on Development and use of decision sup-
which the spreadsheets were originally port systems was optional in the BPG, so
used) were excluded. Minor adaptations to use is a pertinent measure of success in
wording were also made to reflect the en- this study (DeLone & McLean, 1992).
vironment in which application development Because of concerns that perceptions of
and use occurred. The resulting item set current use might include time spent itera-
consisted of 40 items, each scored on a tively developing the systems, intended use
Likert scale of 1 to 7 where (1) was la- was considered more appropriate. Partici-
beled ‘strongly agree’ and (7) was labeled pants were asked to indicate their intended
‘strongly disagree.’ use of the system over the next four quar-
In addition to the participants’ assess- terly decisions in the BPG. This item was
ments of system quality, the system quality based on Amoroso and Cheney’s (1992)
of each UDA was assessed by two inde- item to measure intended use and was
pendent assessors using the same set of measured on a five- point scale ranging from
items. Both assessors were IS academics (1) ‘rarely’ to (5) ‘often.’ The timing of
with substantial experience teaching data collection for this study means that
spreadsheet design and development. The intended use would reflect responses to the
two final sets of assessments were highly success of the IS during the preceding four
correlated (r = 0.73, p = 0.000). weeks.
Copyright © 2003, Idea Group Publishing. Copying without written permission of Idea Group Publishing is prohibited.
Information Resources Management Journal, 16(1), 24-45, Jan-Mar 2003 33
Copyright © 2003, Idea Group Publishing. Copying or distributing in print or electronic forms without written permission of Idea
Group Publishing is prohibited.
34 Information Resources Management Journal, 16(1), 24-45, Jan-Mar 2003
Table 2: Summary of the Information from the Measurement Models Used to Specify
Parameters in the Structural Models
c
Composite variable; * Two items; s Single item
tor score weights reported by Amos 3.6. explained (Hair et al., 1998), and therefore
The loading of each composite variable on an indication of the success of the model in
its associated latent variable and the error explaining these variables.
associated with using the composite vari- If the hypothesized model is a valid
able to represent the latent variable were representation of end user-developed ap-
estimated as described by Hair et al. plication success, all proposed relationships
(1998). Table 2 provides a summary of the in the model (the relationships reflected in
information from the measurement models H1 to H9) should be significant. All of the
used to specify parameters in the struc- hypotheses specify a direction for the pro-
tural models. posed relationship so a one-tailed t-value
of 1.645 indicates significance at the 0.05
Structural Model Evaluation level (Hair et al., 1998).
Copyright © 2003, Idea Group Publishing. Copying without written permission of Idea Group Publishing is prohibited.
Information Resources Management Journal, 16(1), 24-45, Jan-Mar 2003 35
Table 3: Goodness of Fit Measures Model Coefficients, Standard Errors, and T-Values for the
Model
Path To Estimate Standard t-value
From error
System Quality Perceived System Quality -0.179 0.144 -1.240
Perceived Information Quality User Satisfaction 0.643 0.095 6.798***
Perceived System Quality User Satisfaction 0.310 0.105 2.955**
Perceived Information Quality Intended Use -0.113 0.258 -0.439
Perceived System Quality Intended Use -0.111 0.195 -0.568
User Satisfaction Intended Use 0.843 0.336 2.513**
Intended Use Perceived Individual Impact -0.183 0.118 -1.547
User Satisfaction Perceived Individual Impact 1.131 0.197 5.735***
Perceived Individual Impact Organizational Impact -0.022 0.058 -0.376
** p < 0.01 (one tailed test) *** p < 0.001 (one tailed test)
Figure 4: Structural Equation Model Showing the Standardized Path Coefficient for Each
Hypothesized Path and the R2 for Each Dependent Variable
R2=0.031 R2=0.272
Perceived
System -0.18 -0.086 Intended
System
Quality Use
Quality
-0.19 R2=0.577 R2=0.002
-0.09 Perceived
0.61** -0.04 Organizational
Individual
0.34** Impact
Impact
0.84***
Perceived
User
Information
0.70*** Satisfaction
Quality
R2=0.607
Copyright © 2003, Idea Group Publishing. Copying or distributing in print or electronic forms without written permission of Idea
Group Publishing is prohibited.
36 Information Resources Management Journal, 16(1), 24-45, Jan-Mar 2003
and user satisfaction and individual impact and McLean’s original model, whilst not
(H8). These are illustrated in Figure 5. formally explored in this paper was ad-
dressed in post hoc analysis. When the
User Satisfaction Reflects Perceived model was altered to include a two-way
Information Quality and Perceived relationship between use and satisfaction
System Quality and then tested using AMOS, there was
The findings that perceived informa- an identification problem, which meant that
tion quality had a large positive influence the model could not be uniquely estimated.
on user satisfaction, and that perceived It hence could not be accepted. This post
system quality had a significant positive in- hoc analysis does not, however, preclude a
fluence on user satisfaction, are consistent more complex relationship, which should
with the findings of Seddon and Kiew be tested in future research: user satisfac-
(1996) for organizational systems. Seddon tion may explain intended use, while actual
and Kiew (1996) suggested that user sat- use may affect subsequent user satisfaction.
isfaction might be interpreted as a response
to three types of user aspirations for a sys- User Satisfaction Influences Perceived
tem: information quality, system quality, and Individual Impact
usefulness. Perceptions of information qual- User satisfaction strongly influenced
ity and system quality should then explain the perceived impact of the UDA on the
a large proportion of variance in user satis- individual user (R2 = .577). Again, this find-
faction. ing is consistent with the results of studies
conducted with organizational systems
User Satisfaction Influences Intended (e.g., Gatian, 1994; Gelderman, 1998;
Use Roldán & Millán, 2000). In this study, the
User satisfaction had a significant more satisfied the user developers were
positive influence on intended use. Thus, with their systems, the more strongly they
the more satisfied with an application an agreed that the system helped them per-
end user was, the more they intended to form well in the business game.
use the application in the future. This is con-
sistent with Baroudi et al.’s (1986) findings Hypothesized Relationships Not
in the organizational domain. Supported by This Research
The issue of a two-way relationship
between use and satisfaction, as in DeLone The hypothesized paths that were not
Figure 5: Relationships Between IS Success Factors Supported by this Research in the UDA
Domain
Perceived
System Intended
Quality Use
0.34** 0.61**
User
Satisfaction
0.70*** 0.84***
Perceived Perceived
Information Individual
Quality Impact
Copyright © 2003, Idea Group Publishing. Copying without written permission of Idea Group Publishing is prohibited.
Information Resources Management Journal, 16(1), 24-45, Jan-Mar 2003 37
supported by this study were: system system quality on use via user satisfaction
quality→perceived system quality (H1); was not significant. The indirect influence
perceived information quality→use (H4); of perceived information quality on in-
perceived system quality→use (H5); use tended use has been demonstrated in re-
→ individual impact (H7); and individual im- search on other types of systems (Klobas
pact →organizational impact (H9). & Clyde, 2000; Klobas & Morrison, 1999).
These observations confirm the need for
System Quality Does Not Influence Perceived further research on how perceived quality
System Quality affects intended system use, with the me-
The lack of relationship between sys- diation of attitudes including (but not lim-
tem quality and perceived system quality ited to) user satisfaction.
in this study provides justification for the The lack of evidence for any linear
concerns expressed in the literature about influence (either direct or indirect) of per-
the ability of end users to make realistic ceived system quality on intended frequency
judgments of system quality (Edberg & of use may point to a different influence
Bowman, 1996). function. Users may need to use a poor
The lack of relationship between sys- quality system more frequently to meet their
tem quality and perceived system quality needs. Alternatively, they may choose to
might be due to two factors. Firstly, end use a high quality system more frequently
user developers’ perceptions of system because it meets their needs well. Further
quality might be compromised if they lack research is needed to understand reasons
the knowledge to make realistic judgments. for differences on intended frequency of
Secondly, their judgment might be clouded use.
by their close involvement with both the
application development process and with Intended Use Does Not Influence Perceived
the application itself. Cheney, Mann, and Individual Impact
Amoroso (1986) argued that end user de- No significant relationship was found
velopment can be considered as the ulti- between intended use and perceived indi-
mate user involvement. End user develop- vidual impact. This is consistent with
ers are not only the major participants in Gelderman’s (1998) and Roldán and
the development process but also often the Millán’s (2000) observations in the organi-
primary users of their applications. Appli- zational domain and Seddon’s (1997) con-
cations can come to be viewed as much tention that the causal relationship between
more than merely problem-solving tools. use and individual impact proposed by
DeLone and McLean may not exist.
Perceived Information Quality and In this study, anticipated higher fre-
Perceived System Quality Do Not Directly quency of use over subsequent decision
Influence Intended Use
periods was not associated with any in-
Neither perceived information qual-
crease in perceptions that using the sys-
ity nor perceived system quality influenced
tem would have greater impact on success
intended use directly. Post hoc analysis
in the business game. If we assume that,
(Baron & Kenny, 1986) showed that infor-
given the close proximity between future
mation quality has a significant (p < 0.05)
use and survey completion, intended use is
indirect effect on use via user satisfaction,
a good surrogate for past use in this case,
but that the indirect effect of perceived
we need to explain why higher frequen-
Copyright © 2003, Idea Group Publishing. Copying or distributing in print or electronic forms without written permission of Idea
Group Publishing is prohibited.
38 Information Resources Management Journal, 16(1), 24-45, Jan-Mar 2003
cies of use are not associated with higher impact measures and obtained inconsistent
perceived individual impact. One reason results.
was identified earlier: higher frequency of Whilst changes in quantitative indica-
use may reflect an inefficient system and tors of organizational effectiveness would
therefore low productivity rather than fre- provide a clear signal of organizational im-
quent use to obtain substantive benefits. In pact, more subtle impacts may be involved.
the UDA domain, an additional issue is that DeLone and McLean (1992 p. 74) recog-
time spent using the system may be con- nized that difficulties are involved in “iso-
founded with time spent on iterative en- lating the effect of the I/S effort from the
hancement of the system. In their 18 month other effects which influence organizational
study of 51 UDAs, Klepper and Sumner performance.” Again, this issue is likely to
(1990) found that evolutionary change oc- be magnified in the UDA domain, where
curred in nearly all the UDAs. Frequency system use may be very local in scope. Any
of use may be a less valuable indicator of changes in organizational impact for a par-
system success in the UDA domain than ticular organization would be the result of
in the organizational domain, unless re- the combined individual effects of the
searchers are able to differentiate time UDAs in the organization, which may well
spent on development and time spent on be of varying quality. Individual UDAs could
unproductive work from time spent using have potentially conflicting effects on each
the system to obtain information or to as- other’s use as well as on organizational
sist directly with decision-making. effectiveness, making it difficult to detect
a systematic effect.
Individual Impact does not Influence In the study in which they reported a
Organizational Impact relationship between individual impact and
Individual impact did not have a sig- organizational impact, Kasper and Cerveny
nificant influence on organizational impact. (1985) used objective measures for both
The participants in the study evidently felt constructs. It is possible that perceived in-
their UDAs were contributing to their indi- dividual impact is not a realistic indicator
vidual performance, yet this was not re- of actual individual impact, but rather is bi-
flected in the game outcome. The relation- ased because of factors not included in this
ship between individual impact and organi- model, distorting its relationship with orga-
zational impact is acknowledged to be com- nizational impact. This would suggest that
plex (Ballantine et al., 1998; Shayo et al., user developers are not only poor judges
1999). Organizational impact is a broad of the quality of their systems, but also poor
concept, and there has been a lack of con- judges of the impact of their systems on
sensus about what organizational effective- their own performance.
ness is and how it should be measured
(Thong & Chee-Sing, 1996). Roldán and Demonstrating UDA Impact and
Millán (2000) used four measures of indi- Success Within the DeLone and
vidual impact and four measures of orga- McLean Framework
nizational impact in their investigation of the
applicability of DeLone and McLean’s The four hypothesized DeLone and
model in the executive IS domain. They McLean model paths that were supported
tested relationships between each possible in this study suggest that the impact of a
pair of individual impact and organizational UDA is mediated via user satisfaction.
Copyright © 2003, Idea Group Publishing. Copying without written permission of Idea Group Publishing is prohibited.
Information Resources Management Journal, 16(1), 24-45, Jan-Mar 2003 39
Copyright © 2003, Idea Group Publishing. Copying or distributing in print or electronic forms without written permission of Idea
Group Publishing is prohibited.
40 Information Resources Management Journal, 16(1), 24-45, Jan-Mar 2003
posal to separate impact measures from sults reported here and the criticisms of the
one another and from use: there was no DeLone and McLean model offered by
evidence of correlation between use, indi- Seddon and Ballantine and his colleagues.
vidual impact, or organizational impact. This Behavioral intention models may also
study does not, however, support Seddon’s be useful in understanding UDA success.
proposal for two separate sub-systems; The most popular use model in recent IS
rather, it suggests that user satisfaction is a literature, the Technology Acceptance
key indicator of subsequent outcomes, in- Model (Davis, 1986), has been used con-
cluding use and individual impact. A single sistently to demonstrate that perceived use-
model that explains user satisfaction is fulness of a system is associated with its
therefore more appropriate than Seddon’s use (Adams, Nelson, & Todd, 1992; Davis,
proposed dual system model. 1989, 1993; Taylor & Todd, 1995). It makes
The DeLone and McLean model was intuitive sense to propose that perceived
also analyzed critically by Ballantine et al. usefulness is associated with actual use-
(1998) who, like Seddon, proposed but did fulness and therefore with the impact of
not test an alternative. The Ballantine model an IS. Several richer use models have been
suggested that a three-dimensional model developed from Ajzen and Fishbein’s work
of success may be more appropriate, but on the social psychology of human behav-
again the present study does not support ior (the Theory of Reasoned Action,
such a separation. (Fishbein & Ajzen, 1975); the Theory of
A different approach has been fol- Planned Behavior (Ajzen, 1991)). These
lowed by Goodhue and colleagues models characterize use as a human be-
(Goodhue, 1988; Goodhue, 1995; Goodhue, havior influenced by beliefs about, and at-
Klein, & March, 2000; Goodhue & Thomp- titudes to, the outcomes of use, and useful-
son, 1995). Drawing on the job satisfaction ness as one of the desired outcomes asso-
literature, they proposed that an explana- ciated with use. One such model, the
tion of IS success needs to recognize the Planned Behavior in Context (PBiC) model
task for which the technology is used and (Klobas & Clyde, 2000; Klobas &
the fit between the task and the technol- Morrison, 1999), has been used to demon-
ogy. They proposed a Technology to Per- strate that users’ attitudes to a range of
formance Chain that is consistent with individual impacts (outcomes), including but
DeLone and McLean’s model in that both not limited to usefulness, influence their in-
use and user attitudes about the technol- tention to use Internet-based ISs. Provided
ogy lead to individual performance impacts. there is a relationship between the outcomes
Reflection on Goodhue’s concept of task- of use that are valued by individual users
technology fit suggests that the lack of ob- and the impact of systems on individuals
served relationship between use and im- and organizations, the PBiC and other use
pact in the study reported here may be ex- models based on Ajzen and Fishbein’s work
plained by the need to use the system for may contribute to more satisfactory expla-
more tasks (learning and development) than nations of IS success. Further research in
the functional tasks on which impact (per- this direction is recommended.
formance) measures were based. None-
theless, Goodhue’s model does not resolve
the questions of relationship between use
and user attitudes raised by both the re-
Copyright © 2003, Idea Group Publishing. Copying without written permission of Idea Group Publishing is prohibited.
Information Resources Management Journal, 16(1), 24-45, Jan-Mar 2003 41
CONCLUSIONS APPENDIX
This study has provided the first em- Items used to measure constructs in the DeLone and
McLean model
pirical test of an adaptation of the DeLone
and McLean model in the UDA domain. Information Quality
The model was only partially supported by Do you get the information you need in time?
the data. Of the nine hypothesized relation- Does the system provide output that seems to
be just about exactly what you need?
ships tested by SEM, four were found to
Does the system provide the precise informa-
be significant and the remainder not sig- tion you need?
nificant. The analysis provided strong sup- Does the system’s information content meet your
port for relationships between perceived needs?
system quality and user satisfaction, per- Is the information provided by your system un-
derstandable?
ceived information quality and user satis- Is the information provided by your system
faction, user satisfaction and intended use, complete?
and user satisfaction and perceived indi-
vidual impact. System Quality and Perceived System Quality
Economy
It is notable that the model paths that
The system increased my data processing ca-
were supported in this study are those that pacity
reflect user perceptions rather than objec- Portability
tive measures. User satisfaction reflects a The system can be run on computers other than
user’s perceptions of both quality of the the one presently used
The system could be used in other similar orga-
system itself and the quality of the infor- nizational environments, without any major
mation that can be obtained from it. In- modification
tended ongoing use of the IS reflects user Reliability
satisfaction, and the impact that an indi- Unauthorised access is controlled in several parts
of the system
vidual feels an IS has on their work re-
The data entry sections provide the capability
flects their satisfaction with the IS. How- to easily make corrections to data
ever, no significant paths were found in- Corrections to errors in the system are easy to
volving the objectively measured constructs make
system quality and organizational impact. Understandability
The same terminology is used throughout the
System quality did not influence perceived system
system quality, and perceived individual Data entry sections are organized in such a way
impact did not influence organizational im- that the data elements are logically grouped to-
pact. gether
The data entry areas clearly show the spaces
This study indicates that user percep-
reserved to record the data
tions of IS success play a significant role in Data is labeled so that it can be easily matched
the UDA domain. Further research is re- with other parts of the system
quired to understand the relationship be- The system is broken up into separate and inde-
tween user perceptions of IS success and pendent sections
Each section has a unique function
objective measures of success, and to pro- Each section includes enough information to help
vide a model of IS success appropriate to you understand its functioning
end user development. The documentation provides all the information
required to use the system
The documentation explains the functioning of
the system
Copyright © 2003, Idea Group Publishing. Copying or distributing in print or electronic forms without written permission of Idea
Group Publishing is prohibited.
42 Information Resources Management Journal, 16(1), 24-45, Jan-Mar 2003
Copyright © 2003, Idea Group Publishing. Copying without written permission of Idea Group Publishing is prohibited.
Information Resources Management Journal, 16(1), 24-45, Jan-Mar 2003 43
Copyright © 2003, Idea Group Publishing. Copying or distributing in print or electronic forms without written permission of Idea
Group Publishing is prohibited.
44 Information Resources Management Journal, 16(1), 24-45, Jan-Mar 2003
Copyright © 2003, Idea Group Publishing. Copying without written permission of Idea Group Publishing is prohibited.
Information Resources Management Journal, 16(1), 24-45, Jan-Mar 2003 45
Val Hobbs is a senior lecturer in the School of Information Technology at Murdoch University.
She has undergraduate degrees in computer science and ecological science, and completed
her PhD at Aberdeen University. Her main research interests and publications are in the
fields of end user computing, information technology education, database modelling and
design, and knowledge management.
Jane Klobas is an associate professor in the School of Media and Information at Curtin
University of Technology and Visiting Professor at Bocconi University, Milan. She has a
PhD in the psychology of information system and information resource use. Her research
concerns evaluation and use of information systems and information resources, and
incorporates elements of psychometrics and social and economic impact studies. She has
published widely on information management and applications of educational technology.
Copyright © 2003, Idea Group Publishing. Copying or distributing in print or electronic forms without written permission of Idea
Group Publishing is prohibited.