~ Pergamon

Omega, Int. J. Mgmt Sci. Vol. 23, No. 6, pp. 587-605, 1995
0305-0483(95)00035-6
Copyright © 1995 ElsevierScienceLtd
Printed in Great Britain.All rights reserved
0305-0483/95 $9.50 + 0.00

The Effects of Self-efficacy on

Computer Usage
M IGBARIA

Claremont Graduate School, Claremont, Calif., USA

J IIVARI
University of Jyvaskyla, Jyvaskyla, Finland

(Received January 1995," accepted after revision June 1995)

This paper examines the effect of self-efficacy, belief in one's capabilities of using a computer in the
accomplishment of specific tasks, on computer usage. It introduces an extended technology acceptance
model (TAM) that explicitly incorporates self-efficacy and its determinants (experience and
organizational support) as factors affectingcomputer anxiety, perceivedease of use, perceivedusefulness
and the use of computer technology. A survey of 450 microcomputer users in Finland found strong
support for the conceptual model. In accordance with TAM, perceived usefulness had a strong direct
effect 6n usage, while perceived ease of use had indirect effect on usage through perceived usefulness.
Self-efficacyhad both direct and indirect effects on usage, demonstrating its importance in the decision
to use computer technology. It also had a strong direct effect on perceivedease of use, hut only an indirect
effect on perceived usefulness through perceived ease of use. Computer experience was found to have
a strong positive direct effect on self-efficacy, perceived ease of use, perceived usefulness and usage.
Organizational support and computer anxiety had only indirect effects on usage, mainly through
perceived usefulness. Implications of these findings are discussed for researchers and practitioners.

Key words--self-efficacy, acceptance, usage, Finland, globalization, cultural differences.

INFORMATION TECHNOLOGY (IT) with its The a c c e p t a n c e a n d use o f c o m p u t e r s by

c a p a c i t y to process, store a n d t r a n s m i t infor- i n d i v i d u a l s a p p e a r to be limited due to fear o f
m a t i o n has a significant p o t e n t i a l i m p a c t on c o m p u t e r s , confidence a n d ability, resistance to
o r g a n i z a t i o n a l effectiveness a n d p r o d u c t i v i t y new technology, perceived difficulty o f use, not
[13, 17, 31, 32, 64, 87]. D e s p i t e the realization u n d e r s t a n d i n g the i m p o r t a n c e o f technology,
that I T is key to the success a n d survival o f a n d lack o f m o t i v a t i o n to a d o p t a new
c o m p a n i e s in a highly c o m p e t i t i v e e n v i r o n m e n t , t e c h n o l o g y [20, 45, 91]. G r e a t e r a t t e n t i o n needs
the p o t e n t i a l benefits o f c o m p u t e r s as aids to to be p a i d to the factors that cause i n d i v i d u a l
m a n a g e r i a l decision m a k i n g m a y n o t be fully resistance to c o m p u t e r usage.
realized due to p o o r a c c e p t a n c e by users [88]. IS research has a t t e m p t e d to identify n u m e r -
I n d i v i d u a l s are s o m e t i m e s unwilling to accept ous factors affecting computer usage
a n d use available systems a n d express less t h a n [20, 29, 30, 51, 63, 104]. A n i n d i v i d u a l ' s per-
enthusiastic response to new t e c h n o l o g y intro- ceived ability to a d o p t c o m p u t e r t e c h n o l o g y
d u c e d by c o m p a n i e s , even if the system m a y successfully has been shown to be a m a j o r factor
increase their p r o d u c t i v i t y [1 l, 102]. A Fortune affecting his o r her willingness to accept new
article states t h a t " M a n y w o r k e r s are suspicious t e c h n o l o g y [23,45, 59]. A m o n g the various
o f new t e c h n o l o g y , even hostile to it" [32] (p. 44). theoretical m o d e l s d e v e l o p e d to examine individ-

587
588 Igbaria, livari--Effects of Self-efficacy on Computer Usage

ual reactions to computing technology, the incorporating the external factors affecting
technology acceptance model (TAM) of Davis computer usage and by testing the model in a a
et al. [20] has emerged as especially promising European country. Specifically, it seeks to extend
[1, 69]. Although research on TAM has provided previous research by incorporating self-efficacy
insights into computer usage, it has focused on and its determinants (experience and organiz-
perceived ease of use and perceived usefulness as ational support) as the external factors affecting
the determinants of usage rather than on the computer anxiety, perceived ease of use,
external factors affecting these determinants. perceived usefulness and the usage of computer
Additionally, TAM suggests that individuals will technology. It tests the model using a survey
use computer technology if they believe it will targeted to professional and managerial users of
result in positive outcomes. It does not explicitly computers in Finland.
consider how individuals' expectations of their
capabilities influence their behavior. On the
RESEARCH BACKGROUND
other hand, Social Cognitive Theory claims that
beliefs about outcomes may be insufficient to
Social Cognitive Theory and self-e~icacy
influence behavior if individuals doubt their
capabilities to successfully undertake behaviors The present study is influenced by the Social
[6-8]. Bandura [6] argues that self-efficacy, in Cognitive Theory (SCT) which is a widely
addition to outcome expectations, must be accepted and empirically validated theory of
considered. He states that "individuals can individual behavior based on the work of
believe that a particular course of action will Bandura [6-8]. SCT incorporates two specific
produce certain outcomes, but if individuals expectations: (1) outcome expectations; and (2)
entertain serious doubts about whether they can expectations related to self-efficacy. Outcome
perform the necessary activities such information expectations are similar to the perceived
does not influence their behavior" [6] (p. 193). usefulness construct developed by Davis [19],
This argument emphasizes the impact of the where individuals tend to undertake behaviors
individual's cognitive state on outcomes and the they believe will help them perform their job
importance of understanding both self-efficacy better. Wood and Bandura state that "self-
and outcome expectations. The perceived efficacy refers to beliefs in one's capabilities to
usefulness construct measured by Davis [19] and mobilize the motivation, cognitive resources, and
Davis et al. [20] reflects beliefs (or expectations) courses of action needed to meet given
about outcomes. Self-efficacy, the belief that one situational demands" [101] (p. 408). SCT claims
has the ability to perform a particular action, is that both expectations are basic determinants of
an important construct of the Social Cognitive user behavior. Bandura [6-8] suggests that
Theory. Therefore, this research is aimed at perceived self-efficacy plays an important role in
investigating the role of both outcome expec- affecting motivation and behavior. The individ-
tations and self-efficacy in computing behavior. uals' perceived abilities to attain the standards
Self-efficacy has been shown to be associated they have been pursuing have an impact on
with an individual's performance in computer individual cognitive and behavioral reactions
training and technology acceptance [12, 15, 16, (i.e. motivation and performance). Those
23, 38, 45, 67, 98]. Studies have found evidence individuals who distrust their capabilities are
of a relationship between self-efficacy and (a) easily discouraged by failure, whereas those who
registration in computer courses at university are highly assured of their efficacy for goal
[45], (b) adoption of high technology products attainment will intensify their efforts when their
[44], (c) innovation [12], and (d) performance in performances fall short and persevere until they
software training [16, 38, 98]. Given the import- succeed. Bandura also identified several sources
ance of self-efficacy for predicting and improving of information about self-efficacy expectations,
work performance and behavior [7, 8, 36], these among them are enactive mastery (personal
studies argued the need for further research to experience) and verbal persuasion (e.g. perceived
examine fully the role of self-efficacy in encouragement and support from others).
computing behavior. Furthermore, prior re- Despite the acceptance of SCT within the
search on TAM is confined to North American psychological and organizational behavior liter-
samples. The present study extends TAM by ature, and given its importance for predicting
 Omega, Vol. 23, No. 6 589

and improving work performance and behavior intentions [20]. The importance of subjective
[7,8,15,16,36], it has rarely been used norms can be assumed to be related to an
within the IS context. While outcome expec- individualism-collectivism dimension in Hofst-
tations have been researched by IS researchers ede's framework. This leads to an interesting
[1, 19, 20, 51, 91], more research is needed to question as to what extent the TAM may be valid
explore fully the role of self-efficacy in computing in more collective societies than USA. Therefore,
behavior. this study examines the determinants of
computer technology acceptance in Finland,
Cultural context and external validity which differs culturally from North America in
A central concern in scientific research is several respects. In his analysis of 50 countries
external validity and "a key dimension of and 3 regions, Hofstede [48] listed the rank
external validity is international" [86] (p. 28). numbers of the countries from high to low (I for
Hofstede [47-49] recognizes that many popular the largest, 53 for the smallest) and found that
management and motivation theories such as both USA and Finland have small power
Herzberg's two-factor theory [43], Maslow's distances (scores 40 and 33, and ranks 38 and 46,
hierarchy of needs [68] and McGregor's theories respectively), but differ in individualism (scores
[70] reflect the North American culture and 91 and 63, and ranks 1 and 17, respectively),
argues that their applicability in other cultures is uncertainty avoidance (scores 46 and 59, and
questionable. A rudimentary survey of IS ranks 43 and 31/32, respectively) and especially
research indicates that almost all existing masculinity (scores 62 and 26, and ranks 15 and
empirical IS research has been conducted in 47, respectively). Because the empirical analysis
North America using American subjects. The of the long-term orientation of 23 countries did
applicability of this body of IS research findings not include Finland, it is omitted here.
to other cultures is unknown. The European Lachman et al. [56] propose that any
Journal of i n formation Systems claimed that "the comprehensive cross-national research on or-
social and cultural characteristics of European ganizations and management should incorporate
institutions can be studies as distinct from, or resource availability since resources affect
perhaps in contrast to, North American or organizational structure and behavior. Ameri-
Japanese institutions" [60] (p. 1); it also cans have more access to new technology
suggested that research addressing European (particularly microcomputers) than other users.
concerns should be conducted. Additionally, The Scandinavian countries (Denmark, Finland,
Aharoni and Burton [2] in a special issue of Norway and Sweden) form an interesting
Management Science suggested that more comparison to the United States because IT
research is needed to address the generalizability spending per capita is high. According to the
of management science, where our knowledge in recent OECD report concerning the year 1989
many ways is specific and limited to a given [77] they, together with Switzerland, occupied the
country or a culture. They concluded that since first five positions in per capita IT spending, with
"[T]he world has increasingly become a global the United States being sixth. Finland is
village, and large, multinational enterprises particularly interesting from the viewpoint of the
operate in a globally integrated fashion" (p. 1), present paper because the proportion of
we need to examine whether the findings of spending on PCs and work stations was clearly
studies conducted in one region, mainly, North the highest among the OECD countries.
America, are valid in other cultures. Despite
these concerns, cross-cultural studies in the IS
CONCEPTUAL MODEL AND RESEARCH
field are quite rare [85].
HYPOTHESES
Hofstede argues in several contexts that
popular motivation theories are culture-bound. In addition to the SCT [6-8], the theoretical
As will be explained below, TAM is derived from grounding for this research comes from Fishbein
the theory of reasoned action (TRA) [24]. A and Ajzen's [24] theory of reasoned action
conspicuous difference between the two is that (TRA), Ajzen's [3, 4] theory of planned behavior
TAM omits subjective norms, mostly because of (TPB); and Davis et al. [20] technology
methodological reasons and partly because they acceptance model (TAM). SCT provides a
were not significant in explaining behavioral theoretical basis for describing behavioral and
590 Igbaria, Iivari--Effects of Self-efficacy on Computer Usage

affective reactions to computing technology. self-efficacy is an important motivational vari-

This theory posits that expectations are the able, which influences individual affect, effort
primary determinants of behavior and affective persistence and motivation. The relationship
outcomes. Specifically, it states that environmen- between self-efficacy and perceived usefulness is
tal and personal factors such as verbal (i.e. social) meant to present the effect of self-efficacy on
persuasion and experience influence expectations motivation as well as on outcome expectations.
that subsequently affect individuals' outcomes. Additionally, individuals who feel less capable of
TAM adapted the generic Fishbein and Ajzen's handling a situation may resist it because of
[24] TRA model to the particular domain of user feelings of inadequacy or discomfort which may
acceptance of computer technology, replacing result from expected changes. On the other hand,
the TRA's attitudinal determinants (derived individuals with high self-efficacy will perceive
separately for each behavior) with two variables: the system to be easy and useful due to the effect
perceived ease of use and perceived usefulness, of self-efficacy on the degree of effort, the
employed in computer usage contexts. It argues persistence and the level of learning which takes
that behavior (computer usage) is strongly place [6] and will be less resistant to changes.
affected by perceived usefulness [1, 20]). Adams Individuals' perceived ability to use a product
et al. [1] also reported that perceived ease of use successfully affects their evaluative and behav-
has direct effects on both perceived usefulness ioral response to the product [23]. Therefore, self
and usage. Additionally, Davis et al. [20] suggest efficacy is likely to affect beliefs and behavior.
that there are external variables that affect both Specifically, it will affect system usage directly
perceived ease of use and perceived usefulness. and indirectly through perceived ease of use and
Figure 1 presents the model examined in this perceived usefulness.
study. It integrates these existing theories and The detailed hypotheses to be introduced
extends TAM by incorporating self-efficacy below are based on the idea of structural
explicitly in the model. equation modeling techniques [25]. The specific
structural equation modeling technique to be
The importance of self-efficacy applied in this paper will be introduced in greater
Self-efficacy is associated with beliefs and detail in the Methods section. The path
behavior [6-8, 34, 36]; it also has a critical coefficient corresponding to the arrows in Fig. 1
influence on decisions involving computer usage represents the direct effects between the involved
and adoption [15, 20, 23, 45, 59]. Individuals variables. An indirect effect represents those
who consider computers too complex and believe effects through the intervening variables; it is the
that they will never be able to control these product of the path coefficients along an indirect
computers will prefer to avoid them and are less route from cause to effect via tracing arrows in
likely to use them. Gist [35] also suggests that the headed direction only. When more than one

Self-Efflcacy~ , b :,el Perceived

Easeof
Computer Use
Experience I \
Computerl ~ , b , , SystemUsage
Anxiety I ~
2b, c,d, e

Organlzatlonsl~f
Support F Perceived
Usefulness

Fig. 1. The determinants and consequences of self-efficacy: a computer usage model. Note: the numbered
arrows correspond to the hypotheses described in the body of the paper.
 Omega, Vol. 23, No. 6 591

indirect path exists, the total indirect effect is to have only an indirect effect on usage through
their sum. The sum of the direct and indirect self-efficacy and other mediator variables. For
effect reflects the total effect of the variable on the example, Hill et al. [45] found that computer
endogenous variable. experience per se does not directly affect
The hypotheses below are mainly expressed in subsequent behavior regarding use of computer
terms of direct effects. When a significant direct technology, but indirectly through changes in
effect is not assumed to exist, the hypothesis only perceived efficacy. Thus, experience is hypoth-
concern the total effect. esized to have an indirect effect on computer
usage through self-efficacy, perceived ease of use
The determinants o f self-ej~ieacy and perceived usefulness. Therefore,
Bandura [6-8] identifies several determinants
Hypothesis la predicts that computer experi-
of self-efficacy, including enactive mastery skills
ence will have a positive direct effect on
and verbal persuasion. From a review of the IS
self-efficacy.
literature, these two variables are operational-
Hypothesis Ib predicts that computer experi-
ized as prior experience and support, respect-
ence will have a negative direct effect on computer
ively. Computer experience is hypothesized to
anxiety.
affect self-efficacy expectations positively with
Hypothesis lc predicts that computer experi-
perceived successes in task performance raising
ence will have a positive direct effect on perceived
mastery expectations and failures lowering them.
ease o f use,
He asserts that experience is particularly
Hypothesis ld predicts that computer experi-
influential because of its direct, personal nature.
ence will have a positive direct effect on perceived
He also hypothesized that verbal persuasion
usefulness.
positively affects self-efficacy, where perceived
I4ypothesis le predicts that computer experi-
encouragement and support from others raises
ence will not have a significant direct effect on
efficacy expectations. He also states that
computer usage but the total effect will be positive
self-efficacy expectations induced through verbal
and significant.
persuasion are likely to be weaker than those
derived from personal experience. Based on the Verbal persuasion, i.e. support and encourage-
self-efficacy paradigm, the weaker effect of verbal ment, was also hypothesized to affect beliefs,
persuasion on self-efficacy is due to the attitudes and behavior. Individuals rely, in part,
assumption that verbal persuasion does not on the opinion of others as well as the support
provide a direct, experiential base. These two and encouragement they receive in forming
factors may also be directly related to behavior judgments about their own abilities. Addition-
and motivation. Specifically, prior experience is ally, since individuals need more resources to
hypothesized to affect behavior and motivation help them become more proficient, it is expected
to the extent that the individual is able to assess that higher organizational support would result
the skill level exhibited in doing the task. This is in higher judgments of self-efficacy on the part of
consistent with the TAM and TRA models, individuals. The availability of assistance to
where these three factors are classified as external individuals who need it is likely to increase their
factors affecting both perceived ease of use and ability to perform a task. Moreover, support was
perceived usefulness. believed to be an indication of organizational
The TRA model, as well as TAM, propose that norms regarding use, and this would positively
external factors, such as experience, will affect influence outcome expectations and beliefs in
behavior through their effect on beliefs. addition to self-efficacy. Davis et al. [20] also
Computer experience has been found to be emphasized that perceived usefulness and
associated with self-efficacy, computer anxiety, perceived ease of use were affected by manage-
perceived ease of use and perceived usefulness ment support. Further, in testing a subset of the
[34-38, 51, 58, 98]. Gist et al. [37, 38] discovered model proposed by Triandis [94, 95]. Thompson
that computer experience is likely to improve a et al. [91] reported that facilitating conditions,
person's perceptions and beliefs about using the including support for users, will have a direct
technology by increasing their beliefs in their effect on utilization. Organizational support has
ability to master the challenge and to reduce any been associated with greater system usage while
fears they may have. Further, experience is likely lack of organizational support has been a critical
592 Igbaria, livari--Effects of Self-efficacy on Computer Usage

barrier to the effective utilization of computers behavior-outcome relationships, such as effort

[30, 63]. Trevino and Webster [93] also suggested to performance, should be considered analogous
that management support is positively related to to self-efficacy, because "expected performance
perceived ease of use while Igbaria [51] found outcomes depend heavily on the type of
that it was negatively related to computer behaviors an individual chooses to execute" [36]
anxiety. In accordance with TAM, the effect of (p. 185). Bandura also distinguishes between
the organizational support on usage is supposed self-efficacy and 'outcome judgment', which is
to be indirect. Therefore, similar to perceived usefulness. He states that,
"In any given instance, behavior would be best
Hypothesis 2a predicts that organizational predicted by considering both self-efficacy and
support will have a positive direct effect on outcome beliefs" [7] (p. 140). Empirical support
self-efficacy. for the effect of both self-efficacy and outcome
Hypothesis 2b predicts that organizational beliefs on computer acceptance was found by
support will have a negative direct effect on Hill et al. [36]. They observed that self-efficacy
computer anxiety. with respect to computers plays an important
Hypothesis 2c predicts that organizational role in determining an individual's decision to use
support will have a positive direct effect on ease of them. However, Gist [36] argues that the
use. complexity of the task involved will affect the
Hypothesis 2d predicts that organizational strength of the relationship between self-efficacy
support will have a positive direct effect on and performance. This suggests that perceived
perceived usefulness. difficulty of performing a task (low perceived
Hypothesis 2e predicts that organizational ease of use) may mediate the relationship
support will not have a significant direct effect on between self-efficacy and performance. Specifi-
computer usage but the total effect will be positive cally, we hypothesize that self-efficacy will affect
and significant. computer usage as well as perceived ease of use
and perceived usefulness. Therefore,
The outcomes of self-ej~cacy
Hypothesis 3a predicts that self-efficacy will
Anxiety is a "generalized emotional distress"
have a negative direct effect on computer anxiety.
[75] experienced by an individual. According to
Hypothesis 3b proposes that the individual's
Bandura, self-efficacy is negatively affected by
self-efficacy will have a positive direct effect on
emotional arousal or anxiety. Miura [71] states
his~her perceived ease of use.
that less anxious users tend to feel more
Hypothesis 3c proposes that the individual's
efficacious. However, anxieties are generally
self-efficacy will have a positive direct effect on
divided into two categories which are trait-based
his~her perceived usefulness.
(i.e personality tendencies that are stable over
Hypothesis 3d proposes that the individual's
time and situations) or state anxiety (a transitory
self-efficacy has a positive direct effect on his~her
response to a specific situation). While trait-
use of computer technology.
based anxiety may be considered as antecedent to
low self-efficacy, state anxiety is probably caused A number of studies have documented the
by low self-efficacy. Computer anxiety is a form importance of computer anxiety as a key variable
of state anxiety and is an irrational emotional related to perceived usefulness and usage
distress which is experienced by an individual [33, 52, 73]. Computer anxiety was found to have
when using or considering the use of computer a negative impact on constructs similar to
technology. Bandura [6] and Stumpf et al. [86] perceived usefulness [52]. People who are less
found that individuals experience anxiety in anxious (computerphrenics) are much more
attempting to perform behaviors they do not feel likely to interact with computers than people
competent to perform. who are more anxious (computerphobics). Davis
Bandura [7, 8] argues that self-efficacy beliefs et al. [20] also suggested that computer anxiety
function as proximal determinants of behavior should be brought into future analysis when
(here, computer usage) and motivation (per- examining the factors affecting computer usage.
ceived usefulness and perceived ease of use). Gist Computer anxiety has been linked with negative
links self-efficacy to expectancy, where expec- beliefs (perceived usefulness) about computers,
tations influence action. She suggests that the problems in playing with them, and avoidance of
 Omega, Vol. 23, No. 6 593

the technology [52, 97]. Individuals who feel The finding that the use of computer systems
comfortable with the machine are more likely to is driven to a large extent by perceived usefulness
produce desired consequences. On the other [1, 20] can be explained by motivation theory
hand, individuals who experience high levels of which argues that if an individual perceives an
anxiety are likely to behave more rigidly than activity to be instrumental for achieving valued
individuals whose level of anxiety is relatively outcomes, he or she will be more likely to accept
low. Therefore, the new technology. It has been pointed out that
perceived usefulness appears to exhibit a stronger
Hypothesis 4a predicts that computer anxiety and more consistent relationship with usage
will have a negative direct effect on perceived ease behavior than other variables reported in the
of use. literature including various attitudes, satisfac-
Hypothesis 4b predicts that computer anxiety tion and perception measures [19]. Furthermore,
will have a negative direct effect on perceived the IS literature more generally suggests that
usefulness. perceived usefulness or equivalent measures are
Hypothesis 4c predicts that computer anxiety positively associated with system usage
will not have a significant direct effect on computer [45, 50, 80, 81]. Davis et al. in testing the effect of
usage but the total effect will be negative and perceived usefulness on both attitudes and
significant. behavior, stated that attitude "was generally not
Based on TAM, perceived ease of use and found to intervene between beliefs and inten-
perceived usefulness are thought to be potentially tions" [20] (p. 994). Mathieson [69] found that
important determinants of system usage. Davis perceived usefulness is the major determinant of
et al. [20] found perceived ease of use to have a people's intentions to use computers, and
strong effect on perceived usefulness. Their work Thompson et al. [91] observed that job fit,
was replicated by Adams et al. [1] who reported paralleling the definition of perceived usefulness,
that both are important in affecting individuals' had a strong effect on PC use. Therefore,
decision to use the system (mainly in their first Hypothesis 6predicts that perceived usefulness
study) and that perceived ease of use strongly will be positively related to usage of computer
affects perceived usefulness. Mathieson [69] also technology.
found that perceived ease of use explains a
significant amount of the variance of perceived To summarize, we suggested above a number
usefulness and that perceived usefulness and of hypotheses related to the extended technology
perceived ease of use contribute to behavior. acceptance model incorporating self-efficacy. In
Additionally, several studies on innovation accordance with TAM, perceived usefulness was
adoption and diffusion have examined the considered the major determinant of computer
complexity of innovations [92, 103]. These usage followed by perceived ease of use.
studies imply high complexity is linked to low Self-efficacy was proposed to be an antecedent of
perceived ease of use [72] on the adoption of new perceived ease of use and usefulness that had
technologies [82]. Perceived complexity has been mainly indirect effects on usage through ease of
generally recognized as a factor inhibiting the use and perceived usefulness. We also examined
diffusion of technology [45, 55], because in- the determinants of self-efficacy, i.e. experience
creased complexity of an innovation requires and support. The conceptual model illustrating
increased effort on the part of the adopter, thus the research hypotheses is shown in Fig. 1.
decreasing the likelihood of adoption
[22, 46, 54]. Thompson et al. [91], for example,
METHOD
observed that perceived complexity has a direct
effect on PC use. Therefore, Sample and procedure
Hypothesis 5a predicts that perceived ease of The data for this study were gathered by means
use will have a positive direct effect on perceived of a questionnaire survey administered in
usefulness. Finland during Spring 1993. Initially, the top 120
Hypothesis 5b predicts that perceived ease of companies in Finland in terms of net sales [89]
use will have a positive direct effect on system were selected for the survey. Because of mergers,
usage. bankruptcies, and problems in making contact,
594 Igbaria, livari--Effects o f Self-efficacy on Computer Usage

Table 1. Profile of the participants

the participants were drawn from a sample of 86
Age: Average = 38.92 SD = 8.32 Median = 39.0
corporations. Of the 86 companies contacted, 81 Organizational Average = 10.45 SD = 8.32 Median = 7.5
agreed to participate in this study. Within each tenure:
Job tenure: Average = 5.46 SD = 5.32 Median = 3.0
company, contact persons were identified and Gender: Male = 53.6% Female = 46.4%
were informed by telephone of the purpose of this Education:
study and were asked to describe the type of Less than high school 9.8%
Trade school 4.9%
individual who would be participating. The High School 4.7%
contact persons were usually IS managers. Some college 45.3%
Bachelor's Degree 10.3%
Contact persons in the participating companies Graduate Degree 23.9%
were asked to distribute 10 surveys to individuals Other 1.1%
who were not EDP professionals or intermediary Organizational level:
Non-management/professional staff 51.9%
users (secretaries) and who used computers at First level supervisor 16.7%
work. Participation in the study was voluntary Middle management 25.2%
Top management/executives 3.1%
and people were assured that their individual Unclassified 3.1%
responses should be treated confidentially. A Division~Functional area:
Accounting and finance 30.3%
total of 806 questionnaires were sent to the Marketing and sales 14.1%
contact people in the 81 participating companies Technical areas 17.7 %
General mangement and personnel 13,6%
(two companies wished fewer than 10 question- Others 24.6%
naires). It was up to the contact person to decide Industry:
whom would receive the questionnaires. The Manufacturing 50.0%
Merchandising 30.0%
exclusion of incomplete and returned question- Other 20.0%
naires resulted in a final sample of 450 users from
68 companies, a response rate of 55.8%. The high
response rate may be due to the fact that the
managers encouraged participation, the survey
was translated into the local language, and the engineering, IS (17.7%), and others (24.6%). The
issue being examined was of current concern. majority of were middle levels, and the rest were
The number of employees in the 86 companies first level supervisors and executives. Of the 450
varied between 89 and 28,859; and the participants, 53.6% were males and 46.4% were
corporation's net sales ranged from 639 million females. Age ranged from 21 to 61 years, and the
Finnish marks ($127 million) to 57 billion mean age of the respondents was 39 years
Finnish marks ($11.4 billion) [90]. Small (SD = 8.32). Approximately 45% had com-
companies were mainly in the merchandising pleted some college work, and an additional 35 %
industry. When size was measured in terms of of the participants were college graduates. The
employees, there were 23 manufacturing and rest of the respondents had high school diploma
only 2 merchandising companies among the top or lower. The average length of service in the
25 companies in the initial sample of these current organization (organizational tenure) was
companies, whereas among the smallest 25 10.45 years (SD = 8.31) and the length of tenure
companies there were 17 merchandising and 4 in their current job averaged 5.46 years
manufacturing companies, and 4 in other (SD = 5.32). Table 1 summarizes the demo-
industries or could not be classified [90]. The graphic characteristics of our sample.
average size of companies was 6469 employees in The success of our sample is reflected in the
manufacturing and 1580 in merchandising in the non-significant differences between users in
initial sample. different industries, organizational level, and
The individual respondents were largely functional areas. In view of the high association
employed in two major industries: manufactur- between size and industry, the non-significance
ing (50%) and merchandising (30%). They held differences of industry suggests that the company
professional/non-managerial (51.9%) and man- size is not a significant confounding factor in the
agerial positions in a wide range of functional sample. Further analyses indicated that there
areas including accounting and finance (30.3%), were no significant differences among users
marketing and sales (14.1%), personnel and employed in different industries (manufacturing,
general management (13.6%), technical fields, merchandising, and others) and functional areas
such as operations and production, R&D, (technical vs non-technical) in terms of their
 Omega, Vol. 23, No. 6 595

demographic characteristics (i.e. age, organiz- agreement or disagreement with the following
ational tenure, and job tenure). four statements concerning microcomputers on a
The only demographic variables on which 5-point Likert-type scale ranging from (1)
users differed significantly were gender and strongly disagree to (5) strongly agree: "Using
education in technical vs non-technical func- microcomputers improves my job performance";
tional areas (F = 17.11, P < 0.001 and F = 5.21, "Using microcomputers increases my pro-
P < 0.05, respectively), a greater number of ductivity in the job"; "Using microcomputers
females than males as well as a smaller number provides me with information that would lead to
of educated participants were non-technical better decisions"; and "Using microcomputers
users. enhances my effectiveness in the job".
Perceived ease of use. Based on [19] and [20]
Measures perceived ease of use refers to the degree to which
SCT and TAM were used to explain a specific computer technology is perceived as relatively
behavior (usage) toward a specific target easy to understand and use. Individuals were
(computers) within a specific context (users in asked to indicate the extent of agreement or
Finland). The measures correspond to the the use disagreement with the following four statements
of computers in their organizations in Finland. concerning microcomputers on a 5-point scale
Usage of the computer technology. Based on ranging from (1) strongly disagree to (5) strongly
several studies [14,21, 53] two indicators of agree: "Learning to use microcomputers is easy
microcomputer usage were included in this for me"; "2 find it easy to get microcomputers to
study: (1) perceived daily use of microcomputers; do what I want to do"; "It would be easy for me
and (2) perceived frequency of use of microcom- to become skillful at using microcomputers":
puters. Individuals were asked to indicate the and "I find microcomputers easy to use."
amount of time spent on the microcomputer per Computer anxiety. This refers to the tendency
day, using a 6-point scale ranging from (1) of an individual to be uneasy, apprehensive,
'almost never' to (6) 'more than 3 hours per day'. and/or phobic towards current or future use of
Frequency of use has been suggested by computers in general. The computer anxiety
Raymond [78] and used by Igbaria et al. [53] and items were selected from the original scale
provides a slightly different perspective of use developed by Raub [79] and validated by Igbaria
than time. Frequency of use was measured on a and Chakrabarti [52]. The instrument asks
6-point scale ranging from (1) 'less than once a individuals to indicate their agreement or
month' to (6) 'several times a day'. These disagreement with 3 statements reflecting
indicators are typical of the kinds of self-reported anxiety, apprehension, confusion, hesitation etc.
measures often used to operationalize system in using computers in general (e.g. "I hesitate to
usage, particularly in cases where objective usage use a computer for fear of making mistakes I
metrics are not available. Objective use logs were cannot correct"). The response options, an-
not practical in the present study since chored on a 5-point Likert-type scale, range from
participants used different microcomputers as (1) strongly disagree to (5) strongly agree.
well as different applications for different tasks. Self-efficacy. Self-efficacy was measured using
Self-reported usage should not be regarded as a two-item scale adapted from [45]. Individuals
precise measures of actual usage, although were asked to indicate the extent of their
previous research suggests they are appropriate disagreement or agreement with the following
as relative measures [10]. two statements on a 5-point scale scale ranging
Perceived usefulness. This measure is defined from (1) strongly disagree to (5) strongly agree:
as "the prospective user's subjective probability "I will understand how computers work"; and "I
that using a specific application system will am confident that I could learn computer
increase his or her job performance within an applications."
organizational context" [20] (p. 985). The items Computer experience. Computer experience
used to construct the perceived usefulness scale was assessed by seven items asking respondents
were adapted from prior research [19, 20, 50], to indicate the extent of their experience in using
with appropriate modifications to make them application systems, different types of computer
specifically relevant to microcomputers. Individ- software packages (e.g. spreadsheet, word
uals were asked to indicate the extent of processing), financial modeling, languages (3rd
596 Igbaria, livari--Effects of Self-eJ~cacy on Computer Usage

and 4th generation) and participation in construct in Fig. 1, there is a related

non-technical analysis (feasibility studies and measurement model, which links the construct in
requirements analysis) and technical design of the diagram with a set of items. The measurement
computerized information systems. The overall model consists of the relationships between the
computer experience was the sum of these observed variables (items) and the constructs
experiences. which they measure. The characteristics of this
Organizational support. The measure of model demonstrate the construct validity of the
organizational support, developed by Igbaria research instruments, i.e. the extent to which the
[50], incorporated the general support, which operationalization of a construct actually
includes top management encouragement and measures what it purports to measure. Two
allocation of resources. Individuals were asked to important dimensions of construct validity are
indicate the extent of agreement or disagreement (a) convergent validity, including reliability, and
with the following four statements concerning (b) discriminant validity.
organizational support on a 5-point scale The test of the measurement model includes
ranging from (1) strongly disagree to (5) strongly estimation of the reliability coefficients (the
agree: "Management is really keen to see that we composite reliability) of the measures, as well as
are happy with using our microcomputers"; an examination of the convergent and discrimi-
"Management has provided most of the nant validity of the research instruments. In
necessary help and resources to get us used to the determining the appropriate minimum loadings
microcomputer quickly"; "I am always sup- required for the inclusion of an item within a
ported and encouraged by my boss to use the scale, we used Fornell's [25] recommendation to
microcomputer in my job"; and "I am convinced retain items that loaded highly (0.70 is considered
that management is sure as to what benefits can to be a high loading since the item explains
be achieved with the use of microcomputers". almost 50% of the variance in a particular
construct) on their respective constructs. Fornell
Data analyses and Larcker's [27] criterion that an average
The hypothesized relationships among the extracted variance should be 0.50 or more was
study variables depicted in Fig. 1 were tested by used to assess the average variance extracted for
means of Partial Least Squares (PLS). PLS is a all constructs. We also used the guidelines
second-generation multivariate technique that recommended by Hair et al. [41], in determining
facilitates testing of the psychometric properties the relative importance and significance of the
of the scales used to measure a variable, as well factor loading of each item, i.e. loadings > 0.30
as estimating the parameters of a structural are considered significant; loadings > 0.40 are
model, i.e. the magnitude and direction of the considered more important; and loadings 0.50 or
relationships among the model variables greater are considered to be very significant.
[25, 62, 100]. It is a powerful analytical technique Finally, the criteria suggested by Nunnally [76]
in testing structural equation models, and is were applied to determine the adequacy of the
particularly applicable in research areas where reliability coefficients obtained for each measure.
theory is not as well developed as that demanded To assess discriminant validity of the
by LISREL [26]. As suggested by Lohmoller [61] measures, i.e. the degree to which items
(p. 7) "PLS methods are more close to the data, differentiate among constructs or measure
more explorative, more data analytic". Of distinct concepts, we examined the correlations
particular relevance to this study is the fact that between the measures of potentially overlapping
PLS does not depend on having multivariate constructs [40]. If the items comprising a
normally distributed data (distribution-free). construct correlate more highly with each other
Finally, it can be used with non-interval scaled than with items measuring other constructs in the
data, and importantly, with small samples. model [28, 40], the measure is determined to have
PLS recognizes two components of a causal adequate discriminant validity.
model: the measurement model and the PLS is also used to test the structural model.
structural model. Figure 1 represents the A structural model is a regression-based
structural model being examined. The model technique, with its roots in path analysis, and
describes the relationships or paths among often loosely termed as a causal modeling
theoretical constructs. Furthermore, for each technique. It is a relatively new approach to
 Omega, Vol. 23, No. 6 597

Table 2. Assessment of the measurement model

testing multivariate models with empirical data
The composite
[100]. The structural model consists of the reliability Average variance
unobservable constructs and the theoretical Variables (~ coefficients) extracted/explained
relationships among them (the paths). It Computer experience 1.00 1.00
Organizational support 0.86 0.60
evaluates the explanatory power of the model Self-efficacy 0.73 0.58
and the significance of paths in the structural Computer anxiety 0.82 0.61
Perceived ease of use 0.94 0,79
model which represent hypotheses to be tested. Perceived usefulness 0.96 0.86
Together, the structural and measurement System usage 0.94 0.88
models form a network of constructs and
measures. The item weights and loadings indicate
the strength of measures, while the estimated analyses so that conclusions would not be
path coefficients indicate the strength and the confounded by gender and education differences
sign of the theoretical relationships. across functional areas and industries. Industry
The evaluation of the structural model was and functional area differences in the set of the
conducted with the overall sample. The constructs under investigation were examined
computer program used for this analysis was using a multivariate analysis of variance
LVPLS 1.6 (Latent Variables Path Analysis (MANOVA) with interaction terms. Industry
using Partial Least Squares), developed by and functional areas were the independent
Lohmoller [61, 62]. To test the estimated path variables, the study constructs were the depen-
coefficients, t-statistics were calculated using a dent variables, and the interactions were gender
nonparametric test of significance known as and education. The results show that both F-tests
jackknifing [96, 99]. For more information on for industry and functional areas were insignifi-
PLS, the interested reader can refer to [25] and cant ( F = 1.45 and 0.60, respectively). This
[62]. indicates that no significant differences were
The path coefficient of an exogenous variable found in the study constructs across functional
represents the direct effect of that variable on the areas and industry.
endogenous variable. An indirect effect rep-
resents those effects interpreted by the interven- RESULTS
ing variables; it is the product of the path
coefficients along an indirect route from cause to The measurement model
effect via tracing arrows in the headed direction Table 2 gives the results of the measurement
only. When more than one indirect path exists, model. It displays the composite reliability
the total indirect effect is their sum. The sum of coefficients describing the internal consistency of
the direct and indirect effects reflect the total the measurement model and the average variance
effect of the variable on the endogenous variable extracted. The data show that the composite
[5, 83]. reliabilities of the constructs range from 0.73 to
We also examined the study constructs across 0.96, which satisfy Nunnally's [76] guidelines.
industries and functional areas (technical vs Furthermore, Table 2 demonstrates satisfactory
non-technical). As indicated earlier, there was a convergent validity of the constructs. Results, in
significant relationship between gender and general, show that the convergent validity of our
education and functional areas among the survey measures was strong, as recommended by
participants. Therefore, it was necessary to Fornell and Larcker [27]. Average variance
control for both gender and education in these extracted for all constructs exceeded 0.50. The

Table 3. lntercorrelations among the study constructs

Variables 1 2 3 4 5 6 7
1. Computer experience 1.00
2. Organizational support 0.08 0.60
3. Self-efficacy 0.33 0.11 0.58
4. Computer anxiety -0.26 -0.07 -0.14 0.61
5, Perceived ease of use 0.30 0.14 0.38 0.88 0.79
6. Perceived usefulness 0.22 0.20 0.22 -0.15 0.52 0.86
7. System usage 0.33 0.11 0.15 -0.14 0,27 0.45 0.89
Note: all correlations are significant at 0.05 lower. The diagonals represent the average variance
extracted.
598 Igbaria, livari--Effects of Self-efficacy on Computer Usage
Table 4. Prediction of self-efficacy and Computer anxiety
Self-efficacy Computer anxiety
Variables Direct Indirect Total
Computer experience 0.33* -0.24* -0.02 -0.26*
Organizational support 0.07* - 0.02 0.00 0.02
Self-efficacy - 0.7* - 0.07*
R2 0.12" 0.07*
*P _< 0.01.

item reliabilities of all the constructs (not anxiety is negatively correlated with usage,
presented here) were also significant. perceived usefulness and perceived ease of use
The results of the test conducted to assess (r = - 0.14, - 0 . 1 5 , P < 0.01 and - 0 . 3 3 ,
discriminant validity of the measures revealed no P < 0.001, respectively). Finally, perceived ease
violation of the criteria for discriminant validity. of use is strongly correlated with perceived
The intercorrelations among the items within the usefulness and usage ( r = 0 . 5 2 and 0.27,
construct were found to be stronger than between P < 0.001, respectively).
items from other constructs (not shown here).
Tests o f the structural model
Further, Table 3 presents the intercorrelations
among the study variables. In all of the 21 entries The effects of the external factors on
examined, the squared correlations, representing self-efficacy were examined. Data in Table 4
the shared variance among variables, were found show that, consistent with Hypotheses 1a and 2a,
not to exceed the average variance explained. both antecedent variables explained 12% of the
This suggests that our constructs are distinct and variance and had significant direct effects on
unidimensional. Further, confirmatory factor self-efficacy. The strongest effect is noted for
analysis was used to confirm the results of the computer experience (7 = 0.33, P < 0.001). A
measurement model. Results of the confirmatory smaller but significant direct effect was observed
factor analysis using the probability levels for the for organizational support (7 = 0.07, P < 0.01).
Z2 test confirmed the existence of the seven Consistent with Hypothesis 3a, Table 4 also
constructs/factors. The resulting factor pattern shows that self-efficacy had a positive direct effect
was consistent with the seven factors derived in on computer anxiety ((/~ = - 0.07, P < 0.001).
PLS. In summary, the convergent and discrimi- Computer experience, as expected (Hypothesis
nant validity of our instruments was satisfactory. lb), had a strong direct effect on computer
Intercorrelations in Table 3 also show that anxiety (7 = - 0 . 2 4 , P < 0.001). On the other
self-efficacy is positively correlated with com- hand, support had no significant effect on
puter anxiety, perceived ease of use, perceived computer anxiety (Hypothesis 2b).
usefulness and usage (r = - 0 . 1 4 , 0.38, 0.22, Table 5 shows that self-efficacy, consistent
P < 0.001, and 0.14, P < 0.01, respectively). with Hypothesis 3b, had a strong positive direct
Moreover, experience was found to be strongly effect on perceived ease of use (/3 = 0.29,
correlated with self-efficacy (r = 0.33, P < 0.001). It also shows that computer anxiety,
P < 0.001). Organizational support is signifi- consistent with Hypothesis 3b, had a strong
cantly correlated with self-efficacy (r = 0.11, negative direct effect on perceived ease of use
P < 0.01). While perceived usefulness, perceived (/~ = - 0.26, P < 0.001). Additionally, consist-
ease of use and experience strongly are correlated ent with Hypothesis 1c, computer experience had
with system usage (r = 0.45, 0.27, and 0.33, a strong direct effect on perceived ease of use
P < 0 . 0 0 1 , respectively). Further, computer (7 = 0.14, P < 0.001). Organizational support,

Table 5. Prediction of perceived ease of use and perceived usefulness

Perceived ease of use Perceived usefulness
Variables Direct Indirect Total Direct Indirect Total
Computer experience 0.14" 0.16 0.30* 0.06* 0.16 0.22*
Organizational support 0.09* 0.03 0.12" 0.14" 0.06 0.20*
Self-efficacy 0,29* 0.02 0.31" 0.01 0.16 0.17"
Computer anxiety -0.26* -0.26* -0.04 -0.13 -0.17"
Perceived ease of use 0.49* 0.49*
R" 0.26* 0.30*
*P < 0.01.
 Omega, 1Iol. 23, No. 6 599

Table 6. Prediction of system usage

significant. These results suggest that perceived
Variables Direct Indirect Total
ease of use plays a very important role in
Computer experience 0.24* 0.11 0.35*
Organizational support 0.01 0.09 0.10"
mediating the relationships between experience,
Self-efficacy 0.03 0.08 0.11 * anxiety and self-efficacy and perceived useful-
Computer anxiety - 0.02 - 0.07 - 0.09*
Perceived ease of use 0.01 0.20 0.21"
ness.
Perceived usefulness 0.41 * 0.41 * Consistent with Hypothesis 6, perceived
R~ 0.26*
usefulness had a strong direct effect on system
*P < 0.01.
usage ( f l = 0.41, P_<0.001) (Table 6). The
results showing the direct and indirect effects of
consistent with Hypothesis 2c, had a significant external factors on computer usage are also
but not a strong direct effect on perceived ease of presented in Table 6. Contrary to Hypothesis le,
use (7 = 0.09, P < 0.001). These variables, computer experience had a strong positive direct
combined, explained 26% of the variance of ease effect on usage (7=0.31, P_<0.001) and,
of use. Note that experience had a strong indirect consistent with Hypothesis 2e and 4c, organiz-
effect on perceived ease of use, through ational support and computer anxiety did not
self-efficacy and computer anxiety. have significant direct effects on usage, although
The results reported in Table 5 also show that the total effects were significant. This indicates
the model variables explained a significant that self-efficacy, computer anxiety, perceived
variation in perceived usefulness (R2= 0.30, ease of use and usefulness partially mediated the
P < 0.001). Consistent with Hypothesis 5a, effect of experience and organizational support
perceived ease of use had a very strong direct on self-reported usage behavior. Inconsistent
effect on perceived usefulness (fl = 0.49, with Hypothesis 3d self-efficacy had no direct
P_< 0.001). Additionally, organizational sup- effect on usage, but it had a strong indirect effect
port, consistent with Hypothesis 2d, had a on usage, mainly through perceived ease of use
significant direct effect on perceived usefulness and usefulness. Contrary to Hypothesis 5b
(7 = 0.14, P < 0.001). In accordance to Hypoth- perceived ease of use did not have a direct effect
esis ld computer experience had a significant on usage, but a very strong indirect effect through
direct effect on perceived usefulness (~ = 0.06, perceived usefulness. Finally, Table 6 shows that
P _< 0.001). While computer anxiety and self- 26% of the variance (P < 0.001) of system usage
efficacy, contrary to Hypotheses 3c and 4b, had was explained.
insignificant direct effects on perceived useful- In summary, self-efficacy had direct effects on
ness, they had strong indirect effects through computer anxiety and perceived ease of use.
perceived ease of use. Also, the total effects are Also, perceived usefulness had the strongest

Table 7. Summary of the results

Hypothesis la: Computer experience will have a positive direct effect on self-efficacy (S)
Hypothesis lb: Computer experience will have a negative direct effect on computer anxiety (S)
Hypothesis lc: Computer experience will have a positive direct effect on perceived ease of use (S)
Hypothesis ld: Computer experience will have a positive direct effect on perceived usefulness (S)
Hypothesis le: Computer experience will not have a significant direct effect on computer usage but the total effect wilt be positive and
significant (NS)
Hypothesis 2a: Organizational support will have a positive direct effect on self-efficacy (S)
Hypothesis 2b: Organizational support will have a negative direct effect on computer anxiety (NS)
Hypothesis 2c: Organizational support will have a positive direct effect on ease of use (S)
Hypothesis 2d: Organizational support will have a positive direct effect on perceived usefulness (S)
Hypothesis 2e: Organizational support will not have a significant direct effect on computer usage but the total effect will be positive
and significant (S)
Hypothesis 3a: Individual's self-efficacy will have a negative direct effect on his/her computer anxiety (S)
Hypothesis 3b: Individual's self-efficacy will have a positive direct effect on his/her perceived ease of use (S)
Hypothesis 3c: Individual's self-efficacy will have a positive direct effect on his/her perceived usefulness (NS)
Hypothesis 3d: Individual's self-efficacy has a positive direct effect on his/her use of computer technology (NS)
Hypothesis 4a: Computer anxiety will have a negative direct effect on perceived ease of use (S)
Hypothesis 4b: Computer anxiety will have a negative direct effect on perceived usefulness (NS.
Hypothesis 4c: Computer anxiety will not have a significant direct effect on computer usage but the total effect will be negative and
significant (S)
Hypothesis 5a: Perceived ease of use will have a positive direct effect on perceived usefulness (S)
Hypothesis 5b: Perceived ease of use will have a positive direct effect on system usage (NS)
Hypothesis 6: Perceived usefulness will be positively related to usage of computer technology (S)
S, supported; NS, not supported.

OME 23/6~B
600 Igbaria, livari--Effects of Self-efficacy on Computer Usage

direct effect on system usage. Some of the The results generally confirm earlier research
external factors had direct, indirect and total on TAM. They indicate the high internal
effects on system usage. Specifically, we observed consistency of the measures of perceived ease of
that computer experience had strong direct use and perceived usefulness and their satisfac-
effects on self-efficacy, computer anxiety, per- tory convergent and discriminant validity.
ceived ease of use and usage and an additional Consistent with the research in North America,
indirect effect on perceived usefulness. Computer and in accordance to Davis [19] and Davis et al.
anxiety played an important role in directly [20], perceived usefulness was found to have a
affecting perceived ease of use, and had stronger direct effect on usage than perceived
significant total effects on perceived usefulness ease of use. The result showing a strong direct
and usage. Organizational support also had effect of perceived ease of use on perceived
strong direct effects on self-efficacy, perceived usefulness was consistent with the research in
ease of use and perceived usefulness, and a North America, as was the lack of an effect of
significant total effect on usage. We also found perceived ease of use on usage. These results
that perceived ease of use had a strong direct strengthen the external validity of T A M in other
effect on perceived usefulness. Table 7 sum- countries.
marizes the results of our analyses. Our paper also shed some light on partially
inconsistent earlier results based on TAM [1]. IS
research suggests that individuals will use
DISCUSSION AND CONCLUSIONS
computing technology if they believe it will have
The purpose of this study was to explore fully positive outcomes (perceived usefulness). How-
the role of self-efficacy in computing behavior. It ever, this study based on SCT, suggests that we
examined the determinants and the outcomes of need to understand both individuals' expec-
self-efficacy in a computing behavior context tations of their capabilities and beliefs about
among computer users in Finland. The concep- outcomes in examining the use of computer
tual model and related hypotheses are depicted in technology. Beliefs about outcomes may not be
Fig. 1. All the identified variables turned out to sufficient to affect behaviors if individuals doubt
have a significant total effect on usage. They their capabilities to successfully use computer
supported most of the hypotheses (cf. Table 7 technology [15].
shows that 14 of the 20 hypotheses and Moreover, consistent with SCT, computer
sub-hypotheses were supported). The results of experience and support affect self-efficacy. The
this research provide support for the SCT results also support Bandura's [6-8] conjecture of
perspective on computing behavior. Self-efficacy experience as the most influential determinant of
was found to play an important role in shaping self-efficacy. This suggests that as users evaluate
individuals' beliefs and behaviors. The respon- their experiences with the system, perceptions of
dents in this study with high self-efficacy used computer self-efficacy and outcome expectations
computers more, derived more from their use of change. The above findings nevertheless empha-
computers and leveraged management support size the importance of external variables
better. Additionally, perceived usefulness (out- (experience and organizational support) more
come expectations) was found to have a than the original TAM which assumes the
significant impact on computer usage. Self- influence of external factors to be channeled
efficacy had a significant positive direct effect on through ease of use and perceived usefulness.
computer anxiety and perceived ease of use; the Contrary to TAM, computer experience was
total effect on perceived usefulness and usage was found to have a strong direct effect on usage, its
also significant. The results suggest that total effect being the second highest after
self-efficacy, at least when measured as in this perceived usefulness. The indirect effect, chan-
study, indirectly affects computer usage through neled through self-efficacy, anxiety and perceived
the beliefs concerning perceived ease of use and ease of use, is quite easy to understand. The high
usefulness. The results supported the direct effect direct effect may partly be explained by the fact
on perceived usefulness, while that on usage was that experience was measured in this study as the
not significant. In accordance with Hypothesis 6, number of computer experiences with different
perceived usefulness was found to have a applications and development processes. How-
significant effect on usage. ever, the detailed analysis of usage patterns of
 Omega, Vol. 23, No. 6 601

software packages revealed that this feedback benefits. Therefore, to promote the effective use
from usage to experience mainly concerned the and acceptance of computer technology in the
first two items of computer experience, with the long term, organizations need to positively
use of financial modeling and 3rd and 4th influence user perceptions and beliefs regarding
generation languages being low. the benefits and the usefulness and eventually the
The significance of experience has also obvious usage of computer technology in the workplace.
practical implications. Our results indicate that Management can influence users' beliefs by
perceived usefulness and perceived ease of use are encouraging, showing an interest in, and being
not enough for usage; users must have sufficient aware of the problems encountered by the users.
experience to use the system. The means of Such efforts to strengthen organizational sup-
increasing computer experience are vital for port should yield dividends in the form of
usage. Increased training and educational increased self-efficacy, more favorable beliefs,
programs may foster a feeling of self efficacy, that and eventually increased computer usage.
is+ the belief that one can develop the experience This paper also examined the impact of
necessary to effectively use computers and computer anxiety on perceived ease of use,
strengthen confidence in one's ability to master perceived usefulness and usage. A strong
and use them in one's work. Such training and negative effect was also found between computer
educational programs might emphasize the user anxiety and perceived ease of use. Computer
friendliness of currently available microcom- anxiety also had significant total effects on
puters, and the availability of easy to use perceived usefulness and usage. These effects are
software packages which require little or no mainly indirect through perceived ease of use.
knowledge of programming. In view of the Since this form of anxiety is not a permanent,
primacy of perceived usefulness, it is vital, unchanging characteristic of the individual it can
however, that educational programs emphasize be altered through adequate training offered at
the application of technology to actual work the user's level of ability and confidence. By
situations. carefully selecting software or situations, individ-
The findings from this study corroborate the uals who experience anxiety can encounter
importance of organizational support for positive reinforcement and control over their
computer usage. This research suggests that computer interactions. Efforts to reduce com-
support influences behavior indirectly, through puter anxiety could be instrumental in improving
its influence on self-efficacy and beliefs. Our managers' perceptions and behaviors about
study shows that higher organizational support microcomputers and thereby increasing the
resulted in higher judgments of self-efficacy on likelihood of their using them.
the part of the individuals because they would The results confirming TAM give considerable
have more resources to help them become more additional credence to the external validity of
proficient. Since support can be considered an TAM. There is, however, need for further
indication of organizational norms regarding research on the above model. The present study
use, it positively influences outcome expectations was based on data from one country. Hofstede's
(perceived usefulness). Our results confirm the analysis confirms that Finland is culturally a
importance of support in affecting perceived ease Scandinavian country with a relatively low
of use and perceived usefulness in addition to power distance, relatively high individualism,
self-efficacy. Organizational support is influen- low masculinity, and somewhat higher uncer-
tial in promoting the usage of the system by tainty avoidance than Denmark, Norway and
offering substantial educational programs which Sweden [48]. Since Finland is more feminine and
increase the sense of self-efficacy and perceptions a slightly more collective society, perceived
of ease of use, providing a wider selection of usefulness may not be the dominant factor
different types of software tools potentially affecting usage. An individual's abilities and
useful in users' jobs supporting a wider variety of experiences as well as organizational support are
business tasks, and encouraging users to use the likely to play major roles in affecting usage.
system. In addition, managers or key personnel The present study tested TAM in the case of
who strongly influence the immediate environ- one broad example of computer technology. In
ment are most likely to have a great impact on the view of the variety of modern computer
user's belief in the system's usefulness and technologies, we invite further research applying
602 lgbaria, Iivari--Effects of Self-efficacy on Computer Usage

the model to different technologies. The use of the model by taking into account more
self-report scales to measure the study variables organizational determinants of usage such as
suggest the possibility that common method organizational position, social pressure and task
variance may account for some of the results characteristics.
obtained. In order to pursue further investigation In addition to usage, the present study was able
of the conceptual model, it would be appropriate to explain a significant amount of variance in
to develop more direct and objective measures perceived usefulness (30%) and perceived ease of
for the use of microcomputer technology. use (26%). The results help one to understand
Alternatives to self-report data sources would better what affects computer usage directly and
certainly reduce the likelihood of obtaining indirectly. Together with results of previous
spurious relationships. Additionally, the use of research they have practical significance. As
the Hill et al. [45] measure has some recognized pointed out by Davis et al. [20], TAM has many
weaknesses that serve as limitations to our study. practical implications concerning the desired
These items are aimed at measuring self-efficacy quality of systems and the systems development
strength (i.e. the degree of confidence) and not process. The finding that perceived usefulness
self-efficacy magnitude (i.e. whether individuals had a strong positive effect on usage suggests that
can perform at specific levels on a specific task) individuals are likely to accept a new technology
[57]. They state that the most common method of because of the functions it performs for them.
measuring self-efficacy is to ask subjects to rate Therefore, IS managers should also emphasize
their self-efficacy strength. They also found that the functionality of the system and should convey
very few researchers have used a combination of to end-users in a realistic way what the proposed
the magnitude and strength indexes. Since the system will consist of and what functions will be
self-efficacy magnitude and strength measures available to them. Additionally, the results of this
had weaker predictive validities than the paper confirms the findings of prior research
self-efficacy composites, researchers need to [19, 20, 69] that perceived usefulness is more
consider using the self-efficacy composite influential than perceived ease of use in
measures in future empirical settings, Addition- determining usage. This suggests that users may
ally, Compeau and Higgins [15] suggest that be willing to grapple with difficult interfaces in
research should focus on tasks that required the order to access functionality which facilitates the
use of computers, instead of using computers as performance of a task. The primacy of perceived
an end in itself. The non-random selection of usefulness raises several questions and concerns
participants in the study also may have brought about recent trends in systems development
in some bias, even though we found no significant approaches. The evolution of user interfaces
differences between users in different industries, options [65], for example, has increased the
organizational levels, and functional areas. interest in user interface issues in systems
The model explained 26% of the variance in development [18, 42, 74]. In many applications
usage. Generally speaking, this figure is higher half or more of the code is devoted to the user
than the variances explained in the five interface [39, 66]. Davis [19] evaluated, however,
applications experienced by Adams et al. [1] that the human factors tradition, which has
which ranged from 4 to 35%, but considerably dominated research on user interfaces, has
lower than the variances of the six systems tended to overemphasize perceived ease of use
explained by Davis [19], Davis et al. [20] and and to overlook usefulness. One may also
Mathieson [69] which ranged from 31 to 71%. wonder whether prototyping, that is generally
The relatively low figure of the present study may advocated in the above user interface-oriented
partly be explained by its focus on microcom- approaches, may additionally bias the situation.
puters in general. The fact that a user may Boehm et al. [9], for example, found that
perceive variation in the ease of use and prototyped products rated somewhat lower on
usefulness between different applications can be functionality, but higher on perceived ease of use
expected to weaken the association between and ease of learning. In order to compensate for
perceived ease of use, perceived usefulness and these potential biases, its is important to focus on
overall usage. The figure, 26%, nevertheless the functionality of the system throughout the
implies that there are other factors involved that development process. The measure for perceived
influence usage. A further step might be to extend usefulness may be used in early evaluations of the
 Omega, Vol. 23, No. 6 603

functionality of the system. Sophisticated user 21. DeLone WH (1988) Determinants of success
for computer usage in small business. M I S Q. 12,
interfaces may also help a user to link his/her 51-61.
requirements to the functionality of the tools and 22. Dickerson MD and Gentry JW (1983) Characteristics
the concepts upon which such functionality may of adopters and non-adopters of home computers.
J. Consumer Res. 10, 225-235.
depend. 23. Ellen PS, Bearden WO and Sharma S. (1991) Resistance
to technological innovations: an examination of the
role of self-efficacy and performance satisfaction. J.
Acad. Mktg Sci. 19, 297-307.
REFERENCES 24. Fishbein M and Ajzen I (1975) Belief, Attitude,
Intentions and Behavior: An Introduction to Theory and
1. Adams DA, Nelson RR and Todd PA (1992) Perceived Research. Addison-Wesley, Boston, Mass.
usefulness, ease of use, and usage of information 25. Fornell CR (Ed.) (1982) A Second Generation of
technology: a replication. M I S Q. 16, 227-247. Multivariate Analysis, Volume I and H: Methods.
2. Aharoni Y and Burton RM (1994) Is management Praeger Special Studies, New York.
science international: in search of universal rules. Mgmt 26. Fornell CR and Bookstein FL (1982) Two structural
Sci. 40, 1-3. equation models: LISREL and PLS applied to
3. Ajzen I (1985) From intentions to actions: a theory of consumer exit-voice theory. J. Mktg Res. 19,
planned behavior. In Action Control: From Cognition to 440-452.
Behavior (Edited by Kuhl J and Beckmann J), 27. Fornell CR and Larcker DF (198 I) Structural equation
pp. 11 39. Springer, New York. models with unobservable variables and measurement
4. Ajzen I (1989) Attitude structure and behavior. In error. J. Mktg Res. 18, 39-50.
Attitudes Structure and Function (Edited by Pratkanis 28. Fornell CR, Tellis GL and Zinkhan GM (1982) Validity
AR, Breckler SJ and Greenwald AG), pp. 241-274. assessment: a structural equation approach using
Erlbaum, Hillsdale, N.J. partial least squares. In AMA Educators' Proceedings,
5. Alwin DE and Hauser RM (1975) Decomposition of pp. 405-409. American Marketing Association,
effects in path analysis. Am. Sociol. Rev. 40, 37-47. Chicago.
6. Bandura A (1977) Self-efficacy: toward a unifying 29. Franz CR and Robey F (1986) Organizational context,
theory of behavioral change. Psychol. Rev. 84, 191-215. user involvement, and the usefulness of information
7. Bandura A (1982) Self-efficacy mechanism in human systems. Dec. Sci. 17, 329-356.
agency. Am. Psychol. 37, 122-147. 30. Fuerst W and Cheney P (1982) Factors affecting the
8. Bandura A (1986) The explanatory and predictive scope perceived utilization of computer-based decision
of self-efficacy theory. J. Soc. Clin. Psychol. 4, 359-373. support systems in the oil industry. Dec. Sci. 13,
9. Boehm BW, Gray TL and Seewald T (1984) 554-569.
Prototyping versus specifying: a multiproject exper- 31. Fortune (1993) The new computer revolution. Cover
iment. IEEE Trans. Engng Mgmt SE-10, 290-303. Story, 14 June, 56-81.
10. Blair E and Burton S (1987) Cognitive process used by 32. Fortune (1993) Making high tech work for you. Special
survey respondents to answer behavioral frequency Issue: 1994 Information Technology Guide, Autumn.
questions. J. Consumer Res. 14, 280-288. 33. GilroyFDandDesaiHB(1986)Computeranxiety: sex,
11. Bowen W (1986) The puny payoff from office race and age. Int. J. Man-Mach. Stud. 25, 711-719.
computers. Fortune 26 May, 20-24. 34. Gist ME (1987) Self-efficacy: implications for organiz-
12. Burkhardt M and Brass DJ (1990) Changing patterns ational behavioral and human resource management.
or patterns of change: the effects of a change in Acad. Mgmt Rev. 12, 472-485.
technology on social network structure and power. 35. Gist ME (1989) The influence of training method on
Admin. Sei. Q. 35, 104-127 self-efficacy and idea generation among managers. Pets.
13. BusinessWeek (1993) The technology payoff: a special Psychol. 42, 787-805.
report. 14 June, 56-68. 36. Gist ME (1992) Self-efficacy: a theoretical analysis of its
14. Cheney P and Dickson GB (1982) Organizational determinants and malleability. Acad. Mgmt Rev. 17,
characteristics and information systems success: an 183-211.
extrapolation investigation. Acad. Mgmt J. 25, 37. Gist ME, Rosen B and Schwoerer C (1988) The
170-184. influence of training method and trainee age on the
15. Compeau D and Higgins C (1991) A social cognitive acquisition of computer skills. Pers. Psychol. 41,
theory perspective on individual reactions to computing 255-265.
technology. In Proceedings o f the Twelfth International 38. Gist ME, Schwoerer C and Rosen B (1989) Effects of
Conference on Information Systems (Edited by Gross JI, alternative training methods on self-efficacy and
Benbasat I, DeSanctis G and Beath CM), New York, performance in computer software training. J. Appl.
pp. 187-198. Psychol. 74, 884-891.
16. Compeau D and Higgins C (1996) Application of social 39. Gould JD, Boies SJ and Lewis C (1991) Making usable,
cognitive theory to training for computer skills. Inform. useful, productivity-enhancing computer applications.
Syst. Res. In press. Commun. A C M 34, 74-85
17. Curley KF (1984) Are there any real benefits from office 40. Grant RA (1989) Building and testing a causal models
automation. Bus. Horiz. 4, July-August, 37-42. of an information technology's impact. Proceedings of
18. Curtis B and Hefley B (1992) Defining a place for the Tenth International Conference on Information
interface engineering. IEEE Software March, 84-86. Systems, Boston, Mass., pp. 173-184.
19. Davis FD (1989) Perceived usefulness, perceived ease of 41. Hair JF, Anderson RE, Tatham RL and Black, W.C.
use, and user acceptance of information technology. (1992) Multivariate Data Analysis with Readings, 3rd
M I S Q. 13, 983-1003. edition. Macmillan, New York.
20. Davis FD, Bagozzi RP and Warshaw PR (1989) User 42. Hartson HR and Boehm-Davis D (1993) User interface
acceptance of computer technology: a comparison of development processes and methodologies. Behav.
two theoretical models. Mgmt Sci. 35, 982-1003. Inform. Technol. 12, 98-114.
604 Igbaria, livari--Effects of Self-efficacy on Computer Usage

43. Herzberg F (1966) Work and the Nature of Man. World, theory of planned behavior. Inform. Syst. Res. 2,
Cleveland, Ohio. 173-191.
44. Hill T, Smith ND and Mann MF (1986) 70. McGregor D (1960) The Human Side of Enterprise,
Communicating innovations: convincing computer McGraw-Hill, New York.
phobics to adopt innovative technologies. In Advances 71. Miura IT (1987) The relationship of computer
in Consumer Research (Edited by Lutz R J), Vol. 3, self-efficacy expectations to computer interest and
pp. 419-422. Association for Consumer Research, course enrollment in college. Sex Roles 16, 303-311.
Provo, Utah. 72. Moore GC and Benbasat I (1991) Development of an
45. Hill T, Smith ND and Mann MF (1987) Role of efficacy instrument to measure the perceptions of adopting
expectations in predicting the decision to use advanced information technology innovation. Inform. Syst. Res.
technologies: the case of computers. J. Appl. Psvchol. 2, 192-222.
72, 307-313. 73. Morrow PC, Prell ER and McElroy JC (1986)
46. Hirschman EC (1980) Innovativeness, novelty seeking, Attitudinal and behavioral correlates of computer
and consumer creativity. J. Consumer Res. 7, 283-295. anxiety. Psychol. Rep. 59, 1199-1204.
47. Hofstede G (1980) Culture's Consequences, Abridged 74. Nielsen J (1992) The usability engineering life cycle.
edition, Sage, Beverly Hills. Computer 25, No. 3, 12-22.
48. Hofstede G (1991) Cultures and Organizations: 75. Nietzel MT, Berstein DA and Russel RL (1988)
Software for the Mind. McGraw-Hill, London. Assessment of anxiety and fear. In Behavioral
49. Hofstede G (1994) Management scientists are human. Assessment, A Practical Handbook (Edited by Bellack
Mgmt Sci. 40, 4-13. AS and Herson M), 3rd edition. Pergamon Press,
50. Igbaria M (1990) End-user computing effectiveness: a Toronto.
structural equation model. Omega 18, 637-652. 76. Nunnally JC (1978) Psychometric Theory, 2nd edition.
51. Igbaria M (1993) User acceptance of microcomputer McGraw-Hill, New York.
technology: an empirical test. Omega 21, 73-90. 77. OECD (1992) Information Technology Outlook 1992.
52. Igbaria M and Chakrabarti A (1990) Computer anxiety OECD, Paris.
and attitudes towards microcomputer use. Behav. 78. Raymond L (1985) Organizational characteristics and
Inform. Technol. 9, 229-241. MIS success in the context of the small business. MIS
53. Igbaria M, Pavri F and Huff S (1989) Microcomputer Q. 9, No. 1, 37-52.
application: An empirical look at usage. Inform. Mgmt 79. Raub AC (1981) Correlates of computer anxiety in
16, 187-196. college students. Unpublished Ph.D. dissertation,
54. Iivari J. (1986) An innovation research perspective on University of Pennsylvania, Philadelphia, Pa.
information system implementation. Int. J. Inform. 80. Robey D (1979) User attitudes and management
Mgmt 6, 123-144. information systems use. Acad. Mgmt J. 22, 527-538.
55. LaBay DG and Kinnear TC (1981) Exploring the 81. Robey D and Zeller RF (1978) Factors affecting the
consumer decision process in the adoption of solar success and failure of an information system for product
energy systems. J. Consumer Res. 8, 271-278. quality. Interfaces 8, 70-75.
56. Lachman R, Nedd A and Hinings B (1994) Analyzing 82. Rogers EM (1983) Diffusion of Innovation. Free Press,
cross-national management and organizations: a New York.
theoretical framework. Mgmt Sci. 40, 40-55. 83. Ross DR (1975) Direct, indirect, and spurious effects:
57. Lee C and Bobko P (1994) Self-efficacy beliefs comments on causal analysis of interorganizational
comparison of five measures. J. Appl. Psychol. 79, relations. Admin. Sci. Q. 20, 295-297.
364-369. 84. Rosenzweig PM (1994) When can management science
58. Lee DS (1986) Usage patterns and sources of assistance research be generalized internationally. Mgmt Sci, 40,
to. personal computer Users. MIS Q. 10, 313-325. 28-39.
59. Leonard-Barton D and Kraus WA (1985) Implement- 85. Straub DW (1994) The effect of culture on IT diffusion:
ing new technology. Harvard Bus. Rev. 63, November- E-mail and FAX in Japan and the U.S. Inform. Syst.
December, 102-110. Res. 5, 23-47.
60. Liebenau J and Smithson S (1991) Editorial. Eur. 86. Stumpf SA, Brief AP and Hartman K (1987)
J. Inform. Syst. 1, 1-2. Self-efficacy expectations and coping with career-re-
61. Lohmoller JB (1981) LVPLS 1.6: Latent Variables Path lated events. J. Vocat. Behav. 31, 91-108.
Analysis with Partial Least Squares Estimation. 87. Sullivan-Trainor ML (1991) Is technology worth it?
University of Federal Armed Forces, Munich. Computerworld 13 May, 81-82, 86.
62. Lohmoller JB(1989) Latent Variable Path Modeling 88. Swanson EB (1988) Information System Implemen-
with Partial Least Squares. Physica, Heidelberg. tation. Irwin, Homewood, Ill.
63. Lucas HC (1978) Empirical evidence for a descriptive 89. Talouselama(1992) 500suurintayritystaSuomessa(The
model of implementation. MIS Q. 2, No. 2, 27-41. top 500 companies in Finland), No. 20, pp. 38-57.
64. Maglitta J (1991) It's reality time. Computerworld 29 90. Talouselama (1993 ) 500 suurinta yritysta Suomessa (The
April, 81-83. top 500 companies in Finland), No. 21, 36-55.
65. Mandelkern D (1993) Graphical user interfaces: the 91. Thompson RL, Higgins CA and Howell JM (1991)
next generation, introduction. Commun. ACM 36, No. Personal computing: toward a conceptual model of
4, 36-39. utilization. MIS Q. 15, 125-143.
66. Markus A and van Dam A (1991) User-interface 92. Tornatzky LG and Klein KJ (1982) Innovation
development for the nineties. Computer 24, No. 9, characteristics and innovation adoption-implemen-
49-57. tation: a meta-analysis of findings. IEEE Trans. Engng
67. Martocchio JJ and Webster J (1992) Effects of feedback Mgmt EM-29, 28-45.
and cognitive playfulness on performance in microcom- 93. Trevino LK and Webster J (1992) Flow in computer-
puter software training. Pers. Psychol. 45, 553-578. mediated communication: electronic mail and voice
68. Maslow AH (1970) Motivation and Personality. Harper mail evaluation and impacts. Commun. Res. 19,
& Row, New York. 539-574.
69. Mathieson K (1991) Predicting user intentions: 94. Triandis HC (1971) Attitude and Attitude Change.
comparing the technology acceptance model with the Wiley, New York.
 Omega, Vol. 23, No. 6 605

95. Triandis HC (1980) Values, attitudes, and interpersonal tion-H (Edited by Joreskog KG and Wold H.),
behavior. In Nebraska Symposium on Motivation, pp. 1-54. North-Holland, Amsterdam.
Beliefs, Attitudes, and Values, pp. 195-259. University 101. Wood RE and Bandura A (1989) Impact of conceptions
of Nebraska Press, Lincoln. of ability on self-regulatory mechanisms and complex
96. Tukey JW (1958) Bias and confidence in not-quite large decision making. J. Pets. Soc. Psychol. 56, 407-415.
samples. Annls Math. Statist. 29, 614. 102. Young TR (1984) The lonely micro. Datamation 30,
97. Webster J (1989) Playfulness and computers at work. 100-114.
Unpublished Ph.D. dissertation, New York University, 103. Zaltman G, Duncan R and Holbek J (1977) Innovations
New York. and Organizations. Wiley, New York.
98. Webster J and Martocchio JJ (1992) Microcomputer 104. Zmud RW (1979) Individual differences and MIS
playfulness: development of a measure with workplace success: a review of the empirical literature. Mgmt Sci.
implications. M1S Q. 16, 201-224. 25, 966-979.
99. Wildt AR, Lambert ZV and Durand RM (1982)
Applying the jackknife statistics in testing and
interpreting canonical weights, loadings and cross-load- ADDRESS FOR CORRESPONDENCE: Professor MagM Igbaria,
ings. J. Mktg Res. 19, 9% 107. Programs in Information Science, Claremont Graduate
100. Wold H (1982) Soft modeling the basic design and School, 130 East Ninth Street, Claremont, CA
some extensions. In Systems Under Indirect Observa- 91711-6190, USA.