Sie sind auf Seite 1von 57
Testing the Determinants of Microcomputer Usage via a Structural Equation Model MAGID IGBARIA, TOR GUIMARAES, AND GORDON B. DAVIS Maaib IcBaria is a Professor of MIS in the College of Business and Administration at Drexel University and a Vi 1g Professor of Decision Sciences at the University of Hawaii. He holds a B.A. in statistics and businéss administration and an M.A. in information systems and operations research from Hebrew University. He received his Ph.D. in computers and information systems from Tel Aviv University. Formerly, he lectured at Tel Aviv University, Hebrew University, and Ben-Gurion University in Israel, and acted as the administrative director of the Center of Management Informa- tion Systems (CEMIS) at Tel Aviv University. He has published articles on manage- . Ment of IS functions, cconomics of computers, computer performance evaluation, charging of computer services, compumetrical approaches in IS, IS carcers, interma- tional IS, and microcomputers in business in Applied Statistics, Communications of the ACM, Computers and Operations Research, Decision Sciences, Information ‘Systems Research, Informatign & Management, INFOR, International Journal of Man-Machine Studies, “Journal of Engineering and Technology Management, Omega, MIS Quarterly, and others. His current research interests focus on international information systems, management of information systems, carcer development of IS professionals, and end-user computing. Tor Guimaracs holds the J.E, Owen Chair of Excellence in Information Systems at Tennessee, Technological University. He has a Ph.D. in MIS from the University of Minnesota and an M.B.A. from California State University, Los Angeles. He was a professor and chairman of the MIS department at St. Cloud State University, and before that an assistant professor and director of the MIS Certificate Program at Case Westem Reserve University. He has spoken at numerous meetings sponsored by professional organizations including ACM, IEEE, ASM, DPMA, INFOMART, and Sales and Marketing Executives. He has consulted on several IS topics with many leading organizations, including TRW, American Greetings, AT&T, IBM, and the Department of Defense. He has published over sixty articles in leading journals such as Communications of the ACM, Decision Sciences, Information Systems Research, MIS Quarterly, Omega, Computers and Operations Research, Information & Man. agement, and Data Base. GorDON B. Davis is Honeywell Professor of Information Systems at the University of Minnesota Carlson School of Management where he was one of the founders of the MIS program. He received his Ph.D. in accounting from Stanford University. He has published several texbooks and journal articles, and has spoken at numerous profes- sional meetings nationally and internationally. Journal of Management Information Systems 1 Spring 1995, Vol 11, No.4, pp. 87=114 Copyright © 1995 ME. Sharpe, Inc, ABSTRACT: The, objective of this study was to develop and test an integrated concep- tual model of microcomputer usage. This study used the technology acceptance model and sought to extend it by investigating the impact of the extemal factors (i.c., individual, organizational, and system characteristics) on the user acceptance of microcomputer technology. The paper reports the results of a field study investigating the determinants of microcomputer usage. The analyses of the measurement model confirm the existence of two distinct constructs of (1) beliefs—perceived usefulness and perceived ease of use; (2) organizational support—management support and end-user computing (EUC) support; and (3) microcomputer usage—perceived usage and variety of use, The tested conceptual model confirms the effects of individual, organizational, and system characteristics on perceived ease of use and perceived usefulness. The model also confirms the influence of perceived ease of use on perceived usefulness, and the effects of perceived usefulness on perceived usage and variety of use. Results confirm several previously proposed notions, including the effects of individual, organizational, and system characteristics on ease of use and usefulness; the influence of ease of use on usefulness, and the effects of perceived usefulness on usage and variety of use, The results demonstrate the utility of investigating factors contributing to microcomputer usage and the external factors affecting endogenous variables such as system usefulness. The importance of EUC support and management support is corroborated, as well'as the need for designing mechanisms such as training programs and newsletters to improve user perceptions of microcomputers, KEY WORDS ANDPHRASES: microcomputer usage, perceived usefulness, questionnaire analysis, technology acceptance model. I ORGANIZATIONS ARE ACCELERATING THEIR INVESTMENTS IN VARIOUS computer applications such as word processors, spreadsheets, communications (mainly elec- tronic mail), graphics, and database management. The increasing demand for micro- computers has created a major shortage for PC makers and retailers [77]. Though microcomputers were viewed largely as a technological curiosity in the 1970s, the rapid growth of microcomputers in organizations represents a significant development in the field of MIS [9, 19, 35, 44, 48, 78]. Microcomputers have been hailed as a revolution that will change the nature of professional work (28, 35, 48]. Today, microcomputers make up an increasingly important component of the computing resources in most organizations (38, 44, 45]. The expanding role of microcomputers as components of individual, departmental, and corporatewide data processing and decision support systems is an important reason for gaining a better understanding of the factors affecting microcomputer usage. BusinessWeek [13] reported that computer tech- nology increases company productivity and moves it to a higher productivity growth path. A recent estimate in the United Kingdom predicted that the ratio of computer terminals or microcomputers (largest component) to office workers was already approaching an average of 1:1 [74], indicating that almost all knowledge workers will have their own microcomputer both to perform stand-alone tasks and to use network services. Inspiteof this heavy dependence on microcomputers, Thompson etal. [78] conclude that “we know relatively little about the forces that influence the utilization of personal computing” [p. 124]. They identify the need for greater attention to the usage of new technology, particularly the factors that cause individual resistance to computer usage, While system usage may not ensure improved user task performance or return on investment, it remains a critical variable for organizations to derive any benefit from new technology. Therefore, itis important to understand the factors. affecting computer usage. Our main objective in this paper is to develop a more complete understanding Of the factors affecting user acceptance of this technology. Specifically, we present the test results of an integrated model of the antecedents and consequents of user beliefs toward microcomputers, using multivariate analysis and a structural equation approach, A central focus of MIS implementation research has been to examine the factors affecting user acceptance of computer technology. Several models have been devel- oped to investigate the impact of these factors on the acceptance of computer technology. Among these factors are individual, organizational, and system charac- teristics [17, 29, 30, 41, 52, 75, 85]. Close examination of these studies and others, as reported by Davis et al. [17], reveals a number of problems that make it difficult to interpret and compare results across studies. The problems derive from differences in the conceptualization and operationalization of the measures, lack of. specificity of the different belief and attitude measures, and inadequate theoretical and psychometric Justifications, Therefore, more research, guided by theory development and empirical evidence, isneeded to examine the factors affecting the user acceptance of microcom- puter technology. This paper empirically assesses how well the general model predicts and explains usage of microcomputer technology. Two models have been used by MIS researchers as a theoretical foundation for research on the factors affecting the user acceptance of computer technology: Fishbein and Ajzen’s [22] theory of reasoned action (TRA) and the technology acceptance model (TAM) by Davis et al. [17]. Hubona and Cheney [40] discussed both models, their relationship, and implications in more detail. The TAM adapted the generic TRA model to the particular domain of user acceptance of computer technology, replacing the TRA’s attitudinal determinants, derived separately for each behavior, with a set of two variables (perceived ease of use and perceived usefulness). Both models were found to predict intentions and usage satisfactorily; the TAM, however, was found to beamuch simpler, easier to use, and more powerful model of the determinants of user acceptance of computer technology. In addition, the TAM attitudinal determinants outperformed the TRA’s much larger set of measures. Based on such background, this research uses the TAM to examine the factors affecting user acceptance of computer technology. Although Adams, Nelson, and Todd [1], Davis et al. [17], and Mathieson {55] provided insights into the acceptance of computer technology using the TAM, theirresearch focused only on the determinants of usage rather than the external factors affecting these determinants (¢.g., perceived usefulness and ease of use). Thus, the Present study extends previous research by investigating the network of multivariate relationships among external factors (e.g., individual, system, and organizational characteristics), perceived usefulness, and perceived ease of use, and system use. We developed and tested an integrated model of microcomputer usage, which ‘User Charncterlstcs: ‘User Teatning {User Computer Pxperlence ‘Qreanizational Support: UC Support ‘Management Support Figure 1. Microcomputer Usage: The Conceptual Model (Numbered arrows correspond to the propositions described in the text.) incorporates those variables found consistently to explain and predict the microcomputer technology. The analysis uses a structural equation modeling ap with partial least squares (PLS), PLS is a second-generation multivariate a1 technique used to estimate the parameters of causal models. PLS embraces a and empirical variables simultaneously, and recognizes the interplay of the: dimensions of theory development. The causal modeling technique, often | structural equation modeling, accommodates a priori knowledge derived from and/or previous empirical findings, and “because these methods can combine : as confront theory with empirical data, they offera potential for scientific explz that goes far beyond description and empirical association” [23]. ‘The study was designed to address four interrelated research questions ; managers: (1) What is the impact of the external factors on microcomputer usa} ‘What is the impact of perceived ease of use and perceived usefulness on micr puter usage? (3) What is the impact of perceived ease of use on perceived usefu (4) Do perceived ease of use and perceived usefulness mediate the effects external factors on microcomputer usage? The research described here used tt multivariate technique to test these four research questions, as presented in our | The Theoretical Framework FIGURE 1 PRESENTS THE MODEL OF COMPUTER USAGE of microcomputers exe in this study. The model is based on two models that predict and explain micr puter usage. The first model is the technology acceptance model (TAM), dev by Davisetal. [17] to predict people’s computer acceptance. Briefly, the TAMa the generic theory of reasoned action (TRA), developed by Fishbein and Ajze replacing the TRA’s attitudinal determinants, derived separately for each be (usage), with a set of two variables (perceived ease of use and perceived usefi employed in many computer technology acceptance contexts. Perceived ease and perceived usefulness are thought to be potentially important determing computer technology acceptance. Davis et al. proposed that external factors d affect perceived usefulness and perceived ease of use, and these, in turn, are important determinants of computer technology acceptance. They also proposed that perceived ease of use directly affects perceived system usefulness. ‘The second model is the theory of planned behavior (TPB), developed by Ajzen [2, 3]. The TPB was developed to predict behavior across many settings and was applied to computer acceptance by Mathieson [55, p. 185]. He proposed that behavior is affected by attitudes toward use, subjective norms (SN), and perceived behavioral control (PBC). The author describes in more detail that this model treats the perceived behavioral contro! variables (skills, opportunities, and resources needed to use the system) differently from the TAM model. Further, Mathieson [55, p. 185] reports that “Intention was predicted by attitude and PBC, but not by SN. This suggests that social pressure does not influence individual's decision to use a spreadsheet. Attitude has a slightly stronger effect on intention than PBC.” Mathieson [55] compares these two models and finds that both models are able to predict intentions and system usage satisfactorily. However, TAM has a slight empir- ical advantage; it is simpler and easier to apply, “but only supplies very general information on users’ opinions about a system. TPB provides more specific informa- tion that can better guide development” [p. 173]. Based on this background, this research integrates the TAM (mainly the effect of perceived usefulness and perceived ease of use on computer technology acceptance).and the TPB (mainly the effect of PBC on technology acceptance). Although Davis et al. [17] proposed that the external factors have only indirect effects orl computer acceptance, they did not test it empiri- cally. Further, Mathieson [55] provides insights into the acceptance of computer technology using both TAM and TPB. However, his research focused only on the determinants of usage rather than on the external factors affecting these determinants (€.g., perceived usefulness and ease of use). Thus, the present study seeks to extend previous research by bringing together TAM and TPB and investigating the effects of the external factors (¢.g., individual, system, and organizational characteristics) on Perceived usefulness, perceived ease of use, andcomputer technology acceptance. We developed an integrated model of microcomputer usage, incorporating those variables found consistently to explain and predict the use of microcomputer technology, particularly TAM and TPB, and using a structural equation modeling approach with Partial least squares (PLS). Figure 1 presents the model of microcomputer usage examined in this study. The model is developed based on the TAM [1, 17,70], the theory of planned behavior [55], and models on user acceptance of computer technology [15, 18, 21, 29,41, 46, 52, 78, 85]. A review of the relevant literature suggests that usage of microcomputers is affected directly and/or indirectly by: (1) external variables (¢.g., individual, system, and organizational characteristics); (2) perceived ease of use; and (3) perceived usefulness, The model includes five sets of variables: (1) individual characteristics: user training and computer experience (18, 44, 52, 55, 85]; (2) organizational charac- teristics: organizational support [15, 52, 78]; (3) system characteristics: system quality [30, 43, 53, 67]; (4) beliefs: perceived ease of use and perceived usefulness [16, 17, 41, 55, 75, 78); and (5) microcomputer usage. - Based on these five sets of variables, the model in figure 1 graphically depicts fi Propositions tested in this study. The propositions are grouped into two sets. Proposition Set 1: External Factors ‘User beliefs are determined by individual characteristics, organizational character tics, and system characteristics. All of the following propositions, related to the imp of the external factors on usage and beliefs, are expected to be true based on pr research, The TRA, as well as the TAM models, propose that external factors, such individual, system, and organizational characteristics, will affect behavior (i.e., usa; through their effects on beliefs. On the other hand, the theory of planned behav (TPB), discussed by Ajzen [2, 3], hypothesizes a direct effect of perceived behav control, among others, on intention. PBC deals with individual's perceptions of “ presence or absence of requisite resources and opportunities” [4, p. 457] necessary perform the behavior. It includes skills, opportunities, and resources needed to use | system. This relationship was empirically supported by Mathieson [55] in predict technology acceptance. Additional support for the direct effect of the external fact ‘on acceptance was also found by Thompson et al. [78] using a subset of the mo proposed by Triandis [79]. 1 Individual Characteristics Ja. User training will directly and indirectly affect usage through beliefs. 1b. Computer experience will directly and indirectly affect usage through belie Based on the theoretical model of Zmud [85], individual characteristics have be reported to play an important role in the eventual success of MIS. The acceptance computer technology depends on the technology itself and the level of skill or expert of the individual using the technology [57]. User training and experience, representi individual skills and expertise, were found to be related to user beliefs and usage. 7 impact of beliefs and usage has been well documented [41, 69, 85]. These studies others suggest that user training plays a very important role in influencing user beli toward the system and that training programs are likely to increase user confidence their ability to master and use computers in their work [31]. Recent research [18, 58] provides evidence that the type of microcomputer training will influence sub quent task performance and users’ belief that they can develop skills necessary to computers. In addition, opportunities to gain experience using computers and inf mation systems are thought to improve users’ beliefs about the system [41, 65]. U training and computer experience were also found to be positively related to syst usage (19, 30, 44, 46, 48, 69]. It was found that lack of training is a major reason the lack of MIS success. In addition, they found that prior computer experie promoted increased computer usage. Therefore, it is proposed that individual char teristics will directly and indirectly affect usage through beliefs. Specifically, Propose that computer experience and user training will be positively associated with beliefs (perceived ease of use and perceived usefulness) and usage. Organizational Characteristics 2. Organizational support will directly and indirectly affect usage through beliefs. Based on the TAM, Davis et al. [17] proposed that beliefs (perceived ease of use and perceived usefulness) were affected by organizational support. Organizational support has been emphasized by several researchers as a potential determinant ‘of system success [21, 30, 44, 52]. High levels of organizational support are thought to promote more favorable beliefs about the system among users as well as MIS staff [42, 52]. Organizational support was found to be associated with favorable beliefs and with greater system usage; furthermore, lack of organizational support is considered as a critical barrier to the effective utilization of computers (30, 41, 52]. Further, a subset of the model proposed by Triandis [80] was tested by Thompson et al. [78]. They Teported that facilitating conditions will have a direct effect on behavior. Thompson et al. [78] included support for users as one of the facilitating conditions that can influence utilization. Organizational support was found to be directly related to system usage [52]. Therefore, we propose that organizational support is Positively, related to beliefs about microcomputers and uset acceptance of the technology (i.e. usage). System Characteristics 3. The quality of the system will direcily and indirectly affect usage through beliefs. Based on the TAM, Davis etal. [17] proposed that system quality, asan external factor, has an impact on beliefs. These relationships are well documented [7, 8, 10, 20, 56]. Models of MIS success have also identified system quality as a relevant system characteristics to user perception and beliefs about the technology (30, 42, 52], which, in turn, affected system usage [67]. It is suggested that system quality plays a ‘prominent role as a determinant of system usage [30, 44, 52, 60, 76]. System quality may function asa source of information relevant to feelings of self-efficacy, judgments of mastery, and self-determination, and hence as determinants of perceived ease of use and usefulness. Systems with higher user-perceived quality are likely to be used more frequently than those of lower quality. Shneiderman [71] stressed the importance of a system’s user interface quality to the acceptance of new technology into organi- zations. Therefore, we propose that system characteristics are positively related to beliefs and user acceptance of microcomputer technology (i.c. usage). Proposition Set 2: Beliefs Usage of microcomputers is affected by user beliefs about the usefulness and system ease of use. All of the following usage Propositions are expected to be true b Previous research, Davis [16] found that beliefs (Perceived ease of use and perceived usefuln highly correlated with perceived usage. Adams et al. [1] and Davis et al. [1 found that both are correlated (only in the first study), and Swanson [75] fou the value dimension (similar to perceived usefulness) and the accessibility din (similar to perceived case of use) are correlated with. self-reported information ¢ (0.20 and 0.13, respectively). In addition, it was suggested that ease of use: play important role in influencing decisions to use or not use the system [12, 71]. F more, the self-efficacy paradigm suggests that Perceived ease of use and pe usefulness are determinants of user behavior (i.e. system usage) [6, 16]. Perceived Ease of Use 4a, Perceived ease of use is positively related to microcomputer usage. 4b. Perceived ease of use is positively related to perceived usefulness. Based on the TAM [1, 17, 70], perceived ease of use is related to both system and perceived usefulness. They proposed that perceived ease of use is direc indirectly related to beliavior (i.e., usage) through its effect on perceived usef These effects have been tested and found to be significant (1, 17]. Thus, itis pr that perceived ease of use is related to perceived usefulness and system usage. F it is proposed that perceived ease of use Will have both direct and indirect eff usage through its impact on perceived usefulness. Perceived Usefulness 5. Perceived usefulness is positively related to microcomputer usage. User acceptance of computer systems is driven to a large extent by per usefulness (1, 17]. Perceived usefulness influences user: ‘acceptance of systems the reinforcement value of outcomes. Adams et al. (1] and Davis et al. [17] fou Perceived usefulness is a major determinant of behavior (i.e., usage). Th literature also reported that perceived usefulness is Positively associated with s usage [16, 21, 41, 52, 54, 66, 75]. Therefore, we Propose that perceived usef will be directly related to user acceptance of microcomputer technology (i.e., u The technology acceptance model [1, 17), conceptual models of MIS succes 85] and related empirical research [14, 41, 67] support user beliefs as a intervening variable through which the effects of exogenous variables on syste are transmitted. Thus, the sets of individual characteristics, organizational char istics, and system characteristics are Posited to influence microcomputer usag directly and indirectly through their effects on beliefs, Although most of the relationships discussed above have been researched e no single study has examined all of these relationships in an integrated model foc ‘on microcomputers. Thus, the present: study is expected to contribute to MIS lite in three ways. First, unlike prior research that has examined a limited numt variables, this study tests an integrated model of the antecedents and consequen Table 1 Profile of Respondents Age Mean = 29.1 Median = 27.0 Range = 21-56 Gender Male = 68% Female = 32% Education: ‘Some graduate schoo! 84.0% Graduate degree 16.0% Organizational level: Professional staff . 57.2% First-level supervisor 21.7% Department heads and middle managers 18.3% Strategic management (executives) 2.8% beliefs. Second, with limited exceptions, the prior studies have generally examined only bivariate relationships of individual, organizational, and system characteristics. This study examines the effects of these variables on microcomputer ‘usage. Last, the present study utilizes a structural equation analytical technique that permits a concur- Tent assessment of the adequacy of the measurement model and the conceptual model used to assess microcomputer usage. Method Sample and Procedure A QUESTIONNAIRE, DESCRIBED AS A SURVEY OF MICROCOMPUTER USAGE, was distrib- uted to part-time M.B.A, students at an castem university who were employed full-time ina variety of manufacturing, service, merchandising, financial services, and government organizations. They held professional and managerial positions in a wide range of functional areas including accounting, finance, marketing, general manage- ment, information systems, and engineering. Out of 280 questionnaires distributed, 236 were completed and returned, representing a response rate of 84 percent. The exclusion of responses from incomplete questionnaires resulted in a final sample of 214 participants. Of the 214 respondents, 68 percent were males, and 32 percent were females. The average age of the respondents was 29.1 years (see Table 1). Operational Measures of the Study Variables System Usage Based on several studies [e.g., 14, 19, 44, 73], four indicators of microcomputer usage were included in this study: (1) perceived daily use of microcomputers; (2) perceived frequency of microcomputer use; (3) the number of software packages used | participants; and (4) the number of business tasks the system is used for. Foll Lee [48], and Igbariaet al. [44], individuals were asked to indicate the amount spent on the microcomputer per day, using a six-point scale ranging from (1) “ never” to (6) “more than 3 hours per day.” Frequency of use has been suggest Raymond [64] and used by Igbaria et al. [44], and provides a slightly dif perspective of usage than time. Both can also be couched as alternate measures same concept. Frequency of use was measured on a six-point scale ranging frc “Jess than once a month” to (6) “‘several times a day.” In a microcomputer environment, users have a wide choice of software pac! In such an environment a good indication of overall usage and the variety of performed on the microcomputer can be provided by measuring the num different packages used [44, 62]. A list of ten packages was given and respor were asked to indicate whether they used each one of these packages. In additic number of business tasks performed by the participants can be another indicator microcomputer usage [14, 44]. For the purpose of this study, eight tasks were d and the participants were asked to indicate whether they used the microcompt perform these tasks. These indicators are typical of the kinds of self-reported measures often u operationalize system usage, particularly in cases where objective usage metr notayailable. Objective usage logs were not practical in the present study since partic used different microcomputers as well as different applications for different task reported usage should not be regarded as precise measures of actual usage, al previous research suggests they are appropriate as relative measures [11]. Beliefs Based onthe TAM [1, 17], the measure of beliefs consists of two distinctive me: perceived ease of use and perceived usefulness. Four items that describe the ex which the users believe the microcomputer will enhance their jobs is labele¢ ceived usefulness.” The remaining four items describe problems and difficul using the microcomputer, due to its hardware limitations, and excess efforts 1 to perform the tasks. The second construct was reversed to reflect ease of learni remembering how to use a microcomputer in a generic sense. It was labeled “per ease of use.” This distinction between usefulness and ease of use reinforces prior MIS re results. Adams etal. [1], Davis [16], Davis etal. [17], Igbaria [41], Larcker and [47] and Swanson [75] developed different indicators to measure these variable indicators were found to be very similar to those used previously. Howev perceived ease of use indicator reflects more individual characteristics than ¢ teristics of the system being used. Thus, our model considered both as ¢ constructs, and their predictors and outcomes were examined separately. Perceived usefulness: This measure is defined as “the prospective user’s sut probability that using a specific application system will increase his or | performance within an organizational context” [16]. The items used to construct the perceived usefulness scale were adapted from prior research [41]. They are similar to scales used by Adams et al. [1], Davis [16], and Davis et al. [17], with appropriate modifications to make them specifically relevant to microcomputers. Individuals were asked to indicate the extent of agreement or disagreement with four statements concerning microcomputers on a five-point Likert-type scale ranging from (1) “strongly disagree” to (5) “strongly agree.” Perceived ease of use: This refers to the degree to which prospective users expect the computer system to be free of effort [16, 17]. The perceived ease of use items were adapted from prior research [39, 41] with appropriate modifications to make them relevant to microcomputers specifically operationalized. The four-item instrument is similar to the scale developed by Davis et al. [16] and validated by Adams et al. [1] and Mathieson [55]. Individuals were asked to indicate the extent of agreement or disagreement with four statements concerning microcomputers ona five-point Likert- type scale ranging from (1) “strongly disagree” to (5) “strongly agree.” Organizational Support Following Igbaria [41] and Igbaria and Chakrabarti [42], the measure of organizational support encompassed two broad categories of support: (1) end-user support, which includes the availability of system development assistance, specialized instruction, and guidance in using microcomputer applications; and (2) management support, which includes top management encouragement, and allocation of resources, The scale consisted of eight items, four representing end-user support and four representing management support. Respondents were asked to express their agreement/disagree- ment with eight statements concerning organizational support on a five-point Likert- type scale ranging from (1) “strongly disagree” to (5) “‘strongly agree.” System Quality Prior research [42, 43, 52, 53] has identified five different dimensions of system quality: functionality, equipment performance, interaction, environment, and the quality of the user interface. The scale used for this study included five items that measure these various dimensions. The response options, anchored on a five-point Likert-type scale, range from (1) “poor” to (5) “excellent.” Computer Experience and Training Computer experience was assessed by asking respondents to indicate the extent of their experience in using five different generic types of computer software and computer languages, and in developing computerized information systems, The response options ranged from (1) “none” to (5) “extensive.” In addition, the respon- dents were asked to report the extent of training on microcomputers they had received from four sources: college courses, vendor training, in-house training, and self-train- ing. For each source, this was measured with a five-item scale ranging from (1) “nc to (5) “extensive” [58]. Data Analysis The first stage of the analysis involved assessment of the Teliability of the meas uused to operationalize the variables in the study. We used partial least squares (Pl which is a multivariate technique that facilitates testing of the psychometric proper of the scales used to measure,a variable, as well as estimating the parameters ¢ structural model—that is, the magnitude and direction of the relationships among model variables (23, 50, 83]. The technique is: Particularly applicable in research ar where theory is not as well developed as that demanded by LISREL (25, 41]. suggested by Lohmoller [51, p. 7], “PLS methods are more close to the data, m explorative, more data analytic.” Of particular relevance to this study is the fact t PLS does not depend on having multivariate normally distributed data (distributi free). Finally, itcan be used with non-interval-scaled data and, importantly, with sn samples. PLS recognizes two components of a causal model: the measurement model ¢ the structural model. Figure 1 represents the structural model being examined. 1 model describes the relationships or paths among theoretical constructs. Furth more, for each construct in figure 1, there is a related measurement model, wh: links the’ construct in the diagram with a set of items. For example, perceiv usefulness is a composite of four related items. The measurement model consi of the relationships between the observed variables (items) and the constructs th measure, The characteristics of this model demonstrate the construct validity the research instruments—that is, the extent to which the operationalization o construct actually measures what it purports to measure. Two important dim sions of construct validity are (1) convergent validity, including reliability, a (2) discriminant validity. To test the model, the sample was split into two subgroups. The initial model v tested using the first split sample. Revisions were made based on the initial results ‘suggested by Grant [32]. The revised model was tested with the holdout sample. T sample was split by putting all responses from questionnaires with even identificati numbers into one sample (1), and all the responses with odd identification numb into a second sample (S2). There was no a Priori reason to expect systema differences between the two samples and both are comparable. The initial analy. was conducted with sample $1. Missing values were replaced with the mean of t nonmissing items for the same constructs. The resulting sample sizes were N1 = 1 (for sample $1) and N2 = 105 (for sample $2), The test of the measurement model includes estimation of the reliability cocfficies (Cronbach's alpha) of the measures, as well as an examination of the convergent a discriminant validity of the research instruments, In determining the appropriz minimum loadings required for the inclusion of an item within a scale, we us Fornell’s [23] recommendation to retain items that loaded highly (0.70 is consider to be a high loading since the item explains almost 50 percent of the variance in a Particular construct) on theirrespective constructs. Fornell and Larcker's (47) criterion that an average extracted variance should be 0.50 or more was used to assess the average variance extracted for all constructs. We also used the guidelines recom- mended by Hair ct al. [37] in determining the relative importance and significance of the factor loading of each item—that is, loadings greater than 0.30 are considered significant; loadings greater than 0.40 are considered more important; and loadings 0.50 or greater are considered to be very significant. Finally, the criteria suggested by Nunnally [59] were applied to determine the adequacy of the reliability coefficients obtained for each measure. To assess the discriminant validity of the measures—that is, the degree to which items differentiate among constructs or measure distinct: concepts—we examined the correlations between the measures of potentially overlapping constructs [41]. If the items comprising an instrument that measures a construct correlate more highly with each other than with items measuring other constructs in the model {27, 32], the measure is determined to have adequate discriminant validity. PLS is also used to test the structural model. A structural model isaregression-based technique, with its roots in path analysis, and often loosely termed as acausal modeling technique. Itis a relatively new approach to testing multivariate models with empirical data [83]. The structural model consists of the unobservable constructs and the theoretical relationships among them (the paths). It evaluates the explanatory power of the model and the significance of paths in the structural model, which represent hypotheses to be tested. Together, the structural and measurement models form a network of constructs and measures. The item weights and loadings indicate the Strength of measures, while the estimated path coefficients indicate the strength and the sign of the theoretical relationships. 7 ‘The evaluation of the structural model was conducted with the overall sample. The computer program used for this analysis wasLVPLS 1.6 (latent variables path analysis using partial least squares), developed by Lohmoller [50, 51]. For more information on PLS, the interested reader is referred to Fornell [23], Lohmoller [SO], and Wold [84]. The assessment of the measurement model and the structural model should be simultaneously examined. The initial evaluation was completed with the orig- inal S1 sample; the measurement model was revised and retested using sample S1. The holdout sample (S2) was used for retesting the revised measurement and structural model. The estimated path coefficients of both models were tested with t-statistics calculated using a nonparametric test of significance known as jackknifing (24, 81, 82). ‘The path coefficient of an exogenous variable represents the direct effect of that variable on the endogenous variable. An indirect effect represents those effects interpreted by the intervening variables; itis the product of the path coefficients along an indirect route from cause to effect via tracing arrows in the headed direction only, When more than one indirect path exists, the total indirect effect is their sum. The sum of the direct and indirect effect reflects the total effect of the variable on the endogenous variable (5, 68). Table 2 Assessing the Measurement Model (Sample S1) Variables Reliability of a scale” Portion of the variance extracted” End-user training 0.70 0.39 ‘Computer experience 0.77 0.42 Organizational support Management support 0.89 0.66 EUC support 0.81 0.52 System quality 0.79 0.44 Beliefs about microcomputers Perceived ease of use 0.82 0.63 Perceived usefulness 0.86 0.60 Microcomputer usage 0.82 0.54 2 EA; * Reliability = GA) — +. Gay +t (e) pee 2 » portion of variance extracted = a Results ‘THE RESULTS OF THE TESTS OF THE MEASUREMENT MODEL using S1 are reporte« Table 2. The internal consistency of the measurement model was assessed by c¢ puting the composite reliability. These reliability coefficients are displayed for all study variables using S1 in Table 2. The table indicates that some construct revisi were needed in order to meet established guidelines. The loadings for the consin (EUC support, management support, perceived case of use, perceived usefulness, microcomputer usage) revealed that both aspects of organizational support (E support and management support) and beliefs (perceived case of use and percei usefulness) met all three criteria for validity and reliability. The average extrac variance of system quality was slightly below 0.50. Since all the factor loadings ¥ very significant, and the reliability of the construct was considered high (0.79), construct was retained, Multidimensionality appears to be an issue with microcomputer usage. Two ind tors—frequency of use and time of uso—were proposed to measure “perceived usa} and were so labeled. The remaining two indicators—number of tasks and numbe applications—measure different aspects of system usage representing the importa of use, as well as the variety of use. These were labeled “variety of use.” In microcomputer environment, it was found that it is very important to examine Table3 Assessing the Revised Measurement Model (Sample S1) Variables Reliability of a scale” Portion of the variance extracted” End-user training 0.73 0.49 Computer experience 0.80 0.50 Organizational support ‘Management support 0.89 0.66 EUC support 0.81 0.52 ‘System quality 0.79 0.44. Beliefs about microcomputers Perceived ease of use 0.82 0.53 Perceived usefulness 0.86 0.60 Microcomputer usage Perceived usage 0.91 0.81 Variety of use 0.82 : 0.70 f = OE * Reliability = | (Ay +E var (e;) > * Portion of variance extracted = perceived usage as well as the variety of use [9, 44, 48, 62]. Therefore, the model was revised to incorporate both constructs. The remaining constructs were also analyzed. The loadings of one item of user training (vendor training) and computer experience (using fourth-generation lan- guages) were very low, and were thus removed. Note that the majority of participants had very little experience in 4GL. This may be due to relatively low availability or it may reflect the low use of such languages. Further, since the participants are part-time M.B.A. students, most of them had formal university training instead of vendor training. Some of these indicators have low loadings due to a floor effect (the mean response of vendor training was 1.74 with a low standard deviation of 1.11 compared with the second lowest, in-house training, with an average of 2.38 and standard deviation of 1.20. Similarly, experience with fourth-generation languages had a very low average). Testing the Revised Model The original model was revised to incorporate the constructs changed as a result of the evaluation of the measurement model. The revised model was retested using the same sample (S1); the results are presented in Table 3, which shows that the composite Table4 Discriminant Validity Using Sample S1 Variables 1 2,3 4 5 6 7 8 1. Usertraining 0.490 2. Computer 0.176 0.500 experience 3. EUC support 0.008 0.008 0.520 4, Management 0.040 0.001 0.465 0.660 support 5. System quality 0.078 0.073 0.097 0.040 0.440 eee 0.102 0.040 0.101 0.027 0.152 0.530 of use 7. Perceived - 0.109 0.084 0.158 0.017 0.160 0.402 0.600 usefulness 8, Perceived usage 0.078 0.073 0.101 0.055 0.073 0.063 0.162 0.810 9, Variety ofuse 0.176 0.119 0.050 0.008 0.084 0.046 0.090 0.189 0. Diagonals represent the average variance extracted, while the other matrix entries represent tt shared variance. reliabilities of all of the constructs were satisfactory (ranging from 0.73 to 0.91) the item reliabilities of the constructs were significant. Only one construct training) did not meet Fornell and Larcker’s [26] recommendation for the extr variance. Since almost all the conditions were met except one, which is slightly than the requirements, and the indicators were found to represent the various structs, we decided not to revise the model further and considered it satisfactory few low values suggest that further research is required to refine the measurem these constructs and improve convergence. ‘The results from discriminant analysis are shown in Table 4. The survey dis: nated adequately between proposed cause and effect constructs. The shared vat (the squared correlations) was less than the amount of variance extracted t indicators measuring the constructs. In other words, the constructs were corn more highly with their indicators than with other constructs in the model. In sum the convergent validity of our survey measures using S1 was adequate. A\ variance extracted for most of the constructs exceeded 0.50 (except three cons that were slightly below 0.50), the item reliability (factor loadings) exceede (except one item that was slightly below 0.50), and the composite reliability scales exceeded 0.80. Furthermore, the survey discriminated adequately be proposed cause and effect constructs. Given the revised measurement model, the results of the conceptual mode reported using the original sample (S1). The results of testing the structural Table S The Direct Effects of the Microcomputer Usage Model (Sample S1) Perceived ease Perceived Perceived Variety of use usefulness usage of use User training 0.17"** 0.09" 0.09*"* 0.27"** Computer 013°" 0.03" 0.17"** 0.19°"* experience Organizational support: c 0.35** 0.17" EUC support 0.2 Management "0.04" 0.23" 0.09" support ‘System quality O27" 0.12" 0.05 ot" Beliefs about microcomputers: Perceived ease oso" ont 0.10" of use Perceived 0.3" o.16"* usefulness : Fr 0.26" 04s" 0.270" 0.26" . * p$0.05; ** p $0.01; *** p< 0.001. using SI are presented in Table 5. The data in Table 5 show that the exogenous variables explained 26 percent of the variation in perceived ease of use, 48 percent in perceived usefulness, 27 percent in perceived usage, and 26 percent in variety of use. Table 5 shows that of the twenty-four direct paths tested in the structural model, twenty-three were found to be statistically significant. Results show that EUC support, system quality, computer experience, and user training are positively related with Perceived ease of use. Perceived ease of use is the strongest determinant of perceived usefulness, followed by EUC support and management support. Results also show that computer experience had an indirect effect on perceived usefulness through perceived ease of use. The remaining constructs have direct and indirect effects on perceived usefulness. . Table 5 also shows that perceived usefulness had a significant effect on per- ceived usage and variety of use. We also found that user training, computer experience, EUC support, management support, and perceived ease of use had direct effects on perceived usage as well as on variety of use. Note that system quality was found to have only an indirect effect on perceived usage through perceived usage and perceived ease of use. This is consistent with the findings of Igbaria et al. [43] who suggested that improved system quality may serve to generate more favorable beliefs and thereby contribute to increased microcom- puter usage. System quality has been found to enhance perceived usefulness [10] and to affect perceived ease of use [8, 10, 20, 56]. Table 6 Assessing the Revised Measurement Model (Sample $2) Variables Reliability of a scale" Portion of the variance 5 extracted? End-user raining 0.89 0.73 Computer experience 0.80 0.51 Organizational support Management support 0.83 058 EUC support 0.88 0.65 System quality 0.85 0.53 Beliefs about microcomputers Perceived ease of use 0.85 0.60 Perceived usefulness 0.82 054 Microcomputer usage Perceived usage 0.80 0.68 Variety of use : 0.82 0.69 = , EA, * Relsbitiy == = (A)? +E var(e), P On? * Porton of variance erated =—t- Retesting the Revised Structural Model Given the revised measurement model, the hypothesized structural model can evaluated using a different data sct [61]. The initial evaluation was completed v the original split-half data (S1); the holdout sample (S2) was used for testing revised model. The revised model, consisting of the nine constructs, was tested u: sample S2. In general, the convergent validity of the survey constructs using holdout sample was adequate (see Table 6). Average variance extracted for constructs exceeded 0.50. Item reliabilities for the constructs were significant. ' composite reliability of all the constructs exceeded 0.80. Furthermore, Table 7 sh that the survey discriminated adequately between proposed cause and effect c structs. Examination of the data in Table 8 using sample S2 shows that the model explai 36 percent and 37 percent of the variance in perecived ease of use and percei usefulness, respectively. Table 9 shows that the model explained 31 percent anc * percent of the variance in perceived usage and variety of use, respectively (in all ca P $0.01). Tables 8 and 9 indicate that of the twenty-five direct paths in the rev: structural model, twenty-one were significant at the 0.05 level or lower. Table 7 Discriminant Validity Using Sample S2 Variables 1 2 3 4 5 6 7 8 9 1. User training 0.730 2.Computer 0.132 0.510 experience 3.EUC support 0.020 0.049 0.650 4.Management 0.023 0.045 0.268 0.560 ‘support 5. System quality 0.059 0.033 0.005 0.058 0.530 6. Perceived 0.143 0.088 0.088 0.121 0.215 0.600 ease of use 7. Perceived 0.131 0.038 0.007 0.023 0.135 0.327 0.540 usefulness 8. Percelved 0.307 0.200 0.037 0.082 0,072 0.232 0.149 0.680 usage 9. Variety of use 0.254 0.097 0.032-'0.033 0.055 0.111 0.097 0.166 0.690 Diagonals represent the average variance extracted, while the other matrix entries represent the shared variance, Perceived Ease of Use Table 8 shows that all five external variables had significant directeffects on: perceived ease of use. The strongest effects are noted for system quality, user training, both aspects of organizational support—EUC support and management suppor—and com- puter experience. Perceived Usefulness The path analysis results for perceived usefulness show that ‘perceived case of use had a strong positive effect on perceived usefulness. Four of the five external variables also had significant direct effects on perceived usefulness: user training, system quality, EUC support, and computer experience, However, management support had no significant direct effect. System Usage The data in Table 9 show that, consistent with the model's predictions, perceived usefulness had a positive effect on perceived usage and variety of use. Other significant and direct predictors of perceived usage and variety of use, respectively, were user training, computer experience, and EUC support. Table 9 also shows that system Table8 Results of the Structural Model Predicting Perceived Usefulness and Perceived Ease of Use (Sample S2) Antecedent Perceived case Perceived variables of use usefulness Standardized Standardized Indirecteffects Total effects path path coefficients coefficients (direct (direct effects) effects) User training 0.22%" 0.17" O.11 0.28°** Computer 0.10" 0.04 0.05 0.09" experience Organizational support EUC support 0. 0.08" 0.08 Management on 0.04 0.07 o.nte ‘support System quality 0.36" 0.12% 0.18 0.30" Perceived ease _ 0.49%" 0.4g"* of use Fr 0.36"** 0.37"** * pS.0.05; ** p $0.01; *** p $0,001. quality and management support had positive direct effects only on perceived usag The data reported in Tables 8 and 9 also illustrate the decomposition of total effec into direct and indirect [5, 68]. An indirect causal effect represents the effect antecedent variables mediated by the intervening variables in the model—that perceived ease of use and perceived usefulness. User training and system qual showed strong direct and indirect (through perceived ease of use) effects on perceiv usefulness. Further, computer experience had small but significant direct and indir effects on perceived usefulness through perceived case of use. Note that while El support had both direct and indirect effects on perceived usefulness, managemt support had only an indirect effect on perceived usefulness.” Table 9 also shows that perceived ease of use had only an indirect effect on perceit usage, mainly, through perceived usefulness. System quality was also found to hz an indirect effect on variety of use through perceived case of use and perceit usefulness, but had both direct and indirect effects on perceived usage. User traini computer experience, and EUC support had both direct and indirect effects on b perceived usage and variety of use through perceived ease of use and percei usefulness. Management support was found to have a significant éffect on percei usage; however, it had insignificant direct and indirect effects on variety of use. Fig 2 summarizes the results of the microcomputer usage model. Table 9 — Results of the Structural Model Predicting Microcomputer Usage: Perceived Usage and Variety of Use (Sample S2) Antecedent Perceived usage Variety of use variables Standardized Indirect Total ‘Standardized Indirect ‘Total path effects effects path effects effects coefficients coefficients Girect Girect effects) effects) Usertraining 0.39" 0.04 as 0.95" 0.08 oat Computer O11" 0.01 0.12"** 0.22"** 0.03 0.25°** experience + Organizational support EUC 0.03" 0.02 0.05"** 0.14%" 0.05 o.19"* support Manage- 0,06" 0.02 0.08" 0.04 0.02 0.02 ment ‘support ‘System 0.05" 0.05 ~0.10"" 0.01 0.10 0.09"" quality Beliefs about microcomputers Perceived 0.05 0.05 0.10"** 0,09"** 0.11 0.20°** ease of use Perceived 0.10°** 0.10°** 0.22"** 0.22"** usefulness Fr oat 0.45" *p $0.05; ** p $0.01; *** ps 0.001. Discussion, Managerial Implications, and Study Limitations ‘THE PRIMARY OBJECTIVE OF THIS STUDY WAS TO INVESTIGATE the use of a structural equation model to measure the determinants of microcomputer usage. Overall, some support was found for the structural model tested in this study. The measurement model has provided new insights regarding some of the main constructs. Two separate components of organizational support (management support and EUC support) were examined. That emphasizes the need for managers to focus on two different support areas and to coordinate the efforts of top management and the personnel who support end users. It is essential for the MIS to provide effective user support and to encourage end users to use their systems. ‘Two distinct constructs related to beliefs were also incorporated: perceived ease of use and perceived usefiulness. These are consistent with Davis [16], who found these User Characteristics: User Training User Computer Experience Organizational Support: EUC Support Management Support Figure 2. Results of the Microcomputer Usage Model (Numbered arrows correspond to direct effects.) constructs to be significantly correlated with self-reported usage. Davis et al. [17] also found that perceived ease of use had a direct effect on perceived usefulness and that both will be affected by extemal variables such as user training, system quality, and organizational support. Others also developed measures similar to perceived ease of use and perceived usefulness [29, 47, 75]. Since these two constructs are shown to be statistically distinct dimensions and they have been defined and validated by several researchers [16, 17, 47, 75], researchers should use these two constructs instead of the overall user beliefs measure. This use should facilitate comparison ofresults across studies. ‘Note that perceived ease of use and perceived usefulness were significantly intercorrelated (r= 0.63 and 0.57 for samples $1 and S2, respectively). These correlations are similar to the findings of Davis [16] in Study 1, but higher than those found in Study 2. Further, ease of use.and usefulness are significantly correlated with self-reported usage (r = 0.32 and 0,34 for usage and 0.39 and 0.48 for variety of use, respectively). These results are consistent with those reported by Davis [16] and Adams et al. [1]. Two different constructs for microcomputer usage were also observed. Usage was defined in two different ways: as perceived usage and as variety of use. While perceived usage refers to the amount of time spent interacting with a microcomputer and frequency of use, variety of use refers to the importance of use and the collection of specific software packages used. These results are similar to the findings of Igbaria etal. [44] and Pentland [62]. Both are’self-reported measures and were found to be correlated (r = 0.44 and 0.41 for samples $1 and S2, respectively). More research is needed to investigate the relationships between these two constructs using more ‘objective measures, such as monitoring the system to gather information regarding the number of software functions actually used, Another objective measure might be actual usage, as represented by frequency of use and length of time of use. It is acknowledged that some users may overstate perceived use, as reported by Pentland [62], and that objective measures may provide different results. ‘The tested conceptual model confirms the effects of individual, organizational, and ‘system characteristics on perceived ease of use and perceived usefulness. The model also confirms the influence of perceived ease of use on perceived usefulness, and the effects of perceived usefulness on perceived usage and variety of use. Note that the effects of ease of use on perceived usage and variety of use were found to be channeled through perceived usefulness, thus demonstrating the key intervening role of perceived usefulness [17]. As proposed, user training was found to have direct and indirect effects on all four endogenous variables: perceived ease of use, perceived usefulness, perceived usage, and variety of use. This finding is consistent with the results from Nelson and Cheney [58], Igbaria [41], and Schewe [69] showing strong correlations between training and beliefs (perceived case of use and perceived usefulness) as well as microcomputer acceptance (in this study, measured as perceived usage and variety of use). This underscores the importance of user training in influencing users’ perceptions and eventually their use of the system. The importance of prior computer experience in promoting increased microcom- puter usage was also corroborated. Computer experience influences perceived usage and variety of use directly as well as indirectly through perceived ease of use and

Das könnte Ihnen auch gefallen