(Thompson Et Al., 1991) Personal Computing Toward a Conceptual Model of Utilization

Personal Computing: Toward a Conceptual Model of Utilization Author(s): Ronald L. Thompson, Christopher A. Higgins and Jane M. Howell Source: MIS Quarterly, Vol. 15, No. 1 (Mar., 1991), pp. 125-143 Published by: Management Information Systems Research Center, University of Minnesota Stable URL: http://www.jstor.org/stable/249443 . Accessed: 01/02/2015 04:36 Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp

. JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected].

.

Management Information Systems Research Center, University of Minnesota is collaborating with JSTOR to digitize, preserve and extend access to MIS Quarterly.

http://www.jstor.org

This content downloaded from 193.255.248.150 on Sun, 1 Feb 2015 04:36:34 AM All use subject to JSTOR Terms and Conditions

Utilizationof Personal Computers

Personal Computing: Toward a Conceptual Model of Utilization1 By: Ronald L. Thompson School of Business Administration University of Vermont Burlington, Vermont 05405 Christopher A. Higgins School of Business Administration University of Western Ontario London, Ontario CANADAN6A 3K7 Jane M. Howell School of Business Administration University of Western Ontario London, Ontario CANADAN6A 3K7

Abstract Organizations continue to invest heavily in personal computers for their knowledge workers. When use is optional, however, having access to the technology by no means ensures it will be used or used effectively. To help us gain a better understanding of factors that influence the use of personal computers, researchers have recently adapted the theory of reasoned action proposed by Fishbein and Azjen (1975). This study uses a competing theory of behavior proposed by Triandis (1980). Responses were collected from 212 knowledge workers in nine divisions of a multi-national firm, and the measures and research hypotheses were analyzed using partial least squares (PLS). The results show that social norms and three components of expected consequences (complexity of use, fit between the job and PC capabilities, and long-term con'This research has been supported in part by: the Social Sciences and Humanities Research Council of Canada, the National Centre for Management Research and Development (Canada), and the School of Business Administration at the University of Western Ontario. This article was developed in part from a paper presented at the Administrative Sciences Association of Canada (ASAC) annual meeting in Montreal in June 1989.

sequences) have a strong influence on utilization. These findings confirm the importance of the expected consequences of using PC technology, suggesting that training programs and organizational policies could be instituted to enhance or modify these expectations. Keywords: Personal computing, information technology utilization, attitudes, behavior ACM Categories: K.0, K.6.0, K.8, K.m

Introduction Many information systems (IS) researchers have stressed the need to build IS research on a cumulative tradition, using referent disciplines and theoretical arguments as a foundation (Goodhue, 1988; Keen, 1980; Robey, 1979). The value of referent disciplines, Keen (1980) argues, is to keep IS researchers from "falling into the framework of the month trap" (p.11). Research investigating the relationship between attitudes and computer utilization is one area where many IS researchers have been remiss, to date, in using existing models or theories, particularly those from the social psychology literature (Davis, et al., 1989; Goodhue, 1988; Robey, 1979). This lack of theoretical justification provides a potential explanation for the mixed empirical support found for the hypothesis that attitudes influence computer use (Davis, et al., 1989; Lucas, 1975; 1978; Pavri, 1988; Robey, 1979; Schewe, 1976; Schultz and Slevin, 1975; Swanson, 1982). IS researchers have recently adopted Fishbein and Azjen's (1975) theory of reasoned action into the context of information technology use (see Davis, et al., 1989; Pavri, 1988). This theory, widely tested in sociological and psychological research, has been found to be lacking in certain respects. Triandis (1980) has proposed a theory that incorporates many of the same concepts and contructs but also modifies and redefines them. For example, while Fishbein and Azjen's theory considers all beliefs that a person has about an act or behavior, Triandis makes a distinction between beliefs that link emotions to the act (occurring at the moment of action) and beliefs that link the act to future consequences. He argues that behavioral intentions are determined by feelings people have toward the behavior

MIS Quarterly/March 1991

This content downloaded from 193.255.248.150 on Sun, 1 Feb 2015 04:36:34 AM All use subject to JSTOR Terms and Conditions

125

Utilization of PersonalComputers

(affect), what they think they should do (social factors),and by the expected consequences of the behavior.Behavior,in turn, is influencedby what people have usuallydone (habits),by their behaviorial intentions, and by facilitating conditions. Despitethe acceptance of Triandis'(1980)theory withinthe psychologicalliterature,it has not been used within the IS context. Accordingly, the purpose of the study described in this article is to conduct an intitialtest of a model of personal computer (PC) utilization using a subset of Triandis'(1980)theoryof attitudesand behavior. Thistheoryimpliesthat the utilizationof a PC by a knowledgeworkerin an optionaluse environment would be influenced by the individual's feelings (affect)towardusing PCs, social norms in the work place concerning PC use, habits associated withcomputerusage, the individual's expected consequences of using a PC, and facilitatingconditions in the environmentconducive to PC use.

Conceptual Model and Research Hypotheses The theoretical grounding for this research comes from the work of Triandis(1971; 1980). In earlier work, Triandis (1971) argued that behavioris determinedby whatpeoplewouldlike to do (attitudes),what they thinkthey should do (social norms), what they have usually done (habits),and by the expected consequences of their behavior. He suggested that attitudes involve cognitive, affective, and behavioralcomponents. The cognitive component of attitudes involves beliefs. In the context of PCs, for example,a person may holda beliefthatPCs make workmore efficient.The affectivecomponentof attitudeshas a like/dislikeconnotation.Thus,the statement "I hate computers"is considered an indicationof the affectivecomponentof attitudes. Behavioralintentionsare simplywhatindividuals intend to do. For example, the assertion "I will start to learn a software package tomorrow" representsa behavioralintention.Thus, attitudes involvewhat people believe (cognitive),feel (affective), and how they would like to behave (behavioral)towardan attitudeobject. Later, Triandis(1980) presented a more comprehensivemodelof interpersonalbehavior.The

majorstatement of this model (see Figure 1) is that social factors, affect, and perceived consequences influence behavioralintentions,which in turn influence behavior. In addition,Triandis states that habits are both direct and indirect determinantsof behavior. He furtheracknowledges that even when intentions are high, behaviormaynotoccurifthe "geography"of the situation(i.e., facilitatingconditions)makes the behaviorimpossible. Thus, if someone intends to use a PC but does not have easy access to one, usage is less likelyto occur. The model includes othervariables,not germaneto this study, such as culture,the social situation,and genetic biologicalfactors that can influence behavior. Inthis study,we test a subset of Triandis'(1980) theory applied to the context of PC use. Specifically,we examined the direct effects of social factors, affect, perceived consequences, and facilitating conditions on behavior. Behavioralintentions were excluded from the model becuase it was actual behavior(i.e., PC utilization)in which we were interested. Habits were excluded because, in the context of PC utilization,habits (i.e., previous use) have a tautological relationshipwith current use. The relevantconstructsare discussed in moredetail below, and a furtherdiscussion of the role of habits is also presented.

Social factors Triandis(1971)arguedthatbehavioris influenced by social norms, which depend on messages receivedfromothersand reflectwhatindividuals thinkthey should do. In his laterwork,Triandis (1980) expanded this term and called it social factors, that is, "the individual'sinternalization of the reference groups' subjective culture,and specific interpersonalagreements that the individualhas made withothers, in specific social situations"(p. 210). Subjectivecultureconsists to do whatis perceived of norms(self-instructions to be correct and appropriateby members of a culturein certainsituations);roles (whichare also concerned with behaviors that are considered correctbut relateto persons holdinga particular positionin a group,society,or socialsystem);and values (abstractcategories withstrong affective components). Empiricalsupport for the relationshipbetween social normsand behaviorcan be foundin many

126 MIS Quarterly/March1991

This content downloaded from 193.255.248.150 on Sun, 1 Feb 2015 04:36:34 AM All use subject to JSTOR Terms and Conditions

Utilization of PersonalComputers

Habit

Facilitating Conditions

Behavior

Figure 1. Factors Influencing Behavior (a subset of the model proposed by Triandis, 1980) studies. Forexample,Tornatskyand Klein(1982), in a meta-analysisof 75 studies of the relationship betweeen innovationcharacteristics and adoption,foundthat compatibilityof the innovation withthe normsof the potentialadoptershad a significantinfluenceon adoption.The relationship is also consistentwiththe theoryof reasoned action proposed by Fishbein and Ajzen (1975), a theorythat has alreadybeen tested withinthe IS context(Davis,et al., 1989; Pavri,1988). More specifically,Pavri(1988) reportsa positive relationshipbetweensocial normsand the utilization of PCs by managers in optional use environments. AlthoughDavis, et al. (1989) reportno significantrelationshipbetweensocial normsand usage, they attributethis unexpected findingto the weak psychometricpropertiesof theirsocial norms scale, and the particularIS context (i.e., use of a wordprocessing system) in whichtheir research was conducted. Consistent with Triandis'(1980)theoryand the evidence supporting this relationship,the hypothesisto be tested is:

H1: There will be a positive relationship between social factors concerning PC use and the utilization of PCs.

Affect Triandis(1971) defines attitude as "an idea, charged withaffect, that predisposes a class of actions to a particularclass of social situations" (p. 2). He acknowledges that attitudeis an imprecise termthat is more useful for discussions where precisionis not necessary. For research involvinga linkbetween attitudesand behavior, however, Triandis(1980) argues for precision through the separation of the affective and cognitive components of attitudes. To do this, Triandis(1980)uses the termaffect,whichrefers to "the feelings of joy, elation, or pleasure, or depression, disgust, displeasure, or hate associated by an individualwitha particularact" (p. 211). Accordingto Goodhue (1988), most IS researchers have not made a distinctionbetween the

MIS Quarterly/March1991 127

This content downloaded from 193.255.248.150 on Sun, 1 Feb 2015 04:36:34 AM All use subject to JSTOR Terms and Conditions

of PersonalComputers Utilization

affective component of attitudes (which have a like/dislikeconnotation)and the cognitive component or beliefs (which are the informationa person holds about an object, issue, or person). Forexample,a close examinationof Schultzand Slevin's (1975) operationalizationof user attitudes (a single construct) toward mainframe systems suggests that many questions tap the cognitiveas well as the affectivecomponents. If these are in fact separate components, combining them into a single component makes it impossible to assess their relative influence. Similarly,Lucas (1978) also used a mixtureof cognitiveand afffectingquestionsto measurethe single construct of attitudes. Controversy also remains among those researchers who acknowledge the differences betweenthe affectiveand cognitivecomponents. Burnkrant and Page (1982)suggest thatalthough there may be a theoretical justification for separatingthe cognitivefromthe affectivecomponent, when it comes to measurement they should be treated as the same construct. Goodhue (1988), on the other hand, states that mixing the measurement of the components withinthe same constructcould introduceadditional bias or random errorbecause the affect towardthe object could influencethe responses to the cognitivequestions. To be consistent with Triandis'(1980) theory,we have made a distinction between the affective and cognitive components of attitudes, leading to the following hypothesis: H2: There will be a positive relationship between affect toward PC use and the utilization of PCs.

Perceived consequences Accordingto Triandis(1971), anotherimportant factorinfluencingbehavioris the expected consequences of the behavior, later re-named perceived consequences (Triandis,1980). He argues that each act is perceived as having potentialconsequences thathave value,together thatthe consequence willoccur. witha probability The perceived consequences construct is consistent withthe expectancy theoryof motivation proposedby Vroom(1964)and developedfurther by Porterand Lawler(1968). The basic premise of expectancy theoryis that individualsevaluate the consequences of their behaviorin terms of

potential rewards and base their choice of behavioron the desirabilityof the rewards.Robey (1979)suggests thatfutureresearchon attitudes should be done withinthe context of the expectancy theoryof behaviorand proposed a model based on this theory. Inherstudyof the optionaluse of a decisionsupportsystem by senior undergraduatestudents, DeSanctis(1983)findsweak-to-moderate support for-her hypotheses derived from expectancy theory. Beatty (1986) also uses expectancy theoryas the basis forher investigationof the use of computer-aideddesign and manufacturing systems. She finds a strongerrela(CAD/CAM) tionship between expectations and actual use. Based on expectancytheory,ifthe expectedconsequences of using a PC are attractive(such as the increasedopportunityforpreferredfuturejob of obtainingthe assignments),and the probability consequences are high, then utilizationof a PC will be greater. Perceivedconsequences are likelyto have many dimensions.Forexample,enhancedjob satisfaction and morejob flexibilitymay be two different constructsthatcould be labelled perceivedconsequences. Triandis(1971) acknowledges that the perceived consequences construct in his model is not unidimensional,possibly having several components.This is consistentwithconceptualargumentsand empiricalfindingsof other researchers,who suggest thereare multiplecomponents(Fishbeinand Azjen, 1975; Lucas 1978; Schultz and Slevin, 1975). Inthis study,three dimensionsof perceivedconsequences are defined. Two of these are nearterm in nature, while the third is more futureoriented.The firstof the near-termconsequences relatesto perceptionsaboutthe complexityof using a PC.

Complexity Accordingto Rogersand Shoemaker(1971)complexityis defined as "the degree to whichan innovation is perceived as relativelydifficultto understand and use" (p. 154). Tornatzkyand Klein(1982) find that the more complex the innovation, the lower its rate of adoption. If PC utilizationcan be viewed withinthe contextof innovationadoption,then these results suggest a negative relationshipbetween complexity and

128 MIS Quarterly/March1991

This content downloaded from 193.255.248.150 on Sun, 1 Feb 2015 04:36:34 AM All use subject to JSTOR Terms and Conditions

Utilization of PersonalComputers

utilization.Withinthe IS literature,Davis, et al. (1989) propose a technologyacceptance model thatincludesa constructthattheytermperceived ease of use. This is defined as the degree to whichthe user expects the system to be free of effort.Intheirstudy they find a positive correlation between perceived ease of use and behavioralintentions.Inthis study,we examined complexityof PC use, the oppositeof ease of use. The related hypothesis is therefore: H3: There will be a negative relationship between the perceived complexity of a PC and the utilization of PCs.

Job Fit The second near-termcomponentrelates to the capabilitiesof a PC to enhance an individual's job performance.Morespecifically,this dimension is definedas perceivedjob fitand measures the extentto whichan individualbelieves thatusing a PC can enhance the performanceof his or herjob (e.g., obtainingbetterinformation fordecision making or reducing the time requiredfor completing importantjob tasks). The positive relationshipbetween perceivedjob fit and PC utilizationhas empiricalsupport. In Tornatskyand Klein's(1982)meta-analysisof innovationadoption,they find that an innovation is more likelyto be adopted when it is compatible with individuals'job responsibilities.Robey (1979) finds that the "performancefactor," as operationalizedby Schultz and Slevin (1975), is the strongest predictorof utilization.Theirconstruct is similarto Floyd's(1986) "system/work fit" (i.e., facilitatingaccomplishment of core tasks, improvingindividualjob productivity,and improvingqualityof workoutput),whichhe found to be positivelyrelatedto the use of mainframebased informationsystems. It is also similarto Davis,et al.'s (1989)"perceivedusefulness"construct(definedas the user's subjectiveprobability that using a specific applicationsystem will increase his or her job performance),which they find to be strongly correlated with utilization. Additionalsupportis offeredby Goodhue(1988), who suggests thatan importantpredictorof use is the correspondencebetweenjobtasks and the capabilitiesof the informationsystem to support the tasks. Cooper and Zmud(1990), in a study of the adoptionof MRPsystems, also findtasktechnology compatibilityto be a majorfactorin

explainingadoptionbehaviors.Buildingon these findings, the hypothesis to be tested is: H4: There will be a positive relationship between perceived job fit and the utilization of PCs. Long-Term Consequences of Use The thirdand finalcomponentof perceivedconsequences includedhere is definedas long-term consequences of use. These are outcomes that have a pay-offin the future,such as increasing the flexibilityto change jobs or increasingthe opportunitiesfor more meaningfulwork.Forsome individuals,the motivationto adoptand use PCs may relate more to buildingor planningfor the futurethan to addressing currentneeds. Empiricalsupportfor this construct is provided by Beatty(1986),who findsa strongpositiverelationship between perceived long-term consequences of use and actual use of CAD/CAM systems. Interviewswithadopters revealed that they believed that use of the system would enhance theircareer mobility,even thoughthey were not convinced it wouldassist them greatly on their current job. Therefore, the next hypothesis is: H5: There will be a positive relationship between perceived long-term consequences of use and the utilization of PCs.

Facilitatingconditions Triandis(1980) states that behaviorcannot occur ifobjectiveconditionsin the environmentprevent it. He defines facilitatingconditions as "objectivefactors,'outthere'in the environment, thatseveraljudges or observerscan agree make an act easy to do" (p. 205). Inthe context of PC use, the provisionof supportforusers of PCs may be one type of facilitatingconditionthat can influence system utilization.By trainingusers and assisting them when they encounterdifficulties, some of the potentialbarriersto use are reduced or eliminated.Schultzand Slevin(1975)consider "support/resistance" (the system has top management, technical, implementation,and organizational support,and not undueresistance) as one factorinfluencingutilization.Robey(1979) finds a positive correlation between "support/resistance"(as defined and measured by

MIS Quarterly/March1991 129

This content downloaded from 193.255.248.150 on Sun, 1 Feb 2015 04:36:34 AM All use subject to JSTOR Terms and Conditions

Utilization of PersonalComputers

Schultz and Slevin, 1975) and use of a system. The next hypothesis is: H6: There will be a positive relationship between facilitating conditions for PC use and the utilization of PCs.

Habits Althoughhabitsare not specificallytested in this study,they are clearlyan importantdeterminant of behaviorand must be acknowledged. According to Triandis (1971), habits are situationbehavior sequences that occur without selfinstruction.The individualis usually not conscious of these sequences. Priorresearchhas shownthathabitsare a strong predictorof behavior.Forexample, Sugar (1967) measuredthe attitudes,norms,and habitsof college students concerningcigarettesmoking.On a separate occasion, the same students were offered a cigarette.The strongest single predictor of behaviorwas habit, followed by norms; the least importantpredictor was attitudes. This wouldbe expected because frequent,repetitive past behavior(i.e., a habit)generally would be highlycorrelatedwithcurrentbehavior.Forthis study, however, including the habit construct presents a majordifficulty. At a conceptuallevel, one could argue that habit should play a role in the utilizationof a PC; the PC might be used for certain tasks simply because it has been used in the past, not necessarily because it is the most efficientor effective approach. At a measurement level, however,difficultiesexist. Triandis(1980) notes that habits can be measured by the frequency of occurrenceof behavior.This is preciselyidenticalto our measure of utilization,which leads to a tautology.Forthis reason, we did not include the habit construct in this study. To summarize,we have adaptedthe theoryof interpersonalbehaviorproposedby Triandis(1980) to the context of PC use by knowledgeworkers in optionaluse environments.This theory suggests thataffect,perceivedconsequences, social factors,facilitatingconditions,and habitsare the primarydeterminantsof behavior.Twomodifications to the theorywere made in orderto test the model withinthe IS context. First,three distinct cognitive components of perceived consequences (complexity,job fit and long-termcon-

sequences) were identified.Second, the habit constructwas excluded from the analysis. The conceptual model illustrating the research hypotheses is shown in Figure 2.

Methods of constructs Operationalization To operationalizethe constructs, we referenced the workof manyresearchers,includingAmoroso (1986), Beatty (1986), Floyd (1986), Howard (1985), and Pavri(1988). We were unableto use exact replicationsof many of the instruments used in these studies because they were originally designed within the context of mainframebased systems. Instead,we developednineitems (using the work of other researchers for guidance), modified 14 items from previous scales, and used nine items directlyfrom prior studies. Table 1 lists, in an abbreviatedformat, all of the measurement scale items that were ultimatelyselected. The instructionsprovidedto the respondentand scale anchorsare also shown for one construct as an example. Social factors were operationalizedby asking respondents: (1) the proportion of their coworkerswho regularlyused a PC; (2) the extent to whichseniormanagementof the business unit supportedPC use; (3) the extent to whichthe individual'sboss supportedthe use of PCs forthe job; and (4) the extent to which the organization supportedthe use of PCs. These questionswere designed to tap the normsforPC use at the peer, superior,and organizationallevel. The fouritems were adopted from Pavri (1988). The affect construct was operationalizedwith three items:(1) PCs made workmoreinteresting; (2) workingwith PCs was fun; and (3) PCs were all right for some jobs but not the kind of job wanted(reversescored).The firsttwo itemswere taken from Howard(1985), while the thirdwas developedspecificallyforthis study.A five-point, Likert-typescale was used, with anchors ranging from stronglydisagree to strongly agree. The complexityscale includedstatements such as "workingwith PCs is complicated" and "it takes too long to learn how to use a PC to make it worththe effort."Respondents were asked to ratethe degree to whichthey agreedordisagreed witheach statement, on a scale from1 (strongly

130 MIS Quarterly/March1991

This content downloaded from 193.255.248.150 on Sun, 1 Feb 2015 04:36:34 AM All use subject to JSTOR Terms and Conditions

Utilizationof Personal Computers

Utilization of PCs ,.

Figure 2. Factors Influencing the Utilization of Personal Computers (adapted from the model proposed by Triandis, 1980)

disagree)to 5 (stronglyagree).The measurement scale was developed on the basis of workconducted by Howard (1985) and Pavri (1988). Neitherof theirinstrumentswere appropriatefor this study, however, and four items were therefore modifiedfor our purpose.

Forjob fit, respondents were asked to indicate the extent to whichthey agreed withstatements relatingto the potentialapplicationof PCs forjobrelated efficiency, effectiveness, quality, and overall performance improvement. The six statements used forthe measures, such as "for

MIS Quarterly/March 1991

This content downloaded from 193.255.248.150 on Sun, 1 Feb 2015 04:36:34 AM All use subject to JSTOR Terms and Conditions

131

of PersonalComputers Utilization

Table 1, Operationalization of Constructs Measure

Construct Social Factors

SF1. SF2. SF3. SF4.

The proportionof departmentalco-workerswho use a PC. The senior management of this business unit has been helpful in introducingPCs. My boss is very supportiveof PC use for my job. In general, the organizationhas supportedthe introductionof PCs. Affect

AF1. PCs make work more interesting. AF2. Workingwith a PC is fun. AF3. PCs are okay for some jobs but not the kind of job I want (reverse scored). Near-TermConsequences: Complexity C01. C02. C03. C04.

Using a PC takes too much time from my normalduties. Workingwith PCs is so complicated, it is difficultto understandwhat is going on. Using a PC involves too much time doing mechanical operations (e.g., data input). It takes too long to learn how to use a PC to make it worththe effort. Near-TermConsequences: Job Fit*

JF1. JF2. JF3. JF4. JF5. JF6. LT1. LT2. LT3. LT4. LT5. LT6. FC1. FC2. FC3. FC4. UT1. UT2. UT3.

Use of a PC will have no effect on the performanceof my job (reverse scored). Use of a PC can decrease the time needed for my importantjob responsibilities. Use of a PC can significantlyincrease the qualityof output of my job. Use of a PC can increase the effectiveness of performingjob tasks (e.g., analysis). A PC can increase the quantityof output for same amount of effort. Consideringall tasks, the general extent to which use of PC could assist on job. Long-Term Consequences PC will increase the level of challenge on my job. PC will increase the opportunityfor preferredfuturejob assignments. PC will increase the amount of varietyon my job. PC will increase the opportunityfor more meaningfulworko PC will increase the flexibilityof changing jobs. PC will increase the opportunityto gain job security. Facilitating Conditions Guidance is available to me in the selection of hardwareand software. A specific person (or group) is available for assistance with software difficulties. Specialized instructionconcerning the popularsoftware is available to me. A specific person (or group) is available for assistance with hardwaredifficulties. Utilization The intensityof job-relatedPC use (minutes per day, at work). The frequency of PC use. The diversityof software packages used for work(numberof packages). Use Use Use Use Use Use

of a of a of a of a of a of a

*The instructionsto the respondents for these items were: "Inthis section we wish to determinehow useful you believe a personal computercould be for your currentjob responsibilities.Please tell us how much you agree or disagree with each of the following statements (1 = strongly diagree; 2= somewhat disagree; 3= neitheragree nor disagree; 4= somewhat agree; 5= stronglyagree)." (Note: the instructionsand scale anchors differedfor other constructs.)

132 MIS Quarterly/March1991

This content downloaded from 193.255.248.150 on Sun, 1 Feb 2015 04:36:34 AM All use subject to JSTOR Terms and Conditions

Utilization of PersonalComputers

my currentposition,use of a PC could substantially decrease the time needed to performmy importantjob responsibilities,"were developed specificallyforthis study, using the workof Floyd (1986) and Schultzand Slevin (1975)as a guide. A five-point,Likert-type scale was employed,with anchorsrangingfromstronglydisagreeto strongly agree. The long-termconsequences of PC use were measuredusing a shortenedversionof the scale developed by Beatty (1986) for her study investigating the use of CAD/CAMsystems. Six items relevantto the use of PCs were identified, and a five-point scale was employed. The respondentswere asked to indicatewhetherthey thought that using a PC would decrease or increase each particular workoutcome,such as the opportunityforpreferredfuturejob assignments. The scale anchors were: (1) extreme decrease, (2) slight decrease, (3) neither increase nor decrease, (4) slight increase, and (5) extreme increase. Facilitatingconditions were operationalizedin terms of technical supportfor PC use, withfour questions adapted from Amoroso (1986). The respondents were asked to give their level of agreementor disagreementwithfourstatements: (1) guidanceis availableforthe selectionof hardware and software; (2) a specific person is availableforassistance withsoftwaredifficulties; (3) specialized instructionis availableforpopular software packages; and (4) a specific person is available for hardwareproblems. A five-point, Likert-typescale was employed, rangingfrom 1 (stronglydisagree) to 5 (stronglyagree). The utilizationof PCs scale was takenfromPavri (1988),who in turnbased his measurementscale on the work of Cheney (1984) and Raymond (1985).Threedimensionswere suggested forthe utilizationof PCs:(1) intensityof use, (2)frequency of use, and (3) diversityof softwarepackages used. Intensityof use was measured by giving the respondenta selectionof fivetimecategories for daily use rangingfromless than 15 minutes (category1) to morethan 120 minutes(category 5). Frequencyof use was operationalizedwith fourcategories, rangingfromless than once per week to several times a day. The diversityof use was calculated by counting those software packages forwhichthe responseforextentof use was "to some extent" or greater (3, 4, or 5 on a five-pointscale).

Sample The studywas conductedin a largemultinational manufacturing organization.The populationof interest was knowledge workers (defined as managers or professionals)who used a PC in their jobs. The study excluded individualswho were requiredto use a PC. Four hundredand fifty-five questionnaires were distributed throughout the organization. A total of 278 questionnaires were returned, for a gross response rate of 61 percent. Thirty-six respondents reported that they did not have access to a PC, nine respondentsdid notanswer questions concerning PC utilization,and 21 individualsindicatedthatthey were requiredto use a PC as part of their jobs. These respondents were removed, leaving a final sample of 212 knowledge workers(a net response rate of 47 percent). The respondents represented a wide varietyof job classificationsacross the nine divisions of this organization.Ninetypercent of the respondentswere male, and more than 70 percent held undergraduateor graduate degrees. The majorityof respondents were between 30 and 50 years of age (60 percent), with the remainder being split roughly between those youngerthan30 and olderthan50. About43 percent of respondentsclassified theirjob function as engineer or professional, 32 percent as managerialor supervisory,20 percent as staff specialist, and 5 percentas executive. Descriptive statisticsrelatingto the respondents'utilization of PCs are shown in Tables 2 and 3.

Procedure A contact person within the participating organization arranged for the distributionof questionnairesto managers and professionals who used PCs to some extent. The survey package contained a cover letter from the organizationas well as a cover letter from the researchers. All respondents were guaranteed confidentialityof individualresponses, and only summary statistics were returnedto the participatingorganization.One follow-upletterwas sent to non-respondents. The datawerecollectedusingthe DISKQmethod developed by Higgins,et al. (1987).This method replaces the traditionalpencil-and-papersurvey withan interactivequestionnairepresentedon a

MISQuarterly/March1991 133

This content downloaded from 193.255.248.150 on Sun, 1 Feb 2015 04:36:34 AM All use subject to JSTOR Terms and Conditions

Utilizationof Personal Computers

Table 2. Intensity and Diversity of Use of Personal Computers by Respondents Diversity of Use**

Measure

Intensity of Use*

Firstquartile; Median: Thirdquartile:

30 - 45 minutes 60 - 90 minutes > 120 minutes

1 package 2 packages 3 packages

Range:

< 15 min. to > 120 min.

1 to 5 packages

Table 3. Frequency of Use of Personal Computers by Respondents Measure

Frequency of Use

Percent

4 49 53

2% 23% 25%

106

50%

1. Once or twice per month 2. Once or twice per week 3. About once per day 4.

Several times per day

*Minutesper day, home and office use. PC. In their research on DISKQ,Higgins, et al. (1987) demonstrated that it was an efficient technique that obtained higher response rates than traditionalmethods. They also showed no statistical differences in answers to socially desirable and highly personal questions. The use of the DISKQmethod could raise the possibilityof non-response(selection)bias. Ifwe had includednon-users in the study, then selection bias woulddefinitelybe a concern. Since we includedonlythose who used PCs, however,we believe the potentialfor bias was minimized.

Data analysis To test the research hypotheses, partialleast squares (PLS) analysis was used. PLS is a regression-based technique, with roots in path analysis (Pedhazur, 1982; Wold, 1985). It has emerged as a powerful approach to studying causal models involvingmultipleconstructswith multiplemeasures. Ithas been used in marketing (Barclay,1986),organizationalbehavior(Higgins and Duxbury,forthcoming;Howelland Higgins, 1990), and MIS(Grant,1989; Rivardand Huff, 1988). Fornell(1982; 1984) refersto techniquessuch as PLS and its close cousin LISREL(Joreskogand Sorbom,1981)as second generationmultivariate analysis techniques. These second generation

134

**Numberof differentsoftware packages used. techniquesare superiorto traditionalregression and factoranalysis because the items measuring a construct(i.e., the measurementmodel)are assessed withinthe context of the theoretical model. Incontrast,computingfactorscores and importingthem intoa regressionmodelassumes thatthe scores are portable,an assumptionthat Fornell(1982) argues is not tenable. PLSallowsone to do a combinedregressionand principalcomponents factor analysis withinthe same statisticaltechnique. The factorstructure is a resultof the simultaneousassessment of item intercorrelations within the context of the hypothesized relationships between the constructs. If the PLS model is run with no hypothesizedpaths, the factorpatternmatrixis identicalto that obtained using principalcomponents analysis. The computer programused forthis analysiswas LVPLS1.6 (LatentVariables Path Analysis using Partial Least Squares), developed by Lohmoller(1981). For more informationon PLS, the interested reader can refer to Fornell(1982; 1984) and Wold (1985).

Test of measures The internalconsistency of the measurement model was assessed by computingCronbach's coefficientsare displayed alphas.These reliability for each of the constructsin Table 4. The Cron-

MIS Quarterly/March 1991

This content downloaded from 193.255.248.150 on Sun, 1 Feb 2015 04:36:34 AM All use subject to JSTOR Terms and Conditions

Utilizationof Personal Computers

Table 4. Factor Pattern Matrixof Measurement Scales Using PLS Analysis Measure

Construct 3 4

Cronbach's Alpha

1

2

C01 C02 C03 C04

.69 .80 .36 .59

.26 .17 .25 .19

.25 .04 .11 .11

.37 .37 .20 .26

.13 .18 .03 .06

.12 .05

.20 .23

.01 -.07

.10 .17

2. Job Fit

JF1 JF2 JF3 JF4 JF5 JF6

.30 .14 .18 .24 .31 .34

.44 .53 .80 .61 .49 .90

.12 .22 .25 .29 .24 .22

.34 .30 .42 .38 .39 .50

.27 .19 .13 .17 .14 .22

.12 .19 .11 .12 .11 .12

.17 .20 .31 .23 .19 .34

.82

3. Long-Term Consequences

LT1 LT2 LT3 LT4 LT5 LT6

.02 .16 .09 .18 .03 .10

.19 .15 .18 .23 .17 .13

.73 .53 .68 .79 .50 .45

.27 .20 .28 .29 .17 .23

.18 .16 .10 .17 .06 .10

.15 .00 .00 .05 -.08 .02

.18 .13 .17 .20 .12 .11

.76

.61

1. Complexity

4. Affect

5. Social Factors

5

6

7

AF1

.40

.40

.40

.65

.27

.16

.21

AF2

.28

.26

.21

.29

.12

.09

.09

AF3

.43

.48

.32

.95

.24

.09

.30

.60

SF1

.12

.11

.07

.22

.78

.18

.30

SF2

.18

.15

.19

.16

.50

.29

.10

SF3 SF4

.17 .05

.19 .15

.30 .15

.23 .08

.68 .54

.31 .58

.11 .11

6. Facilitating Conditions

FC1 FC2 FC3 FC4

.03 .06 .12 .04

.14 .12 .13 .09

.01 .10 .09 .12

.07 .10 .12 .12

.31 .39 .37 .37

.76 .88 .78 .90

.07 .06 .07 .14

.86

7. Utilization

UT1 UT2 UT3

.25 .23 .16

.35 .35 .13

.24 .19 .13

.29 .28 .14

.26 .25 .21

.10 .13 - .01

.90 .85 .51

.64

bach's alphavalues rangefrom.60 (forcomplexity)to .86 (forfacilitatingconditions).The lower reliabilitiesforthe scales can be partlyattributed to the small number of items in the scales because the calculationof Cronbach'salpha is affected by scale length. Given the exploratory

.65

natureof the study,the scales were deemed adequate to continuebut indicatethatfuturestudies should develop stronger measures. The primarycriterionfor discriminantvalidityis that each indicatormust load more highlyon its

MIS Quarterly/March 1991

This content downloaded from 193.255.248.150 on Sun, 1 Feb 2015 04:36:34 AM All use subject to JSTOR Terms and Conditions

135

Utilizationof Personal Computers

associatedconstructthanon any otherconstruct. Table 4 provides the factor pattern matrixthat shows the loadingsof each itemon allconstructs. Forthose unfamiliarwithPLS,the factorpattern matrixcan be interpretedin the same manneras principalcomponents factor analysis. Withone exception,the conditionsfordiscriminantvalidity were satisfied, indicating that the measures distinguishedbetweenconstructs.Thisexception occurredbetween two constructs,social factors and facilitatingconditions,where one item(SF4) loaded slightly higher on facilitatingconditions (see Table 4). To providea comparison with more traditional techniques, the results froman exploratoryfactoranalysis(usingprincipalcomponentsanalysis) are displayedinTable5. Several differencesare apparent.First,the analysis extractedeight factors instead of seven. The extra factor (column 6, items LT5,LT6)appears to be a component of long-term consequences, although an examinationof the questions did not indicate any substantivereasonforthis result.A moreserious problem occurred between social factors and facilitatingconditionswheretwo items(SF2, SF4) loadedmorehighlyon facilitatingconditionsthan on social factors.A thirditem, SF3, also loaded highlyon bothfactors.Thisresultis also apparent in Table 4, although it was not as serious. The problemis likelya result of our operationalizationof facilitatingconditionsas technicalsupport. Itappears that for many respondents, technical supportmay be indistinguishablefromorganizational support(or norms)for PC utilization. These results demonstrate some of the differences between PLS and traditionaltechniques. First, PLS performsa restrictedfactor analysis in the sense that the analysis is limited to the number of user-defined factors (constructs).This is similarto specifyingthe number of factors to be extracted using factor analysis (for example, using the "N=" option on the

FACTORprocedure of SAS). Second, both techniques identifyweaknesses in discriminant validitybetweenthe social factorsand facilitating conditionsconstructs, althoughthe exploratory factor analysis indicates that the weakness is moresevere. As a general comment,we believe that more weight should be given to the results producedby second generationtechniquessuch as PLS. Factor analysis is atheoretical in the sense that the factors are constructed without

136

referenceto the nomologicalnetworkin whichthe factorsare being used. However,factorsget their meaning from the empirical data and the theoretical model in which they are imbedded (Bagozzi and Phillips,1982). Thus, PLS, by explicitlyconsideringthe measurementand structural models simultaneously, provides a more inthe completeanalysisforthe inter-relationships model (Fornell,1982). Table 6 presents the intercorrelationsbetween the constructs obtained fromthe PLS analysis. This table indicates, as expected, some multicollinearitybetween social factors and facilitatingconditionsand between affectand the three perceived consequences constructs.

Results Table 7 shows the path coefficients, which are standardizedregression coefficients, generated fromthe PLS analysis. Jackknifing(Fornelland Barclay, 1983) was used to calculate the statistical significance levels for these coefficients. This is a non-parametrictechnique that does not requirethe usual assumptions of normalityassociated with regression models. The tests of hypotheses providemoderatesupport for our model of PC utilizationbased on Triandis'(1980) theory of behavior.Fourof the six hypothesizedrelationshipswere statistically significant(p < .01), and the amountof variance in utilizationexplained by the model was 24 percent. Support was found for Hypothesis 1, which postulatedthat social factorswouldpositivelyinfluence the utilizationof PCs (path = .22; p < .005). Hypothesis2 was not supported.The path fromaffect to utilizationwas .02, which was not statistically significant. For Hypothesis 3, as predicted,there was a significant,negative relationshipbetween perceptions about complexity of use and the utilizationof PCs (path = -.14; p < .01). Similarly,Hypothesis4, whichstated that jobfitwouldbe a predictorof utilization,was supportedby the results (path = .26; p < .005). For Hypothesis5, the path coefficientbetween longterm consequences and utilizationwas positive and statisticallysignificant(path = .10; p < .01), providingsupportforthe hypothesis.Facilitating conditions(Hypothesis6) had a small, negative whichwas notstatistically influenceon utilization, significant (path = -.04).

MIS Quarterly/March 1991

This content downloaded from 193.255.248.150 on Sun, 1 Feb 2015 04:36:34 AM All use subject to JSTOR Terms and Conditions

Utilizationof Personal Computers

Table 5. Principal Components Analysis of Measurement Scales Measure

Factor 4 5

1

2

3

JF5 JF6

.51 .67 .74 .77 .79 .74

.04 .17 .01 .07 .10 - .01

.03 -.02 .15 .22 .10 .03

.23 .11 -.01 .10 .24 .20

-.08 .15 .14 .05 .01 .19

-10 .19 .06 .05 .06 .04

Facilitating Conditions

FC1 FC2 FC3 FC4

.15 .07 .01 .05

.77 .87 .77 .82

-.21 .04 .07 .12

-.02 -.07 .05 -.02

.04 -.01 .04 .01

Long-Term

LT1 LT2 LT3 LT4 LT5 LT6

.17 - .01 .03

.08 .03 -.05

.61 .61 .70

-.10 .22 -.05

.18 .14

.04 -.12

.76 .47

.11 .01

.15

-.03

.22

Job Fit

Consequences

Complexity

Utilization

Affect

Social Factors

JF1 JF2 JF3 JF4

CO1 C02 C03 C04

.26

-.05 .28

.11

.11

.14

-.03

6

7

8

.01

-.08

.46

-.08

.20 .06

.12 .09

.13

.01

.16

.14

.14 -.06 -.13 -.10

.11 .06 .01 .07

-.04 .04 .06 .01

.04 - .01 .08

.10 .33 .07

.06 .04

.04 02

.15

.08

.10 .08

-.02 '.66'

.05

.741

.08 .03 .12

- .07 -.11 -.06

.26

- .08

.65

.06

.03

-.06

.48

.25

.07

.02

-.03

.61

.05

.01

.03

.73

.03

.01

.05

.04 .04

.01

.01

.20 -.38

.12 .34

.68 .76

.23 -.01

-.12

-.05

.16 -.12

.42 -.02

.05

UT1 UT2

.25 .11

.03 .06

.09 .08

.14 .01

.83 .83

UT3

.07

- .03

.21

.20

.49

AF1 AF2

.27 .12

.13 .10

.28 .15

.08 .21

.10 .05

AF3

.32

.19

.31

.20

-.02

.41

.26

SF1

.17 .05

-.02 - .08

-.11 .07

.17 .07

-.07 .55

.10 .17

.69 .40

.02

.41

SF2 SF3 SF4

-.07

.09 .43

-.02

.10 .14

.05

.34

.32

.17

.03

.12

.03

.72

.01

.02

.01

.38

In this study, a theory proposed by Triandis (1980)was adoptedas a basis forexaminingthe strength of differentcomponents of expected consequences of use on PC utilization,as well

-.09

.20

as the relativeinfluenceof social factors, affect, and facilitating conditions. Specifically, the findingsshowed that social factors, complexity, job fit, and long-termconsequences had significant effects on PC use. There was no evidence

MIS Quarterly/March 1991

This content downloaded from 193.255.248.150 on Sun, 1 Feb 2015 04:36:34 AM All use subject to JSTOR Terms and Conditions

137

Utilizationof Personal Computers

Table 6. Intercorrelations Between Constructs Construct

1

intercorrelations 4 5 3

2

7

6

1.00

1. Complexity 2. Job fit

.28

1.00

3. Long-TermConsequences 4. Affect

.17

.28

1.00

.48

.52

.39

1.00

5. Social Factors

.19

.21

.24

.28

1.00

6. FacilitatingConditions 7. Utilization

.07

.14

.11

.13

.43

1.00

.28

.38

,25

.32

.31

.11

1.00

Table 7. Tests of Hypotheses Based on Path Coefficients Path Coefficient

Hypothesis H1. Social Factors to Utilization

.22**

H2. Affect to Utilization

.02 -.14**

H3. Near-TermComplexityto Utilization H4. Near-TermJob Fit to Utilization

.26** .10*

H5. Long-TermConsequences to Utilization H6. FacilitatingConditionsto Utilization

-.04

R2 = .24 *p < .01;*p

p < .005

thataffectand facilitatingconditions(as defined) influenced PC use. The modelproposedby Triandis(1980) includes habits as one factor influencing behavior. As discussed, we did not include habits withinthis study because there was not a clear distinction between the independentvariable(habitsof PC use) and the dependent variable(PC use). Undoubtedly,the amountof variance explained in utilizationwouldhave increasedif we could have operationalizedhabits without introducingproblems of discriminant validity. We can be reasonably confident, however, that withinthe context of this study, factors other than habits have a significantinfluence on PC utilization. This study was an initialtest of Triandis'theory withinthe IS context. Several limitationsof the work are apparent and must be considered in

138

futuretests of this theory. First,the respondents were from one organization. Hence, the of these resultsto otherorganizageneralizability tions remainsto be determined.A second limitation is that utilizationwas operationalizedbased on the perceptionsof our respondents. A better approach would have been to obtain precise usage statisticsthroughan electronicmonitorto confirm or disconfirm the perceptions of the respondents.Thisapproachhas been suggested by many IS researchers (e.g., Robey, 1979). However, it is currentlydifficultto implement monitorson PCs, essentially makingit feasible only with mainframe-basedsystems. A third limitationrelates to our measurement model. First,there was the problemof discriminant validity between social factors and facilitatingconditions.Triandisdefines facilitating conditions as objective factors that make a

MIS Quarterly/March 1991

This content downloaded from 193.255.248.150 on Sun, 1 Feb 2015 04:36:34 AM All use subject to JSTOR Terms and Conditions

Utilizationof Personal Computers

behavioreasy to do. Forexample, one important facilitatingconditionis the ease withwhichan individualcan access a PC. Others include the ease withwhichsoftwareor hardwareupgrades can be purchased or the extent to which home computersare offeredas partof the job package. In retrospect, it appears that technical support is only one type of facilitatingcondition;others shouldhave been included.Infact,technicalsupportprovidedby the organizationappears to be closely related to social factors. Clearly, if the organizationhas positive normsconcerning PC use, it would be predisposed to providing technical support. Finally,the affect construct needs to be revisited.Whilewe believe that the items chosen in this study measure affect, they do not measure all possible facets of affect towardPC use. Thisscale needs to be bolstered by includingother items. Turningto the results, the relationshipbetween social factorsand utilizationis positiveand significant. This is consistent withTriandis'theory as well as the theoryof reasoned action proposed by Fishbeinand Azjen(1975). However,the nonsignificantrelationshipbetweenaffectand utilization is inconsistent with Triandis'theory. One possible explanationis that PCs do not evoke strong emotions, either positive or negative, among managers or professionals. If PCs are seen simply as tools, and not as technology to be liked or disliked,then affect would not have an impact.This resultis inconsistentwithDavis, et al. (1989),who finda significant,positiverelationshipbetween affect (attitude)and behavioral intentions.The differencesmay be a resultof the varying contexts being studied. Davis, et al.'s research lookedat affectto a particularsoftware package,whilewe examinedthe like/dislikecomponent to using PCs in general. Itis more likely, however, that the observed differences are a result of the different theoretical structures. Davis, et al. measured the indirectinfluenceof affect on usage through intentions, while our model employs a direct linkfromaffect to use. Althoughthe affective component of attitudes was not significantlyrelated to utilization,the cognitivecomponentsas measuredby perceived consequences were significantpredictorsof PC utilization.The negative relationshipbetween complexityand utilizationis consistentwithmany previousstudies (Davis, et al., 1989; Tornatzky

and Klein,1982).The strongpositiverelationship betweenjob fitand utilizationwas also expected and supports previous research (Davis, et al., 1989; Robey, 1979). The relationshipbetween long-termconsequences and utilizationwas also significantbut weaker than the path fromnearterm consequences to utilization.The weaker relationshipfor long-termconsequences is consistent with expectancy theory (Porter and Lawler,1968). A key component of this theory states that payoffs that occur closer to the behaviorare moremotivatingthan payoffsin the future.Thus, near-termconseqeunces wouldbe more likelyto motivate PC use than long-term consequences. Intheirtechnologyacceptance model, Davis, et al. (1989) find a stronger relationshipbetween perceived usefulness and utilizationthan between ease of use and utilization.Ourresultsare similarin that job fit (operationalizedsimilarto Davis, et al.'s perceived usefulness) was a strongerpredictorof utilizationthan complexity (similar to Davis, et al.'s ease of use when reversed scored). The stronger link between perceived usefulness and utilizationthan between affectand utilizationfoundby Davis,et al. is consistent with our results, although, as discussed previously,affect was not significant in our research. The small, negative relationship between facilitating conditions (operationalized as technicalsupport)and PC utilizationis inconsistent withmost previousstudies (Amoroso,1986; Jobber and Watts, 1986; Lucas, 1978). We cannot conclude, however,that there is no relationconditionsand utilization ship betweenfacilitating because we only measured one aspect of facilitatingconditions. As noted earlier, other measures of facilitatingconditionsshould have been used, such as ease of access to a PC and/orease of purchasingsoftwareor hardware Davis,et al. (1989)do not upgrades.Interestingly, find a significant effect of accessibility on behavior in their study. They reasoned that accessibilitywas notan issue forthe respondents in their research. This suggests that the operationalizationof facilitatingconditions must take context intoaccount. Inotherwords,if everyone has a PC on his or herdesk, then facilitatingconditionsoperationalizedas access wouldnot have any variance.

MIS Quarterly/March 1991

This content downloaded from 193.255.248.150 on Sun, 1 Feb 2015 04:36:34 AM All use subject to JSTOR Terms and Conditions

139

Utilizationof Personal Computers

Managerial Implications

concerning the difficultyof using a PC may decrease.

It is interestingto speculate whether any of the variables in this study could be influenced by managementaction. Inthis respect, Cheney, et al. (1986) identifyseveral factors that they believed influencethe "success" of end-usercomputing(EUC).These factorswere then classified as controllable,partiallycontrollable,and uncontrollable.Cheney, et al. suggest thatfocusing on controllableor partiallycontrollablefactorswould lead to a stronger potential to influence EUC

In addition to providing implications for managers, this study has providedpossibilities forfutureresearch.One such possibilityinvolves the role of experience with personal computers and the concept of the relativeinfluence of differentfactorschanging as experience changes. Triandis(1980)arguesthatexperienceinfluences expected consequences of behaviors. The influence of experience on expected consequences could be tested by comparingthe paths in the modelacross samples of experiencedand inexperiencedPC users. Triandis(1971)also proposes that experience changes the influenceof other factors, such as social norms. Withinthis context, it could be hypothesizedthat normsexert a strong influence on utilizationfor inexperienced users but less influenceon utilization for more experienced users.

success.

One partiallycontrollable factor may be the beliefs about the level of correspondence between job tasks and the PC environment(i.e., job fit).Specifically,communicationaimedat increasing the awareness of potentialapplicationsof PC technologyforcurrentjob positionsmayinfluence the perception of job fit. Similarly,education aimed at strengthening the expected consequences of using PCs, such as greater effectiveness and efficiency in performing job functions, could have a positive influence on utilization. Enhancingthe perceived relationshipbetween job tasks and PC utilizationcould be accomplished throughrole models. Inhis workon social learningtheory, Bandura(1986) postulates that role models could positivelyinfluenceinnovation adoption.Empiricalsupportforthis contentionis providedby a multitudeof field and case studies that link the presence of a champion-an individualwho enthusiasticallypromotesan innovation throughcriticalorganizationalstages-and innovationsuccess (Achilladelis,et al., 1971; Burgelman,1983; Ettlie,et al., 1984). Thus, encouraginghighlyregarded,visibleorganizational members to use PCs may be an effective way of championinguse throughoutthe organization. Social factorsmayalso be a partiallycontrollable factor; for example, it may be possible to influence norms by publicizingthe successes of early adopters of technology. Another partially controllable factor may be perceptionsaboutthe complexityof using a PCo Specifically, training aimed at reducing the perceived difficultyof using a PC could have a positive influence on actual use. This is also relatedto experience levels and learningcurves; as individualsgain experiencewithdifferentsoftware packages, for example, their perceptions

In conclusion, this research has provided a numberof contributions,the most significantbeing the testing of Triandis'(1980) theory within the IS context. The results suggest that future research on computer utilizationwithinthe IS context can productivelyuse Triandis'workas a frame of reference.

Acknowledgements The authors would like to thank the associate editor and the anonymous reviewers for their helpfulcomments on this article.

References Achilladelis,B., Jervis, P., and Robertson,A. A Studyof Success and Failurein IndustrialInnovation,Universityof Sussex Press, Sussex, England, 1971. Amoroso, D. Effectiveness of End-User DevelopedApplicationsin Organizations:An EmpiricalInvestigation,unpublisheddoctoral dissertation, Universityof Georgia, Athens, GA, 1986. Bagozzi, R. and Phillips,L. "Representingand Testing OrganizationalTheories: A Holistic Construal,"AdministrativeScience Quarterly (27:3), September 1982, pp. 459-489. Barclay,D. TheImpactof the OrganizationCon-

140 MIS Quarterly/March1991

This content downloaded from 193.255.248.150 on Sun, 1 Feb 2015 04:36:34 AM All use subject to JSTOR Terms and Conditions

Utilization of PersonalComputers

text on Conflictin OrganizationalBuying:A Systems View, unpublisheddoctoraldissertation,Universityof Michigan,AnnArbor,Ml, 1986. Beatty, C.A. The Implementation of TechnologicalChange:A FieldStudyof Computer Aided Design, unpublished doctoral dissertation, Universityof Western Ontario, London,Ontario,1986. Burgelman,R.A. "A Process Modelof Internal CorporateVenturingin the DiversifiedMajor Firm," Administrative Science Quarterly (28:2), June 1983, pp. 223-244. Burnkrant,R.E. and Page, T.J. "AnExamination of the Discriminantand PredictiveValidityof Fishbein'sBehavioralIntentionModel,"Journal of MarketingResearch (19:4), November 1982, pp. 550-561. Cheney, P.H. "Effects of Individual Characteristics,OrganizationalFactors and Task Characteristicson ComputerProgrammer Productivityand Job Satisfaction,"Information and Management, July 1984, pp. 209-214. Cheney, P.H., Mann, R., and Amoroso, D. FactorsAffectingthe Success "Organizational of End User Computing,"Journalof Management InformationSystems (3:1), Summer 1986. Cooper, R. and Zmud, R. "Information Technology Implementation Research: A TechnologicalDiffusionApproach,"Management Science (36:2), February 1990, pp. 123-139. Davis, F.D., Bagozzi, R.P., and Warsaw, P.R. "User Acceptance of ComputerTechnology: A Comparisonof Two TheoreticalModels," ManagementScience (35:8),August1989, pp. 983-1003. DeSanctis, G. "ExpectancyTheory as an Explanation of VoluntaryUse of a DecisionSupport System," Psychological Reports (52:1), February1983, pp. 247-260. Ettlie, J.E., Bridges, W.P., and O'Keefe, R.D. "OrganizationStrategy and StructuralDifferences for RadicalVersus IncrementalInnovation,"ManagementScience (30:6),June 1984, pp. 682-695. Fishbein,M.and Ajzen,I. Belief,Attitude,Intentionsand Behavior:An Introduction to Theory and Research, Addison-Wesley,Boston,MA, 1975. Floyd,S. A CausalModelof ManagerialWorksta-

tion Use, unpublisheddoctoraldissertation, Universityof Colorado,Boulder,CO, 1986. Fornell, C. (ed.). A Second Generation of Multivariate Analysis, Methods: Vol. 1, Praeger, New York,NY, 1982. Fornell,C. "ASecond Generationof Multivariate Analysis:Classificationof Methods and Implications for Marketing Research," unpublished paper, Graduate School of Business Administration, University of Michigan,Ann Arbor,Ml, April1984. Fornell,C. and Barclay,D. "Jackknifing:A Supplementto Lohmoller'sLVPLSProgram,"unpublished paper, Graduate School of Business Administration, University of Michigan,Ann Arbor,Ml, 1983. Goodhue, D. "IS Attitudes:TowardTheoretical and DefinitionClarity,"Data Base (19:3/4), Fall/Winter1988, pp. 6-15. Grant,R. "Buildingand Testing a Modelof an Information Technology's Impact," Proceedings of the Tenth InternationalConference on InformationSystems, December 1989, Boston, MA,pp. 173-184. Higgins, C.A. and Duxbury,L.E. "Work-Family Conflict in the Dual-Career Couple," Behaviorand HumanDecision Organizational Processes, forthcoming. Higgins,C.A.,Dimnik,T.P. and Greenwood,H.P. "TheDISKQSurveyMethod,"Journalof The Market Research Society (29:4), October 1987, pp. 437-445. Howard,G.S. "Measurementof Managers'Attitudes Towardthe Usefulness of Microcomputers: An Instrument for the MIS Researcher's Toolkit,"Proceedings of the AmericanInstituteof DecisionSciences Conference, Las Vegas, NV, November 1985. Howell,J.M. and Higgins,C.A. "Championsof Technological Innovation," Administrative Science Quarterly(35:2), June 1990, pp. 317-341. Jobber, D. and Watts,M. "BehaviouralAspects of MarketingInformationSystems," Omega International Journalof ManagementScience (14:1), 1986, pp. 69-79. Joreskog, K.G. and Sorbom, D. LISREL V: Analysisof LinearStructuralRelationshipsby Maximum Likelihood and Least Squares Methods, National EducationalResources, Chicago, IL, 1981. Keen, P.G.W. "MIS Research: Reference Disciplinesand a CumulativeTradition,"Pro-

MIS Quarterly/March1991 141

This content downloaded from 193.255.248.150 on Sun, 1 Feb 2015 04:36:34 AM All use subject to JSTOR Terms and Conditions

of PersonalComputers Utilization

Conference ceedings of the FirstInternational on InformationSystems, Philadelphia, PA, December 1980, pp. 9-18. Lohmoller,J.B. LVPLS1.6 ProgramManual:Latent Variables Path Analysis With Partial Least-Square Estimation, Universityof the Federal Armed Forces, Munich, Germany, 1981. Lucas, H. "Performanceand the Use of an Information System," Management Science (21:8), April1975, pp. 908-919. Lucas, H. "EmpiricalEvidencefora Descriptive Model of Implementation,"MIS Quarterly (2:2), June 1978, pp. 26-37. Pavri, F.N. An EmpiricalInvestigationof the FactorsContributing to Microcomputer Usage, unpublisheddoctoraldissertation,University of Western Ontario,London,Ontario,1988. Pedhazur,E.J.MultipleRegressionin Behavioral Research, Holt, Rinehartand Winston, Inc., New York, NY, 1982. Porter, L.W. and Lawler, E.E. ManagerialAttitudes and Performance, Irwin-Dorsey, Homewood, IL, 1968. Characteristics and Raymond,L. "Organizational MIS Success in the Context of the Small Business," MISQuarterly(8:1), March1985, pp. 37-52. Rivard,S. and Huff,S.L. "Factorsof Success for End User Computing,"Communicationsof the ACM(29:5), May 1988, pp. 486-501. Robey, D. "User Attitudesand ManagementInformationSystem Use," Academyof Management Journal (22:3), September 1979, pp. 466-474. Rogers, E.M.and Shoemaker,F.F. Communication of Innovations:A Cross-CulturalApproach, Free Press, New York,NY, 1971. Schewe, C.D. "The MIS User: An Exploratory BehavioralAnalysis,"Academy of Management Journal (19:4), December 1976, pp. 577-590. Schultz, R.L.and Slevin, D.P. "Implementation and OrganizationalValidity:An EmpiricalInvestigation"in ImplementingOperationand Research/ManagementScience, R.L.Schultz and D.P. Slevin (eds.), American Elsevier Publishing Co., New York, NY, 1975, pp. 153-182. Sugar, J. "AnAnalysisof the Relationshipof Attitudes and Behavior,"NSF Undergraduate Research Participation,FinalReport,University of Illinois,Champaign,IL, 1967.

Swanson,E.B. "MeasuringUserAttitudesin MIS Research: A Review," Omega International Journalof ManagementScience (10:2),1982, pp. 157-165. Tornatzky, L.G. and Klein, K.J. "Innovation Characteristics and Innovation AdoptionA Meta-Analysis of Findings," Implementation: IEEETransactionson EngineeringManagement (29:1), February1982, pp. 28-45. Triandis,H.C.AttitudeandAttitudeChange,John Wiley and Sons, Inc., New York, NY, 1971. Triandis,H.C. "Values, Attitudes,and Interpersonal Behavior,"Nebraska Symposium on Motivation, 1979: Beliefs, Attitudes, and Values,Universityof NebraskaPress, Lincoln, NE, 1980, pp. 195-259. Vroom,V. Workand Motivation,Wileyand Sons, Inc., New York,NY, 1964. Wold, H. "Systems Analysis by Partial Least Squares," in Measuringthe Unmeasurable, P. Nijkamp,H. Leitnerand N. Wrigley(eds.), Martinus Nijhoff Publishers, Boston, MA, 1985, pp. 221-251.

About the Authors Ronald L. Thompson is assistant professor in the Universityof Vermont'sSchool of Business Administration.He completed his Ph.D. in business administrationat the University of WesternOntarioand formerlytaught MISat the Universityof Calgary's Facultyof Managment. He also has several years' experience in various managementpositionswithina majorCanadian bank. He has presented papers at various national and internationalconferences and has publishedin Informationand Managementand the Journalof Creative Behaviour. His current researchinterestsfocus on the adoptionof informationtechnology by individualsand organizations and the measurement of the contribution of IS to the organization. Christopher A. Higgins is associate professor in the management science and information systems group of the School of Business Administration,the Universityof Western Ontario. He holds a Ph.D. and master's of mathematics from the Universityof Waterlooand has published articles in the Communicationsof the ACM, Administrative Science Quarterly,

142 MIS Quarterly/March1991

This content downloaded from 193.255.248.150 on Sun, 1 Feb 2015 04:36:34 AM All use subject to JSTOR Terms and Conditions

Utilization of PerrsonalComputers

Dynamics,TheJournalof Applied Organizational Psychology, Sloan ManagementReview, ACM Transactionson OfficeInformation Systems, and OfficeTechnologyand People. The primaryfocus of his researchis the impactof technologyon individualsand organizations.Specificresearchinterests include computerized monitoring, organizationalchampions,and the impactof personal computers on work and family life. Jane M. Howell is assistant professor in the organizationbehavior group of the School of

Business Administrationat the University of WesternOntario.Followingfouryears of human resource management experience, she earned her master'sdegree in psychologyat the University of WesternOntarioand then went on to complete a Ph.D. in business administrationat the Universityof BritishColumbia.She has published articles in AdministrativeScience Quarterly, Dynamics,and Business QuarterOrganizational ly. Hercurrentresearch focuses on champions and innovation,transformational leadership,and managerialuse of personal computers.

MIS Quarterly/March1991 143

This content downloaded from 193.255.248.150 on Sun, 1 Feb 2015 04:36:34 AM All use subject to JSTOR Terms and Conditions