9

Click here to load reader

Measurement of Job and Work Involvement

Embed Size (px)

DESCRIPTION

Artigo Científico

Citation preview

  • Journal of Applied Psychology1982, Vol. 67, No. 3, 341-349

    Copyright 1982 by the American Psychological Association, Inc.0021-9010/82/6703-0341S00.75

    Measurement of Job and Work Involvement

    Rabindra N. KanungoFaculty of Management, McGill University, Montreal, Quebec, Canada

    In view of the recent distinction between job and work involvement, this studydeveloped separate measures of the two constructs using three different tech-niques: semantic differential, questionnaire, and graphic. The conceptual basisof the two constructs and the reasons for developing new measures of the con-structs are discussed. Data collected from a heterogeneous sample of 703 em-ployees are analyzed to establish reliability, construct validity, and criterion-related validity of each measure. Relative effectiveness of the three techniquesused to measure the constructs are examined. Results reveal that questionnaireand graphic measures pass the tests of reliability and validity. Semantic differ-ential measures, however, have questionable validity for measuring work involve-ment. Possible uses of these new measures in future research are suggested.

    Past psychological research (Lodahl &Kejner, 1965; Rabinowitz & Hall, 1977;Saleh & Hosek, 1976) in the area of jobinvolvement is fraught with problems of con-ceptual ambiguities and measurement in-adequacies. The major source of conceptualambiguity lies in the use of the construct"job involvement" that carries excess mean-ing. Consequently, the techniques developedto measure the construct suffer from theproblems of construct validity. Without ad-equate construct validity, inferences basedon the data on job involvement provided byexisting instruments are often misleadingand difficult to interpret.

    The excess meanings of the job involve-ment construct can be identified in four dif-ferent ways. First, past conceptualizationsof the construct have confused the issue ofjob involvement with the issue of intrinsicmotivation on the job (Gorn & Kanungo,1980; Kanungo, 1981). The most widelyused measure of job involvement, developedby Lodahl and Kejner (1965), combinesitems representing the two issues. Someitems, such as, "I live, eat and breathe my

    This study was supported by a grant from the For-mation des Chercheures et d'Action Concertee, Gov-ernment of Quebec. I wish to thank Rajan Natarajanfor his assistance in data collection and analysis.

    Requests for reprints should be sent to Rabindra N.Kanungo, Faculty of Management, Samuel BronfmanBuilding, McGill University, 1001 Sherbrooke StreetWest, Montreal, Quebec, Canada H3A 1G5.

    job," represent a person's psychologicalidentification with the job. Other items, suchas, "sometimes I'd like to kick myself for themistakes I make in my work," represent aperson's intrinsic motivation at work for ful-filling self-esteem needs.

    Second, in dealing with the construct, re-searchers have confused the issue of identi-fying the antecedent conditions of job in-volvement with the issue of identifying thestate of job involvement and its subsequenteffects (Kanungo, 1979). Saleh and Hosek's(1976) scale, for instance, contains threecategories of items that describe (a) pre-sumed causal conditions of job involvement(e.g., "how much chance do you get to dothings your own way?"), (b) presumed ef-fects of job involvement (e.g., "I avoid takingon extra duties and responsibilities in mywork"), and (c) the state of job involvementitself (e.g., "the most important things I doare involved with my job"). A third way inwhich the construct carried extra meaningcan be seen in the description of job involve-ment as both a cognitive and a positiveemotional state of the individual. Lodahl andKejner's (1965) scale contains items thatrepresent these two meanings. Items such as,"the major satisfaction in my life comesfrom my job" and "the most importantthings that happen to me involve my work,"are descriptions of affective and cognitivestates, respectively.

    Finally, earlier conceptualizations of jobinvolvement have failed to distinguish two

    341

  • 342 RABINDRA N. KANUNGO

    different contexts in which an individual canshow personal involvement (Kanungo, 1981).The two contexts are (a) specific or partic-ular job context and (b) generalized workcontext. Involvement in a specific job is notthe same as involvement with work in gen-eral. The former is a belief descriptive of thepresent job and tends to be a function of howmuch the job can satisfy one's present needs.But involvement with work in general or thecentrality of work in one's life is a normativebelief about the value of work in one's life,and it is more a function of one's past cul-tural conditioning or socialization. Thus, jobinvolvement is a descriptive belief that iscontemporaneously caused whereas work in-volvement is a normative belief that is his-torically caused.

    Job involvement as a specific belief re-garding one's relationship with one's presentjob is also different from organizational com-mitment (Porter, Steers, Mowday, & Bou-lian, 1974), which refers to a general atti-tude toward an organization as a whole.Work involvement should also be distin-guished from the Protestant Ethic. Belief inthe centrality of work may result from Prot-estant-Ethic-type socialization, but the twoare not identical. Protestant Ethic may noteven be a necessary condition for work in-volvement to develop. It is conceivable thatwork involvement may result from socializa-tion that is not of the Protestant Ethic type.Researchers have confused work involve-ment with the Protestant Ethic in the samemanner as they have confused job involve-ment with intrinsic motivation. A lack ofclear conceptual distinction between job andwork involvement is reflected in the previ-ously developed scales of both job involve-ment (Lodahl & Kejner, 1965; Saleh &Hosek, 1976) and work values (Blood, 1969).These scales have used the words job andwork interchangably, and one is not surewhether the respondents view them synon-ymously. The validity and usefulness of aconceptual distinction between involvementin a particular job and identification withwork in general have been demonstrated re-cently by Gorn and Kanungo (1980).

    In view of the above construct validityproblems that are associated with past re-search, Kanungo (1979) argued for a refor-

    mulation of the construct of involvementthat eliminates the problems of excess mean-ing. According to such reformulation (Kan-ungo, 1979, 1981), involvement either in thecontext of a particular job or with work ingeneral can be viewed as a cognitive or beliefstate of psychological identification. An in-dividual's psychological identification witha particular job (or with work in general)in turn depends on (a) the saliency of his orher needs (both extrinsic and intrinsic) and(b) the perceptions he or she has about theneed-satisfying potentialities of the job (orwork). Viewed in this way, job involvementand work involvement cannot be measuredwith the existing instruments (Blood, 1969;Lodahl & Kejner, 1965; Saleh & Hosek,1976). This necessitates the development ofvalid and reliable new measures of job andwork involvement for use in future research.The present study is an attempt in this di-rection.

    Method

    Item Constructions for Involvement Scales

    For the purpose of obtaining distinct measures of spe-cific job involvement and of general work involvement,three different measurement formatsquestionnaire,semantic differential (Osgood, Suci, & Tanncnbaum,1957), and graphic techniqueswere used. Question-naire items that directly reflected a cognitive state ofpsychological identification were judged and compiledby 10 graduate students after a thorough search of theexisting measures of involvement and alienation in boththe psychological and sociological literature (e.g., Blau-ner, 1964; Clark, 1959; Dubin, 1956; Lodahl & Kejner,1965; Saleh & Hosek, 1976; Wollack, Goodale, Wijting,& Smith 1971). There was complete agreement by the10 judges on 12 items for inclusion in the Job Involve-ment Questionnaire (JIQ) and on 9 items for inclusionin the Work Involvement Questionnaire (WIQ). For theJIQ and WIQ items, 6-point agree-disagree responseformats were used. Subsequent items analyses resultedin dropping 2 items from the JIQ and 3 items from theWIQ scales because of their low interitem and item-total correlations. Thus the final scales contained 10 JIQitems (i.e., "The most important things that happen tome involve my present job"; "To me, my job is only asmall part of who I am"; "I am very much involvedpersonally in my job"; "I live, eat and breathe my job";"Most of my interests are centered around my job";"I have very strong ties with my present job which wouldbe very difficult to break"; "Usually I feel detached frommy job"; "Most of my personal life goals are job-ori-ented"; "I consider my job to be very central to myexistence"; "I like to be absorbed in my job most of thetime"). In addition, there were 6 WIQ items (i.e., "The

  • MEASUREMENT OF JOB AND WORK INVOLVEMENT 343

    most important things that happen in life involve work";"Work is something people should get involved in mostof the time"; "Work should be only a small part of one'slife"; "Work should be considered central to life"; "Inmy view, an individual's personal life goals should bework-oriented"; "Life is worth living only when peopleget absorbed in work").

    Another six graduate students, using available liter-ature and dictionaries for synonyms and antonyms,made an extensive search for key words that clearlyreflected the notion of psychological identification. Thisprocess yielded 11 bipolar items on which all the sixjudges agreed. These items with a 7-point response for-mat were used to construct Job Involvement SemanticDifferential (JISD) and Work Involvement SemanticDifferential (WISD) scales. Three items were rejectedfrom each scale on the basis of interitem and item-totalcorrelations. Thus each scale contained 8 items (involv-ing-noninvolving; important-unimportant; fundamen-tal-trivial; essential-nonessential; identifled-not iden-tified; attached-detached; integrated-nonintegrated;united-disunited). Finally, three graphic items repre-senting the notion of psychological identification wereprepared for use in each of the job and work contexts.Two of these items were finally selected after item anal-yses. In one item, two circles representing self and jobor work were presented with varying degrees of overlap(no overlap to complete overlap). In the other item, ahuman figure (representing self) and an office desk (rep-resenting job or work) were presented with varying dis-tances between them. These two items formed the JobInvolvement Graphic (JIG) scale in the job context andthe Work Involvement Graphic (WIG) scale in the workcontext. Both JIG and WIG items used a 7-point re-sponse format.

    Design of the Questionnaire

    A questionnaire1 containing three parts was designedfor the purpose of testing the reliability and validity ofthe newly constructed job and work involvement scales.Part 1 of the questionnaire contained JISD, JIQ, andJIG scale items. In addition, this part included two otherinstruments. One instrument measured the perceivedimportance of 15 job outcomes by asking the respon-dents' satisfaction with the same 15 job outcomes andoverall satisfaction with their present job. The validityand reliability of these two instruments were establishedin earlier studies (e.g., Corn & Kanungo, 1980; Kan-ungo, Gorn, & Dauderis, 1976), and the instrumentswere used to test the criterion-related concurrent valid-ity of the newly developed involvement scales. Part 2of the questionnaire contained the three work involve-ment scales (WISD, WIQ, and WIG). Part 3 of thequestionnaire was designed to determine the demo-graphic characteristics of the respondents.

    Subjects and Procedure

    The questionnaire was written in both French andEnglish following the translation-retranslation proce-dure (Brislin, Lonner, & Thorndike, 1973) and wasadministered to 900 full-time French- and English-speaking employees enrolled in evening extension courses

    in three different universities in Montreal. These em-ployees belonged to various industrial and governmentalorganizations in and around Montreal. The respondentswere told that participation in the study was optionaland that they could be assured of the confidentiality ofthe data. The questionnaire was completed during theclass hour in groups of varying sizes ranging from 40to 100. The final count revealed that 703 completedquestionnaires (184 in French and 519 in English) werereturned.

    A parallel study was conducted in two of the univer-sities to establish the test-retest reliabilities of the mea-sures included in the questionnaire. One evening exten-sion course (with approximately 50-55 full-timeemployees enrolled in it) that was offered in each of theuniversities was used for this purpose. The questionnairewas administered twice, 3 weeks apart. The respondentswere asked to put their identification numbers on thequestionnaire each time they were tested. Matching ofidentification members revealed that data from 63 re-pondents could be used in the test-retest analysis.

    Results

    Demographic Data

    Both the samples in the original and in thetest-retest study were heterogeneous in com-position. In the original sample, employeesbelonging to public and private sector or-ganizations were equally represented. Al-most half of the employees came from largeorganizations (with more than 700 employ-ees) and the other half came from small- ormedium-size organizations. Fifty-seven per-cent of the respondents were male and 43%were female, with a mean age of 28 years(SD = 6.66) for the total sample. There were37% French Canadian and 41% EnglishCanadian subjects, and the remaining 22%belonged to other ethnic groups. Forty per-cent of the respondents were married, and60% were single. Their education levelsranged from high school to advanced grad-uate degrees, and their income levels rangedfrom less than $10,000 to more than $40,000per year. Almost half the sample had or-ganizational tenure of 2 to 5 years. Of theother half of the sample, approximately 20%had less than 2 years, and 30% had morethan 5 years of organizational experience.The test-retest sample closely resembled theoriginal sample in its composition.

    1 The complete questionnaire is available upon re-quest from the author.

  • 344 RABINDRA N. KANUNGO

    Table 1Reliability Coefficients for Involvement andJob Satisfaction Scales

    Coefficient

    ScaleInternal Test-

    consistency retest

    Job involvement

    JISDJIQJIG

    WISDWIQWIG

    .81

    .87

    .70

    Work involvement

    ,83.75.68

    .74

    .85

    .82

    .78

    .67

    .67

    Job satisfaction

    15-item scale .89 .73Overall job satisfaction

    (single item) .87

    Note. Internal consistency (Cronbach a) coefficients arebased on data from 703 respondents. Test-retest coef-ficients are based on data from 63 respondents. JISD =Job Involvement Semantic Differential scale; JIQ = JobInvolvement Questionnaire; JIG = Job InvolvementGraphic scale; WISD = Work Involvement SemanticDifferential; WIQ = Work Involvement Questionnaire;WIG = Work Involvement Graphic scale.

    Empirical Properties of the InvolvementScales

    Item analyses. The 8 items included inthe JISD and WISD scales had medianitem-total correlations of .75 (range = .64-.82) and .74 (range = .71-.82), respectively.The median item-total correlation for the 10items in the JIQ scale was .68 (range = .59-.74). For the 6 items in the WIQ scale, themedian item-total correlation was .67(range = .54-.74). The intercorrelations foritems in the JIG and WIG scales were .70and .68, respectively. The means and stan-dard deviations for each of the six involve-ment scales were as follows: JISD: M =23.94, SD = 10.07; WISD: M = 20.30, SD= 8.28; JIG: M= 31.31, 57)= 10.61; WIQ:M= 20.70, SD= 5.97; JIG: M= 8.39;SD = 3.01; and WIG: M = 9.04, SD = 2.69.In the case of the JISD and WISD, lowerscores represented higher involvement. Forall other scales higher scores representedhigher involvement.

    Dimensionality of the scales. The scoreson job and work involvement items were fac-tor analyzed separately for each of the threemethods (semantic differential, question-naire, and graphic). The principal-compo-nent analysis followed by a varimax rotationwas used to arrive at factor solutions. Eachanalysis yielded two clear interpretable fac-tors of Job Involvement and Work Involve-ment. For semantic differential items, thefirst factor loaded highly on WISD items(item loadings ranged from .64 to .81 forWISD and from .01 to .15 for JISD), andthe second factor loaded highly on JISDitems (item loadings ranged from ,42 to .81for JISD and from .01 to .15 WISD). Thetwo factors, with eigenvalues of 5.64 and3.40, explained 56.5% of the total variance(and 93.3% of common variance). For ques-tionnaire items, the first factor reflected jobinvolvement (item loadings ranged from .44to .77 for JIQ and from .07 to . 17 for WIQ)and the second factor reflected work involve-ment (item loadings ranged from .40 to .73for WIQ and from .00 to .31 for JIQ). Thetwo factors, with eigenvalues of 5.15 and2.39, explained 47.2% of the total variance(and 93.6% of the common variance). Fi-nally, for the graphic items, again Job In-volvement and Work Involvement emergedas the first (item loadings were .95 to .70 forJIG and .09 and .19 for WIG) and the sec-ond (item loadings were ,70 and .89 for WIGand .21 and .06 for JIG) factors, respec-tively. The eigenvalues were 2.11 and 1.23,explaining 83.5% of the total variance (and100% of the common variance). Followingseparate analyses, item scores from all sixscales put together were again factor ana-lyzed. In spite of the introduction of methodvariance, a two-factor solution clearly re-vealed differential factor loadings on job andwork involvement items. The first factorloadings ranged from .70 to .40 on job andfrom .14 to .03 on work involvement items.The second factor loadings ranged from .79to .34 for work and from .25 to .00 for jobinvolvement items. The eigenvalues were9.68 and 5.18, explaining 41.3% of the totalvariance. These results clearly suggest dis-tinctiveness and unidimensionality of joband work involvement constructs.

  • MEASUREMENT OF JOB AND WORK INVOLVEMENT 345

    Reliabilities of the scales. The internalconsistency and test-retest reliabilities of thesix involvement scales and the job satisfac-tion measures are presented in Table 1. Thereliability coefficients ranged from .67 to .89,suggesting that both reliability of repeatedmeasurements and of internal consistency ofitems were adequate for these scales. Thecorrelation between the two job satisfactionmeasures as parallel form tests was .78.

    Convergent and discriminant validity ofinvolvement scales. Intercorrelations amongthe six involvement scales are presented inTable 2 in the form of a validational matrix(Campbell & Fiske, 1959). From the valid-ity diagonals (boxed correlations in Table 2)it can be seen that all the correlations werestatistically significant (p < .01), suggestingconvergent validity of the scales. However,the magnitude of the correlations suggestthat convergent validities of questionnaireand graphic scales measuring job involve-ment (r = .80) and work involvement (r =.69) were quite high. By comparison, se-mantic differential scales showed a moderateto very weak relationship to other scalesmeasuring job and work involvement. TheJISD scale showed a moderate relationshipto both the JIQ (r = -.33) and the JIG(r = -.44) scales. The WISD scale showed

    a very weak relationship to both the WIQ(r = -.12) and the WIG (r = -.24) scales.

    Assessment of discriminant validities re-quires that monotrait-heteromethod values(agreement between different ways of mea-suring the same trait) should exceed the het-erotrait-heteromethod values (agreementbetween different traits measured in differ-ent ways). Table 2 shows that every boxedcorrelation representing a rnonotrait-hetero-method value is higher than the adjacentnoncircled correlation representing a het-erotrait-heteromethod value.

    A second but more stringent criterion forthe assessment of discriminant validity re-quires that monotrait-heteromethod val-ues should exceed heterotrait-monomethod(agreement between different traits mea-sured the same way) values. This would in-dicate whether common trait variance isgreater than common method variance. Thiscriterion was satisfied in 67% of the cases.A closer inspection of Table 2 reveals thatthe semantic differential format did not meetthis criterion, particularly in measuring workinvolvement. The validity of the WISD scaleis questionable because the correlations be-tween the WISD and the WIQ and WIGinvolvement measures (r& = -.12 and -.24,respectively) did not exceed the correlations

    Table 2Multitrait-Muhimethod Matrix for Job and Work Involvement Scales

    Semanticdifferential

    Scale JISD WISD

    Semantic differential ^JISD ^\WISD 1 .28*\

    QuestionnaireJIQ [~-.33*| -.08WIQ .01 | -.12* |

    GraphicJIG | -.44* | -.09WIG .02 Ljl-24*J

    Questionnaire Graphic

    JIQ WIQ JIG WIG

    .29*""-\

    .so* i .21* r\.

    .33* [~69*n L-36*\^Note. N = 703. Correlations enclosed in boxes represent validity diagonal or monotrait-heteromethod values;correlations enclosed in triangles represent heterotrait-monomethod values. The remaining correlations representheterotrait-heteromethod values. Negative correlations are due to the reverse scoring of scales using semanticdifferential format. JISD = Job Involvement Semantic Differential scale; WISD = Work Involvement SemanticDifferential scale. JIQ = Job Involvement Questionnaire; WIQ = Work Involvement Questionnaire; JIG = JobInvolvement Graphic scale; WIG = Work Involvement Graphic scale.

  • 346 RABINDRA N. KANUNOO

    Table 3Analysis of Variance of Multitrait-Multimethod Matrix

    Source

    Respondent (R)R X TraitR X Method

    Error

    df

    702702

    1, 404

    1, 404

    MS

    2.94.78.48

    .66

    F

    4.45*1.18*

    .72

    VC

    .33

    .31

    .19

    ,38

    VCindex

    .46

    .44

    .33

    Note. VC = variance component.*p < .01.

    between the WISD and the JISD (/ = .28).In the case of the JISD scale, on the otherhand, the validity criterion is met becausecorrelations between the JISD and the JIQand JIG scales (rs = -.33 and -.44, respec-tively) were higher than the correlation be-tween the JISD and the WISD (r = .28).However, because of the moderate relation-ship of the JISD with the other two job in-volvement measures, its use should be dis-couraged.

    The convergent and discriminant validi-ties of the questionnaire and graphic scalesseem to be adequate. In fact, if one removesthe two semantic differential scales (JISDand WISD) from the matrix in Table 2, thepicture becomes clear. For the two job (JIQand JIG) and the two work (WIQ and WIG)involvement scales, the monotrait-hetero-method correlations were .80 and .69, re-spectively. These were substantially higherthan the monomethod-heterotrait correla-tions of .36 and .29 and heteromethod-het-erotrait correlations of .33 and .21, respec-tively.

    The multitrait and multimethod matrixpresented in Table 2 was further reanalyzedin a confirmatory fashion using the analysisof variance (ANOVA) three-way classifica-tion model suggested by Kavanagh, Mack-inney, and Wolins (1971). In this model, thepurpose of the analysis was to estimate threeeffects: (a) respondent (R) effect or the de-gree to which the alternative methods andtraits yield similar involvement scores oragreement within respondents; (b) Respon-dent X Trait (R X T) interaction effect orthe degree of rated discriminations on traitsby respondents; and (c) Respondent X

    Method (R X M) interaction effect or thedegree of disagreement on methods by re-spondents. The three effects corresponded toan estimate of overall convergent validity,discriminant validity, and method bias, re-spectively. The results of the ANOVA are pre-sented in Table 3. Both the main effect ofR and the R X T interaction effect were sta-tistically significant (p < .01), suggestingconfirmatory evidence for convergent anddiscriminant validities of the instruments.

    Following the suggestion of Kavanagh etal. (1971), the variance component and vari-ance component indexes for each of the threeeffects were also calculated to compare theirrelative impacts in the study. The results arealso presented in Table 3. Although the R XM interaction (method bias) effect was notstatistically significant, its variance compo-nent index showed that its relative impactwas not too small to ignore, presumably dueto the use of semantic differential formats.Separate estimations of variance compo-nents and variance component indexes formatrices with and without semantic differ-ential scales are presented in Table 4. Theresults clearly suggest that method bias anderror components were substantially re-duced; convergence and discriminant valid-ities were clearly enhanced when semanticdifferential scales were eliminated from thematrix.

    Criterion-related concurrent validity ofthe scales. The concurrent validity of thejob and work involvement scales was ex-amined by testing three theoretical predic-

    Table 4Comparison of Methods Used to MeasureInvolvement

    Scale

    Source

    SD and SD and Questionnairequestionnaire graphic and graphic

    VC VC VCVC index VC index VC index

    Respondent (R)R X TraitR X Method

    Error

    .25

    .18

    .24

    .51

    .32

    .26

    .32

    .33

    .29

    .27

    .39

    .45

    .42

    .40

    .53

    .44

    .05

    .21

    .71

    .67

    .19

    Note. SD = semantic differential; VC = variance component.

  • MEASUREMENT OF JOB AND WORK INVOLVEMENT 347

    tions derived from the motivational frame-work proposed by Kanungo (1979, 1981).First, it has been suggested that involvementin one's present job stems primarily from theperception of need-satisfying potential of thejob, whereas involvement with work in gen-eral is more a matter of past socialization.Thus it was expected that measures of jobinvolvement compared to measure of workinvolvement would be more strongly asso-ciated with measures of job satisfaction.

    Second, it has been proposed (Gorn &Kanungo, 1980; Kanungo, 1979, 1981) thatjob involvement is more a function of sat-isfaction of salient needs on the job. Thusit was expected that the job involvementmeasures would be more strongly associatedwith the satisfaction of salient rather thannonsalient needs. This pattern of relation-ships was not expected in the case of workinvolvement because work involvement is aculturally conditioned normative belief andis not directly dependent on present job sat-isfaction.

    Finally, it has been argued that employeeswith salient extrinsic needs are as likely tobe involved in their jobs as employees withsalient intrinsic needs, provided they haveequal levels of job satisfaction (Gorn & Kan-ungo, 1980). Thus, controlling for the levelsof job satisfaction, one would not expect anydifference between extrinsic and intrinsicoriented employees in their job involvement,On the same basis, however, it would be dif-ficult to predict how the two groups wouldreact to work involvement measures. Asmentioned before, work involvement is notdependent on present job satisfaction.

    It may be recalled that the questionnaireused in this study included a 15-item jobsatisfaction scale and a single-item overalljob satisfaction index. In order to examinethe relationship of job satisfaction with joband work involvement, the six involvementscales were correlated with the two measuresof satisfaction (results are shown in Table5). Tests of difference between the depen-dent correlations (McNemar, 1969, p. 158)clearly supported the first prediction that jobsatisfaction measures have a stronger rela-tionship to job involvement than to work in-volvement (see Table 5 for t values).

    The second prediction, that job involve-

    Table 5Correlation of Involvement Scales With JobSatisfaction Measures

    Scale

    JISDWISD/(700)

    JIQWIQ((700)JIGWIG((700)

    Job satisfactionscale

    -.27*-.015.96*

    .57*

    .12*12.18*

    .65*

    .24*12.61*

    Overall jobsatisfaction

    -.56*-.OS

    12.79*

    .43*

    .049.64*

    .55*

    .0613.94*

    Note. JISD = Job Involvement Semantic Differentialscale; WISD = Work Involvement Semantic Differen-tial scale; JIQ = Job Involvement Questionnaire; WIQ =Work Involvement Questionnaire; JIG = Job Involve-ment Graphic scale; WIG = Work Involvement Graphicscale.* p < .01.

    ment is a function of salient need satisfac-tion, was tested in the following manner. Theprecetved importance rankings of the 15 jobfactors were analyzed to determine the sa-liency of needs of the respondents. For everyrespondent, salient need was defined as thetwo job outcomes that were ranked by therespondent as 1st and 2nd in order of im-portance. The nonsalient need was definedas the two job outcomes that were rankedby the respondent as 14th and 15th in im-portance. The respondent's satisfaction scoreson the two salient job outcomes were addedto represent a single score for salient needsatisfaction. Likewise, the respondent's sat-isfaction scores on the two nonsalient joboutcomes were added to represent a singlescore for nonsalient need satisfaction. Eachof the six involvement scale scores were thencorrelated with the salient and nonsalientneed satisfaction scores. The results are pre-sented in Table 6. As expected, the job in-volvement measures correlated more stronglywith salient than with nonsalient need sat-isfaction. This pattern, however, was not ob-served for the work involvement measures.

    The third prediction regarding the job in-volvement of intrinsic and extrinsic orientedrespondents was tested by following a pro-cedure suggested by Gorn and Kanungo(1980). A group of intrinsic (n = 76) and

  • 348 RABINDRA N. KANUNGO

    Table 6Correlation of Involvement Scores With Salientand Nonsalient Need Satisfaction

    Scale

    JISDJIQJIGWISDWIQWIG

    Salientneed

    satisfaction

    -.24*.49*.56*

    -.02.10.18*

    Nonsalientneed

    satisfaction

    -.12*.31*.32*

    -.07.10.16*

    /(700)

    3.06*5.47*7.21*

    .05

    .00

    .51

    Note. JISD = Job Involvement Semantic Differentialscale; JIQ = Job Involvement Questionnaire; JIG = JobInvolvement Graphic scale; WISD = Work InvolvementSemantic Differential scale; WIQ = Work InvolvementQuestionnaire; WIG = Work Involvement Graphicscale.* p < .01.

    a group of extrinsic ( = 42) respondentswere chosen on the basis of their perceivedimportance of job outcome rankings. Eachmember of the intrinsic group perceived thetwo intrinsic outcomesinteresting natureof work and responsibilityas being the twomost important job outcomes. Each memberof the extrinsic group, on the other hand,perceived the two extrinsic outcomesmoneyand securityas the two most important joboutcomes. The choice of these outcomes torepresent intrinsic and extrinsic needs de-pended on two criteria. First, these outcomesare clearly distinguishable as intrinsic andextrinsic outcomes; second, within the in-trinsic and extrinsic categories, these out-comes were cited most frequently in theoverall sample as the first or second rankingjob outcome.

    Analysis of covariance was performed sep-arately on scores obtained from each of thesix involvement scales, treating job satisfac-tion score (on the 15-item scale) as the co-variate. The results revealed nonsignificantF values (p > .05) in each case. This con-firms the expectation that when controllingfor job satisfaction, job involvement of in-trinsic and extrinsic individuals will notdiffer.

    Discussion

    This study lends considerable support forthe conceptual distinction between job and

    work involvement proposed by previous re-searchers (Gorn & Kanungo, 1980; Wollacket al, 1971) and provides refinements in thedefinition and measurement of involvementin the two contexts. The results reveal thatall three job involvement scales and two ofthe work involvement scales (WIQ andWIG) have satisfactory psychometric prop-erties. The scales have reasonable levels ofinternal consistency and test-retest reliabil-ity. They seem to pass the tests of unidi-mensionality and of convergent and discrim-inant validity. The tests of criterion-relatedconcurrent validity of these measures alsoadd to their strength.

    This study explored the use of three dif-ferent formats for measuring job andwork involvement. Previous researchers havemainly used the questionnaire format, butfor cross-cultural and comparative researchuse of other formats such as graphic or pic-torial techniques may be more useful. Re-sults of this study show that the two graphicscales (JIG and WIG) correlate highly withtheir respective questionnaire scales (JIQand WIG), suggesting that the former caneasily act as substitute for the latter. For thecomprehension of the construct, graphicscales (as opposed to questionnaires) de-mand very little linguistic competence of therespondent. Hence, they might be more use-ful in cross-cultural and comparative re-search on involvement. They can also bemore effective when administered to less ed-ucated samples, or when time considerationsdo not allow administration of longer ques-tionnaires.

    In contrast to the graphic and question-naire formats, the use of the semantic dif-ferential format, particularly the WISDscale, seems to have questionable validity.Posttest interviews of some respondents re-vealed that they found it difficult to relateto the abstract 7-point scales using wordssuch as fundamental-trivial and essential-nonessential. Evaluating their present jobsin terms of these scales was relatively lessdifficult than evaluating the generalized no-tion of work. This suggests that the semanticdifferential format should be used with cau-tion even when measuring involvement in thepresent job context. Perhaps its usage shouldbe limited to only highly educated samples

  • MEASUREMENT OF JOB AND WORK INVOLVEMENT 349

    and in very specific contexts (Edwards &Waters, 1980).

    The new scales for measuring job andwork involvement can be used in future re-search to achieve several objectives. First,studies that aim at exploring the nature ofjob and work involvement within organiza-tions and at identifying the antecedent andconsequent conditions can use these instru-ments. Second, the instruments can be usedin studies that attempt to relate alienationand involvement in different spheres of life,such as work, family, and community. Third,tests of theoretical predictions derived fromexisting formulations on alienation and in-volvement (e.g., Kanungo, 1979) can be con-ducted more effectively with the use of thesescales. Finally, the use of these scales canestablish more meaningfully the cross-cul-tural validity and generalizability of findingsrelated to job and work involvement.

    References

    Blauner, R. Alienation and freedom: The factoryworker and his industry, Chicago: University of Chi-cago Press, 1964.

    Blood, M. R. Work values and job satisfaction. Journalof Applied Psychology, 1969, 53, 456-459.

    Brislin, R. W., Lonner, W. J., & Thorndike, R. M.Cross-cultural research methodology. New York:Wiley, 1973.

    Campbell, D. T., & Fiske, D. W. Convergent and dis-criminant validation by the multitrait-multtmethodmatrix. Psychological Bulletin, 1959, 16, 81-105.

    Clark, J. P. Measuring alienation within a social system.American Sociological Review, 1959, 24, 849-852.

    Dubin, R. Industrial workers' worlds; A study of thecentral life interests of industrial workers. SocialProblems, 1956,3, 131-142.

    Edwards, J. E., & Waters, L. K. Academic job involve-ment: Multiple measures and their correlates. Psy-chological Reports, 1980, 47, 1263-1266.

    Gorn, G. J., & Kanungo, R. N. Job involvement andmotivation: Are intrinsically motivated managersmore job involved? Organizational Behavior andHuman Performance, 1980, 26, 265-277.

    Kanungo, R. N. The concept of alienation and involve-ment revisited. Psychological Bulletin, 1979,56", 119-138.

    Kanungo, R. N. Work alienation and involvement:Problems and prospects. International Review of Ap-plied Psychology, 1981, 30, 1-15.

    Kanungo, R. N., Gorn, G. J., & Dauderis, J. J. Moti-vational orientation of Canadian anglophone andfrancophone managers. Canadian Journal of Behav-ioral Science, 1916,8, 107-121.

    Kavanagh, M. J., Mackinney, A. C., & Wolins, L, Is-sues in managerial performance: Multitrait-multi-method analysis of ratings. Psychological Bulletin,1971, 75, 34-49.

    Lodahl, T. M., & Kejner, M. The definition and mea-surement of job involvement. Journal of Applied Psy-chology, 1965, 49, 24-33.

    McNemar, Q. Psychological statistics. New York:Wiley, 1969.

    Osgood, C. E., Suci, G. J., & Tannenbaum, P. H. Themeasurement of meaning. Urbana: University of Il-linois Press, 1957.

    Porter, L. W., Steers, R. M., Mowday, R. T., & Boulian,P. V. Organizational commitment, job satisfaction,and turnover among psychiatric technicians. Journalof Applied Psychology, 1974, 59, 603-609.

    Rabinowitz, S., & Hall, D. T. Organizational researchon job involvement. Psychological Bulletin, 1977, 84,265-288.

    Saleh, S. D., & Hosek, J. Job involvement: Conceptsand measurements, Academy of Management Jour-nal, 1976, 19, 213-224.

    Wollack, S., Goodale, J. G., Wijting, J. P., & Smith,P. C, Development of the survey of work values. Jour-nal of Applied Psychology, 1971, 55, 331-338.

    Received July 27, 1981