Examining the causal order of job satisfaction and organizational commitment

Examining the causal order of job satisfaction and organizational commitment

Journal of Management, March, 1992 by Robert J. Vandenberg, Charles E. Lance

Four hypotheses have been advanced regarding the causal relationship between job satisfaction and organizational commitment: (a) satisfaction causes commitment, (b) commitment causes satisfaction, (c) satifaction and commitment are reciprocally related, and (d) no causal relationship exists between the two constructs. These four hypotheses were represented by separate structual models in a longitudinal research design. Using a sample of management information systems professionals, the models were tested using a combination of pseudo-generalized least squares, and full information maximum-likelihood estimation procedures. The latter procedures controlled for the unmeasured causal variables problem characterizing past studies. Results supported the commitment-causes-satisfaction model.
Job satisfaction is a pleasurable or positive emotional state resulting from the apprasial of one's job and job experiences (Locke, 1976). On the other hand, organizational commitment refers to the strength of individuals' indentification with and involvement in a particular organization (Mowday, Porter, & Steers, 1982).
Knowledge of the causal priority between these two constructs has both
theoretical
and practical implications (Curry, Wakefield, Price, & Mueller, 1986).
Typically,
job satisfaction is presented as causlally antecedent to organnizational
commitment
in conceptual models, such as those for turnover (e.g., Mowday et al,
1982) and work adjustment (e.g., Vandenberg & Scarpello, 1990). Further,
certain
interventions (e.g., job enrichment, realistic job previews) are designed to
increase
satifaction, and thus, commitment, under the rational that such increases result in enhanced employee psychological well-being, productivity, and
retention.
Despite its importance and the belief that job satisfaction (JS) is a cause oforganizational commitment (OC), the causal relationship between the two constructs is not clearly understood. First, little research has focused specifically on this issue. Second, research findings on this issue have been mixed (e.g., Bateman & Strasser, 1984; Curry et al., 1986; Dossett & Suszko, 1990; Farkas & Tetrick, 1989; Lance, 1991; Wlliams & Hazer, 1986). Third, the appropriateness of the methods used by the past researchers to address causal priority may be questioned.
It is critical to clarify the causal priority between JS and OC. It has been assumed that JS causally precedes OC (particularly in tests of theoretical models), but there have been few empirical tests of the validity of this assumption. In reality, some other causal sequence may more correctly describe the relationship between the two constructs. Thus, current theoritical models may misspecify the
JS-OC causal relationship. The purpose of this study is to provide a more
thorough assessment of the causal relationship between JS and OC than that provided through previous research.
Conceptual Models
The research literature on the job satisfaction - organizational commitment
(JS -OC) relationship suggests four competing models: (a) JS is causally antecedent to OC (JS [right arrow] OC, (b) OC is causally antecedent to JS (OC [right arrow] JS), (c) JS and OC are reciprocally related (JS [left arrow] [right arrow] OC), and (d) no causal relationship exists between JS and OC.
Model 1 (JS [right arrow] OC). Of the four models, the most widely accepted among researchers is that JS is causally antecedent to OC (Mowday et al.,1982). Perhaps the most prominent argument favoring this causal order is based upon the
notion that JS is determined by olny a subset of personal and organizational
factors
(e.g., job and job facets) that determine OC. JS is viewed as one of the
relatively
micro determinants of OC that is more macro in its orientation of the individual to the organization (Bluedorn, 1982; Farrell & Rusbult, 1981; Rusbult &
Farrell, 1983; Williams & Hazer, 1986). Further, proponents of this position
argue that job satisfaction reflects immediate affective reactions to the job and job facets (Locke, 1976). Thus, it forms soon after organizational entry. On the other hand, due to its macro orientation, organizational commitment is thought to develop more slowly, and after the individual possesses a firm understanding of not only the job and job facets but also of organizational goals and values, performance expectations and their consequences, and the implications of maintaining membership in the organization (Mowday et al., 1982). The latter understanding is not immediate and requires exposure to a variety of organizational components outside of the job. Consequently, OC is seen as forming and stabilizing sometime after organizational entry with the more immediate formaton of JS acting as one of its many determinants.
Model 2 (OC [right arrow] JS). Other researchers have argued that OC is
causally
antecedent to JS (Bateman & Strasser, 1984). Briefly, the rationale is based
upon a behavioral commitment perspective whereby the act of joining an arganization and the conditions surrounding that act (e.g., whether a person joined when
other employment opportunities were available) determine individuals' attitudinal commitment to the organization (Salancik & Pfeffer, 1978; Staw, 1980). Stronger attitudinal commitment to the organization results from joining that organization when other attractive employment alternatives were available. The latter process is a function of the degree or amount of cognitive dissonance experienced by the
individual after joining the organization. The stronger the cognitive
dissonance,
the greater the need to reduce it, and one means of dissonance reduction is
rationalizing
the choice (e.g., enhancing its positive aspects over those of the unchosen
alternatives, convincing oneself that he/she must really like or be
psychologically
attached to the choice given its selection over other choices, etc.).
Subsequently,
"commitment initiates a rationalizing process through which individuals 'make sense' of their current situation by developing attitudes (satisfaction) that
are consistent with their commitment" (Bateman & Strasser, 1984:97). OC is antecedent to JS, therefore, because it is the basis for developing other attitudes, such as JS.
Model 3 (JS [left arrow] [right arrow] OC). Both Model 1 and Model 2 above
are theoretically defensible, and ave some empirical support. This suggests, as others (Farkas & Tetrick, 1989; Lance, 1991; Williams & Hazer, 1986) have noted, that JS and OC may be reciprocally related. The expectation of a reciprocal relationship between the two constructs is also maintained on the basis of previous findings supporting reciprocal relationships between other similar constructs (e.g., JS and job perceptions,
James & Jones, 1980; James & Tetrick, 1986).
Model 4: No causal relationship. One argument for the absence of a JS - OC causal relation in that JS and OC are correlated due to the effects of common causal variables (Lance, 1991). For example, organizational factors, such as
job and role characteristics (James & James, 1989), and personality traits, such as locus of control (Spector, 1982) or positive-negative affectivity (Brief, Burke,
George, Robinson, & Webster, 1988), could influence both JS and OC. Thus,
observed
JS - OC correlations may reflect the fact that JS and OC share common
antecedents,
but are not causally related.
Empirical Evidence
Six studies have focused directly on testing tha causal relationship between JS and OC (Bateman & Strasser, 1984; Curry et al., 1986; Dossett & Suszko, 1990; Farkas & Tetrick, 1989; Lance, 1991; Williams & Hazer, 1986). In reanalysis of previously published cross-sectional data, Williams abd Hazer (1986) found support for Model 1 (JS [right arrow] OC). Moreover, although not focused specifically on the JS - OC relationship, Vandenberg and Scarpello (1990) also found support for Model 1 in cross-sectional examination of a group of newcomers and in a group of tenured employees. Using a cross-lagged panel design and longitudinal data, however, Bateman and Strasser's (1984) findings favored Model 2 (OC [right arrow] JS). Dossett ans Suszko (1990) also found support for Model 2 among a sample of 890 manufacturing employees, in a double cross-validation of structural equation models. In contrast, through a longtitudinal replication of Williams and Hazer's (1986) study, Farkas and Tetrick (1989:855) advocated Model 3 (JS [left arrow] [right arrow] OC) on the basis that their results
suggested "that commitment and satisfaction may be either cyclically or reciprocally related." Lance
(1991) also found support for Model 3 in a cross-sectional study of 1870
telecommunications
employees, but noted that the JS [left arrow] [right arrow] OC relationship,
though
reciprocal, was asymmetric. That is, JS had a much stronger effect on OC than did
OC on JS. finally, in a replication of Bateman and Strasser's (1984) study,
Curry et al. (1986) supported Model 4 (no causal relationship).
The above studies, however, may be criticized, both specifically and
collectively,
in terms of their ability to adequately test the JS - OC causal relation.
First,
Bateman and Strasser (1984) failed to correct their structural parameter
estimates
for measurement error. Second, the primary basis for establishing causal
priorities
between JS and OC in both Bateman and Strasser (1984) and Curry et al. (1986) was the presence of significant cross-lagged effects. Rogosa (1980:257) noted severe limitations to cross-lagged correlation designs for identifying causal priority among variables and has recommended that this design "be set aside as a dead end." Third, only Dossett and Suszko (1990), and Lance (1991) explicitly tested
Model 3 (JS [left arrow] [right arrow] OC). Consequently, other findings
favoring either the
JS [right arrow] OC or OC [right arrow] JS positions may have obscured reciprocal relations.
Finally, and most importantly, no research to date on the JS - OC relationship has addresed bias in structural parameters introduced by the unmeasured variables problem. According to James (1980:415) "the operative question is not whether ons has an unmeasured variables problem but rather the degree to which the unavoidable unmeasured variables problem biases estimates of path coefficients (a form of structural parameter) and provides a basis for alternative explanations of results."
There are three general approches for dealing with the unmeasured variables problem. One is randomization, but random assignment of individuals to JS or OC levels is not feasible. Second, all relevant causes of JS and OC could conceivably be identified and measured, but this is not a reasonable research expectation. In a third, statistical approach, the goal is to obtain estimates of the causal JS - OC relation while controlling statistically for the effects of unmeasured causal variables. James and Singh (1978) and Pindyck and Rubinfeld (1976) have shown that this type of statistical control is possible using longitudinal data and pseudo-generalized least squares estimation procedures. This was the approach taken in the present study.
In summary, this study was designed to meet four needs for research on the causal relationship between JS and OC: (a) correction of structural model parameter estimates for measurement error, (b) an explicit test of the reciprocal JS - OC relationship, (c) competitive tests among the four models of JS - OC causal relations described above, and (d) statistical correction of causal parameter estimates for biasing effects of unmeasured relevant causal viriables.
Participants were 100 management information systems professionals from a multinational, software research and deveolpment firm located in the southeast. The sample was randomly selected from a total population of 455 employees. although all 455 employees completed surveys at Time 1 (see explanation below), business constraints prevented use of the total population at Time 2. Of the selected individuals, 100% completed the survey at Time 2. The selected group did not differ significantly from the total population on any demographic characteristic, such as age, gender, year of education, and organizational and job tenure. Fifty-nine percent were female with a median age of 34 years. Median years of education were 16.
With respect to bonus equity, research supports the role of rewards (pay, promotion, recognition, benefits, ect.) and their equitable distribution as important events or conditions (as opposed to agent) affecting employees' satisfaction with the job and job facets (Locke, 1976:1321-1324). Porter and Lawler (1968) indicate that the perceived equity of rewards is a direct antecedent to job satisfaction. Others note, though, that one must distinguish between perceptions of distributive justice on the one hand and procedural justice on the other hand in determining effects on satisfaction (Folger & Konovsky, 189). Namely, equitable perceptions of distributive justice relate most strongly to satisfaction and do not relate to institutional perceptions, such as organizational commitment. As with the corporate values, the performance bonus system was quite salient to the current sample and had a strong link to the job from the employees' perspective. Bonuses in this organization were quite large, but their determination and distribution were secretly conducted. A strongly held belief among employees was that it was the particular job and little else that determined both whether a bonus was given and the bonus amount. Thus, a primary criterion used by employees to effictively evaluate their jobs in this organization was its associated (perceived) bonus amount. Job dissatisfaction occurred to the extent a poor bonus amount was received. The employees could transfer dissatisfaction. We felt that bonus equity for this sample had avery prominent role in determining the job satisfaction of these employees and thus disigned a bonus-equity measure to act as the instrument for job satisfaction. further, items for this measure reflected for the most part perceptions of distributive justice rather than procedural justice.
Value congruence. Value congruence was measured by summing scores on three items written around the stated corporate values. Items were anchored by 5-point scales of agreement. The items were "The company really means it when they say: (a) people are the key, (b) the customer is king, and (c) we are committed to excellence in everything we do." Coefficient alphas were .46 and .52 for Times 1 nad 2, respectively. The stability coefficient was .70.
FIML final estimation. LISREL VI (Joreskog & Sorbom, 1986) was used to obtain FIML estimates of the models' parameters. The FIML estimates were obtained by (a) using the sample covariance matrix, (b) using p-GLS estimates as initial estimates, and (c) restricting parameter estimates to be equal across measurement waves to reflect the assumption of stationarity. Further, to guard against overcorrecting parameter estimates for attenuation due to unreliability, values of the latent-to-manifest variable parameters were fixed to the highest estimated reliability coefficients obtained for each variable.
Less clear was the difference between Model 2 (OC[right arrow] JS) and model 3 (JS Mathematical Expression Omitted). Three considerations, however, favored
Model 2 over Model
3 with respect to its fit to the data. First, Model 2 was the more parismonious model of the two. Indeed, it provided the most parsimonous solution to the data. (TABULAR DATA OMITTED)
Second, the coefficients associated with the paths from JS to OC in the
reciprocal
model (Model 3) were not significantly different from zero. Thus, from a
theory-trimming
perspective, one could eliminate those paths from further consideration. The nonsignificant paths explains the greater parsimony of Model 2 over Model 3. Finally, Model 2 did not statistically differ from the upper-bound model [differece MATHEMATICAL EXPRESSION OMITTED]. In contrast, Model 3 resulted in asignificantly worse fit compared to the upper-bound model [MATHEMATICAL EXPRESSION OMITTED]. Thus, overall results supported a recursive model in which OC is causally antecedent to JS.
At the suggestion of one reviewer, we also conducted cross-lagged analyses by regressing Time 2 satisfaction and commitment measures (JST2 and OCQT2)respectively) on the Time 1 measures (JST1 and OCQT1)(see Bateman & Strasser, 1984). Although the limitations of this approach are now well known (Rogosa, 1980; Rogosa, 1988; Williams & Podsakoff, 1989:264-272), researchers have sought to establish causal prioties between variables on the basis of difference in the sizes of cross-lagged regression coefficients (e.g., Bateman & Strasser;
1983; Bateman & Strasser, 1984; Currey et al., 1986). In the present case,
differences in standardized stability coefficients (MATHEMATICAL EXPRESSION OMITTED) paralled correlational results shown in Table 1. Differences in cross-lagged coefficients (MATHEMATICAL EXPRESSION OMITTED) also support our earlier conclusions. Thus, although these results should be interpreted with caution (Rogosa, 1988), they nevertheless corroborate results of the direct tests of causal order shown in Table 2, which indicate support for Model 2 (OC RIGHT ARROW JS).
Final parameter estimates for Model 2 (OC RIGHT ARROW JS) are shown in Figure
2.2 Results in Figure 2 indicate that OC and bonus equity provided a
parsimonious account of employee JS in the present sample. This finding is supported by
(a) sturctural parameter estimates representing the links of JS with OC and
bonus
equity were strong and statistically significant, (b) residual variances were
small,
and (c) the correlation between unmeasured causes of JS (contained in
MATHEMATICAL
EXPRESSION OMITTED) also was small. The larger residual variances for OC and a statistically significantcorrelation between unmeasured causes of OC (MATHEMATICAL EXPRESSION OMITTED) indicated thatthere were other important determinants of OC not accounted for in this study. Recall however, that biasing effects due to unmeasured variables were statistically controlled for in estimates of the causal parameters. Finally, assumptions concerning the roles of valye congruence and bonus equity as instruments for OC and JS,
respectively, appeared to be reasonable ones given the strength and statiscal significance of their parameter estimates.
Discussion
In this study, we extended research on the JS - OC causal relation by: (a)
disattenuating structural parameter estimates for measurement error, (b) testing explicitly for a possible reciprocal relationship between JS and OC, (c) comparing the fits of four competing casual models of the JS - OC relationship, and (d) correcting parameter estimates for bias introduced by unmeasured relegant causal variables, Consistent with Bateman and Strasser, (1984), and Dossett and Suszko (1990), the results supported the idea that OC is casually atecedent to JS.
The third issue concerns testing more elaborate models of relationships
between
JS and OC. Future research should seek to identify additional instruments to identify models which test for possible reciprocal effects between JS and OC
One may never be completely assured that variables selected theoretically as
instrumental variables strictly statisfy all the rwquired assumptions. We mad every effort in the present study to identify unique instruments for JS and OC that had conceptual justification for this role. Empirically, there were at least three indications that we were successeful in our efforts: (a) correlations between insturments and appropriate endogenous variables were higher than correlations with the other endogenous variables (e.g., correlations of bonus equity with satisfaction were higher than with commitment, see Table 1), (b) estimated direct effect of instruments on respective endogenous variables were strong (see Figure 2), and (c) excellent overall fit for Model 2 (see Table 3) suggested the absence of specification errors (e.g., direct effects of bonus equity on commitment). Nevertheless, the dynamics of the presents sample's work environment may have inflated the
relationship between the instrument and the focal construct. Thus, we need to broaden the search for unipqe instruments of both constructs that are less
sample-specific and more generalizable across time, settings, and samples. Future research should also seek to identify appropriate common causal determinants of JS and OC and integrate them into more elaborate causal models (Mathieu & Zajac, 1990). This would permit more meaningful comparison of results across studies.
Finally, the present findings have both theoretical and practical implications
Most researchers accept the position that JS is causally antecedent to OC.
Based largely on the use of static correlations between commitment and satisfaction or the use of cross-sectional designs, this position finds some research support in tests of broad theoretical modwls (e.g., lance, 1991; Marsh & Mannari, 1977; Vandenberg & Scarpello, 1990; Williams & Hazer, 1986). However, of the studies that specifically focus on the JS - OC relationship and that have used stronger research designs, none have supported the satisfaction-causes-commitment model. Colectively, findings suggest that we need to crarefully exmamine our premises concerning the JS - OC relationship in our theoretical models. Also,
both Bateman and Strasser's (1984:109-110) and Curry et al.'s (1986:848)
conclusions
remain valid concerning the practical implications of identifying the
appropriate
causal relationship between the JS and OC. Namely, many interventions are based upon theoretical models that include commitment and satisfaction as component constructs. Whitout understandin the causal relationship between the two constructs, these interventions may not have intended effects.
Reference
Bateman, T.S., & Strasser, S. 1983. A cross-lagged regression test of the relationships between job tension and employee satisfaction. Journal of Applied Psycholoby, 68:439-445. Bateman, T.S., & Strasser, S. 1984. A longitudinal analysis of the antecedents of organizational comitment. Acamemy of Management Journal, 27:95-112.
Bluedorn, A.C. 1982. A unified model of turnover from organizations. Human
Relations, 35:135-153.
Brief, A.P., Burke, M.J., George, J.M., Robinson, B.S., & Webster, J. 1988.
Should negativity remain an unmeasured variable in the study of job stress? Journal of Applied Psychology, 73: 193-198. Curry, J.P., Wakefield, D.S., Price, J.L., & Mueller, C.W. 1986. On the causal order of job satisfaction
and organizational commitment. Academy of Management Journal, 29:847-858.
Dossett, D.J., & Suszko, M. 1990. Re-examing the causal direction between job
satisfaction and
organizational commintment. Paper presented at the annual meeting of the
Society for
industrial and Organizational Psychology, Miami, Fl.
Farkas, A.J., & Tetrick, L.E. 1989. A three-wave longitudinal analysis of the
causal ordering of satisfaction and commitment on turnover decisions. Journal of Applied Psychology, 74:855-868.
Farrell, D., & Rusbult, C.E. 1981. Exchange variables as predictors of job
satisfaction, job commitment and turnover: The impact of rewares, costs, alternatives and investments. Organizational Behavior and Human Performance, 27:78-95.
Folger, R., & Konovosky, M.A 1989. Effects of procedural and distributive
justice on reactions to pay raise decisions. Academy of Management Journal, 32:115-130.
James, L.A., & James,L.R. 1989. Intergrating work environment preceptions:
Explorations into the measurement of meaning. Journal of Applied Psychology, 74:739-751.
James,L.R., Mulaik, S.A., & Brett, J.M. 1982. Causal analysis: Assumptions,
models, and data. Beverly Hills, C.A: Sage.
James, L,R., & Singh, K. 1978. An introduction to the logic, assumptions, and
basic analytic procdedures of two-stage least squares. Psychological Bulletin, 85:1104-1122. James, L.R., & Terick, L.E., 1986. Confirmatory analytic tests of three causal models relating job perceptions to job satisfaction. Journal of Applied Psychology, 71:77-82.
Joreskog, K.G., & Srbom, D. 1986. LISREL VI: Analysis of linear structural
relationships by maxium liklihood. Chicago: National Ecucation Resources. Katz, K., & Kahan, R.L. 1978. The social psychology of organizations(2nd ed.). New York: Wiley.
Lance, C.E 1991. Evaluation of a structural model relating job satisfaction,
organizational commitment, and precursors to voluntary turnover. Multivariate Behavioral Research, 26: 137-162.
Locke, E.A., 1976. Nature and causes of job satisfaction. In M.D. Dunette, (ED)Handbook of industrial and organizational psychology: 1297-1349. Chicago: Rand McNally.
Marsh, R.M., & Mannari, H. 1977. Organizational commitment and turnover: A
prediction study. Administrative Science Quarterly, 22:57-75. Mathieu, J.E., & Zajac, D.M. 1990. A review and meta-analysis of the antecedentscorrelates, and consequences of organizational commitment. Psychological Bulletin, 108: 171-194.
Mowday, R.T., Porter, L.W., & Steers, R.M. 1982. Employee-organization
linkages: The psychology of commitment, absenteesim, and turnover. New York: Academic Pres. Pierce, J.L., & Ducham, R.B. 1976. Econometric models and economic forecasting.New York; McGraw-Hill.
Porter, L.W., & Lawler, E.E. 1968. Managerial attitudes and performance.
Homeood, II: Irwin-Dorsey. Rogosa, D. 1980. A critique of corss-lagged correlation. Psychological Bulletin88: 245-258.
Rogosa, D. (1988). Myths about longitudinal research. In K. W. Schaie, R.T.
Robert J. Vandenberg "Examining the causal order of job satisfaction and organizational commitment". Journal of Management. . FindArticles.com. 28 Jan. 2009. http://findarticles.com/p/articles/mi_m4256/is_n1_v18/ai_12289745