Organizational Behavior

^Academy of Management Joumai1984, Vol. 27, N o . 1 ,95-112.

A Longitudinal Analysisof the Antecedents

of Organizational Commitment^THOMAS S. BATEMAN

Texas A&M UniversitySTEPHEN STRASSER

The Ohio State University

From longitudinal data from 129 nursing departmentemployees, organizational commitment was found to beantecedent to job satisfaction rather than an outcome ofit. Furthermore, several other variables were found to becausally related to satisfaction but not commitment. Im-plications of unsubstantiated assumptions regardingcauses of commitment are discussed.

' For over a decade now organizational researchers have been studyingorganizational commitment in its relationships with various stituational char-acteristics, attitudes, and behaviors of employees. Commitment has beenoperationally defined in many ways, but one major stream of current re-search (Angle & Perry, 1981; Mowday, Steers, & Porter, 1979) has viewedthis construct as multidimensional in nature, involving an employee's loyaltyto the organization, willingness to exert effort on behalf of the organiza-tion, degree of goal and value congruency with the organization, and desireto maintain membership (Porter, Crampon, & Smith, 1976; Porter, Steers,Mowday, & Boulian, 1974). The study reported here attempts to add tothe knowledge of organizational commitment by providing the first longi-tudinal multivariate analysis aimed at deriving causal inferences regardinga number of its presumed antecedents.

Interest in studying organizational commitment has continued for a num-ber of reasons. It consistently has been shown to be related to: (a) employeebehaviors, such as job search activities, turnover, absenteeism and, to alesser extent, performance effectiveness (Abelson & Sheridan, 1981; Angle& Perry, 1981; Bluedorn, 1982; Farrell & Rusbult, 1981; Marsh & Man-nari, 1977; Morris & Sherman, 1981; Porter et al., 1976; Porter et al., 1974;

'The authors wish to thank Michael Abelson and two anonymous reviewers for their time and helpfulsuggestions on earlier drafts of this paper. This research was supported by research grants from TulaneUniversity's Bio-Medical Research Program and the School of Business Administration. The authorsof this paper wish to be considered as equal contributors.

95

96 Academy of Management Journal March

Steers, 1977); (b) attitudinal, affective, and cognitive constructs such asjob satisfaction, job involvement, and job tension (Hall & Schneider, 1972;Hrebiniak & Alutto, 1972; Porter et al., 1974; Stevens, Beyer, & Trice, 1978;Stone & Porter, 1976); (c) characteristics of the employee's job and role,including autonomy and responsibility (Koch & Steers, 1978), job varietyand task identity (Steers, 1977), and role confiict and ambiguity (Morris& Koch, 1979; Morris & Sherman, 1981); and (d) personal characteristicsof the employee, such as age, sex, need for achievement, and job tenure(Angle & Perry, 1981; Hall & Schneider, 1972; Hrebiniak & Alutto, 1972;Koch & Steers, 1978; Sheldon, 1971; Steers, 1977). This range of relation-ships, coupled with the belief that organizational commitment is a relativelystable attitude over time when compared to job satisfaction (Porter et al.,1974), suggests the importance of pursuing a thorough understanding ofthe operation of this major construct.

Research Problems

Despite the large number of studies that focus on organizational com-mitment, a number of research issues remain. First, past research has notempirically established the causal relationships between commitment andthose situational variables and attitudes presumed to be its antecedents.Second, the relationship of commitment to external environmental factors,such as perceived availability of other jobs, has gone relatively unstudied.Third, there remain a number of internal organizational variables—mostnotably, job tension and certain leadership behaviors—that need furtherintegration into organizational commitment research. These shortcomingsin the extant literature have important theoretical and applied significance,and they form the basis for this investigation.

The first research problem, that of the lack of clear causal specification,is due to cross-sectional designs employed by the vast majority of studies.Consequently, data analytic plans have relied primarily on zero-order cor-relations and multiple regression equations that are concurrent rather thanpredictive in nature. Only Porter and his colleagues (Porter et al., 1974,1976) have employed multiple measurements in longitudinal designs. How-ever, even in these investigations the number of variables studied was quitelimited, and there was no specification of any causal ordering, particular-ly with respect to the antecedents of commitment.

Typically, researchers have relied on explanations that derive from, forexample, exchange theory (March & Simon, 1958) and the concept of "sidebets" (Becker, 1960) to justify presumed causal linkages from characteristicsof the employee and his/her work environment to organizational commit-ment (Marsh & Mannari, 1977; Steers, 1977). Longitudinal research remainsto be conducted to confirm the antecedent nature of these variables. Jobsatisfaction, too, often is considered to be an attitudinal cause of commit-ment (Bluedorn, 1982; Price & Mueller, 1981) on the basis of Porter et al.'s(1974) argument that job satisfaction is a less stable and more rapidly formed

1984 Bateman and Strasser 97

construct than commitment. However, the validity of this perspective isnot altogether clear. In fact, there exists a viable alternative perspectivethat suggests that commitment to an organization may be a cause ratherthan a result of job satisfaction.

Building on the work of Bem (1967), Salancik and Pfeffer (1978) sug-gested that commitment to a course of action may determine subsequentattitudes. Commitment initiates a rationalizing process through which in-dividuals "make sense" of their current situation by developing attitudesthat are consistent with their commitment. Numerous laboratory and fieldstudies have confirmed this causal sequence. (See Kiesler, 1971, and Salancik,1977, for reviews.) In the specific case of commitment to an organization,a recent study (O'Reilly & Caldwell, 1981) showed that certain aspects ofthe initial decision to join an organization were related to subsequent com-mitment and satisfaction. Although these researchers defined commitmentas a behavior (i.e., job choice) rather than a cognitive construct, as in Porteret al. (1974), it is not unlikely that a cognitive outlook such as commit-ment is rationalized by subsequent attitudes of job satisfaction. Clearly,longitudinal empirical assessments aimed at establishing causality arerequisite.

The second research issue alluded to earlier relates to understanding theimpact of external environmental factors on organizational commitment.Only recently have researchers begun to look directly at this issue. Blue-dorn (1982) and O'Reilly and Caldwell (1981) found an inverse relation-ship between commitment and environmental job alternatives, and Farrelland Rusbult (1981) found no direct path between these two variables. Thepotential significance of an environmental alternatives variable should con-tinue to be pursued in any research on organizational commitment. It seemslikely that employees' perceptions of potential market alternatives will af-fect the value they place on their current organizational affiliation. Previouswritings have invoked market opportunities as a possible explanation forvariation in commitment levels (Sheldon, 1971), but few studies have em-pirically assessed the causal impact of this factor.

The third issue pertains to the scope of the internal organizational vari-ables previously studied in commitment research. For example, only Hre-biniak and Alutto (1972) have investigated the relationship between jobtension and commitment, yet the sustained interest in job tension and stress(Abelson & Sheridan, 1981; Brief, Schuler, & Van Sell, 1981) makes thisan important area for further inquiry. In addition, there has been only aminimal amount of integration between organizational commitment andresearch on leadership. To date the focus has been on the initiating struc-ture and consideration dimensions of leader behavior (Morris & Sherman,1981) to the exclusion of leader reward and punishment behaviors. Recentresearch (Podsakoff, Todor, & Skov, 1981; Sims, 1980) suggests that leaderreward and punishment behaviors are powerful predictors of other out-come variabies such as job satisfaction and performance. It would seemjustifiable to expect that the receipt of rewards and/or punishment from


one's organizational superior would also have an effect on commitment,and therefore to include these particular leader variables in any predictivemodel of organizational commitment.

The study reported in this paper attempts a resolution of these short-comings in the organizational commitment literature via a multivariate,longitudinal study of the presumed causes of commitment. A set of rele-vant organizational variables, personal characteristics and attitudes of theemployee, and perceptions of environmental alternatives were assessed attwo separate times in their relationships with commitment. Multiple regres-sion analyses were performed to validate the explanatory power of the totalset of predictor variables, as is customarily done in other studies. Subse-quently, in a methodological advance over previous studies, the longitudinalnature of the data was utilized in tests directed towards drawing inferencesregarding the presumed causal orderings surrounding the organizationalcommitment construct.

Method

Research Setting and Sample

This study was originally designed primarily as a predictive study of em-ployee turnover. It included the impact of organizational commitment, jobsatisfaction, and other predictor variables. It also was originally intended,however, that causal analysis focusing on organizational commitmentmodels would be conducted. Data were collected from nursing departmentemployees from four hospitals located in a large southern city. Two of thehospitals were investor owned, one was a university teaching hospital, andthe fourth was a Veterans' Administration hospital. Self-administered ques-tionnaire responses were solicited from all nursing-related jobs, includingnurses' aides, licensed practical nurses (LPNs), and registered nurses (RNs),representing all types of nursing units (emergency, pediatrics, surgery, etc.).As such, the total sample was representative across nursing department tasksas well as types of hospitals.

Procedure

Questionnaires were originally distributed to approximately 900 nursingemployees. Included with each questionnaire was a cover letter explainingthe importance of the research, the support of hospital administration, andthe strict confidentiality of the responses. Participants were asked to com-plete the questionnaire on their own time and return it, with the last fourdigits of their Social Security numbers as a means by which only the re-searchers would identify them, to a collection box at the sample sites. Thisbox was collected by research assistants on a daily basis over a period of10 days. A total of 374 identifiable questionnaires were completed at time1 ( i) for a response rate of 42 percent.


A second administration of the same questionnaire was conducted ap-proximately five months later ( 2). A cover letter explaining that this wasthe' final stage of the study was again distributed with the questionnaireto all nursing department employees at the four hospitals. Of the approxi-mately 900 questionnaires that were again distributed, 412 were returnedfor a time 2 response rate of 46 percent.

Respondent identification, using the partial Social Security number andother demographic information in several cases in which the numbers wereidentical, revealed that 129 employees had completed usable questionnairesat both times. The results reported in this paper are based on the responsesfrom only these repeat respondents (Bateman & Strasser, 1983). The averageage of the respondents was 35.2 years; the median tenure in nursing andin their current job was 10.83 and 3.85 years, respectively; and 95 percentof the participants were female.

In an effort to assess the representativeness of the repeated measures groupto the rest of the study sample, /-test comparisons were run on a set ofstudy variables. Here, the repeated measures group (« = 129) was comparedto the one-time respondents at ti and 2 (total /i = 528). No differencesacross these three groups would indicate that the repeated measures sam-ple, on which causal analyses were to be conducted, would be representa-tive of over 70 percent (129-1-528) of the total study population of 900.Using a two-tailed ^test at a = .05, the repeated measures group and theone time respondents showed no differences in age, career tenure, satis-faction, commitment, centralization/decentralization, stress, or need forachievement at either ti or tz. The only significant difference to emergewas job tenure at ti but not at t2. Although comparisons of the repeatedmeasures group to the nonrespondents was not possible, the available datastrongly suggest that the repeated measures group is quite similar to a largeproportion of the study population.

Measures

The questionnaires measured 13 variables. The focal outcome variablewas organizational commitment. The other 12 included 4 demographic and8 nondemographic predictors.

Organizational commitment was measured with a 15-item scale developedby Porter et al. (1974). The psychometric properties of this scale have beenthoroughly investigated (Mowday et al., 1979). In the present study, thescale demonstrated internal consistency reliabilities of a = .90 at time 1 anda = .89 at time 2, and a test-retest reliability of .65.

Leader reward and punishment behaviors were measured with the scalesoriginally developed by Johnson (1973). The psychometric properties andconceptual distinctiveness of these scales have been reported elsewhere (Sims& Szilagyi, 1975). In this study, Cronbach's alpha for leader reward behaviorwas .96 and .95 for the two waves of data, and the leader punishmentbehavior scale's alpha was .69 at time 1 and .68 at time 2. The test-retest

100 Academy of Management Journai March

reliabilities were .54 and .37 for leader reward and punishment behavior,respectively.

Job characteristics were measured with the Job Diagnostic Survey (Hack-man & Oldham, 1975). The reliability of the motivational potential scale(MPS) was a = .71 at time 1 and a = .71 at time 2, and the test-retest reli-ability was .48.

Centralization was measured with a 6-item scale of perceived participa-tion in decision making used by Morris and Steers (1980) as adapted fromVroom (1960). Coefficient alpha for the present study was .81 at time 1and .81 at time 2; the test-retest reliability was .61.

Need for achievement was measured with Steers' (1975) short 5-item scale.Cronbach's alpha at time 1 and time 2 were .43 and .60, respectively. Thetest-retest reliability was .69.

Perceived environmental alternatives were measured with three items as-sessing the chances of finding an acceptable alternative job, the desirabil-ity of another job in another organization, and a comparison of the desirabil-ity of another job to the present job. Internal reliabilities were a = .65 anda = .58 at times 1 and 2; the test-retest reliability was .53.

Job tension was assessed via the job related tension scale developed byKahn, Wolfe, Quinn, Snoek, and Rosenthal (1964). Items from the scaletap role stressors such as confiict, ambiguity, and overload. In this study,the coefficient alphas were .74 at time 1 and .73 at time 2; the test-retestreliability was .68.

Job satisfaction was measured with the Job Descriptive Index (JDI) de-veloped by Smith, Kendall, and Hulin (1969). The reliabilities and valid-ities for the several subscales (satisfaction with work, co-workers, supervi-sion, pay, and opportunities for promotion) are extremely well established,and the scale's usefulness for longitudinal research has been recommended(Schneider & Dachler, 1978). In the present study, internal reliabilities onthe overall (combined) scale were a = .64 at time 1 and a = .66 at time 2;the test-retest reliability was .68.

Age, job tenure, career tenure, and education level were recorded fromfour single-item self-report responses.

Analysis and Results

Relationships among the focal variables were assessed in several stages.Table 1 presents the means and standard deviations of the nondemographicvariables. These variables all proved to be relatively stable over the five-month time interval. Table 2 displays the static zero-order correlationsamong all variables at time 1 and time 2, and the test-retest reliabilities.As can be seen, many of the correlations are significant at both time periods.

Multiple regression was next employed as a test of the predictive powerof the variables presumed to be antecedent to commitment. This basic ap-proach, which indicates the relative importance of individual predictors whenthe others are statistically controlled for and which yields summary statistics


Table 1Means and Standard Deviations of the Study Variabies

(A^= 129)

Focal outcome variable:Organizational commitment

Nondemographic causal predictors:Leader reward behaviorLeader punishment behavior <•Job characteristicsCentralizationNeed for achievementEnvironmental alternativesJob tensionJob satisfaction

Time 1

X

72.4

79.530.3

124.320.225.4

9.525.231.4

S.D.

15.9

20.55.7

72.86.03.72.66.2,8.7

Time 2

X

71.4

81.332.0

120.420.826.09.2

25.431.8

S.D.

15.9

20.56.0

66.55.94.52.65.98.3

indicating overall explanatory power, typifies the approach taken in thepublished empirical tests of commitment models (Marsh & Mannari, 1977;Steers, 1977). In the present study several multiple regression analyses weremade possible by the collection of two waves of data. An analysis of thetime 1 data was followed by a second analysis applied to the time 2 dataas a second (replication) test of the power of the set of predictor variables.These static regressions were run for the purpose of comparing the overallpredictive power {R^) of the present set of variables to the predictivepower of other studies. Additionally, a final multiple regression was per-formed using the time 1 predictor variables as "independent" variables ina regression equation predicting the time 2 commitment values as the "de-pendent" variable. The impact of the predictors was assessed hierarchi-cally, after the time 1 value of commitment was controlled for. This time-lagged regression analysis has the effect of removing some method vari-ance inherent to single-administration self-report methodology. Further-more, because of the temporal ordering of the predictor (time 1) variablesand the criterion (time 2 commitment) variable, a multivariate test that ismore representative of the presumed causal priorities is created.

The two regressions on the two separate waves of static data yield multi-ple Rs of .75 {R^=.51) and .79 (i?^=.63), which compare favorably tosome studies (Marsh & Mannari, 1977) and rival others (Steers, 1977). Thereare two significant contributors to the prediction of organizational com-mitment that reliably arise in both regressions—job satisfaction (/i /3 = .24,;7<.O5; ti 0=.53, pK.OOl) and environmental alternatives (ti i8 = -.29,p<.001; 2 iS =-.28,/X.OOl). The time-lagged stepwise regression, how-ever, indicates the inability of these variables to predict, longitudinally,subsequent commitment. After entering the ti commitment variable topartial out the test-retest correlation, the entire set of variables did not addsignificantly to prediction (A/?^= .055, ^9,94= 1.42, p = .20). Marginal sta-tistical significance of leader reward behavior and job satisfaction are re-vealed when stepwise regression is used. When this analysis is repeated


I

ao

orre

l

UOISU3X

UOIID2HDJIU3Q

ID!tU3Wd

)U3UI3A3m3y

qof

vp mO O

1 1

-.01

-.11

S 31 1

O

.32*

'

*

19*

1

-.04

1

00

o1

.03

S

16*

1

**

1

.57*

^

**

32*

1

***

.61

<N

fn1

.60*

"

1

•

hhk1 1

**

1 1

.15*

.61*

"

00

*

1

**

1

.54*

'

1

*

57*

'**

-.3

6.1

4*

! ^

58*

1

•**

.63

00

.30*

"

ON « Q r-

|vi u — o 00"^ ^ ^ ' ^ *^

r " r

* * «* * * *

r- p- 00 v o l o l — ' 00 fn —

* * * • * * « * «

f n VO VO I I fS T f i - i v o — - T t r - r -o o » n p ^ f s ^ f s ^ ^ — . r N j ^ ^ — i

§

o l O s ^ f s r r s m— p p p p pr *

pr

«vo

I • «

2 «

« ^w < Z

EI -g .2i a S -2 E Ife fc I . | | 2 Ito M C C -S > iu a> V u A c o

• J .J U H »5 W U

•o-=

•§1

J'ooo8tS VI VI VI


using pairwise rather than listwise deletion of missing data to capture moreof the study sample, these findings become nonsignificant at/?:< .05. Again,when these predictors are entered as a block after accounting for the effectof t\ commitment, the result remains nonsignificant (A/?^= .061, 9,100 =1.48, p<.\l) when the pairwise analysis is used. Despite a high level ofconcurrent validity, the lack of true (longitudinal) predictive validity castssome doubt on the presumed causal impact of these variables.

The next stage of analysis addressed the issue of causal priorities betweenorganizational commitment and each of the variables that are predictedto be causes of commitment. Eight two-variable systems (each separate non-demographic predictor, coupled with commitment) were created and testedvia cross-lagged regression analysis (Rogosa, 1980). With this procedure,contextual information surrounding a given two-variable system is initial-ly provided by six correlations—two static correlations for time 1 and time2 (rx,y, and r^^y^, two autocorrelations (ry_^^^ and Tyjy ) that indicate the test-retest reliabilities or stabilities of the two variables, and two cross-laggedcorrelations (r jyj and /"yj j) between the time 1 value of one variable andthe subsequent time 2 value of the other variable.

Causal analysis and tests of spuriousness typically are conducted by sta-tistical correction and comparisons of the magnitudes of the cross-laggedcorrelations (Kenny, 1975). Recent criticism of cross-lagged correlations(Rogosa, 1980), however, alternatively suggests the use of structural regres-sion models for the analysis of longitudinal panel data. This approach wasapplied in the present study. For a given pair of variables, x and y, thecausal influence from A: to j ' is represented by the regression parameter ofthe path from x at time 1 to j at time 2. In like manner, the causal influ-ence from J to X is represented by the regression parameter of the pathfrom prior ^ to a subsequent x. Thus, where:

the parameters ^i and 71 represent the time-lagged influence of a variableon itself, and ^2 and 72 represent the time-lagged causal effects betweenX and y. Under the usual assumptions governing regression analysis Ginear-ity, additivity, etc.) a nonzero value of a relevant parameter is indicativeof a significant causal effect.

Essentially, then, the time-lagged explanatory power of a potentiallycausal variable is determined only after the "dependent" variable's time1 influence upon "itself" is first controlled for. An additional advantageto analyzing the data over two time periods is that some of the methodvariance of a single collection of self-report data is removed. Finally, thereader should be cautioned that these procedures do not ultimately "prove"causality. However, confidence in making causal inferences is substantial-ly strengthened if one of the pertinent regression parameters is significant.


and reciprocal causation also can be indicated if both ^2 and 72 aresignificant.

Table 3 indicates the cross-lagged correlations and regression coefficients(static correlations and test-retest reliabilities are found in Table 2). Althoughthe static correlations generally are highly significant, only two of the eightpairs of regression parameters show indication of any causal ordering.Leader punitive behavior is shown to have a negative causal impact on com-mitment, and commitment is indicated to have a positive causal impact onjob satisfaction. Furthermore, given the weakness of the leader punitivebehavior effect, especially within the context of unstable and nonnegativestatic correlations and low test-retest reliabilities of the leader punitive be-havior scale, in total there is little to argue for the presumed antecedentnature of this set of predictor variables. Particularly noteworthy, however,is that job satisfaction, most commonly considered a cause of organiza-tional commitment, is shown to result from commitment. Subanalyses onthe five facets of satisfaction revealed that only one—satisfaction withpay—did not show at least a trend in this direction. Satisfaction with op-portunities for promotion was significant at/7<.07; and the other threesubscales—satisfaction with work, co-workers, and supervision—were allstrongly significant resuhs of commitment.

Table 3Cross-Lagged Correlations and Regression Parameters for

Commitment and Predictor Variables

Predictor Variables (x)with Commitment (y)

Job satisfactionEnvironmental alternativesJob tensionNeed achievementCentralizationMotivating potential scoreLeader reward behaviorLeader punishment behavior

^Two-tailed ^tests of statistical*/>:S.O5

Cross-Lagged

.39*-.43*-.27*

.08

.22*

.27*

.26* >•

.01

significance.

Correlations

''yi'<2

.54*'-.39*'- .33*'

.09

.25*'

.22*'

.33*'

1 .

> •

>•

1*

>•

.01

Time-LaggedRegression

02

.02-.10

.00

.00-.11

.00-.14-.14*

Parameters^

72

.25**-.13-.04

.00-.03

.00

.04-.08

A final set of cross-lagged regression analyses was performed using sat-isfaction, rather than commitment, as the focal variable. This was donefor three a priori reasons and one post hoc reason. First, there are strongarguments for expecting satisfaction to result from many of the variablespresumed also to be antecedent, to commitment. Second, the strong rela-tionship between commitment and satisfaction is well established, and thepossibility that this may obscure other relationships with commitment wasreinforced by the multiple regressions in which satisfaction was such a


consistently powerful and significant predictor. Third, testing these causallinkages might afford more contextual information within which resultsfrom the commitment analyses could be more appropriately viewed. Finally,the results that reveal commitment as a cause of satisfaction suggest thepursuit of satisfaction as a relatively more important outcome variable thancommitment in this study.

Table 4 shows these results. Five of the eight two-variable analyses sug-gest job satisfaction as an outcome variable. Organizational commitmentand the motivating potential of the job have positive causal impacts; andjob tension, leader punitive behavior, and environmental alternatives havenegative impacts. Thus, the overall pattern of causal conclusions from theseresults suggests that several variables are antecedent not to commitmentbut to job satisfaction, and that commitment is one of the several causesof satisfaction. ^

Table 4Cross-Lagged Correlations and Regression Parameters for

Job Satisfaction and Predictor Variables

Predictor Variables (x)with Job Satisfaction (y)

Organizational commitmentEnvironmental alternativesJob tensionNeed achievementCentralizationMotivating potential scoreLeader reward behaviorLeader punishment behavior

Cross-Lagged

'''iy2

.54***-.36***-.43***

.11

.47***

.52***

.32***-.01

Correlations

^''2yi

.39***-.24***-.29***

.03

.36***

.33***

.40***-.04

Time-LaggedRegression Parameters'

P2

.25**-.14*-.23**

.10

.12

.18*-.14-.21**

72

.02-.05-.07

.02

.00-.05

.09-.15

°Two-tailed ^tests of statistical significance.

***p£.OOl

In order to assess the possibility of artifactual results in Tables 3 and4 due to aggregating these data across four hospitals, additional analyseswere conducted. For each statistically significant result reported, dummyvariables representing each hospital were entered into the regression equa-tion as blocking variables. The results were unaffected except for one minorexception: the gamma for leader punishment behavior in Table 4 changesfrom -.15 to -.19, now significant a.tp<.05.

Finally, the complete set of ti predictor variables were combined in atime-lagged multiple regression with 2 job satisfaction as the dependentvariable. After the initial entry of ti job satisfaction into the equation, theset of other ti variables added significantly to the prediction of subsequent( 2) satisfaction (A/? = .158, ^9,9, = 4.627, p<.0001). Significant individ-ual variables included (in order of entry) organizational commitment(j8 = .25, A/? = .O44, 2,98 = 8.64, /7<.O1); leader punishment behavior

106 Academy of Management Journai March

(/3 = -.22, Ai?^=.O43, 3.97 = 9.28, p<.01); job tension (/3 = -.17,A/?^=.O23, 4,98 = 5.09, /?<.O5); leader reward behavior (|8 = -.2O,A/?^=.O21, 5,95 = 4.73, /7<.O5); and motivating potential of the job(j3=.15, A/?^=.O15, F6,94 = 3.55, p<.01). The negative coefficient forleader reward behavior is apparently indicative of a (net) suppression ef-fect that can occur when the correlation among two predictor variables (inthis case ti satisfaction and leader reward behavior) and the dependentvariable ( 2 satisfaction) are all positive (Cohen & Cohen, 1975). Multi-coUinearity among the predictors served to diminish the predictive powerof environmental alternatives. However, these results are not contrary tothe significant bivariate findings in Table 4; they serve only to indicate therelative power of the causal predictors.

Discussion

The multiple regression results clearly show that the presumed predic-tors of commitment account for a large proportion of the variance in thisconstruct. The explanatory power of the variables used in this study, asnoted earlier, compares quite favorably with previous organizational com-mitment research (Marsh & Mannari, 1977; Morris & Sherman, 1981; Steers,1977). The stability of this set of predictors over time is also evident. Ofthe 12 variables investigated, however, only 2 account for most ofthe vari-ance in organizational commitment: overall job satisfaction and environ-mental alternatives. This outcome is reliably demonstrated in both the time1 and time 2 regressions, and it is consistent with the findings of Farrelland Rusbult (1981) in the survey component of their research.

These results differ from previous research, in part, in terms of the minorrole played by personal characteristics of the employee in explaining vari-ation in commitment. Age and education, significant predictors in the multi-variate models offered by Morris and Sherman (1981) and Steers (1977),are not significant in this investigation when the other predictors are sta-tistically controlled. Nor is need for achievement, which arises as an im-portant variable in Steer's (1977) study, a significant contributor to the ex-plained variance here. However, this may simply be due to weak psycho-metric properties of the scale. Furthermore, age has significant zero-ordercorrelations with commitment at both time periods. The subsequent lackof significance in the multiple regression analyses probably is due to thestrong predictive power of satisfaction and environmental alternatives, whichwere not in the models of Morris and Sherman (1981) and Steers (1977).

The results of the initial time-lagged multiple regression analyses indicatea relative inability to predict organizational commitment with variables otherthan prior commitment, suggesting that the theoretical causal presumptionsbehind the set of predictor variables may be invalid. The cross-lagged bi-variate regression results shed additional light on this suspicion of incor-rect causal specification. Two important findings arise. First, with the


possible exception of leader punitive behavior, none of the predictorsemerges as being antecedent to organizational commitment. This finding(nonfmding) runs counter to what previous researchers have believed tobe the causal priorities among these variables—theoretical models of or-ganizational commitment validated only by static correlational analysis(Marsh & Mannari, 1977; Steers, 1977) would have predicted otherwise.Second, there is evidence that overall satisfaction is not a cause of com-mitment but rather a result of it. Although the marginal relationship (A/? )revealed in the stepwise multiple regression of commitment on the predic-tor variables suggests a need for some restraint and further test, the un-equivocal result of the bivariate analysis brings into question argumentsmade by previous researchers who view commitment as a time-lagged out-come of employee satisfaction (Marsh & Mannari, 1977; Porter et al., 1974).Perhaps employees become committed to the organization before attitudesof satisfaction can meaningfully emerge. This is consistent with previouswork that suggests that the employee may come to develop attitudes thatare consistent with his or her existing level of commitment to the employ-ing organization (Staw, 1980). In fact, organizational commitment may startquite early, perhaps even as a function of pre-entry experiences (O'Reilly& Caldwell, 1981; Schein, 1968). To the extent that this occurs, the stabil-ity of the commitment construct noted by Porter et al. (1974) also becomesmore understandable.

When comparing the cross-lagged bivariate regressions surrounding or-ganizational commitment with those surrounding job satisfaction (Tables3 and 4), at least two primary points surface. First, the comparison of jobsatisfaction and commitment as dependent variables is useful because itsuggests that the lack of causal findings pertaining to commitment is notdue to inherent limitations of the methodological and analytical proceduresemployed. Had a similar pattern of insignificant results emerged with thecross-lagged bivariate analyses surrounding job satisfaction, this possibil-ity would have been more compelling. As it was, though, 5 of the 8 bi-variate regressions (more accurately, 8 of 12—from subanalyses) were sta-tistically significant.

Nonetheless, one cannot altogether rule out the possibility that commit-ment may have been shown to be caused by some of the study variablesif some different time lag had been used. Perhaps five months is more ap-propriate for uncovering the causes of satisfaction than of commitment.The choice of an "appropriate" lag is an empirical question, but little hasbeen written about guidelines in choosing the best lag. However, Sims andWilkerson (1977) found that the cross-lagged test was fairly robust withrespect to "missing" the appropriate lag. This conclusion, coupled withthe several significant outcomes (one of which included commitment as acausal variable), helps to provide confidence in the utility of the study design.

The second point is based on the conceptual overlap between existingmodels of job satisfaction and of commitment. A general trend underly-ing past research into the antecedents of organizational commitment and


of job satisfaction is that characteristics of the job, work setting, and theindividual are presumed to be causal to the formation of each of these at-titudes. For example, Hackman and Lawler (1971) report a positive rela-tionship between job characteristics and employee satisfaction for employeeswith higher need strength. Similarly, Steers (1977), Buchanan (1974), andKoch and Steers (1978) show job characteristics such as task identity andjob challenge to be related to organizational commitment. In essence, whenthe literatures surrounding job satisfaction and commitment are compared,one sees a similar set of (presumed) antecedent factors emerging. Indeed,the conclusion that those factors that cause job satisfaction are similar tothose causing commitment is appealing. However, the results of this in-vestigation do not support this assertion. The set of predictor variables gen-erally is shown to be antecedent to job satisfaction but not to commitment.

The temporal relationship of environmental alternatives to satisfactionis especially noteworthy because it supports Bluedorn's (1982) placementof this variable in his turnover model. Yet, despite the strong demonstra-tion of the negative relationships between environmental alternatives andcommitment and satisfaction, and the causal priority from alternatives todissatisfaction, questions remain regarding the role of alternatives in theprocess of attitude formation and in explaining behaviors such as turnover.For example, the existence of job alternatives may provide an explanationfor some of the relationships between demographic characteristics of theemployee and commitment and turnover (Price, 1977). Table 2 correlationsprovide possible support for this notion with respect to age, as suggestedby Sheldon (1971), and possibly career tenure but not the other demographicvariables. Furthermore, the timing of the existence of job alternatives mayaffect even the direction of the relationships. O'Reilly and Caldwell's (1981)data suggest that the presence of alternatives at the time of job choice (i.e.,before entry into the organization) has a positive impact on subsequent com-mitment to the organization; however, the availability of job alternativesafter organizational membership has been established may exert a negativeimpact on commitment (O'Reilly & Caldwell, 1981) and satisfaction (pres-ent study). Without doubt, the recent attention paid to environmental al-ternatives (Farrell & Rusbult, 1981; O'Reilly & Caldwell, 1981; Pfeffer &Lawler, 1980), after a long period of neglect, is worthy of continued pur-suit in uncovering some of these process issues.

Conclusions

Virtually all of the published articles in the organizational commitmentliterature have contained static correlational relationships between com-mitment and its presumed antecedents. And, as with so many other streamsof organizational literature, researchers appropriately have called for longi-tudinal designs in order to demonstrate causality more clearly. As an ini-tial response to this call, the longitudinal data presented here revealed sig-nificant findings. The study variables causally predicted job satisfaction


much better than they predicted subsequent commitment. The findings alsosuggest that commitment may be a construct that is neither simultaneouswith nor a consequence of job satisfaction. Rather, organizational com-mitment appears to be one of the many causes of satisfaction.

Future commitment research must continue to move away from staticcorrelational studies. More panel or other longitudinal designs are neededto: (1) provide replication and generalization tests of the findings reportedhere and (2) continue to specify the causes of commitment. More specifi-cally, an important generalization issue emerges when considering the longi-tudinal results of this investigation. The present study focused on the causalantecedents of a psychological measure of organizational commitment. Analternative measure of commitment, based on exchange theory, has a dif-ferent conceptual foundation and set of presumed antecedents (Hrebiniak& Alutto, 1972; Stevens et al., 1978). For example, Stevens et al. (1978)found job involvement, attitude toward change, work overload, and skilllevels of subordinates to have significant static relationships with exchangecommitment. Hrebiniak and Alutto (1972) also found two of the variablesin the present study—job tension and job satisfaction—to be significantstatic predictors of exchange commitment. In the present investigation, notime-lagged causal impact of these variables on psychological commitmentwas found. The causal effects of these and other variables on exchangecommitment remain to be tested via longitudinal designs. In short, threerelated avenues for future longitudinal research are suggested: (1) the pur-suit of other causal antecedents of psychological commitment; (2) the test-ing of hypothesized antecedents of exchange commitment; and (3) the si-multaneous investigation of both commitment constructs within the samestudies. If such a protocol is followed, the antecedents of both psychologi-cal and exchange commitment can be meaningfully compared for their con-gruence or noncongruence.

Appropriate causal specification will improve the models of commitment,which apparently have inadequately and/or erroneously identified its an-tecedents; empirically unsubstantiated causal assumptions presently needto be viewed with more caution. For example, the temporal ordering fromcommitment to satisfaction suggested by the present data is the reverse ofthat suggested by others (Marsh & Mannari, 1977; Porter et al., 1974), al-though it is consistent with the theoretical discussions of Bem (1967) andSalancik and Pfeffer (1978). More importantly, until other antecedents ofcommitment are more reliably established, organizational interventionsaimed at increasing commitment and its consequential beneficial employeebehaviors may not realize their intended effects.

The present findings suggest that the interventions implied by modelsof commitment—for example, improving the job itself or reducing job ten-sion—may result in higher satisfaction but not commitment. The costs ofsuch interventions thus will not be salvaged through their intended gains.If future research continues to fail to demonstrate longitudinally the causesof commitment other than demographic variables such as age and education


(Steers, 1977) or the existence of other job alternatives at the time of jobchoice (O'Reilly & Caldwell, 1981), then it may be that employee commit-ment can be influenced only through job selection techniques. Subsequentinterventions may be incapable of increasing commitment levels and there-fore may be a waste of organizational resources.

On the other hand, it is known that commitment does decrease prior toturnover (Porter et al., 1974, 1976). If further research can reveal othercauses of commitment that the organization can infiuence directly, a numberof benefits will accrue. First, the costs associated with misspecified inter-ventions can be reduced. Second, appropriate interventions may have theirintended results. Finally, direct improvements in commitment levels mayhave not only positive behavioral consequences but, according to the pres-ent results, the indirect outcome of increased employee satisfaction as well.

References

Abelson, M. A., & Sheridan, J. E. Catastrophe model of employee withdrawal leading to job termi-nation among nursing home staff. Paper presented at the National Academy of Management Meet-ings, San Diego, 1981.

Angle, H. L., & Perry, J. L. An empirical assessment of organizational commitment and organiza-tional effectiveness. Administrative Science Quarteriy, 1981, 26, 1-13.

Bateman, T., & Strasser, S. A cross-lagged regression test of the relationship between job tension andemployee satisfaction. Journal of Applied Psychology, 1983, 68, 439-445.

Becker, H. S. Notes on the concept of commitment. The American Joumai of Sociology, 1960, 66, 32-42.

Bem, D. J. Self-perception: An alternative interpretation of cognitive dissonance. Psychological Review,1967, 74, 183-200.

Bluedorn, A. C. A unified model of turnover from organizations. Human Relations, 1982, 35, 135-153.

Brief, A., Schuler, R., & Van Sell, M. Managing job stress. Boston: Little, Brown, 1981.

Buchanan, B. Building organizational commitment: The socialization of managers in work organiza-tions. Administrative Science Quarterly, 1974, 22, 533-546.

Cohen, J., & Cohen, P. Applied multiple regression/correlation analysis for the behavioral sciences.Hillsdale, N.J.: Lawrence Erlbaum, 1975.

Farrell, D., & Rusbult, C. E. Exchange variables as predictors of job satisfaction, job commitmentand turnover: The impact of rewards, costs, alternatives, and investments. Organizational Behaviorand Human Performance, 1981, 27, 78-95.

Hackman, J. R., & Lawler, E. E. Employee reactions to job characteristics. Journal of Applied Psy-chology, 1971, 55, 259-286.

Hackman, R., & Oldham, G. Development of the Job Diagnostic Survey. Journal of Applied Psychology,1975, 60, 159-170.

Hall, D. T., & Schneider, B. Correlates of organizational identification as a function of career patternand organizational type. Administrative Science Quarteriy, 1972, 17, 340-350.

Hrebiniak, L. G., & Alutto, J. G. Personal and role-related factors in the development of organiza-tional commitment. Administrative Science Quarterly, 1972, 17, 555-573.

Johnson, R. D. An investigation of the interaction effects of ability and motivational properties ontask performance. Unpublished dissertation, Indiana University, 1973.

Kahn, R. L., Wolfe, D. M., Quinn, R. P. , Snoek, J. D., & Rosenthal, R. A. Organizational stress:

Studies in role conflict and ambiguity. New York: Wiley, 1964.

Kenny, D. A. Cross-lagged panel correlation: A test for spuriousness. Psychological Bulletin, 1975,

82, 887-903.

Kiesler, C. A. (Ed.). The psychology of commitment: Experiments linking behavior to belief. NewYork: Academic Press, 1971.


Koch, J. D., & Steers, R. M.'Job attachment, satisfaction and turnover among public sector employees.Joumai of Vocational Behavior, 1978, 12, 119-128.

March, J. G., & Simon, H. A. Organizations. New York: Wiley, 1958.

Marsh, R. M., & Mannari, H. Organizational commitment and turnover: A prediction study. Admin-istrative Science Quarterly, 1977, 22, 57-75.

Morris, J. H., & Koch, J. L. Impacts of role perceptions on organizational commitment, job involve-ment and psychosomatic illness among three vocational groupings. Journal of Vocational Behavior,1979, 14, 88-101.

Morris, J. H., & Sherman, J. D. Generalizability of an organizational commitment model. Academyof Management Joumai, 1981, 24, 512-526.

Morris, J. H., & Steers, R. M. Structural influences on organizational commitment. Joumai of Voca-tional Behavior, 1980, 17, 50-57.

Mowday, R. T., Steers, R. M., & Porter, L. The measurement of organizational commitment. Jour-nal of Vocational Behavior, 1979, 14, 224-247.

O'Reilly, C. A., Ill, & Caldwell, D. F. The commitment and job tenure of new employees: Some evi-dence of post-decisional justification. Administrative Science Quarterly, 1981, 26, 597-616.

Pfeffer, J., & Lawler, J. Effects of job alternatives, extrinsic rewards and behavioral commitmenton attitude toward the organization. Administrative Science Quarterly, 1980, 25, 38-56.

Podsakoff, P. M., Todor, W. D., & Skov, R. Effects of leader reward and punishment behaviorson subordinate performance and attitudes. Proceedings of the Academy of Management, 1981,151-155.

Porter, L. W., Crampon, W. J., & Smith, F. W. Organizational commitment and managerial turn-over: A longitudinal study. Organizational Behavior and Human Performance, 1976, 15, 87-98.

Porter, L. W., Steers, R. M., Mowday, R. T., & Boulian, P. V. Organizational commitment, jobsatisfaction, and turnover among psychiatric technicians. Journai of Applied Psychoiogy, 1974, 5,603-609.

Price, J. L. The study of turnover. Ames, Iowa: Iowa State University Press, 1977.

Price, J. L., & Mueller, C. W. A causal model of turnover for nurses. Academy of Management Jour-nal, 1981, 24, 543-565.

Rogosa, D. A critique of cross-lagged correlations. Psychological Bulletin, 1980, 88, 245-258.

Salancik, G. R. Commitment and control of organizational behavior and belief. In B. Staw & G. Salancik(Eds.), New directions in organizational behavior. Chicago: St. Clair Press, 1977, 1-54.

Salancik, G. R., & Pfeffer, J. A social information processing approach to job attitudes and task design.Administrative Science Quarterly, 1978, 23, 224-235.

Schein, E. Organizational socialization and the profession of management. Industrial ManagementReview, 1968, 9, 1-6.

Schneider, B., & Dachler, H. A note on the stability of the job descriptive index. Journal of AppliedPsychology, 1978, 63, 650-653.

Sheldon, M. E. Investments and involvements as mechanisms producing commitment to the organiza-tion. Administrative Science Quarterly, 1971, 16, 143-150.

Sims, H. P., Jr. Further thoughts on punishment in organizations. Academy of Management Review,1980, 5, 133-138.

Sims, H. P., Jr., & Szilagyi, A. D. Leader reward behavior and satisfaction and performance. Organi-zational Behavior and Human Performance, 1975, 14, 426-438.

Sims, H. P., Jr., & Wilkerson, D. Time-lag in cross-lag correlation: A computer simulation. DecisionSciences, 1977, 8, 630-644.

Smith, P. C , Kendall, L. M., & Hulin, C. L. The measurement of satisfaction in work and retire-ment. Chicago: Rand McNally, 1969.

Staw, B. M. Rationalization and justification in organizational life. In B. M. Staw & L. L. Cummings(Eds.), Research in organizational behavior. Greenwich, Conn.: JAI Press, 1980, 45-80.


Steers, R. M. Effects of need achievement on the job performance-job attitude relationship. Journalof Applied Psychology, 1975, 60, 678-682.

Steers, R. M. Antecedents and outcomes of organizational commitment. Administrative Science Quarter-ly, 1977, 22, 46-56.

Stevens, J. M., Beyer, J. M., & Trice, H. M. Assessing personal, role, and organizational predictorsof managerial commitment. Academy of Management Journal, 1978, 21, 380-396.

Stone, S. F., & Porter, L. W. Job characteristics and job attitudes: A multivariate study. Journal ofApplied Psychology, 1976, 60, 57-64.

Vroom, V. H. Some personality determinants of the effects of participation. Englewood Cliffs, N.J.:Prentice-Hall, 1960.

*Thomas S. Bateman is Assistant Professor, Department of Man-agement at Texas A&M University.

Stephen Strasser is Assistant Professor, Hospital and Health Ser-vices Administration Division, Ohio State University.

Organizational Behavior

Technology

job satisfaction porter

employees job

job involvement

job variety

causes of commitment

stone porter

membership porter

job tenure angle perry