Professional Documents
Culture Documents
(2)
2. Two-factor
(AC-ORG and
AC-SUP1 factor)
775.53
(1)
3. Two-factor
(AC-ORG and
AC-GR1 factor)
589.04
(1)
4. Two-factor
(AC-SUP and
AC-GR1 factor)
829.69
(1)
5. One-factor 1177.72
(24)
6. Null 2425.37
.94
3. AC-GR 3.34 .87 .41
.25
.86
4. Perceived organizational
support
2.72 .85 .40
.30
.22
.90
5. LMX: aect 3.18 1.19 .35
.70
.16
.35
.93
6. LMX: loyalty 3.35 1.12 .36
.46
.12 .26
.61
.93
7. LMX: contribution 3.49 .94 .19
.17
.12 .24
.21
.35
.69
8. LMX: professional
respect
3.21 1.20 .29
.61
.07 .28
.67
.54
.17
.95
9. Perceived work group
cohesiveness
3.47 .89 .16
.27
.41
.29
.35
.36
.19
.29
.89
Note. a coecients are reported on the diagonal. LMX, leadermember exchange.
*
p < :05.
**
p < :01.
Table 4
Study 1: Regression analysis for aective commitment variables
Variable AC-ORG AC-SUP AC-GR
Perceived organizational support .35
.07 .13
LMX: aect .03 .49
.11
LMX: loyalty .04 .04 ).06
LMX: contribution .08 .01 .03
LMX: professional respect .09 .26
).13
Perceived work group cohesiveness ).01 ).04 .37
Note. Entries are standardized regression coecients. For AC-ORG, F 6; 193 7:26
; for AC-SUP,
F 6; 193 36:00
.
*
p < :001.
54 C. Vandenberghe et al. / Journal of Vocational Behavior 64 (2004) 4771
3. Study 2: Relationships with intent to quit and turnover
Field theory (Lewin, 1943) asserts that individuals behavior is primarily inu-
enced by those elements from the environment which are perceived as being proxi-
mal and salient (Mathieu, 1991; Mathieu & Hamel, 1989). Along this line, it is
likely that the notion of salience is determined by the type of behavior one is willing
to predict. For example, the organization focus may be viewed by an employee as
salient when it comes to predict turnover behavior. This is because turnover is the
kind of behavior for which cognitive deliberations about the viability of ones mem-
bership in the organization are activated (Hom, Caranikas-Walker, Prussia, & Gri-
eth, 1992). In order to decide to pursue or to stop organizational membership, the
employee will primarily evaluate the current status of his/her relationship with the
organization, hence his/her level of commitment to the organization. In contrast,
it is likely that the nature of relationships with, and level of commitment to, other
foci (e.g., supervisors or work groups) will be less relevant for predicting turnover.
The match among entities for determining the strength of the commitmentbehavior
relationship is thus critical. We suggest that the old conception of Lewins (1943)
proximality/salience of behavior corresponds to what current scholars more com-
monly consider to be a levels-of-analysis issue. Indeed, as shown by Meyer and Hers-
covitch (2001), if the level captured by the commitment construct does not t the
level which is tapped by the behavior, the strength of the relationship is likely to
be diminished.
Consistent with this point of view, organizational commitment has been repeat-
edly reported to relate negatively to intended and actual turnover (Mathieu & Zajac,
1990; Tett & Meyer, 1993). As evidenced by the turnover literature, its inuence on
turnover behavior is largely mediated by turnover intentions (Hom et al., 1992; Hom
& Grieth, 1991; Price & Mueller, 1986; Sager, Grieth, & Hom, 1998; Tett &
Meyer, 1993). Conversely, and as mentioned above, due to the fact that supervisors
and work groups address a dierent level of analysis, we do not expect commitment
to the supervisor and to the work group to be as inuential on the employee with-
drawal process from the organization. We thus posit the following:
Hypothesis 2a: Aective organizational commitment has a signicant indirect ef-
fect on actual turnover through intent to quit.
The tenets of eld theory (Lewin, 1943) suggest that the elements from the envi-
ronment that are not salient and proximal to the individuals action will not aect
his/her behavior directly (Mathieu, 1991). As far as organizational turnover is con-
cerned, such foci as the supervisor and the work group may not be so psychologically
salient as the organization because it remains possible to maintain membership in the
organization when commitment to these foci is low. For example, in case of low
commitment, the employee may wish to change supervisors or work groups, or avoid
being too much in contact with these target entities (Eisenberger et al., 2002;
Malatesta, 1995). Support for this contention can be found in several studies. First,
re-analyzing Beckers (1992) data, Hunt and Morgan (1994) found support for a
key-mediating construct model in which commitment to constituencies that are
cognitively distant from the organization, such as the supervisor, the work group,
C. Vandenberghe et al. / Journal of Vocational Behavior 64 (2004) 4771 55
and top management, exerted its inuence on work behavior through organizational
commitment. As Hunt and Morgan (1994) noted, their ndings may be due to the
fact that the outcomes of interest were mainly directed toward the organization
(e.g., intent to quit and organizational citizenship behavior). That is, these outcomes
referred to a level to whom organizational commitment is naturally tied. Similarly,
Wayne et al. (1997) found organizational commitment to be more strongly related to
intentions to quit than LMX, a variable thought to lead to aective commitment
to the supervisor. Finally, Bishop et al. (2000) reported organizational commitment
to be more strongly linked with intended turnover than team commitment. All these
arguments suggest that commitments to the supervisor and to the work group may
(a) have less direct inuence on turnover than organizational commitment and (b)
mainly exert their eect indirectly, through commitment to the global organization.
This leads us to hypothesize that:
Hypothesis 2b: Aective commitment to the supervisor and to the work group
has a signicant indirect eect on intent to quit through aective organizational
commitment.
3.1. Method
3.1.1. Sample and procedure
A random sample of 704 Alumni from a Belgian university who graduated be-
tween 1990 and 1996 was selected. Note that this sample was independent from
the sample used in Study 1. As part of a larger survey, respondents received a ques-
tionnaire including, among others, the three-foci commitment and intent to quit
measures. Questionnaires were sent directly to the subjects private addresses. A cov-
er letter accompanying the questionnaire explained the objective of the study, as-
sured that responses would be condential, and asked participants to send back
their completed questionnaire to the researchers oce using a pre-stamped enve-
lope. This letter also mentioned that questionnaires were coded so as to allow re-
searchers to send a reminder memo to non respondents after a 2-week period, and
to track respondents organizational membership status after 18 months. In total,
408 subjects responded to the survey. Excluding the incomplete questionnaires,
397 responses were usable for conducting the analyses (a 56.4% response rate).
Among respondents, 353 could be contacted 18 months later to determine if they
voluntarily left their organization during that time lag. Because we were interested in
voluntary turnover, we excluded from the analysis 37 respondents who were laid o,
leaving a nal sample of 316 employees. As respondents were asked to report the
name of their organization, we were able to contact their employer in order to verify
the accuracy of their membership reports. We did so for a random subset of 50 em-
ployees. No discrepancy was detected between employees reports of their member-
ship and reason of turnover (voluntary vs. involuntary) if any, and the
corresponding employers views.
The nal sample of respondents had an average age of 29.5 years SD 4:6 and
an average organizational tenure of 3.5 years SD 4:3 at Time 1. Among partic-
ipants, 67.4% were male and 92.0% were employed full-time. Respondents worked in
56 C. Vandenberghe et al. / Journal of Vocational Behavior 64 (2004) 4771
banking and insurance (16.7%), industries (13.1%), research (11.2%), consulting
(8.7%), public administration (8.3%), computer science (7.4%), public health
(6.7%), transportation and communication (3.8%), law (2.9%), environmental agen-
cies (2.6%), advertising (1.9%), construction (1.6%), trade (1.3%), teaching (1.0%),
media and real-estate (.6%), leisure and sports (.3%), or were unclassied (11.9%).
3.1.2. Measures
Commitment. Aective commitment to the organization, the supervisor, and the
work group was measured using the same scales as in Study 1.
Intent to quit. We measured intent to quit with two items: I often think about
quitting this organization and I intend to search for a position with another em-
ployer within the next year. These items were adapted from Hom and Grieth
(1991) and Jaros (1997), respectively.
Turnover. Stayers were rated as 1 while voluntary leavers were rated as 2. The
turnover rate in the sample was 20.0%.
3.2. Results and discussion
Data were analyzed using the structural equations modeling approach using the
LISREL 8.3 package (J ooreskog & S oorbom, 1993). The dichotomous nature of the
turnover variable violated a key assumption of standard structural equation model-
ing procedures which presume interval-level scaling of variables. To circumvent this
statistical problem, we followed Bollens (1989) recommendations and used, as input
for model estimation, a matrix of polychoric correlations and its associated asymp-
totic covariance matrix in place of the traditional covariance matrix of observed vari-
ables, and analyzed the data by the weighted-least-squares method of estimation.
A similar approach for correcting data for noninterval scaling has been used in prior
turnover research (Jaros, Jermier, Koehler, & Sincich, 1993; Sager et al., 1998).
Table 5 presents the descriptive statistics and the correlations among latent vari-
ables in the conrmatory factor analytic model. As can be seen, intent to quit was
more strongly associated with aective organizational commitment than with com-
mitments to the supervisor (r :71 vs. ).56, t313 4:11, p < :01) and to the
work group (r :71 vs. ).55, t313 7:66, p < :01). Likewise, turnover was
Table 5
Study 2: Descriptive statistics and intercorrelations among latent variables
Variable M SD 1 2 3 4 5
1. AC-ORG 3.18 .90 .89
2. AC-SUP 3.41 .90 .57
.88
3. AC-GR 3.37 .81 .72
.63
.83
4. Intent to quit 2.40 1.19 ).71
).56
).55
.82
5. Turnover 1.20 .40 ).15
).06 ).23
.47
Note. N 316. a coecients are reported on the diagonal. For Turnover, stayers were coded as 1 and
leavers as 2. All intercorrelations are estimated by LISREL.
*
p < :001.
C. Vandenberghe et al. / Journal of Vocational Behavior 64 (2004) 4771 57
more strongly tied to intent to quit than to commitment to the organization (r :47
vs. ).15, t313 6:47, p < :01), to the supervisor (r :47 vs. ).06, t313 5:81,
p < :01), and to the work group (r :47 vs. ).23, t313 7:79, p < :01).
Fit indices for structural models are presented in Table 6. The Hypothesized mod-
el, which proposes organizational commitment to be indirectly related to turnover
via intent to quit (Hypothesis 2a) and commitments to the supervisor and to the
work group to inuence intent to quit through organizational commitment (Hypoth-
esis 2b), yielded a good t to the data, as evidenced by its sizeable GFI, AGFI, TLI,
and CFI values. Although not hypothesized in this study, signicant bivariate rela-
tionships between commitments to the supervisor and to the work group and intent
to quit have been reported in previous research (Becker et al., 1996), leaving the pos-
sibility that these attitudes could also have direct eects on intent to quit. In order to
test these plausible connections, we estimated two models in which the paths from
these constructs to intent to quit were alternatively freed (cf. Table 6). Alternate
model 1, which added a direct eect of commitment to the supervisor on intent to
quit, was a signicant improvement over the Hypothesized model, Dv
2
1 58:80,
p < :001, indicating that this eect was not trivial. In contrast, Alternate model 2,
which added a path from commitment to the work group on intent to quit, did
not dier signicantly from the Hypothesized model, Dv
2
1 2:73, ns, signaling
that the path of interest was not signicant. Finally, in order to test whether the path
from organizational commitment to intent to quit was of the same magnitude than
the path from commitment to the supervisor to intent to quit, we estimated a model
in which these paths were constrained to equality (Alternate 3, cf. Table 6). Contrast-
ing this model with the Alternate model 1 resulted in a signicant decrease in t,
Dv
2
1 19:20, p < :001, revealing that the paths in question were not equal.
The standardized weighted-least-squares estimates of the structural parameters
for the retained Alternate model 1 are reported in Fig. 1. As can be seen, aective
commitment to the supervisor (b :12, p < :001) and to the work group (b :68,
p < :001) were signicantly related to organizational commitment which in turn in-
uenced intent to quit (b :57, p < :001), with the latter being signicantly asso-
ciated with turnover (b :38, p < :001). Although not hypothesized, commitment to
Table 6
Study 2: Fit indices for structural models
Model v
2
df GFI AGFI TLI CFI Model Comparison Dv
2
Ddf
Hypothesized 987.34
(1)
Alternate 2 984.61
(1)
Note. N 316. GFI, goodness-of-t index; AGFI, adjusted goodness-of-t index; TLI, Tucker and
Lewis index; and CFI, comparative t index.
*
p < :001.
58 C. Vandenberghe et al. / Journal of Vocational Behavior 64 (2004) 4771
the supervisor had also a signicant direct eect on intent to quit (b :26,
p < :001).
The estimation of direct eects was supplemented by the calculation of the indi-
rects eects predicted by Hypotheses 2a and 2b. We used the technique recommended
by Sobel (1982, 1986) to estimate the indirect eects of organizational commitment
on turnover (Hypothesis 2a) and of commitment to the supervisor and to the work
group on intent to quit (Hypothesis 2b). The indirect eect of organizational commit-
ment on turnover was ).25 SE :03, and its 95% condence interval ().19/).31)
did not contain zero. This indicates that the eect was signicant, which supports Hy-
pothesis 2a. Similarly, the indirect eect of commitment to the supervisor on intent to
quit was ).09 SE :02 and its 95% condence interval ().05/).13) did not contain
zero. Finally, the indirect eect of commitment to the work group on intent to
quit was ).42 SE :03 and its 95% condence interval ().36/).48) did not contain
zero. The latter results lended support to Hypothesis 2b. Worth of noting also is that
the three commitment variables exerted a signicant indirect eect on turnover [).16
SE :02 for commitment to the supervisor, ).17 SE :02 for commitment to the
work group, and ).25 SE :03 for commitment to the organization].
In brief, the results of this study demonstrated that (a) aective organizational
commitment exerted the strongest direct eect on intent to quit, (b) aective commit-
ment to the supervisor had both a direct and indirect eect on intent to quit, and that
(c) aective commitment to the work group exerted an indirect eect on quitting in-
tentions. The fact that the impact of organizational commitment was stronger than
that of commitment to the supervisor suggests that the organization was psycholog-
ically more relevant in the eyes of employees when they were to generate their mem-
bership decisions.
4. Study 3: Relationships with job performance
Prior research regarding the ability of organizational commitment to predict
job performance has provided mixed results (Cohen, 1991; Mathieu & Zajac,
Fig. 1. Study 2: Standardized path coecients for Alternate model 1. For the sake of parsimony, param-
eters for the measurement portion and disturbance terms are not presented. *p < :001.
C. Vandenberghe et al. / Journal of Vocational Behavior 64 (2004) 4771 59
1990; Randall, 1990). In his quantitative review of the relevant literature, Riketta
(2002) reported a true correlation of .20 between attitudinal organizational commit-
ment and performance. He also found that the way commitment was measured (via
the ACS or the OCQ) did not inuence the strength of the relationship. Thus, the
reasons for such a weak relationship must be searched elsewhere than in dierences
in conceptualization and/or measurement of commitment. Among these, the nature
of performance as well as the foci of commitment seem to be important ones. As far
as in-role activities are concerned, it seems that organizational commitment is not
strongly related to performance (Becker & Kernan, 2001; Bishop et al., 2000; Elle-
mers et al., 1998; Keller, 1997; Settoon et al., 1996). In contrast, when organizational
citizenship behavior is used as criterion, organizational commitment appears to have
a more powerful inuence (Organ & Ryan, 1995; Riketta, 2002). This dierential re-
lationship may be due to the fact that the performance of in-role activities is more
dependent on regular interactions with, and feedback from, proximal entities such
as the work group and supervisor (Becker et al., 1996), these foci being psycholog-
ically more salient and relevant than the distal organization for such purposes. In
contrast, organizational citizenship behavior is concerned with behavior that benets
the organization as a whole, hence renders the organization a more salient entity to
employees.
As supervisors have the formal responsibility to monitor, direct, and provide feed-
back to, employees in the performance of their in-role duties (Eisenberger et al.,
2002), supervisors facilitate the acceptance of performance norms by employees
(Becker et al., 1996; Siders et al., 2001). Due to these interactions, the supervisor
should represent the most salient commitment focus when prediction of job perfor-
mance is at stake. Support for this contention has been provided by Becker et al.
(1996) and Becker and Kernan (2001) who found that commitment to the supervisor
was more strongly associated with performance than was overall commitment to the
organization. Similarly, in traditional work settings, commitment to the work group
may also be less important than commitment to the supervisor for predicting job
performance. This is because co-workers are not formally in charge of monitoring,
guiding, and rewarding employee performance. Note that this may not be the case
in self-directed work teams, in which the completion of tasks and responsibility
for end products are collectively shared by the team (Bishop & Scott, 2000; Bishop
et al., 2000). To summarize, the supervisor should be a more salient and relevant
commitment focus than the organization and the work group when job performance
is the outcome of interest.
Hypothesis 3a: Aective commitment to the supervisor has a signicant direct
eect on job performance.
The role of aective commitments to the organization and to the work group in
the prediction of job performance should be indirect. Indeed, they may act on per-
formance through commitment to the supervisor, the latter representing the person
who is formally responsible for driving employees in performing competently on the
job. The organization is a more distant entity in this case because it provides the
more general rules and strategic orientation within which the actions of individual
employees should take place. Similarly, the work group is psychologically less salient
60 C. Vandenberghe et al. / Journal of Vocational Behavior 64 (2004) 4771
because co-workers may only provide informal aid in the performance of duties.
However, we could expect aective commitment to the organization to be more
strongly related to commitment to the supervisor than commitment to the work
group would be, because supervisors are representatives of the organization toward
employees (Eisenberger, Fasolo, & Davis-LaMastro, 1990; Eisenberger et al., 1986)
while co-workers do not possess such formal status and prestige (Eisenberger et al.,
2002). Consequently, the indirect eect of organizational commitment on job perfor-
mance should be stronger than the corresponding indirect eect of commitment to
the work group.
Hypothesis 3b: Aective commitments to the organization and to the work
group have a signicant indirect eect on job performance via commitment to
the supervisor.
Hypothesis 3c: The indirect eect of aective commitment to the organization on
job performance is stronger than the corresponding indirect eect of commitment to
the work group.
4.1. Method
4.1.1. Sample and procedure
We obtained the formal agreement of the nursing director of a Belgian hospital to
conduct an attitude survey among its nursing sta. As part of a larger study, all
nurses received a questionnaire including the measures of aective commitment to
the organization, to the supervisor and to the work group. A cover letter accompa-
nying the questionnaire explained the objective of the study, assured that responses
would be kept condential, and asked participants to return their completed
questionnaire to the researchers oce using a pre-stamped envelope. The letter also
mentioned that questionnaires were coded in order to match responses with supervi-
sor-rated performance appraisals. Nurses voluntarily completed the survey during
their regularly scheduled work hours.
Of the 530 nurses who were contacted, 278 returned their questionnaire to the
researchers, with 270 of them providing usable returns, for a 50.9% response rate.
Six months later, head nurses were contacted to provide performance ratings for
their ward nurses on a 4-item scale designed by the researchers. Due to missing
data on commitment measures N 19 and performance ratings (2 head nurses
for 23 nurses) as well as turnover among nurses N 34, the number of valid
commitment-performance pairings was reduced to 194. These responses involved
performance assessments by 32 head nurses, for an average number of perfor-
mance ratings per unit of 6.06 (range 116). In the nal sample of nurses, aver-
age age was 35.4 years SD 8:3 and average organizational tenure was 10.3
years SD 6:7. Among respondents, 82.9% were female and 58.0% worked
full-time.
4.1.2. Measures
Commitment. Aective commitment to the organization, the supervisor, and the
work group were measured using the same scales as in the rst two studies.
C. Vandenberghe et al. / Journal of Vocational Behavior 64 (2004) 4771 61
Performance. Supervisor-rated performance appraisals were obtained 6 months
after Time 1 data collection. Based on a panel discussion with the nursing manage-
ment sta, four criteria were selected as being central aspects of nurses job in this
hospital. Head nurses rated the performance of their sta on the following criteria:
(a) quality of care, (b) quality of contacts with patients, (c) work-related helping be-
havior directed toward colleagues, and (d) work-related helping behavior directed
toward the head nurse. Ratings were provided using a 10-point scale anchored ex-
tremely poor (1) and outstanding (10). Performance data were submitted to a
principal components analysis. Results suggested a single factor, as indicated by a
strong loading of items on the rst factor (mean loading: .85), accounting for
71.5% of the total variance, a clear break in the scree plot, and eigenvalues less than
1.0 for the remaining potential factors. We thus combined the items to create a single
score of performance for each nurse. In addition, as head nurses assessed the perfor-
mance of several nurses (range 116), we standardized the performance ratings
within raters, i.e., subtracting the mean of each unit from the score of any particular
nurse. These standardized performance ratings were then used as input in the LIS-
REL analyses. Such a procedure allowed us to control for any potential response
bias in supervisor ratings.
4.2. Results and discussion
The relationships among commitment variables and job performance were ana-
lyzed using the structural equations modeling approach with the LISREL 8.3 pack-
age (J ooreskog & S oorbom, 1993). The covariance matrix of observed variables was
used as input for model estimation through maximum likelihood. Table 7 displays
the descriptive statistics and the intercorrelations among latent variables in the con-
rmatory factor analytic model. Of greatest interest, aective commitment to the su-
pervisor was signicantly associated with job performance (r :24, p < :01).
Fit indices for structural models are presented in Table 8. The Hypothesized mod-
el yielded a good t to the data, as indicated by a TLI of .89 and a CFI of .91,
these indices being particularly suited for the comparison of competing models
(Medsker et al., 1994). As some studies have reported signicant relationships
between organizational commitment and performance (Angle & Lawson, 1994;
Table 7
Study 3: Descriptive statistics and intercorrelations among latent variables
Variable M SD 1 2 3 4
1. AC-ORG 2.79 .89 .79
2. AC-SUP 3.29 .94 .46
.87
3. AC-GR 3.67 .84 .36
.32
.83
4. Job performance 7.63 1.14 .05 .24
.11 .86
Note. N 194. a coecients are reported on the diagonal. All intercorrelations are estimated by
LISREL.
*
p < :01.
**
p < :001.
62 C. Vandenberghe et al. / Journal of Vocational Behavior 64 (2004) 4771
Mayer & Schoorman, 1992; Meyer, Paunonen, Gellatly, Gon, & Jackson, 1989)
and between team commitment and performance (Bishop & Scott, 2000; Ellemers
et al., 1998), we tested for such possibilities by relaxing the corresponding paths in
the Hypothesized model. Alternate model 1, which freed the path from organiza-
tional commitment to performance, did not improve over the Hypothesized model,
Dv
2
1 :34, ns, nor did Alternate model 2, which relaxed the path from work group
commitment to performance, Dv
2
1 :05, ns (cf. Table 8).
Fig. 2 presents the standardized parameter estimates for the Hypothesized model.
As can be seen, organizational commitment (b :40, p < :001) and work group
commitment (b :18, p < :05) were signicantly related to commitment to the su-
pervisor, with the latter being signicantly associated with performance (b :25,
p < :01), hence providing support to Hypothesis 3a.
We estimated the indirect eects of aective commitments to the organization and
to the work group on job performance using the procedure outlined by Sobel (1987).
The indirect eect of organizational commitment on performance was .08 SE :03
and its 95% condence interval (.02/.14) did not contain zero, suggesting that the
eect was signicant. The indirect eect of commitment to the work group on
Table 8
Study 3: Fit indices for structural models
Model v
2
df GFI AGFI TLI CFI Model Comparison Dv
2
Ddf
Hypothesized 372.51