The descriptive statistics and scale intercorrelations are presented in Table 1.
Unlike previous research, significant positive correlations were obtained between
organizational affective and continuance commitment (r D :18; p < :05) and
between work group affective and continuance commitment (r D :29; p < :01).
Significant correlations were obtained between the social interaction and affective
commitment measure within each constituency, and each measurewas significantly
correlated across each constituency (e.g., social interaction for all three constituencies).
These correlations did not exceed multicollinearity guidelines and were used
in further analyses (Kenny, 1979).
Before testing the model of social interaction and affective commitment, a factor
analysiswas conducted to determine the participant’s ability to distinguish between
the three constituencies while using similar measures. The 51 social interaction
items (three 17-item measures) were factor analyzed with a maximum likelihood
extraction and an oblique rotation. The solutionwas set to three factors to represent
the three constituencies (work group, department, and organization). The resulting
matrix was significant, Â2(1125) D 2766:81; p < :001, accounted for 54.3% ofthe variance, and each of the items loaded on the hypothesized constituency factor.
For example, all of the work group items loaded on one factor. An examination
of the individual items revealed that five items loaded across factors. These five
items were dropped from all three scales. A second factor analysis was significant,
Â2(525) D 1678:25; p < :001, accounted for 63.4% of the variance, and revealed
three distinguishable factors representing the three constituencies. Each of the
items loaded on the expected factor, e.g., the work group items loaded on the
work group factor.2 The Cronbach’s alpha reliability estimates of the work group,
department, and organizational subscales remained high, ®’s D :96; :95; and .95,
respectively. The three 12-item scales were used in all subsequent analyses.
The 24 items from the three affective commitment scales were factor analyzed
using a maximum likelihood extraction and an oblique rotation. The solution was
set to three factors to reflect the three constituencies. The resulting matrix was
significant, Â2(207) D 605:28; p < :001, and accounted for 46.2% of the variance.
Although the work group items loaded on one factor, the department and
organizational items were less clear. Corresponding items from the department and
organizational measures loaded together to form a unique factor. This item was
dropped from all three scales and the factor analysis was repeated. The result was
2 Details for the factor analyses can be obtained from the first author.a significant three factor solution, Â2(150) D 463:92; p < :001, that accounted
for 46.0% of the variance and had a distinguishable factor for each constituency.
Each of the items loaded on the expected factor, e.g., the work group commitment
items loaded on the work group factor. The Cronbach’s alpha reliability estimates
for the 7-item work group, department, and organizational scales remained high,
®’s D :83; :85; and .84, respectively. The three 7-item scales were used in all
subsequent analyses.
The strong positive correlations between the social interaction and affective
commitment measures suggested the participants might not have distinguished
between the two constructs. Therefore, we conducted three factor analyses using
a maximum likelihood extraction and an oblique rotation. For the work group, the
matrix was significant, Â2(134) D 347:32; p < :001, accounted for 58.0% of the
variance, and had a two distinguishable factors representing social interaction and
affective commitment. Similarly, the department and organization matrices were
significant and had two distinguishable factors representing social interaction and
affective commitment, Â2(134) D 320:06; p < :001, which accounted for 57.0%
of the variance and Â2(134) D 315:61; p < :001, which accounted for 56.2%
of the variance, respectively. These results confirm the participants did indeed
distinguish between the social interaction and affective commitment constructs
both within and across constituencies.We tested the model using path analysis. Previous research (e.g., Mathieu &
Zajac, 1990) has found that age and tenure are significantly, although weakly,
correlated with affective commitment (Meyer & Allen, 1997). To control for the
influence of these variables on the hypothesized relationships, they were entered
first into the regression equations.
Organizational affective commitment regressed on the control variables resulted
in a significant R; :26; p < :01, with tenure significant, ¯ D :33; p < :01, but
age was not significant (see Table 2). In the second step, organizational affective
commitment was regressed on the control variables, department affective commitment
and organizational social interaction. As hypothesized, department affective
commitment, and organizational social interaction were significantly related to
organizational affective commitment (see Table 2).
The regression of department affective commitment on the control variables
was not significant, R D :17; ns, although tenure was significantly related to
department affective commitment, ¯ D :20; p < :05. As seen in Table 2, when
work group affective commitment and department social interaction were added
to the equation, the result was significant. Work group affective commitment and
department social interaction were significantly related to department affective
commitment, ¯ D :33; p < :001, and ¯ D :55; p < :001, respectively. Similarly,
when work group affective commitment was regressed on the control variables,
the result was not significant, but tenure was significant ¯ D :24; p < :05. When
work group social interaction was added to the equation, the overall result wassignificant, R D :66; p < :001. Work group social interaction was significantly
related to work group affective commitment, ¯ D :64; p < :001.
For the overall test of the model fit, the chi-square indicated the hypothesized
model was an acceptable fit to the data, Â2(13) D 17:39; ns, but Q D :88 indicating
a revised model should be considered. For the prediction of organizational
affective commitment, the addition of a negative path from department social interaction
significantly increased the relationship, R D :77, ¯ D ¡:23; p < :01
(see Table 3). For department affective commitment, the addition of a negative
path from work group social interaction significantly increased the relationship,
R D :75; ¯ D ¡:18; p < :01. Adding the two negative paths, work group interaction
to department affective commitment and department interaction to organizational
affective commitment, improved the model fit, Q D:97; Â2(11)D4:20; ns.
Overall, there was support for the model (see Fig. 3).The purpose of the next set of analyses was to test the relationships between
affective and continuance commitment components. These analyses were conducted
by first determining to which constituency the employee was affectivelycommitted. Participants were classified as having attained affective commitment
to a constituency if they scored above the cutoff of .5 standard deviation above the
mean. Therefore, each participant received a dichotomous classification of either
having or not having affective commitment to each of the three constituencies
(work group, department, and organization). Based on Lewin’s (1943) theory of
psychologically proximal units and Lawler’s (1992; Mueller & Lawler, 1999) assertions
regarding nested collectives, participants were assumed to have affective
commitment to proximal constituencies prior to developing affective commitment
to distal constituencies. Employing this assumption, the following classification
scheme was used. Participants classified as lacking affective commitment to any of
the constituencies were placed in the “low affective commitment” group (n D 73).
Participants above the cutoff score for work group affective commitment, but below
the cutoff score on the department and organizational affective commitment
were placed in the “work group affective commitment” group (n D 15). Those
who scored above the cutoff on work group and department affective commitment,
but below the cutoff on organizational affective commitment were placed inthe “department affective commitment” group (n D 15). Participants who scored
above the cutoff score on all three affective commitment measures were classified
as having “organizational affective commitment” (n D 16). The remaining
participants were not classified because their results on the affective commitment
measures did not conform to any of the previous patterns. These participants typically
worked for the organization far longer than they had worked in the work
group and, therefore, would not be expected to exhibit this pattern.
Once the commitment constituency was identified, the hypothesized relationships
between affective commitment and continuance commitment were examined.
These hypotheses suggested affective commitment would be significantly greater
than continuance commitment for constituencies to which the employee has attained
affective commitment. For those constituencies to which an employee does
not have affective commitment, continuance commitment will be significantly
greater than affective commitment (see Fig. 2 for specific hypotheses). The differences
between the means of the commitment variables were tested using paired
one-tailed t tests. To protect against experimentwise error rate, alpha was adjusted
using the modified Bonferroni test (Keppel, 1982). The results of these analyses
are shown in Table 4. The results are quite similar to the predictions. Eight of the
12 comparisons were significant at p < :0125. As hypothesized for employees
classified as having low affective commitment to any constituency, continuance
commitment significantly exceeded affective commitment for the department and
organization foci. Employees who attained work group affective commitment had
work group affective commitment significantly greater than continuance commitment,
as hypothesized. For those who attained department affective commitment,
affective commitment was significantly greater than continuance commitment for
the work group and department foci, as hypothesized. Employees who attained organizational
affective commitment, affective commitmentwas significantly greaterthan continuance commitment for all three commitment foci. Of the four remaining
pairs, the comparisons of work group commitments for the low affective constituency
and organizational commitments for thework group constituency were in
the hypothesized direction, but failed to reach significance. Only the comparisons
of department commitments for the work group constituency and organizational
commitments for the department constituency were in a direction counter to the
hypotheses.