Escolar Documentos
Profissional Documentos
Cultura Documentos
298
Xu & Bassham
299
300
Xu & Bassham
301
Procedures
An initial email invitation and follow-up emails that encouraged
presidential assistants to participate in the online survey hosted by a midwestern state university were sent to the compiled list of presidential
assistants. Participation was voluntary with no compensation.
Participants were directed to an index page, which gives an
introduction and informs them of the voluntary nature of the survey.
There were no risks associated with this survey. The program used to
build the survey was unable to trace respondents or associate information
with them. Participants that agreed to take the survey consciously
decided to do so by choosing a hyperlink to the survey instrument from
this index page. Completion of the survey of 49 questions took
approximately 10-15 minutes.
The online survey was posted for a maximum of five weeks. A final
notice, announcing the deadline of the online survey, was sent by email
to the presidential assistants prior to the closure of the online survey.
Data Analysis
The psychometric characteristics of the scale were evaluated using
reliability analysis. Confirmatory Factor Analysis (CFA) was conducted
with LISREL-SIMPLIS 8.8 (Jreskog & Dag, 2006) software program,
using maximum likelihood estimation of the sample covariance matrix.
Model fit was evaluated using the minimum fit function 2. As
2
values are potentially inflated by large sample sizes, fit was also
examined using four practical fit indices. They were the root mean square
error of approximation (RMSEA), the comparative fit index (CFI), and
the standardized root mean square residual (SRMR). The ideal cut-off
point used for RMSEA was .08 (Browne & Cudeck, 1993), SRMR was
.08 (Hu & Bentler, 1995), and CFI was .95 (Hu & Bentler, 1995). When
all three criteria are met, the model is regarded as a good fit to the data.
However, a less strict standard is used in this study to judge model fit:
the cut-off point for RMSEA is .08 (Browne & Cudeck, 1993), SRMR,
.08 (Hu & Bentler, 1995), and CFI is .90.
RESULTS
Descriptive Statistics
Table 1 shows the means, standard deviations, skewness, and kurtosis
values for each item. According to Curran, West, and Finch (1996), for
univariate normality, skewness and kurtosis values of 0 to 2, and 0 to 7,
respectively, can be taken as demonstrating sufficient normality. On the
basis of the values shown in Table 1, the data appear to show sufficient
normality. In addition, Table 1 presents the correlation matrix of 24
items.
302
ACS 2
ACS 3
ACS 4
ACS 5
ACS 6
ACS 7
ACS 8
ACS 1
1.000
ACS 2
0.366
1.000
ACS 3
0.375
0.243
1.000
ACS 4
0.327
0.171
0.305
1.000
ACS 5
0.439
0.297
0.340
0.225
1.000
ACS 6
0.489
0.291
0.421
0.327
0.546
1.000
ACS 7
0.444
0.677
0.319
0.321
0.472
0.540
1.000
ACS 8
0.552
0.409
0.416
0.298
0.730
0.646
0.575
1.000
CCS 1
-0.084
0.046
-0.050
-0.043
-0.024
-0.030
0.035
-0.035
CCS 2
0.020
-0.025
0.183
0.061
0.027
0.072
0.039
-0.084
CCS 3
0.332
0.085
0.358
0.364
0.235
0.281
0.278
0.221
CCS 4
0.038
0.045
0.005
0.170
-0.015
0.034
-0.015
-0.021
CCS 5
-0.032
-0.053
-0.049
0.024
-0.106
-0.148
-0.033
-0.211
CCS 6
-0.029
-0.158
-0.042
0.062
-0.080
0.000
-0.112
-0.198
CCS 7
-0.071
-0.138
-0.052
-0.040
-0.130
-0.088
-0.118
-0.222
CCS 8
0.036
0.038
0.145
0.009
0.070
-0.005
0.113
-0.041
NCS 1
0.126
0.090
0.241
0.140
0.171
0.119
0.145
0.143
NCS 2
0.132
0.103
0.168
0.063
0.125
0.107
0.130
0.151
NCS 3
0.200
0.230
0.208
0.096
0.340
0.271
0.224
0.446
NCS 4
0.168
0.127
0.238
0.164
0.135
0.155
0.130
0.185
NCS 5
0.364
0.187
0.314
0.368
0.273
0.307
0.311
0.342
NCS 6
0.072
0.113
0.151
0.086
0.131
0.106
0.110
0.114
NCS 7
0.209
0.137
0.153
0.175
0.103
0.105
0.168
0.071
NCS 8
0.271
0.224
0.332
0.187
0.275
0.321
0.342
0.309
Mean
SD
Skewness
Kurtosis
5.10
5.28
3.72
2.93
5.30
5.11
5.21
5.33
1.317
1.184
1.462
1.396
1.301
1.372
1.229
1.204
-1.659
-2.105
-0.396
0.562
-2.113
-1.665
-1.890
-2.067
2.133
4.466
-0.811
-0.726
3.656
1.776
3.172
3.746
Note: ACS = Affective Commitment Scale, CCS = Continuance Commitment Scale, NCS
= Normative Commitment Scale
Reliability Analysis
Three scales of TCM were subjected to reliability analysis. The
Cronbachs alphas of ACS, CCS and NCS were .848, .746 and .658
respectively. Judging from the small to moderate correlations with the
rest of the CCS items, item CCS03 (Too much in my life would be
disrupted if I decided I wanted to leave my organization now) should be
removed. So should the item NCS03 (Jumping from organization to
Xu & Bassham
303
CCS 2
CCS 3
CCS 4
CCS 5
CCS 6
CCS 7
CCS 8
CCS 1
1.000
CCS 2
0.171
1.000
CCS 3
-0.046
0.335
1.000
CCS 4
0.284
0.215
0.202
1.000
CCS 5
0.183
0.536
0.199
0.340
1.000
CCS 6
0.196
0.423
0.143
0.340
0.418
1.000
CCS 7
0.147
0.384
0.100
0.174
0.361
0.470
1.000
CCS 8
0.094
0.398
0.371
0.244
0.494
0.355
0.252
1.000
NCS 1
0.108
0.184
0.105
-0.072
0.136
0.101
0.106
0.242
NCS 2
0.113
0.106
0.088
0.055
0.069
-0.070
0.070
-0.009
NCS 3
-0.100
-0.090
-0.014
-0.047
-0.184
-0.213
-0.197
-0.096
NCS 4
0.026
0.131
0.245
0.009
0.082
0.052
0.063
0.194
NCS 5
0.011
0.004
0.381
0.032
-0.090
-0.073
-0.086
0.032
NCS 6
-0.050
0.063
0.055
-0.042
0.025
-0.051
0.007
0.038
NCS 7
0.085
0.225
0.188
0.038
0.172
0.210
0.172
0.295
NCS 8
0.040
0.046
0.296
-0.037
-0.036
-0.018
-0.001
0.107
4.37
3.95
4.07
4.23
3.86
3.05
3.96
4.05
1.757
1.488
1.598
1.511
1.524
1.663
1.590
1.610
Skewness
-0.706
-0.211
-0.421
-0.590
-0.293
0.304
-0.375
-0.485
Kurtosis
-0.953
-0.972
-0.946
-0.646
-0.964
-1.189
-0.963
-0.933
Mean
SD
Note: ACS = Affective Commitment Scale, CCS = Continuance Commitment Scale, NCS
= Normative Commitment Scale
organization does not seem at all unethical to me). Items NCS05 (If I got
another offer for a better job elsewhere I would not feel it was right to
leave my organization) and NCS08 (I do not think that wanting to be a
company man or company woman is sensible anymore) were weakly
correlated with the rest of the NCS items, with correlations ranging from
.06 to .42 and .10 to .27 respectively.
304
NCS 2
NCS 3
NCS 4
NCS 5
NCS 6
NCS 7
NCS 8
NCS 1
1.000
NCS 2
0.230
1.000
NCS 3
0.057
0.130
1.000
NCS 4
0.318
0.168
0.103
1.000
NCS 5
0.139
0.041
0.092
0.422
1.000
NCS 6
0.396
0.349
0.090
0.285
0.143
1.000
NCS 7
0.544
0.085
-0.065
0.364
0.117
0.317
1.000
NCS 8
0.159
0.230
0.066
0.192
0.257
0.214
0.119
1.000
3.64
4.49
4.96
3.84
3.73
4.14
3.13
4.08
SD
1.355
1.566
1.520
1.551
1.636
1.431
1.325
1.431
Skewness
0.100
-0.806
-1.300
-0.363
-0.246
-0.504
0.282
-0.460
-0.765
-0.565
0.378
-1.020
-1.154
-0.652
-0.530
-0.648
Mean
Kurtosis
Note: ACS = Affective Commitment Scale, CCS = Continuance Commitment Scale, NCS
= Normative Commitment Scale
Initial Analysis
Initially, a three-factor CFA model was tested. In this model, the
ACS, CCS, and NCS items loaded on their intended factors according to
Allen and Meyers (1990) 24-item model, allowing for correlation among
latent factors. Table 2 presents standardized factor loadings of the items
and inter-correlations among factors. The chi-square statistic for this
model was 2 (249, =279) = 838.95, p < .01. An unacceptable datamodel fit was suggested by the RMSEA (.093, 90% CI = .086, .100),
SRMR (.11), and CFI (.84) (see Table 3). All the items loaded
significantly onto their intended factors; however, the NCS items
generally had lower loadings than items loading onto other factors. These
lower loadings of the NCS items were consistent with its lower
Cronbachs alpha (.658).
This study considered as candidates for deletion those items that had
low factor loadings on their intended factor, in combination with high
modification indices and standardized residuals. At the same time, the
wordings of those items were considered for ambiguity in an attempt to
delete fewer items. As a result, items CCS03 and NCS03 were deleted.
Items NCS05 and NCS08 were not deleted, but were moved from the
Xu & Bassham
305
NCS scale to the ACS scale due to their high correlation with ACS items
and weak association with NCS items. The strong correlations of NCS05
and NCS08 with the ACS items indicated that the item wordings were
ambiguous. Particularly, in item NCS05, if I got another offer for a
better job elsewhere I would not feel it was right to leave my
organization, the word feel suggested an affective connotation. Further,
TABLE 2
Item
A1
A2
A3
A4
A5
A6
A7
A8
C1
C2
C3
C4
C5
C6
C7
C8
N1
N2
N3
N4
N5
N6
N7
N8
A
C
C-L
C-H
N
F1
A
F2
C
F3
N
.64
.52
.51
.39
.75
.74
.70
.87
Modified Three
Factor Solution
F1
A
.65
.35
.15
.57
.39
.53
.61
.37
.44
F3
N
.65
.52
.52
.41
.74
.74
.71
.86
.26
.70
.36
.42
.73
.63
.53
.61
1.0
-.08
F2
C
.27
1.0
F1
A
F2
C-L
F3
C-H
.27
.68
.27
.41
.74
.64
.55
.59
.42
.67
.75
.64
.55
.61
.74
.31
.50
.45
.50
.45
.52
.70
.42
.32
F4
N
.65
.52
.52
.41
.74
.74
.71
.86
.74
.31
1.0
-.12
1.0
.52
.70
.42
1.0
1.0
.30
1.0
.05
-.17
.32
1.0
.98
.37
1.0
.29
1.0
Note: F = Factor; A = ACS; C = CCS; N = NCS; C-L = Continuance Affective Scale Low
Alternative; C-H = Continuance Commitment Scale High Sacrifice
306
The four-factor model was tested in which the CCS scale was split
into two subscales, one assessing the personal sacrifice associated with
leaving the organization and the other an awareness of the lack of job
alternatives (Hackett, et al. 1994). The chi-square statistic for this model
was 2 (203, =279) = 515.91, p < .01. A good fit was suggested based
on RMSEA (.075, 90% CI = .067, .083), SRMR (.071) and CFI (.90).
However, the correlation between the two constructs under CCS was
as high as .98, indicating that the two constructs were actually the same
one. Thus, this study accepted the three-factor modified TCM model.
DISCUSSION
This study is one aspect of a larger study of the effect of presidential
assistants career stages on organizational commitment. Little research
Xu & Bassham
307
has been done with the most prominent office in colleges and
universities, i.e. assistants to presidents, that hold significant power
(Fisher, 1985; Lingenfelter, 2004). Literature suggested that presidential
assistants work as a sounding board for presidents, to whom the
presidents can express roughly formed versions of thoughts and ideas
with confidence (Fisher, 1985; Giddens, 1971; Lingenfelter, 2004).
Statistical analysis of the presidential assistant data on the original
TCM model suggests modification of the scale is necessary. As a result,
two items were removed, with one from the CCS scale and the other one
from the NCS scale. In addition, two items were moved from the NCS to
the ACS scale. Reliability analysis of the ACS and CCS scales shows
acceptable Cronbachs alphas; whereas reliability of the NCS scale is
weak. The CFA test suggests the new model fits the data well. The
correlation between the ACS and NCS factors for the original and
modification scales is not high enough to cause a differentiation problem,
indicating that the two scales do not overlap significantly. Further, test of
the four-factor model indicates that the two CCS subscales, personal
sacrifices associated with voluntary turnover and lack of job alternatives,
do not discriminate from each other.
Bergman (2006) together with others (Carson & Bedeian, 1994;
Cohen, 1993, 1996; Yousef, 2000) called for scale revision of Allen and
Meyers Three-Component conceptualization of organizational commitment. Despite the popularity of the model, research shows there are
consistent issues with the factor structure and item discrimination. The
current study responds to this need by validating further the factor
structure of the TCM model in a new population, presidential assistants
in higher education in the United States.
The results of this study further strengthen the need for organizational
commitment researchers to rewrite at least some of the NCS items to
make them more construct specific. However, this study does not suggest
the necessity to split the CCS scale into two subscales, as some literature
indicated (Carson & Carson, 2002). This inconsistency calls for further
validation of the CCS scale.
There are a few limitations associated with this study. First,
presidential assistants responded to the survey from their individual
perceptions that included personal definitions and life experiences, which
could provide socially desired answers. Second, the parameters of data
collection through an online survey may limit the response rate because
not all presidential assistants accept this medium. Third, this was a crosssectional study of presidential assistants at a particular time. Some of the
organizational commitment components may show changes over time.
Hence, a longitudinal study may result in different conclusions.
308
Xu & Bassham
309
310
Xu & Bassham
311
Copyright of North American Journal of Psychology is the property of North American Journal of Psychology
and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright
holder's express written permission. However, users may print, download, or email articles for individual use.