Multiple-group Analysis
In this study, the conceptual model (see in Figure2) encompasses three latent variables. They
are Job Autonomy, Job Anxiety and Job depression. To test the validity and reliability of each
latent variable, this study uses the exploratory factor analysis (EFA). The test result shows
that setting three latent factors are sufficient (See in Table 7). The chi-square statistic is
244.32 on 25 degrees of freedom. The p-value is below 0.01. All the coding for Variables
measurement is described in Table 2.
Table 7 Factor Analysis
Factor1 Factor2 Factor3
aut1 0.77
aut2 0.79
aut3 0.84
aut4 0.81
aut5 0.59
JA1 0.66
JA2 0.76
JA3 0.71
JD1 0.66
JD2 0.77
JD3 0.77
Measurement equivalence will be tested by using multiple-group comparisons with nested
models. Out of the variety of possible fit indices, this study will present the chi-square
statistics, the comparative fit index (CFI) and the root mean square error approximation
(RMSEA) (Steiger, 1989). Although values greater than .90 are considered to represent a
good fit in terms of the CFIs and values greater than .80 are considered acceptable Although
values greater than .80 are considered acceptable, the RMSEA should be less than .05 for a
good fit and less than .08 for a still reasonable fit of the data to the model (Browne and
Cudeck, 1993). This study will also use ΔCFI as an indicator in the comparison of models.
According to Cheung and Rensvold (2002), values greater than 0.01 indicates a significant
drop in fit (see Table 8). The multiple-group comparison showed a good fit, indicating that
model 4 (very strong invariance) can be accepted for all versions (see Table 8). Under the
constraint of equal factor loadings (measurement weights), Intercepts and means, a significant
increase of the chi-square statistic could be observed, although all other fit statistics point
toward a satisfying fit for the multiple-group comparison. The assumption of equal factor
variances is also supported by the ΔCFI.
Table Multiple-Group Comparisons of Measurement Invariance
χ2 df p-value CFI RMSEA BIC ΔCFI
Model 1: configure invariance 878.54 294.00 0.00 0.99 0.04 210944.48
Model 2: weak invariance (equal loadings) 1198.75 334.00 0.00 0.98 0.04 210905.01 0.01
Model 3: strong invariance (equal loadings + 2279.91 374.00 0.00 0.96 0.06 211626.48 0.02
intercepts):
Model 4: equal loadings + intercepts + means 4135.09 394.00 0.00 0.92 0.08 213301.82 0.04
79
Moderating effect analysis
In this study, the SEM approach, put forward by Marsh, Wen, and Hau (2004), is used
to test the moderating role of financial strain on the relationships between Job
autonomy and workplace wellbeing dimensions. A series of SEM pathway analysis
are conducted among three different groups (whole dataset, the self-employed and
employees) to test the significance of job autonomy’s association with workplace
wellbeing (see Table 10). ANOVA is used to test the difference among models (Table
11). A summary of Goodness of model fit has been manifested by testing Chi-square,
the degree of freedom, CFI, TLI, RMSEA, SRMR. For the CFI and TLI values
greater than .90 are considered to represent a good fit regarding the CFIs, and values
greater than .80 are considered acceptable. Although values greater than .80 are
considered acceptable, the RMSEA should be less than .05 for a good fit and less
than. 08 for a still reasonable fit of the data to the model, and Values for the SRMR
range from zero to 1.0 with well-fitting models obtaining values less than .05
(Byrne,2013), however values as high as 0.08 are deemed acceptable (Hu and Bentler,
1999)(see Table 11).The moderating effect test is conducted by introducing
interaction term (Job Autonomy* Poverty) into the SEM model and has been tested in
three groups (the whole dataset, the self-employed and employees) as well.
Results
Multiple group Analysis
With a strong constraint on loadings and Intercept, this study measures mean differences
of three latent variables (Job Autonomy, Job Anxiety and Job depression) to investigate
the hypothesis among four groups. Firstly, this study checks the differences
80
of latent means among the self-employed groups between the lower-income and
higher-income self-employed, this study sets the group 1 as the baseline group, which
is the self-employed with income below the poverty line, to compare with group 3.
The results in Table 9 revealed that job autonomy has no significant difference in
inter-groups of the self-employed (p>0.05). In the employee group, job autonomy of
individuals with higher income is significantly higher than the employee living in
poverty (p<0.05). Then, this study make the comparisons between the self-employed
and employees at the each financial levels. It can be seen that at both levels, the self-
employed have a significantly greater extent of autonomy than employees (p<0.001),
which confirms that higher job autonomy is an occupational character of being the
self-employed. Therefore, H2a is supported, that when individuals’ incomes are below
the poverty line, the self-employed have higher job autonomy than employees. As for
the workplace wellbeing, the self-employed in poverty have not shown too much
different with the self-employed with higher income (job satisfaction) (p>0.05).
However, within the employee's group, with higher income, the employees are
experiencing a lower level of negative workplace wellbeing (job anxiety and job
depression) (p<0.01) rather than positive workplace wellbeing (job satisfaction)
(p>0.05).
For the intra-group comparison, roughly, when the self-employed and employees with
higher income, the self-employed present a significantly higher job satisfaction and
lower level of job anxiety and job depression(p<0.05). Therefore, these results support
H2b, H2c, and H2d: when the incomes are below the poverty line, the self-employed
have higher job autonomy and higher workplace wellbeing than employees.
81
Table 9 Group Latent Means Differences among Job Autonomy and Workplace Wellbeing
Job Autonomy
Job Anxiety
Job Depression Job satisfaction
∆Estimate Std.err Z-value P(>|z|) ∆Estimate Std.err
Z-value
P(>|z|) ∆Estimate Std.err Z-value P(>|z|) ∆Estimate Std.err Z-value P(>|z|)
Within the self-employed groups
Poverty VS Non-poverty 0.024 1.029 0.304 0.044 0.033
1.356 0.175
0.025 0.026 0.966 0.334 -0.074 0.054 1.370 0.171
-0.024
(Group1 VS Group 3)
Within employees groups
Poverty VS Non-poverty -0.257 0.037 -6.997 0.000 0.181 0.035
5.239 0.000
0.048 0.032 1.501
0.003
-0.090 0.061 -1.469
0.143
(Group1 VS Group 3)
The self-employed VS employees
Poverty
(Group1 VS Group 3)
Non-poverty
(Group2 VS Group 4)
-0.893
0.043 20.543 0.000 0.022 0.043 -0.524 0.000 0.120 0.037 3.254 0.001 -0.238 0.073 -3.250 0.001
-0.605
0.019 31.853 0.000 0.110 0.022 4.923 0.000 0.142 0.018 7.940 0.000 -0.401 0.036
-11.236
0.000
82
Table 10 Correlation between Job Autonomy and Workplace Wellbeing
Job Anxiety
Job Depression
Whole dataset Moderation: the Moderation: Whole dataset Moderation: the Moderation: Whole dataset
self-employed
employee
self-employed
employee
estimate p-valuee estimate p-value estimate estimate estimate p-value estimate p-value estimate p-value estimate
value
Age -0.01 0.00 0.01 -0.06 0.00 0.01 -0.06 0.00 0.01 0.00 -0.01 0.00 0.01
Gender 0.03 0.09 -0.04 -0.03 0.13 0.08 -0.03 0.13 0.08 0.00 0.00 0.72 0.22
Marriage -0.01 0.74 0.01 0.02 0.81 0.01 0.02 0.81 0.01 0.72 0.03 0.32 -0.01
Education -0.04 0.00 0.01 -0.01 0.71 0.01 -0.01 0.71 0.01 0.02 -0.01 0.25 0.03
Job Autonomy -0.15 0.00 -0.12 0.18 -0.00 0.31 -0.18 0.00 -0.31 0.00 -0.01 0.01 0.29
Poverty 0.04 0.00 0.04 0.05 0.00 0.02 0.04 0.00 0.02 0.35 0.08 0.01 -0.07
Autonomy*Poverty -0.01 0.00 -0.02 0.18 -0.00 0.01 -0.02 0.00 -0.01 0.78 -0.01 0.00 -0.01
