Part A
Employment satisfaction plays an important role in one’s wellbeing and productivity. Furthermore, education increases employment satisfaction by increasing hourly income and employment security (Fabra & Camison, 2009) .
Blanchflower & Oswald (1996) found a positive effect of education on employment satisfaction (when controlling for variables such as expectations and income). However, when not controlling for expectations, or income, they found a negative direct effect of education on employment satisfaction. This is due to increased expectations that education places on the individual. When expectations are increased and income and occupation type remain the same, employment satisfaction decreases.
Arrazola et al., (2000) found that income is dependent on one’s level of education, where more skills leads to higher wages. Therefore, as education increases, hourly income increases.
Furthermore, increases in hourly income have shown to have a positive relationship with one’s employment satisfaction due to increasing one’s capacity for the consumption of goods and social prestige (Lydon & Chevalier, 2002). Therefore, as hourly income increases, employment satisfaction increases.
Another variable that effects employment satisfaction is that of the occupation type Clark (1998) found a positive relationship between education and occupation type, where highly educated individuals are more likely to obtain more secure employments that are full-time rather than part-time. The study argued that full-time employments create a higher sense of employment security therefore positively increasing ones’ level of employment satisfaction.
The aim of the study was to examine the effects education (X1) has on employment satisfaction (X4) mediated through occupation type (X2) and hourly income (X3). Furthermore, it aimed to support the empirical research of the direct effects of the included variables. That is (1) that education has a negative direct effect on employment satisfaction, (2) that those with higher levels of education are more likely to work full-time, (3) that those who work full-time earn a higher income, (4) that an increase in hourly income has a positive direct effect on employment satisfaction and finally (5) education should have a positive effect on employment satisfaction when mediated through occupation type and hourly income.
The proposed model
Using Structural Equation Modelling (SEM) a model of the relationships between education, occupation type, hourly income and employment satisfaction was made. SEM uses a series of simultaneous multiple regression analyses to determine the path coefficients of each variable. First, a causal hierarchy was constructed such that the first variable (education) was exogenous and the subsequent variables (occupation type, hourly income and employment satisfaction) were endogenous (see Fig. 1).
The variables were coded in the corresponding manner;
X1: Education = EDUC
X2: Occupation type = OCCU
X3: Hourly Income = HOUR
X4: Employment Satisfaction = SATI
Fig. 1. The causal hierarchy of the education on employment satisfaction.
Fig. 1, demonstrates that education (EDUC) was the exogenous variable, meaning it can have a direct or indirect effect on all subsequent variables but cannot be affected by the other variables in the model. Next in the hierarchy was occupation type (OCCU). The type of occupation one has (either part-time or full-time) cannot influence one’s education, however it can influence one’s hourly income and employment satisfaction directly or indirectly. Following occupation type is hourly income (HOUR). Hourly income was a variable that can only have a direct effect on employment satisfaction and cannot influence one’s level of education or their occupation type. The final variable in the model was employment satisfaction (SATI) the dependant variable (DV).
Participants were Australian citizens between the ages of 25-66 who were employed either part-time or full-time. The ages started at 25 to account for the time it takes to acquire higher education and find work, and limited to 67, the retirement age in Australia. The collection of data was collected through census data and self-report surveys. Education level was measured on a scale (school certificate, higher school certificate, advanced diploma, undergraduate, graduate certificate, graduate diploma, master’s degree and doctoral degree). Occupation type was a dichotomous variable coded 0 for part-time employment and 1 for full-time employment for the participants current occupation. Hourly income was a continuous scale based on the current occupation of the individuals surveyed. Employment satisfaction was measured on a continuous scale for the current occupation they were employed where a higher score equals higher employment satisfaction.
Table 1
Correlations Matrix
EDUC OCCU HOUR SATI
EDUC 1
OCCU .241 1
HOUR .375 .212 1
SATI -.325 .458 .428 1
Statistical analysis and results
A SEM with 3 models was produced where;
• Model 1: Regresses OCCU on EDUC
• Model 2: Regresses HOUR on EDUC and OCCU
• Model 3: Regresses SATI on EDUC, OCCU and HOUR
Fig. 2 Model with standardised path coefficients
Hypothesis 1. Education has a negative direct effect on employment satisfaction
The regression coefficient of education on employment satisfaction supports this hypothesis finding a large negative direct effect of -.658. This indicates that holding constant occupation type and hourly income, individuals who are highly educated tend to report lower levels of employment satisfaction. This shows that an increase in education by one standard deviation unit leads to a decrease in employment satisfaction by .658 units supporting previous literature.
Hypothesis 2. Those with higher levels of education are more likely to work full-time
When regressing education on occupation type, the regression model produced a standardised coefficient of .241 supporting the second hypothesis. This indicating that an increase in one standard deviation in education results in an increase of .241 standard deviation units of occupation type. Therefore, those with higher levels of education tend to have full-time occupations than those with lower education qualifications.
Hypothesis 3. Those who work full-time earn a higher hourly income
The regression equation found support for the third hypothesis with the direct effect of occupation on hourly income having a small positive effect of .129 standard units. This indicates that holding constant one’s level of education, individuals who work full-time occupations tend to earn .129 standard unit’s higher levels of income than those employed as part-time.
Hypothesis 4. An increase in hourly income has a positive direct effect on employment satisfaction
The results also found support for the fourth hypothesis, finding a direct effect of hourly income on employment satisfaction with a large positive effect of .570 standard units. This shows that while holding constant education and occupation type, individuals who earn more hourly income tend to report higher levels of employment satisfaction. Those who earn a higher income hourly, show an increase in employment satisfaction by .570 standard units.
Hypothesis 5. Education has a positive effect on employment satisfaction when mediated through occupation type and hourly income
Finally, the model found support for the fifth hypothesis. The indirect effect of education on employment satisfaction via occupation type and hourly income was .018, indicating that individuals who are highly educated tend to have greater employment satisfaction because they are more likely to be working full-time of which are also more likely to be making a higher hourly income. Therefore, while allowing education, occupation type and hourly income to vary, those with higher education tend to have higher employment satisfaction when mediated through occupation type and hourly income. Although education has a direct negative impact on employment satisfaction, it is countered when mediated through occupation type and hourly income.
Other findings.
Effects on hourly income
The direct effect of education on hourly income was a moderate positive effect of .344 standard units. Therefore, holding constant occupation type, individuals who are highly educated tend to earn a higher hourly income.
An indirect effect of education on hourly income via occupation was .031 standard units. Therefore, when allowing occupation type to vary, those who are highly educated tend to make a higher hourly income because higher education leads to more full-time occupations and full-time occupations tend to have higher hourly incomes.
Total effects on hourly income
The total effect of education on hourly income was equal to the direct effect of education on hourly income (.344) and the indirect effects on hourly income (.031), which equates to .375 (the correlation between education and hourly income).
The other total effect calculated was occupation type on hourly income. The direct effect of occupation type on hourly income was .129 which does not equal the correlation of occupation type and hourly income (.212). This means that a spurious component is occurring and accounts for .083 units. The spurious component is a component of the correlation that is occurring because of some higher variable (education) accounting for the covariation that exists between the direct effect and the mediating variable. This indicates that a standard deviation increase in education leads to a common increase in occupation type and hourly income such that the total effect of occupation on hourly income has changed and is smaller (in relation to the correlation).
Effects on employment satisfaction
Results also show a direct effect of occupation type on employment satisfaction having a moderate to mostly large effect of .496 standard units. This indicates that while holding constant education and hourly income, individuals who have full-time occupations tend to report higher levels of employment satisfaction. This shows that those who work full-time show an increase in employment satisfaction by .496 standard deviation units compared to those who work part-time.
The indirect effect of education on employment satisfaction via occupation type was .120, indicating that individuals with higher levels of education tend to have greater employment satisfaction because higher education leads to full-time work (which is more secure).
The indirect effect of education on employment satisfaction via hourly income was .197, indicating that individuals with higher levels of education tend to have greater employment satisfaction because higher education leads to greater hourly income.
The indirect effect of occupation type on employment satisfaction via hourly income was .074, indicating that individuals who work full-time tend to have slightly higher employment satisfaction because they have a higher hourly income.
Conclusion
The SEM supported the previous literature finding that (1) education has a large negative direct effect on employment satisfaction, (2) that those who are educated tend to have more secure full-time occupations, (3) that an increase in hourly income had a large positive direct effect on employment satisfaction and (5) that education had a small positive effect on employment satisfaction when mediated through occupation type and hourly income.
Part B
Suppose a prior variable (X0) exists to influence X1, the first variable in your model (such that your first variable is not truly exogenous). What impact might this prior variable (X0) have on your account of the model in Part A?
To see the effect a prior exogenous variable would have on the model, the role an exogenous variable must be outlined. The exogenous variable is the variable that is not caused by other variables in the model and has the largest influence on the model via the subsequent endogenous variables. Endogenous variables are effected by the proceeding variables that are either the exogenous variable or other endogenous variables. Therefore, disturbances/spurious components will exist among them. As such if an exogenous variable (X0) were introduced into the model in Part A, that exogenous variable would then have the largest influence on the model and effect the direct and indirect path coefficients of that model. As education now turns into an endogenous variable its path coefficients will be influenced in some way by the new exogenous variable.
For example, Socio-economic status (SES) could be identified as an exogenous variable that positively impacts education, occupation type, hourly income and employment satisfaction. It could also be argued that higher SES could lead to higher employment satisfaction, better education, working full-time and higher hourly income having a positive effect on variables. If SES were the exogenous variable, it would then influence the subsequent variables in the model. In Fig. 3, it can be seen that the direct effect of education on occupation type has changed from the model in Part A. In Part A, the standardised direct effect was equal to the total effect of education on occupation which was .241. This was the same as the correlation between the two because education was the exogenous variable and therefore not influenced by any other variables in the model. In the new model however, the direct effect of education on occupation type has decreased to a standardised score of .159 whilst the correlation of the two has remained the same (.241). Therefore, there is a discrepancy of .241 – .159 = .082 that has been developed due to the exogenous variable. Furthermore, when comparing the new model in Part A, it is evident that all of the path coefficients have reduced indicating that the introduction of SES in the model as the exogenous variable has effected both the direct and indirect path coefficients in the model. The total effect on occupational type was .241 in Part A, which is the same as the new model (total effect = direct effect + indirect effect = .159 + [.520 x .159] = .241) however, the exogenous variable impacted the direct effect education had on occupational type. The total effects are however, affected by the correlation the exogenous variable on them.
Table 2
Correlation matrix where SES exists to influence the model
SES EDUC OCCU HOUR SATI
SES 1
EDUC .520 1
OCCU .241 .241 1
HOUR .375 .375 .212 1
SATI .325 -.325 .458 .428 1
Fig. 3. The inclusion of SES as an exogenous variable
Suppose instead of (i) above, another variable (Xnew) is included in your model, between X2 and X3, so that your model is re-specified as a five-variable causal hierarchy. What impact might the inclusion of an additional mediating variable have on the original account of your model?
For this example, lets argue that productivity (PROD) can be introduced between X2: OCCU and X3: HOUR. The inclusion of XNEW generates new path coefficients (from EDUC to PROD and OCCU to PROD). Furthermore, it will generate direct effects from PROD itself to both HOUR and SATI. Consequently, more indirect effects will be generated. The model in Part A had five indirect effects, a model with 5 variables will add an extra 11 indirect effects.
When comparing the new model to Part A, it is evident that the direct effect of education on occupation remains the same (.241) yet all of the direct effects on employment satisfaction have changed. The direct effects have changed because another variable has been added to the regression equation. What remained the same however were the total effects of education on employment satisfaction (see Table 8 in Appendix). The total effect of education and employment satisfaction did not change because the correlation between the two did not change even after adding the new variable. Therefore, the partitioning of the total effect has changed. In Part A, the total effects of education on employment satisfaction was -.324. This new model introduces a variable between occupation type and hourly income changing the partitioning of the path coefficients and keeping the same total effect (-.324 see table in appendix).
Table 3
Correlation matrix where another variable productivity was introduced into the model
EDUC OCCU PROD HOUR SATI
EDUC 1
OCCU .241 1
PROD .355 .675 1
HOUR .375 .212 .120 1
SATI -.325 .458 -.135 .428 1
Fig 4. Adding productivity into the model between occupation and hourly income.
References
Arrazola, M., Hevia , J., Risueno, M., & Sanz, F. (2000). The effects of Human Capital depreciation on experience-earning profiles: Evidence from salaried Spanish men. Serie Economia, 4-100.
Blanchflower, D., & Oswald, A. (1996). The rising well-being of the young. North Carolina: NBER.
Clark, A. (1998). Labour market and social policy – occasional measures of job satisfaction. What makes a good job? OECD.
Fabra, M., & Camison, C. (2009). Direct and indirect effects of education on job satisfaction: A structural equation model for the spanish case. Economics of Education Review, 600-610.
Lydon, R., & Chevalier, R. (2002). Estimates of the effect wages on job satisfaction. CEP-LSE Discussion Papers, 5-31.
Appendix
SPSS Syntax
MATRIX DATA VARIABLES = EDUC OCCU HOUR SATI
/CONTENTS corr mean stddev /N = 200.
BEGIN DATA.
1.0
.241 1.0
.375 .212 1.0
-.325 .458 .428 1.0
0 0 0 0
1 1 1 1
END DATA.
REGRESSION MATRIX = in(*)
/VARIABLES = EDUC OCCU HOUR SATI
/DEPENDENT = OCCU /ENTER EDUC.
REGRESSION MATRIX = in(*)
/VARIABLES = EDUC OCCU HOUR SATI
/DEPENDENT = HOUR /ENTER EDUC OCCU.
REGRESSION MATRIX = in(*)
/VARIABLES = EDUC OCCU HOUR SATI
/DEPENDENT = SATI /ENTER EDUC OCCU HOUR.
SPSS output
Direct and indirect calculations
Table 5
Direct and total effects of education on occupation type.
OCCU
Direct .241
TOTAL .241
Table 6
Direct, indirect and total effects on hourly income from Model 2.
Calculation SD Units
EDUC
Direct .344
Via OCCU .241 x .129 .031
TOTAL .375
OCCU
Direct .129
TOTAL .129
Table 7
Direct, indirect and total effects on employment satisfaction from Model 3.
Calculation SD Units
EDUC
Direct -.658
Via OCCU .241 x .496 .120
Via HOUR .344 x .570 .196
Via OCCU and HOUR .241 x .129 x .570 .018
TOTAL -.324
OCCU
Direct .496
Via HOUR .129 x .570 .074
TOTAL .570
HOUR
Direct .570
TOTAL .570
Table 8
Total effects of education on employment satisfaction for Part B ii.
NEW for Part Bii Calculation SD Units
EDUC
Direct -.510
Via OCCU .241 x .889 .214249
Via PROD .204 x -.614 -.125256
Via HOUR .380 x .505 .1919
Via OCCU and HOUR .241 x .240 x .505 .0292092
Via OCCU and PROD .241 x .626 x -.614 -.092631724
Via PROD and HOUR .204 x -.177 x .505 -.01823454
Via OCCU, PROD and HOUR .241 x .626 x -.177 x .505 -.01348515741
TOTAL -.324 (rounded to 3 dp)
Table 9
Spurious components on employment satisfaction
Variable Correlation with SATI = Total Effect + Spurious Component
EDUC -.325 = -.324 + -.001
OCCU .458 = .570 + -.112
HOUR .428 = .570 + -.142