Exercise 14

If you’ve made it this far, you might want to try out some variations on the model. Let’s look at two different dependent variables - the emotional well-being score, and the overall personal WBI. Simply change the dependent variable, and run the model again.

For emotional well-being, you should get the following sets of beta coefficients and the following estimation of discrepancy:

Table 4.3. Significant beta coefficients, emotional well-being
Independent variable Extrinsic group Intrinsic group
Income 0.301 0.188
Partner -0.053 -0.112
Gender -0.099 -0.065
Squared multiple correlations 0.117 0.069


Table 4.4. Estimation of discrepancy, emotional well-being
Baseline model 3 47.948 0.000


This time all the variables are significant for both groups, but again there is a clear difference in the influence of income on emotional well-being, with it playing almost twice as large a role for the extrinsic group. Interestingly, here, you get a clear difference between the importance of having a partner for the two groups - with the intrinsic group’s well-being more determined by this variable than the extrinsic group’s is. Also, again, the discrepancy between the test model and the baseline model is significant, therefore justifying separating out the two groups and claiming a meaningful difference.

Let’s try with vitality.

Table 4.5. Significant beta coefficients, vitality
Independent variable Extrinsic group Intrinsic group
Income 0.212 0.172
Partner   -0.027
Gender -0.143 -0.084
Squared multiple correlations 0.075 0.043


Table 4.6. Estimation of discrepancy, vitality
Baseline model 3 17.612 0.001


The discrepancy between the two models is still significant, but this time it is less so. This suggests, aspirations are less important in mediating the relationship between income and vitality than they are in mediating the relationship between income and emotional well-being or life satisfaction. It would be interesting to test, using SEM, whether the remaining mediation effect on vitality might be entirely down to the lower emotional well-being of people with more extrinsic values.

It would also be interesting to test these effects for individual nations, as the above calculations were done for the ESS data set as a whole.

If you’d like to try and learn more about SEM and using AMOS in general, you might want to work through the examples in the user guide, which can be found online.

The measurement of well-being is an evolving science. If you have any particular suggestions, please email them to saamah.abdallah@neweconomics.org.