166/2013 Peter J. Hammond, Federica Liberini and Eugenio Proto

Using British Household Panel Survey data, for subjects not reporting the highest permitted satisfaction level, we show that the conditional income distribution given a higher reported level of life satisfaction first-order stochastically dominates the corresponding conditional distribution given any lower satisfaction level. Subjects reporting the highest satisfaction level, however, have an income distribution dominated by distributions for some less satisfied individuals. Interestingly, this "top anomaly" is undetectable by standard ordered probit analysis. An alternative binary probit model for reporting maximal satisfaction suggests a possible explanation: more educated subjects not only tend to have higher income, but are also less likely to report maximal satisfaction.

Behavioural Economics and Wellbeing