Interpreting satisfaction variables: the case of working hours#(in French)

When they are asked about their satisfaction with respect to various domains (job, financial situation, and so on) individuals give responses which may be misinterpreted because they both report their situation and their judgment about it. Although satisfaction is generally correlated with objective characteristics of the individual's situation the correlation may be spurious due to personality traits that are likely to influence expressed satisfaction. We argue that despite their hybrid nature, satisfaction variables may be informative. When they are used jointly with objective data, these variables allow to infer or estimate distributions of other variables, which are unobserved or cannot be easily measured. But this is valid on the condition that we are able to control for personality traits. To show that we take the example of satisfaction with respect to working hours. We rely on a very simple model to prove that preferences about working time may theoretically be inferred from expressed satisfaction provided that we observe the employee's actual working hours. The model allows us to identify the problems posed by the specific nature of satisfaction variables in order to deal with them adequately when empirically estimating the model. We use the 8 waves of the French part of the European Household Panel and estimate the mean value of the employee's preferred working hours. We show that empirical models that ignore the specificity of satisfaction variables, generate strongly biased estimates.