Alternative interpretations of hours information in an econometric model of labour supply

This paper examines the labour supply behaviour of married women in France. Estimating a model with tax parameter variation, careful re-examination of the treatment of the unearned income variable and taking account of education in modelling preferences result in substantially lower elasticities than found in our previous empirical analysis. It turns out that distinguishing between part-time, full-time and long hours gives virtually the same results as treating observed hours as reflecting desired hours. We provide extensive specification diagnostics, including Heckman-Andrews tests, as well as Hausman tests for the comparison of different handlings of the hours information. We also consider different assumptions concerning the perception of the impact of the tax system and provide some evidence in favour of a correct perception.