On the specification of labour supply and household production models

Household labour supply models are typically estimated on data sets containing information on family consumption and male and female hours of market work. The estimating equations are consistent with a theoretical model which assumes the maximisation of a household utility function defined on consumption and male and female leisures measured as non-market time allocations, subject to a pooled income constraint. This canonical model is generally interpreted to imply a common preference ordering achieved by consensus or imposed by a paternalistic head of household. The approach has been criticised for ignoring or abstracting from the individual decision-taking of family members. In this paper we argue that this criticism is misdirected, and that the limitations of the conventional modelling approach stem from the attempt to estimate labour supply parameters on micro data sets with missing information on individual consumptions of pure leisure and domestic goods, as well as market goods, and on domestic production. We show that even if we begin with an individualistic model, the specialisations forced on us by the data on which empirical work is based can lead to the estimating equations of the canonical model. We go on to show that considerable progress can be made by switching to time use data. We present results for a series of systems, including the canonical model, to demonstrate empirically that labour supply parameters are sensitive to the specialisations adopted in response to the limitations of the data used for estimation.