SOME CHARACTERISTICS OF STRUCTURAL LABOUR SUPPLY MODELS

The choice between 'structural' models and data-descriptive, reduced-form representations of labour supply may appear straightforward. Structural models impose restrictions which may be invalid but, in exchange, provide economic interpretation. Where these structural restrictions are testable a standard approach to model selection is available. However, since labour supply is fundamentally dynamic, structural models usually require the separate specification and identification of expectation processes. Moreover, since many of the endogenous choice variables are censored or discrete, the ability to derive explicit reduced forms corresponding exactly to a given structural model is severely limited. In general, however, it would seem unwise to adopt some particular structural model without recourse to the usual battery of misspecification tests that reduced forms can provide; equally it would be sad not to recover structural parameters. The strategy developed in this chapter is to provide a sequential approach to estimation and testing, avoiding where possible unnecessarily strong structural assumptions. The theme is to assume only what is necessary to identify the structural parameters of interest, at the same time allowing the data the chance to reject the structural assumptions in question.

As economic models of labour supply have become increasingly sophisticated their econometric counterparts have become increasingly dependent on the imposition of structural theoretical restrictions in order to identify the parameters necessary to conduct policy and welfare analysis. This has been effectively illustrated by the results on econometric coherency and theory consistency in the analysis of taxation and labour supply. A well-defined econometric model of discrete or censored labour supply decisions requires a unique solution for labour supply for any wage/income combination.