AbstractThe wide variety of observed behaviour that is displayed in survey data suggests that microeconometric models should allow a reasonable degree of flexibility in specification. However, since these models are generally used for policy analysis they are required to satisfy some degree of theoretical consistency. This requirement has generally led to the adoption of rather simple ‘inflexible’ models based on additive preferences or linear labour supply responses. We estimate and test a sequence of models for the labour supply of married working women, emphasizing the importance of testing both the thecretical and statistical assumptions underlying the mod