On the role of unobserved preference heterogeneity in discrete choice models of labour supply

Abstract

The aim of this paper is to analyse the role of unobserved preference heterogeneity in structural discrete choice models of labour supply. Within this framework, unobserved heterogeneity has been estimated either parametrically or semiparametrically through random coefficient models. Nevertheless, the estimation of such models by means of standard, gradient-based methods is often difficult, in particular if the number of random parameters is high. For this reason, the role of unobserved taste variability in empirical studies is often constrained since only a small set of coefficients is assumed to be random. However, this simplification may affect the estimated labour supply elasticities and the subsequent policy recommendations. In this paper, we propose a new estimation method based on an EM algorithm that allows us to fully consider the effect of unobserved heterogeneity nonparametrically. Results show that labour supply elasticities and other post-estimation results change significantly only when unobserved heterogeneity is considered in a more flexible and comprehensive manner. Moreover, we analyse the behavioural effects of the introduction of a working-tax credit scheme in the Italian tax-benefit system and show that the magnitude of labour supply reactions and the post-reform income distribution can differ significantly depending on the specification of unobserved heterogeneity.

Iteration Capping for Discrete Choice Models Using the EM Algorithm

“…The specification of their model is in line with other recent studies employing the latent class discrete choice models (Hansen and Liu, 2011;Pacifico, 2012), which makes this setup particularly interesting from the perspective of applied economic work.…”

Labor Supply Heterogeneity and Demand for Child Care of Mothers with Young Children

“…1 The key innovation of our approach is however the incorporation of unobserved heterogeneity in the flexible form of latent classes, following Train (2008) and Pacifico (2009). We extend the treatment of unobserved heterogeneity beyond the traditional framework of random coefficient models 2 , avoiding restrictive assumptions on the distribution of the population parameters of the utility function.…”

The Effectiveness of Fiscal Stimuli for Working Parents

“…Furthermore, quasi-concavity of preferences need not be imposed ex ante, and therefore coherency of the model does not implicitly limit the range of behavioral responses that can be obtained (MaCurdy et al, 1990). We model unobserved heterogeneity using the latent classes approach as outlined in Train (2008) and Pacifico (2009), and recently applied to a model with maternal labor supply and childcare choices by Apps et al (2012). Latent classes are a flexible way of modelling unobserved heterogeneity, which can prove important for inference of the model (Pacifico, 2009).…”

“…We model unobserved heterogeneity using the latent classes approach as outlined in Train (2008) and Pacifico (2009), and recently applied to a model with maternal labor supply and childcare choices by Apps et al (2012). Latent classes are a flexible way of modelling unobserved heterogeneity, which can prove important for inference of the model (Pacifico, 2009). The identification of the structural parameters benefits from a large reform in childcare subsidies and in-work benefits for working parents in the sample period, which generates large exogenous variation in the budget sets.…”