Detecting Gender Discrimination in Intrahousehold Resource Allocation

The usual methodology for measuring boy-girl discrimination in the intrahousehold resource allocation is the Engel Curve Approach proposed by Deaton (1989). This method is based on the concept of demographic separability in goods, that formalises the idea of certain goods (adult goods) being little or not at all related to some demographic groups (children). The method suggests that, by analysing the consumption on adult goods when a new child is born, it is possible to determine the existence of gender bias. However, in spite of the great popularity of this method, it fails to detect gender discrimination even in societies in which there are huge evidences of its existence. In this paper, I propose to measure gender bias by exploiting the methodological intelligence of the Deaton (1989) procedure, but testing the demographic separability in preferences instead of in goods. To make this concept feasible, I define the system of budget shares as a latent factor model in which the factors represent the underlying motives of the consumption decisions. By testing demographic separability in preferences,the main difficulties faced by the Engel Curve Approach are solved. Finally, this new procedure is illustrated by measuring gender discrimination in the commonly used data from the 1889/90 US Bureau of Labor report, which consists of 1024 budgets of British families. Two consumption drivers are clearly identified: the first one can be associated to basic necessities (e.g food), and the second to luxuries (e.g. alcohol=). In contrast with the results obtained in the literature, a strong evidence of gender discrimination is found.