Two approaches have been used in the literature to detect gender bias in the intrahousehold allocation of consumption or expenditure: The direct comparison of expenditure on males and females where data are available at the level of the individual and the indirect household expenditure methodology, commonly referred to as the Engel curve approach. Since information on the consumption of or expenditure on each individual member of a household is typically not available in household surveys (where generally only total household expenditure on specific items is available), it is usually not possible to directly observe gender bias in the allocation of expenditure within the household. A researcher must therefore use an indirect method. The Engel curve method seeks to detect differential treatment within the household indirectly by examining how household expenditure on a particular good changes with household gender composition. However, the reliability of the Engel curve methodology as a way of detecting gender bias has been called into question because it has generally failed to confirm discrimination even where it is known to exist, Deaton notes that "it is a puzzle that expenditure patterns so consistently fail to show strong gender effects even when measures of outcomes show differences between girls and boys" (240). Case and Deaton (2003, 11) say "it is not clear whether there really is no discrimination or whether, for some reason that is unclear, the method simply does not work." Ahmad and Morduch (2002, 17) say, "coupled with evidence on [significant gender differences in] mortality and health outcomes, the results on household expenditures pose a challenge in understanding consumer behaviour. This article tests two potential reas ons for this puzzle. First, there are two possible channels through which pro-male bias may occur in expenditure on any particular commodity: (1) via zero purchases for daughters and positive purchases for sons and (2) conditional on positive purchase for both daughters and sons, via lower expenditure on daughters than on sons. If gender bias operates through only one of these mechanisms, then averaging across the two mechanisms may lead to the conclusion of no significant gender bias. Second, there is the issue of the effect of aggregation (of expenditure data across individuals within the household) on the ability to detect gender bias in household expenditures. It may be that somehow aggregation mutes gender effects. © 2005 by The University of Chicago. All rights reserved.
