When considering changes in welfare caused by policy interventions it is good to know how individuals within households are affected by a policy. Many sources of data give information about household aggregate expenditure and consumption, but few sources give information about the individual expenditures and consumptions that contribute to these totals. For private goods such as foods the sums of individual consumptions and their associated expenditures are equal to household consumption and expenditure. This research examines the feasibility of inferring individual consumption of private goods from household-level, aggregate data. Initial results were reported in 'Diet Revealed? Semiparametric Estimation of Nutrient Intake – Age Relationships (with Discussion)', Andrew Chesher, Journal of the Royal Statistical Society A, 1997, Vol 160 and in a paper given at the LACEA conference in Rio de Janeiro 2000.

Current research focuses on an identification issue hinging on the extent to which consumption by a person of one type depends upon household composition, that is the extent to which there is interdependency in consumption within the household. It turns out that absence of interdependency in personal consumption is a testable hypothesis using household-level data alone. However while the test can in principle reject the hypothesis it cannot confirm it in the sense that there are certain modes of interdependency that cannot be detected using household-level data alone. In part of this research we consider the construction and properties of tests of the interdependency hypothesis. We study a variety of survey data sets in which individuals' food consumptions are recorded and examine the strength of consumption interdependency and the extent to which individual consumptions inferred from household level data are poor indicators of actual individual consumptions. The possibility of combining information from individual-level and household- level data is among the other research questions being studied.