This article replicates Guyer's finding in Marginal Gains (2004) of a social gradient in expenditure patterns of Ghanaian households using more flexible statistical techniques than those used in the book. We show that similar gradients are found in Côte d'Ivoire and in Kagera, Tanzania, suggesting that Guyer's finding in Ghana is a manifestation of a more general phenomenon. In addition, we examine patterns of measurement error in household expenditure data from Ghana. This reveals a worrying possibility that survey reports of expenditure may reflect respondents' beliefs about what expenditures should be, as well as actual expenditures within the household.