Poverty rates calculated on the basis of household consumption expenditures are routinely compared across countries and time. The surveys which underlie these comparisons typically differ in the types of food and non-food expenditures included, often in ways which are easily overlooked by analysts. With several examples we demonstrate that these commonly occurring variations in expenditure definitions can give rise to marked differences in poverty rates where there are no real differences in well-being. We show that one approach to calculating poverty lines, used with the headcount measure of poverty, can allow comparisons based on data with different definitions of consumption. In addition to allowing comparative poverty analysis using existing survey data, the results suggest that poverty monitoring could be done effectively at lower cost by alternating detailed expenditure surveys with far more abbreviated surveys.