The problem addresses in this article is the bias to income and expenditure elasticity estimates caused by inattention to measurement error and unobserved heterogeneity. We gauge these biases by comparing cross-sectional, pseudo-panel and true panel data from Polish and American expenditure surveys.