In the study of the economic determinants of alcohol consumption, there is an under-appreciated issue of sample selectivity as consumption is often not fully observed. Sample selectivity involves ignoring consumption below a censorship cut-off level. This article estimates a demand system for beer, wine and spirits with and without allowance for selectivity. Then using a simulation approach, it examines the impact of misjudgement in the censorship level on the estimated demand parameters. A mean squared error (MSE) criterion is suggested for determining the appropriate censorship level.