This paper addresses the possibility of biased estimates of the determinants of food expenditures of households eligible for the food stamp program. The source of such bias, sample partitioning, is describes and a correction procedure is utilized. For eligible nonparticipants' food expenditure estimates, bias was found but not for participants