In contrast to the myriad of empirical work on food demand in other countries, very few studies have considered zero expenditures on some food groups. Those which have attempted have been based on techniques which result in endogeneity and inefficient estimates which in turn may misinform policy calibration. Improving on methodological flaw of previous studies, the present study censors zero expenditures in the first stage using simulation based maximum likelihood multivariate probit. In the second stage, Quadratic Almost Ideal Demand System which allows for a more realistic assumption of curvature in Engels curve is estimated. In turn, food expenditure and price elasticities are derived. In view of the high expenditure elasticities, considering a policy option that would enhance rural consumer income is desirable, since it will result in high consumption thereby providing more incentives for food production.