Projections of world food demands hinge critically on the underlying functional form used to predict future demands. Simple functional forms can lead to unrealistic projections by failing to capture changes in income elasticities of demand as consumer becomes wealthier. This paper compares several demand systems in the projection of disaggregated food demand across a wide range of countries with different income levels using a global general equilibrium model. We find that the recently introduced AIDADS system represents a substantial improvement over existing demand systems currently in use in CGE modeling. In particular, our projection results show that for relatively poor regions experiencing rapid income growth, the widely used LES and CDE demand systems tend to over-predict growth in consumer demand, and hence import and output requirements for food products and under-predict that for non-food products, compared to the AIDADS system. On the other hand, for high-income regions with modest income growth, the choice of functional form is less critical.