Estimating Food Demand in Sistan and Baluchestan Using Two Systems of NNDS and QUAIDS

In this study, the neural network demand System and the quadratic almost ideal demand system are used to estimate food demand in Sistan and Bluchestan province. For this purpose, micro data of urban household consumption in 2011 were used. The accuracy of two systems was estimated. The food items were classified into nine groups by COICOP to estimate demand. The NNDS results which are estimated using feedforward artificial and multilayer perceptron shows that MSE of all groups using this system has an equal MSE or even a smaller one comparing to QUAIDS. The results indicated that in most groups Engle curves for the NNDS was closer to the real curves. The expenditure elasticity of NNDS was almost equal to QUAIDS. The results of both systems showed that expenditure elasticity for the groups of bread-grains, milk products, vegetables-cereals, sugar-confection, flavoring-spices and different kind of beverages were less than one and groups of meat, fat-oils and fruit-nuts were more than one.