Bayesian Estimation of a Censored Linear Almost Ideal Demand System: Food Demand in Pakistan

A censored linear almost ideal demand system for food is estimated with a Bayesian Markov chain Monte Carlo procedure, using a sample of urban households from Pakistan. All own‐price elasticities but one are found to be negative, and all total food expenditure elasticities are found to be positive, with a high posterior probability. There is a mix of gross complements and substitutes among the food products, while net substitution is the predominant pattern. Household characteristics play a role in food expenditures, and regional differences exist. These demand elasticities can inform policy deliberations by the national government and international organizations.