Three food demand estimation procedures were compared to find which best measures the structure of food demand. These procedures were Seemingly Unrelated Regression, Block Additive, and Ordinary Least Squares. The focus was on food commodity substitution. The comparison of results was made on the basis of commodity demand and net change in calories consumed as estimated by each procedure. Each procedure projects a similar pattern of commodity demand and net caloric intake change. Overall, the Seemingly Unrelated RegressIcn procedure provides the most reasonable and consistent results.