Disaggregated household meat demand with censored data

Previous research on meat demand has generally used highly aggregated data (across time, products and consumers). However, meat products across species are likely stronger substitutes than some products from the same species. Further, demand for specific meat products would be expected to respond differently to market information about food safety or other events. This study uses monthly consumer panel data collected between 1992 and 2000 to estimate a disaggregated meat product demand system. The use of the expectations maximization algorithm is introduced to estimate a demand system that adjusts for the econometric problem of censored data resulting from purchased shares of some products by individuals often being equal to zero. Results indicate that certain individual meat products have noticeably different own-price elasticities than existing aggregate meat product estimates of their respective species. Some individual meat products have stronger substitutes across species than within species (e.g. beef steak and pork chops are substitutes, but beef roast and ground beef or not substitutes for steak).