The study we consider the HES data of Bangladesh (1995/96) covering both rural and urban areas. From micro data we get all types of information related to expenditure, income or anything else for every household. The study considers households from rural (504) and urban areas (238) of six divisions in Bangladesh namely: Dhaka (226), Khulna (96), Chittagong (134), Rajshahi (176), Barisal (72) and Sylhet (38). A two-stage stratified random sampling technique was followed in drawing the sample for HES 1995-96 under the framework of Integrated Multipurpose Sample (IMPS) design developed on the basis of population and Housing Census 1991. This design consists of 372 Primary Sampling Units (PSU) throughout the country. Each PSU comprises of around 250 households. In first stage, a total of 372 PSUs were drawn from the sample frame with probability proportional to size (PPS). In the second stage, 20 Households were selected from each PSU by systematic random sampling method. Amongst the 372 PSUs selected for the 1995-96 HES, a total of 7,420 households were interviewed. We consider the following information as the variables for the purpose of our study: (1) per capita expenditure on each food item and (2) per capita total income.We consider major 22 food items, which are as follows: (!) Rice, (2) Wheat, (3) Fish, (4) Mosur, (5) Potato, (6) Green Vegetables, (7) Leafy Vegetables (Shak), (8) Chicken, (9) Meat, (10) Egg, (1 1) Green Chili, (12) Onion, (13) Garlic, (14) Salt, (16) Mustard Oil, (16) Dry Fish, (17) Ghee, (18) Milk, (19) Sugar, (20) Molasses, (21) Fruits and (22) Puffed rice. In our study, the Engel function is estimated for the major 22 food items. For the six food items (rice, fish, potato, green vegetables, onion, salt) for which zero expenditure is less than 20%, OLS procedure is applied. The other sixteen food items (Wheat, Mosur, Leafy Vegetables (shak), Chicken, Meat, Egg, Green Chili, Garlic, Mustard Oi l, Dry fish, Ghee, Milk, Sugar, Molases, Fruits, Puffed rice) for which zero expenditure is 20% or more, Tobit model is applied.