Imputed Missing Incomes and Marginal Propensity to Consume Food

This study has demonstrated the use of a type 2 tobit model to impute the incomes for nonresponding households. The imputed incomes generated from the type 2 tobit model for nonresponding households indicate that average incomes of nonresponding households are slightly higher than those of responding households. Furthermore, this study demonstrates that under the assumption that the predicted incomes derived from the type 2 tobit model are accurate, the deletion of nonresponding households would cause an overestimation of the impact of income on food expenditures for households with annual incomes of less than $20,000. Key words: Engle curve, food expenditure, imputed income, switching regression. The estimated relationships between income and food consumption have been used to as-sess impacts on food consumption of different federal food program alternatives such as di-rect income supplementation versus food stamp supplements (West and Price; Neena