The Effects of Income, Assets, Food Programs, and Household Size on Food Consumption

Abstract

The availability of food to American households is a continuing public problem. There is concern about adequate amounts of food for households and about the coverage and high costs of food distribution programs. Yet the relationships between food availability and its determinants, including food distribution programs, have not been examined closely. This is partly because major increases in these programs have occurred recently and partly because data on all major sources of food for households have not been available. Studies by Prais and Houthakker, Friedman, Currie, and others have examined income food expenditure relationships among European and U.S. households. However, these investigations were conducted prior to the implementation of sizable food distribution programs. Specific questions that have not been fully answered are: How is the value of food consumption affected by receipt of "bonus" food stamps and free lunches?' Have the relative impacts of income and household size on food expenditures changed? The purpose of this paper is to examine these questions using a 1972-73 sample of households containing eight-to twelve-year-old children in the state of Washington. The Sample The sample consisted of households containing a total of 995 children between the ages of eight and twelve. The sample was stratified by poverty level and ethnic group and contained larger numbers of Blacks and Mexican-Americans than would have been obtained by a random sample of the state's population." Donald A. West and David W. Price are associate professors and associate agricultural economists two anonymous reviewers, and the editor of the Journal for helpful comments and constructive criticisms. The authors are responsible for errors or deficiencies. I "Bonus" food stamps are defined as the face value of food stamps minus their cost to the recipient. 2 The entire sample is in essence six different samples each representative of six different strata or populations within the The child is the unit of observation, and the sample generalizes to the socioeconomic background of the child rather than the household. Household size in this sample is larger than a random sample of households of the state would yield for three reasons. First, the household was required to have an eight-to twelve-year-old child, which excluded all one-person and most two-person households. Second , Blacks and Mexican-Americans have larger households than Anglos. Third, the child being the unit of observation gave a higher probability of the larger household being selected. Where siblings were included, the household was assigned a multiple weight. There are several advantages of this type of sample. The disproportionately larger percentage of households with low incomes provides a larger number of observations at the low income level, while still retaining an adequate percentage (62%) with incomes above 125% of poverty level. All households are at approximately the same stage in the family life cycle, most heads were thirty to fifty years of age, so relationships can be es timated net of the influence of life cycle (see Price 1969). Forty-nine percent of the sample were Anglos, 26% were Blacks, and 25% were Mexican-Americans (table 1). Thirty-eight percent of the sample were in households with incomes below 125% of the poverty level. Just over one-fourth of the sample were in households receiving food stamps, and more than one-third received free school lunches. Selection of Variables The dependent variable in the analysis-value of food consumed-consists of all food items purchased plus food from other sources. Food from other sources includes food purchased with bonus food stamps, free U.S. Department of Agriculture commodities, free school lunches, gifts of food, food produced in gardens or on farms by the house-state. These strata are income categories above and below 125% of the poverty level for each of three ethnic groups: Mexican-Americans, Blacks, and Anglos. School districts were stratified by size and geographic area within the state. Probability samples of districts and individual schools were then drawn. Within each school, lists of children in each of the six strata were compiled and random samples were drawn from each list.