Household budget allocation : the interrelationship between fixed expenses and the purchase of food

This project was designed to evaluate the hypothesis that the consumption function for food changes at alternative levels of income due to the effect of a group of expenses over which the consumer unit (CU) has little control in the short run. These "fixed" expenses include shelter, utilities, debt repayment in each time period, and other expenses which must be met. The analysis assumes that food expenditure may be influenced by the fixed expenses, since the latter have a first claim on income. Total food expenditure, on the other hand, may be varied, through quality and possibly even quantity adjustments, to permit additional income to be allocated to fixed.

The changes in the marginal utilities were expected to occur at $8,000 and again at $16,000. Below $8,000, food and basic necessities are emphasized. From $8,000 to $16,000, the CU may incur proportionately more fixed expenses than the additional income should precipitate. Above that level, a more equitable relationship between food and fixed may be revealed, more similar to the case of the low income group.

The fixed expenses were expected to be responsive to the take-home pay of the CU, while food was responsive to take-home pay less fixed expenses, referred to hereafter as residual income. The objectives were to compare alternative forms of the Engel curve using residual income and food expenditures, to evaluate the effects of income and a set of socio-economic variables between income levels, to compare distributional effects on different food categories between income • levels, and to estimate elasticities with respect to both food and fixed expenses.

Data were obtained from two sources. The aggregate relationship between food and fixed expenses was estimated from the 1972-73 Bureau of Labor Statistics survey. Distributional effects were examined from data collected in a 1978 survey in Knox County, Tennessee.

Preliminary investigations revealed that only two age groups could be identified with respect to food expenditures, that hetero-scedasticity was a problem, that some groups of dummy variable slope coefficients could be eliminated, and that specified subgroups did not seem to be different in terms of vectors of estimated coefficients.

Engel curves were estimated with OLS. Elasticity estimates were somewhat higher than most earlier estimates, an occurrence that was anticipated because residual income was used in the calculation.

The expense categories fixed and food were estimated by a system of simultaneous equations. Two stage least squares was employed to obtain consistent estimates.

The basic assertion of the research was that fixed and food are interrelated, and that the relationship changes between income levels as specified above. There appeared to be no evidence to support the assertion that changes in food induced changes in fixed expenses that were consistent with this hypothesis. On the other hand, fixed caused no change in food at low incomes, and had a negative effect on food at moderate and high incomes. These effects are in line with expectations at low and moderate incomes. Other variables included in the model were consistent with the idea that various factors are more important at some income levels than at others. Consumer unit size influenced food expenditure strongly.

Elasticities from this BLS data indicated that, with respect to income, fixed expense was elastic and inferior at low income, and elastic and superior at moderate incomes. The other values were inelastic, and some were normal and some inferior. Food was positive and inelastic with respect to residual income, and inelastic and inferior with respect to fixed expenses.

The distributional effects among meats, produce, dairy, grocery, and meals out were estimated. With respect to residual income in the lower incomes, meats decreased and meals out increased. The response to changes in fixed expense was negative, except for meats. A variable designed to reflect different stages in the life cycle seemed to have some explanatory power. In general, there were few dummy variables significantly different from the base groups.

The elasticity estimates were generally less than 1. With respect to residual income, there seemed to be quite wide fluctuations between income groups, and there were indications that some variables were inferior at particular income levels.

From these results, policy implications may be drawn with respect to income assistance designed to accomplish nutritional goals. It may be the case that cash assistance would not be used to purchase food, but would instead be diverted to consumption of items which contribute to the fixed expenses.