A Superior Measure of Consumption from the Panel Study of Income Dynamics

A number of empirical studies have used the Panel Study of Income Dynamics (PSID) to examine the behavior of consumption at the household level [Hall and Mishkin (1982), Altonji and Siow (1985) Hotz, Kydland and Sedlacek (1985), Zeldes (1985) Altonji (1986)]. Because the PSID does not report total consumption, these researchers have used instead food expenditures as a proxy for consumption. The degree to which empirical results using food consumption generalizes to total consumption is not clear. For example, Zeldes (1985) does not reject Euler equation restrictions for high-income family food consumption, while preliminary results from Flavin (1986) using a different panel survey, do reject the restrictions for total consumption. Studies using the PSID need not limit themselves to food consumption. Other components of consumption reported in the survey include the market value of owned homes, rent, utilities, number of automobiles, and a breakdown of food eaten at home and away from home. A linear prediction of total consumption based on these measures provides a more accurate measure of total consumption than simply using food consumption. Following the procedure developed by Hamermesh (1984), the Consumer Expenditure Surveys (CEX) from 1972-1973 and 1983 are used to estimate the coefficients of this prediction. It is shown in the CEX regressions that food alone explains as little as 26 percent of the variance in consumption; adding the additional consumption measures increases the explanatory power to 78 percent. Predicted consumption estimated using 1972-1973 data appears to be stable over time. The regression coefficients reported below provide the researcher with an accurate and easily implemented measure of consumption for the PSID.