The views expressed are solely attributable to the author and do not necessarily reflect those of the Census Bureau or the Commence Department.

One of the primary goals of the Survey of Income and Program Participation (SIPP) is to provide policy makers, researchers, and others with detailed information on the participation in government assistance programs by persons and households in the United States. Given the importance of the means-tested program or welfare statistics, it is important to examine the effects of attrition on program participation estimates from SIPP. In this paper, we use well-established attrition models (e.g. Hauseman and Wise, 1979; Heckman, 1976; Ridder, 1990) to examine the effects of attrition on program participation and benefits received.

SIPP is a longitudinal survey where individuals are interviewed at relatively frequent intervals (every four months) for a period of two and two-thirds years. As with other longitudinal surveys, attrition is a concern in the SIPP. Several studies have extensively examined the characteristics of attritors and nonattritors (Short and McArthur, 1986) and the cumulative sample loss rates in various SIPP panels (Jabine, 1990). Recently, some studies have directly examined the relation between attrition and SIPP estimates. Among these studies, labor income has not been found to contain attrition biases in the 1990 SIPP panel (Lamas, Tin, and Eargle, 1993). Zabel (1993) shows that attrition has no effect on labor force participation and real wages but has an effect on hourly supply of labor. Klerman (1990) indicates that there are some evidence of attrition effects on health insurance coverage for black males.

This paper attempts to bridge this gap by directly examining the effects of attrition on the estimates of determinants of major means-tested government assistance programs--namely, AFDC, food stamps, General Assistance, SSI, WIC, and Medicaid. Specifically, this study uses a model of attrition and program participation to examine the effects of attrition on the benefit estimates and the program participation status of those who are covered by these means-tested programs. This paper also compares the cumulative nonresponse rates of these programs and the characteristics of attritors and nonattritors.

This paper is organized as follows. Section II presents the model used in this study. Section III discusses regression results for food stamps, AFDC, General Assistance, SSI, WIC, and Medicaid. A brief conclusion is given at the end.