Panel Conditioning Effects in the Survey of Income and Program Participation

1984 panel) by the U.S. Bureau of the Census with the principal objective to provide policy-makers with more accurate and comprehensive information on income and participation in government programs than were available through other data sources. The survey results were intended to inform policy in the areas of welfare and tax reform, and improvement to entitlement programs such as Social Security and Aid to Families with Dependent Children (AFDC). Interviews of panel members are conducted every four months for seven or eight consecutive interviews. A new sample or panel is introduced each year. The survey may thus be thought of as a rotating panel survey. An overview of the SIPP program is found in Nelson, McMillen, and Kasprzyk (1985). In a rotating panel design such as the SIPP, i t is possible to compare estimates based on the responses given by two different panels for the same time period. In the absence of nonresponse, such a comparison will indicate the presence of panel conditioning effects. The l iterature concerning panel conditioning has recently been strengthened by a number of carefully designed studies that have been able to identify changes in aggregate responses given by panel members as the number of interviews increases. The causes of the observed panel conditioning are not known. It may be due to a change in the behavior of the respondent as a result of panel membership or a change in how respondents answer the same question over time. Whatever the mechanism at work, panel bias has been observed in a variety of settings. Kalton, Kasprzyk, and McMillen (1989) review a number of these, including studies of consumer expenditur