Volume 33, Number 2 (Spring) 1998: Attrition in Longitudinal Surveys

The Michigan Panel Study of Income Dynamics (PSID) is unique among large-scale, representative socioeconomic panel data sets in the United States in following the descendants of original sample members. Beginning in 1968 with a sample of approximately 5,000 families, interviews have been conducted annually with the original sample members and families formed by members of the original households, usually children, who have left. (1) As the panel has aged these children have made up an increasingly large proportion of the relevant sample for the analysis of many adult outcomes such as welfare dynamics and early labor market transitions.

This process will continue as parents die and are replaced in the sample by their children. But many of the original children in the PSID attrited either when their parents attrited or after they had set up their own households. The cumulative effects of even fairly low yearly attrition rates applied over a sufficiently long panel has led to the loss of roughly half of the children of original sample members by 1989. In this paper we explore the impact of this attrition on estimates based on data from the second generation. The study complements our companion paper on attrition among respondents who were adults in 1968 .2 We focus our attention on two different types of questions about the second generation. The first set of questions explores the impact of attrition on the mean characteristics (or, more generally, the marginal distribution of the characteristics) of the nonattriting second generation. To answer these questions we rely primarily on a comparison of the 1989 characteristics of surviving children in the PSID (who were 20 to 38 by 1989) to a similar sample drawn from the 1989 Current Population Survey.

John Fitzgerald is a professor of economics at Bowdoin College. Peter Gottschalk is a professor of economics at Boston College. Robert Moffitt is a professor of economics at Johns Hopkins University.