Assessing the Quality of Household Panel Data: The Case of the Panel Study of Income Dynamics

Evidence from a number of methodological studies are used to assess the overall quality of data from the Panel Study of Income Dynamics (PSID). Despite substantial cumulative non-response over the nearly two decades spanned by the study, the sample is found to maintain its representation of the nonimmigrant population of the United States. The most important reasons for this result are that the study's following rules insure that the sample replaces itself in the same manner as the population (through the formation of new families by the offspring of old) and that nonresponse is largely unsystematic. Nonresponse also appears to be largely random with respect to parameters in a number of behavioral models. The accuracy of measures is assessed by comparing survey measures with national aggregates and with highly accurate individual validating data. PSID reports of transfer income appear to compare more favorably with program aggregates than do reports from other large-scale surveys such as the Current Population Survey. Finally, although PSID survey measures generally are unbiased when compared to validating data, they contain amounts of measurement-error variance that range from trivially small to very large.