Sample Attrition in the Presence of Population Attrition"

This paper develops a unique method that accounts for sample attrition in the presence of population attrition for use with an attriting panel and repeated cross-section samples. The key idea is to use the availability of representative cross-sections as the basis for weighting the persons in the balanced panel (Hirano, Imbens, Ridder, and Rubin, 2001). In the presence of population attrition, however, only the first period cross-section is representative. The paper proves that a population attrition process can be identified and estimated without knowing which missing observations are due to population attrition (although much more restrictive assumptions are needed about the determinants than is necessary to handle sample attrition alone). By virtue of the this technique, we can obtain weights for the persons in the balanced panel. The method is applied to a panel sample of immigrants in the United States where selective return migration is the source of population attrition.