Panel Attrition from the PSID: Household Income, Marital Status, and Mortality

This paper is concerned both with the determinants of panel attrition from the Panel Study of Income Dynamics and with its consequences for estimation of dynamic behavioral models which exploit the panel or longitudinal information. In particular we assess the consequences of attrition for behavioral models on (1) household income dynamics (2) adult mortality risk and (3) marriage formation and dissolution. While we find evidence of significant selectivity in attrition behavior the biases that are introduced by ignoring selective attrition are very mild. (authors)