The Importance of Sample Attrition in Life Cycle Labor Supply Estimation

Abstract

We examine the importance of possible non-random attrition to an econometric model of life cycle labor supply including joint nonlinear taxation of wage and interest incomes and latent heterogeneity. We use a Wald test comparing attriters to nonattriters and variable addition testing based on formal models of attrition. Results from the Panel Study of Income Dynamics are that non-random panel attrition is of little concern for prime-aged male labor supply estimation because the effect of attrition is absorbed into the fixed effects. Attrition is less econometrically influential than research design decisions typically taken for granted; the wage measure or instrument set has a much greater impact on the estimated labor supply function of prime-aged men than how one includes panel attrition.