Sample selection rules and the intergenerational correlation of earnings

Abstract

This paper investigates the sensitivity of estimates of the intergenerational correlation of earnings to different sample selection rules. Recent articles report father–son correlations to be on the order of 0.4. Those estimates, however, are based on samples which exclude observations with low or zero earnings. Since events such as unemployment are common, it is not clear that such episodes should be excluded. We show that estimated correlations are quite sensitive to the selection rule used. The sensitivity of estimates to selection rules suggests one should be cautious about using recent estimates to infer the degree of intergenerational mobility.