Biases in Standard Measures of Intergenerational Income Dependence

Abstract

Estimates of the most common mobility measure, the intergenerational elasticity, can be severely biased if snapshots are used to approximate lifetime income. However, little is known about biases in other popular dependence measures. Using long Swedish income series, we provide such evidence for log-linear and rank correlations, and rank-based transition probabilities. Attenuation bias is considerably weaker in rank-based measures. Life-cycle bias is strongest in the elasticity; moderate in log-linear correlations; and small in rank-based measures. However, with important exceptions: persistence in the tails of the distribution is considerably higher, and long-distance downward mobility lower, than estimates from short-run income suggest.