Spillover Bias in Multigenerational Income Regressions

Abstract

A growing literature studies long-term income persistence across more than two generations. Despite a rich understanding of measurement-related biases for the parent–child model, far less is known for the multigenerational model that captures transmission from parents and grandparents. We show that even using a 25-year income average can result in a spurious grandparent coefficient. Importantly, for a given parental measure, averaging over more years for grandparents increases spillover bias. We propose an instrumental variables (IV) approach that can more effectively mitigate bias with shorter time spans of income. With Norwegian administrative data, we reveal a positive spillover bias in the grandfather coefficients.