Estimating Intergenerational Mobility with Incomplete Data: Coresidency and Truncation Bias in Rank-Based Relative and Absolute Mobility Measures

Most of the household surveys available in developing countries suffer from sample truncation because coresidency is used to define household membership. This paper provides evidence on truncation bias in the rank-based relative and absolute mobility estimates in coresident samples, and compares with the bias in intergenerational regression coefficient (IGRC) and intergenerational correlation (IGC). Using surveys from Bangladesh and India that include non-resident children, we find that the slope estimates are biased downward, while the intercept estimates are biased upward, but expected years of schooling conditional on parental education are overestimated in coresident samples. The downward bias in rank correlation is much smaller than that in IGRC, and comparable to that in IGC. The upward bias in the intercept is the largest in the regression used for IGC. Truncation bias in rank-based absolute mobility estimates is the lowest in most cases. The results add to an emerging body of evidence that the rank-based measures are more robust than the widely-used IGRC and IGC in estimating intergenerational mobility with incomplete data.