How Much Should We Trust TSTSLS Intergenerational Mobility Estimates?: Evidence From A Developing Country
30 Pages Posted: 11 Dec 2022
Date Written: November 24, 2022
Abstract
This paper revisits the Two-Sample Two-Stage Least Squares (TSTSLS) method, which is commonly used to estimate intergenerational mobility in the absence of parental earnings data. First, we decompose the TSTSLS intergenerational earnings elasticity (IGE) into the linked administrative data estimate, a projection bias, and a variance bias. We propose a parsimonious imputation procedure to eliminate the variance bias in the IGE and show that, under plausible conditions, the corrected TSTSLS IGE estimate provides a lower bound for the linked administrative IGE. Furthermore, we demonstrate that the uncorrected rank-rank correlation estimated through TSTSLS only exhibits projection bias, thus providing a lower bound to the linked administrative rank-rank correlation. Second, we use administrative data from a developing country to test our lower bound methodology through an Empirical Monte Carlo approach, confirming its validity. These estimates suggest that the following practices should be implemented when the TSTSLS method is used to estimate intergenerational mobility: i) report the variance bias corrected-IGE; and ii) report the results of the rank-rank cor- relation estimated through TSTSLS. Our empirical results shows that both estimates provide a lower bound for the linked administrative IGE and rank-rank correlation, respectively.
Keywords: Intergenerational Mobility, Linked Administrative Data, Two-Sample Two- Stage Least Squares
JEL Classification: J31, J61, J62
Suggested Citation: Suggested Citation