Working Paper: NBER ID: w11943
Authors: Steven Haider; Gary Solon
Abstract: Researchers in a variety of important economic literatures have assumed that current income variables as proxies for lifetime income variables follow the textbook errors-in-variables model. In an analysis of Social Security records containing nearly career-long earnings histories for the Health and Retirement Study sample, we find that the relationship between current and lifetime earnings departs substantially from the textbook model in ways that vary systematically over the life cycle. Our results can enable more appropriate analysis of and correction for errors-in-variables bias in a wide range of research that uses current earnings to proxy for lifetime earnings.
Keywords: No keywords provided
JEL Codes: D3; J3
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| current earnings (y_it) (E25) | lifetime earnings (y_i) (J31) |
| lifetime earnings (y_i) (J31) | current earnings (y_it) (E25) |
| current earnings (y_it) (E25) | errors-in-variables bias (C20) |
| current earnings (y_it) observed in the twenties (J31) | underestimation of true association with lifetime earnings (y_i) (J31) |
| age (J14) | slope coefficient (t) (C29) |
| current earnings (y_it) as proxies for lifetime earnings (y_i) (J31) | attenuation biases in intergenerational elasticities (D15) |