Working Paper: NBER ID: w26087
Authors: Pamela Giustinelli; Matthew D. Shapiro
Abstract: The Subjective ex ante Treatment Effect is the difference between the probabilities of an outcome conditional on a treatment. The SeaTE yields ex ante causal effects at the individual level. The paper gives an interpretation in two workhorse econometric frameworks: potential outcomes and dynamic programming. It finds large effect heterogeneity of health on work in two surveys of older workers, the VRI and the HRS. It shows how reduced-form estimates of health on work are biased when there is unobserved heterogeneity in taste for work. Using the VRI’s panel structure, it validates the elicited conditional probabilities of work given health.
Keywords: health; retirement; subjective expectations; treatment effects
JEL Codes: C21; C83; D84; J26
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| Health deterioration (I12) | Work probabilities (C29) |
| Health deterioration (I12) | Retirement behavior (J26) |
| Subjective ex ante treatment effect (SEATE) (C21) | Work probabilities (C29) |
| Elicited conditional probabilities (C11) | Realized work outcomes (J24) |
| Health (I19) | Work preferences (J29) |