ECON 151 Lecture Notes - Lecture 48: James Heckman, Signalling Theory, Human Capital
4 Dec 2020
School
Department
Course
Professor
Document Summary
If t is random, 1 gives the reduced form/itt (unbiased) impacts. X must be constant or pre-intervention variables. Additional assumptions: no spill-overs across treatment groups. Itt gives the overall treatment effect irrespective of where people participated or not. Itt is known to be the policy relevant parameter. T is used as an instrument for participation d: ( t is an iv) Relevance: t is relevant as being assigned to treatment increase your chance to take the treatment. Independence: by definitione ( |t ) = 0 because of randomization. Yi = 0 + 1di + i structural model where d is treated (in ps) and t is treatment, d is endogenous and t is iv where d is instrumented by t. with fist stage/ differential take-up: The iv/late effect is estimated like usually: iv = late = wald ratio. Itt and late have the same t-values: if one is significant, the other.
