Identification and Estimation of the Marginal Treatment Effect without Instrumental

Speaker：Pan Zhewen, Associate Researcher, Zhejiang University of Finance and Economics

Host：Cheng Mingmian, Assistant Professor, Lingnan College

Time and Date：14:30, Oct. 27, 2023

Venue：Wang Daohan Meeting room, Lingnan hall 

Language: English + Chinese

Abstract：

This paper proposes a method of defining, identifying, and estimating the marginal treatment effect (MTE) without imposing instrumental variable (IV) assumptions of independence, exclusion, and separability. Under the novel definition of MTE based on normalized treatment error that is statistically independent of covariates, we find that the relationship between MTE and standard treatment parameters holds as well in the absence of IV. We provide a set of sufficient conditions ensuring identification of such defined MTE in an environment of essential heterogeneity. The key conditions include a linear restriction on potential outcome regression functions, a nonlinear restriction on the propensity score, and a conditional mean independence restriction that leads to additive separability. We prove identification following the notion of semiparametric identification based on functional form. Consistent semiparametric estimation procedures are suggested, and an empirical application to Head Start is provided to illustrate the usefulness of the proposed method in analyzing heterogenous causal effects when IV is elusive.