Marginal Treatment Effects from a Propensity Score Perspective

Area of Science:

• Econometrics
• Causal Inference
• Biostatistics

Background:

• The marginal treatment effect (MTE) is a key concept in causal inference.
• Existing methods for MTE analysis can be computationally intensive and lack intuitive interpretation.

Purpose of the Study:

• To propose a novel propensity score-based framework for interpreting and analyzing the marginal treatment effect (MTE).
• To develop a redefined MTE that simplifies computation and enhances interpretability of treatment effect heterogeneity.

Main Methods:

• The study redefines MTE as the expected treatment effect conditional on propensity score and a latent variable for unobserved treatment resistance.
• This approach leverages propensity score methods for causal analysis.

Main Results:

• The redefined MTE provides simpler, more intuitive, and computationally easier weights compared to the original MTE.
• This framework directly reveals treatment effect heterogeneity among individuals at the margin of treatment.

Conclusions:

• The redefined MTE serves as a foundational element for constructing standard causal estimands.
• This approach facilitates a broader evaluation of various policy effects by clarifying treatment heterogeneity.