Publication Type
Working Paper
Version
acceptedVersion
Publication Date
10-2020
Abstract
Credible counterfactual analysis requires high-dimensional controls. This paper considers estimation and inference for heterogeneous counterfactual effects with high-dimensional data. We propose a novel doubly robust score for double/debiased estimation and inference for the unconditional quantile regression (Firpo, Fortin, and Lemieux, 2009) as a measure of heterogeneous counterfactual marginal effects. We propose a multiplier bootstrap inference for the Lasso double/debiased estimator, and develop asymptotic theories to guarantee that the bootstrap works. Simulation studies support our theories. Applying the proposed method to Job Corps survey data, we find that i) marginal effects of counterfactually extending the duration of the exposure to the Job Corps program are globally positive across quantiles regardless of definitions of the treatment and outcome variables, and that ii) these counterfactual effects are larger for higher potential earners than lower potential earners regardless of whether we define the outcome as the level or its logarithm.
Keywords
counterfactual analysis, double/debiased machine learning, doubly robust score
Discipline
Econometrics
Research Areas
Econometrics
First Page
1
Last Page
45
Citation
SASAKI, Yuya; URA, Takuya; and ZHANG, Yichong.
Unconditional quantile regression with high-dimensional data. (2020). 1-45.
Available at: https://ink.library.smu.edu.sg/soe_research/2460
Copyright Owner and License
Authors
Creative Commons License
This work is licensed under a Creative Commons Attribution-NonCommercial-No Derivative Works 4.0 International License.
Comments
Submitted to journal