Automatic Debiased Machine Learning of Causal and Structural Effects

2018/09/14 by Victor Chernozhukov, Whitney K Newey, Whitney K. Newey +4 · 1 voice · 22 citations
Economics, Econometrics and Finance · Mathematics · #Econometrics (econ.EM) #FOS: Economics and business #FOS: Mathematics #Statistics Theory (math.ST) #econ.EM #math.ST #stat.TH

paper · pdf · doi:10.48550/arxiv.1809.05224

Econometrica 2022

arxiv created 2022/10/21 · arxiv updated 2022/10/25

Abstract

Many causal and structural effects depend on regressions. Examples include policy effects, average derivatives, regression decompositions, average treatment effects, causal mediation, and parameters of economic structural models. The regressions may be high dimensional, making machine learning useful. Plugging machine learners into identifying equations can lead to poor inference due to bias from regularization and/or model selection. This paper gives automatic debiasing for linear and nonlinear functions of regressions. The debiasing is automatic in using Lasso and the function of interest without the full form of the bias correction. The debiasing can be applied to any regression learner, including neural nets, random forests, Lasso, boosting, and other high dimensional methods. In addition to providing the bias correction we give standard errors that are robust to misspecification, convergence rates for the bias correction, and primitive conditions for asymptotic inference for estimators of a variety of estimators of structural and causal effects. The automatic debiased machine learning is used to estimate the average treatment effect on the treated for the NSW job training data and to estimate demand elasticities from Nielsen scanner data while allowing preferences to be correlated with prices and income.

Citations

• A general theory of hypothesis tests and confidence regions for sparse high dimensional models
• Double/Debiased/Neyman Machine Learning of Treatment Effects
• Honest data-adaptive inference for the average treatment effect under model misspecification using penalised bias-reduced double-robust estimation
• Multivariate regression models for panel data
• The central role of the propensity score in observational studies for causal effects
• Deep Neural Networks for Estimation and Inference
• Convergence of a Block Coordinate Descent Method for Nondifferentiable Minimization

Cited by

• Causal-Policy Forest for End-to-End Policy Learning
• From dense grids to valid inference: Accounting for regularization bias in nonparametric random coefficient models
• Debiased Inference for High-Dimensional Regression Models Based on Profile M-Estimation
• Semi-Supervised Treatment Effect Estimation with Unlabeled Covariates for Prediction-Powered Causal Inference
• Riesz Regression As Direct Density Ratio Estimation
• A Unified Theory for Causal Inference: Direct Debiased Machine Learning via Bregman-Riesz Regression
• genriesz: A Python Package for Automatic Debiased Machine Learning with Generalized Riesz Regression
• Nearest Neighbor Matching as Least Squares Density Ratio Estimation and Riesz Regression
• Direct Debiased Machine Learning via Bregman Divergence Minimization
• Causal Inference in High-Dimensional Generalized Linear Models with Binary Outcomes
• Sensitivity Analysis for Causal ML: A Use Case at Booking.com
• Robust Semiparametric Inference for Bayesian Additive Regression Trees
• A note on the relation between one-step, outcome regression and IPW-type estimators of parameters with the mixed bias property
• Direct Bias-Correction Term Estimation for Propensity Scores and Average Treatment Effect Estimation
• Doubly-Robust LLM-as-a-Judge: Externally Valid Estimation with Imperfect Personas
• Regularizing Extrapolation in Causal Inference
• CausalSent: Interpretable Sentiment Classification with RieszNet
• Deep learning based doubly robust test for Granger causality
• Debiased machine learning for combining probability and non-probability survey data
• Riesz representers for the rest of us
• Regularized Orthogonal Machine Learning for Nonlinear Semiparametric Models
• Kernel Methods for Unobserved Confounding: Negative Controls, Proxies,\n and Instruments
• Structural Representations and Identification of Marginal Policy Effects
• Identification and estimation of causal peer effects using instrumental variables

Discussions

• Newey has used it for average demand elasticity estimation; see S7 of their AutoDML paper which has an equivalent formulation but uses Riesz regression, which is like a balancing form rather than plug [bsky, 4 points, 1 comments]

Related

• Debiased Machine Learning of Conditional Average Treatment Effects and Other Causal Functions
• RieszNet and ForestRiesz: Automatic Debiased Machine Learning with Neural Nets and Random Forests
• Double/Debiased Machine Learning for Treatment and Causal Parameters
• Automatic Debiased Machine Learning for Covariate Shifts
• Automatic debiased machine learning for covariate shifts