CausalMLBook™ | Applied Causal Inference Powered by ML and AI

A comprehensive, rigorous guide to modern causal inference methods powered by machine learning.

By Victor Chernozhukov¹ • Christian Hansen² • Nathan Kallus³ • Martin Spindler⁴ • Vasilis Syrgkanis⁵

¹Massachusetts Institute of Technology ²University of Chicago ³Cornell University ⁴Universität Hamburg ⁵Stanford University

About the Book

This book bridges the gap between machine learning and causal inference, providing rigorous methods for answering causal questions using modern ML tools. Topics span predictive inference, causal identification, double/debiased machine learning, heterogeneous treatment effects, instrumental variables, difference-in-differences, regression discontinuity, and more.

Read the full book and download individual chapters at CausalML-Book.org

Chapters

All chapters are available for free at causalml-book.org.

Preamble

	Chapter
P	Preface
0	Powering Causal Inference with ML and AI

Core Material

	Chapter	Topics
1	Predictive Inference with Linear Regression in Moderately High Dimensions	`Prediction` `Inference`
2	Causal Inference via Randomized Experiments	`Causality` `Inference`
3	Predictive Inference via Modern High-Dimensional Linear Regression	`Prediction`
4	Statistical Inference on Predictive Effects in High-Dimensional Linear Regression Models	`Causality` `Inference`
5	Causal Inference via Conditional Ignorability	`Causality`
6	Causal Inference via Linear Structural Equations	`Causality`
7	Causal Inference via DAGs and Nonlinear Structural Equation Models	`Causality`
8	Predictive Inference via Modern Nonlinear Regression	`Prediction`
9	Statistical Inference on Predictive and Causal Effects in Modern Nonlinear Regression Models	`Causality` `Inference`
10	Feature Engineering for Causal and Predictive Inference	`Causality` `Inference`

Advanced Topics

	Chapter
11	Deeper Dive into DAGs, Good and Bad Controls
12	Unobserved Confounders, Instrumental Variables, and Proxy Controls
13	DML for IV and Proxy Controls Models and Robust DML Inference under Weak Identification
14	Statistical Inference on Heterogeneous Treatment Effects
15	Estimation and Validation of Heterogeneous Treatment Effects
16	Difference-in-Differences
17	Regression Discontinuity Designs

Interactive Labs

All labs run directly in Google Colab — no local setup required.

Available in Python and R

Ch 1 — Prediction & Linear Regression

Lab	Python	R
OLS and Lasso for Wage Prediction
The Gender Wage Gap
Exercise on Overfitting

Ch 2 — Randomized Experiments

Lab	Python	R
Vaccination RCT (Polio 1954)
Covariates in RCT: Precision Adjustment
Reemployment Bonus RCT

Ch 3 — High-Dimensional Linear Regression

Lab	Python	R
Penalized Linear Regressions: Simulation
Case Study: Wage Prediction with ML

Ch 4 — Inference in High-Dimensional Models

Lab	Python	R
Simulation on Orthogonal Estimation
Comparing Orthogonal vs Non-Orthogonal Methods
Testing the Convergence Hypothesis
Heterogeneous Effect of Sex on Wage

Ch 6–7 — DAGs & Structural Equations

Lab	Python	R
Collider Bias (Hollywood)
Causal Identification in DAGs
DoSearch for Causal Identification

Ch 8 — Nonlinear Prediction

Lab	Python	R
ML Estimators for Wage Prediction
Functional Approximations by Trees and Neural Nets

Ch 9 — DML for Causal & Predictive Effects

Lab	Python	R
Effect of Gun Ownership on Homicide
DAG Analysis of 401(k) Impact
DML Inference on 401(k) Wealth Effects
DML for Partially Linear Model (Growth)

Ch 10 — Feature Engineering

Lab	Python	R
Variational Autoencoders and PCA
DoubleML Feature Engineering with BERT		—

Ch 12–13 — IV, Proxy Controls & Weak Identification

Lab	Python	R
Sensitivity Analysis with Sensmakr
Negative (Proxy) Controls
DML for 401(k) with IV
Weak IV Experiments
DML for Partially Linear IV Model

Ch 14–16 — Heterogeneous Effects & Diff-in-Diff

Lab	Python	R
CATE Estimation with Causal Forests		—
CATE Inference: Best Linear Predictors		—
Conditional Average Treatment Effects		—
Difference-in-Differences: Minimum Wage

Software Packages

Companion open-source implementations for the methods covered in the book.

DoubleML

Double/Debiased ML in Python & R

EconML

Heterogeneous Treatment Effects

Stata ML & ddml

Regularized Regression & DML for Stata

PLR, IRM, PLIV, IIVM models. Builds on scikit-learn (Python) and mlr3 (R).

Double ML, Causal Forests, Meta-Learners, IV methods. Part of PyWhy.

lassopack, pdslasso, pystacked, ddml.

Citation

@article{chernozhukov2024applied,
  title   = {Applied Causal Inference Powered by ML and AI},
  author  = {Chernozhukov, Victor and Hansen, Christian and Kallus, Nathan
             and Spindler, Martin and Syrgkanis, Vasilis},
  journal = {arXiv preprint arXiv:2403.02467},
  year    = {2024},
  doi     = {10.48550/arXiv.2403.02467}
}