Topics on econometrics and causal inference

Preface

This is a working notebook rather than a textbook. The chapters are posts I wrote while reading, teaching, and consulting on applied econometrics, often because a colleague or a student asked a question that did not have a tidy answer in the books I had. Each one stands on its own. The order did not follow a plan, and the same topic sometimes appears more than once as my understanding changed.

The common question is the one Angrist and Pischke emphasize: what is being identified, by what variation, under what assumptions. The posts ask that question in different settings: interactions and marginal effects, fixed and correlated random effects, matching and weights, difference-in-differences, synthetic control, count models, TMLE and AIPW, mediation, proximal causal inference, partial interference, conjoint and list experiments, uplift modeling, and policy trees.

Most examples are in R. I use Stata where it is more natural, and sometimes Python or Julia when the package is better there. The two companion books, Introduction to Causal Econometrics and Causal Econometrics with Julia, organize the same material more systematically. This collection is the messier version: useful when I want to see how a specific method behaves on a specific problem.

Where to start

Forty-eight chapters in no particular order is not a reading order. Three routes through the material, for readers who want one:

Foundations. What a coefficient means before any causal language is attached. Interpreting interaction in a regression model → Marginal effects in models with fixed effects → Fixed or random effect, or both? → Matching and weighting, part 1 → part 2.

Modern causal estimation. Nuisance functions estimated by machine learning, with the estimand held fixed. Machine learning for causal effects → Recent causal inference tools, which covers AIPW and double machine learning → Longitudinal modified treatment policies → Intro to proximal causal inference → the proximal simulation.

Panel and policy methods. Designs where the variation comes from timing. Extended TWFE and other DiD estimators → Causal panel, and why DDDiD → Synthetic control → Causal forest in panel data → Policy learning by policytree.

Chapters outside these routes are self-contained and can be read on their own.

How to cite

Ao, Xiang. Topics on Econometrics and Causal Inference. https://xiangao.github.io/blog_book/.

@book{ao_topics_econometrics,
  author = {Ao, Xiang},
  title  = {Topics on Econometrics and Causal Inference},
  year   = {2026},
  url    = {https://xiangao.github.io/blog_book/}
}