No. 282 of 500 No. 53 in Machine Learning

Causal Inference

The Mixtape

By Scott Cunningham

Published
January 26, 2021
Print length
584 pages
Reading level
Intermediate
ISBN-13
978-0300251685
Publisher
Yale University Press
Written for
Analysts, data scientists and graduate students who need credible causal estimates from observational data.

Synopsis

Causal Inference: The Mixtape teaches the research designs that economists and other social scientists use to separate cause from correlation when a randomized experiment is unavailable. Scott Cunningham opens with a review of probability and regression, then builds the conceptual toolkit: directed acyclic graphs for reasoning about confounders and colliders, and the potential outcomes framework for defining what a treatment effect is. Later chapters work through matching and subclassification, regression discontinuity, instrumental variables, panel data, difference-in-differences and synthetic control, each illustrated with published empirical studies, including a randomized HIV-testing experiment used to teach randomization inference.

The approach is applied rather than proof-driven. Every method comes with worked code in both R and Stata, so readers can reproduce published estimates and adapt them to their own data, and the tone stays informal throughout. In a chapter-by-chapter review on the World Bank's Development Impact blog, David McKenzie graded the difference-in-differences chapter highest and the regression review lowest, and noted thin coverage of study design, power calculations and machine-learning approaches to causal questions. For data scientists, the payoff is the judgment to ask whether a model's correlations can support a causal claim before that claim drives a business or policy decision. Yale also agreed to let Cunningham keep a free online edition on his website.

Scorecard

Editorial score

How it scored

Editorial score: 74.1 / 1005 weighted criteria

  • Influence30%

    22.5/30
  • Authority25%

    18.8/25
  • Practical value20%

    16.0/20
  • Currency15%

    9.8/15
  • Readability10%

    7.0/10
  • Editorial score

    74.1/100

How the rank was computed

Rank No. 282 of 500 comes from a composite of 53.7: observable signals 20.4 of 55 points plus editorial review 33.3 of 45 points.

All 11 inputs
InputPoints
Observable signals20.4/55
Reach (Amazon ratings volume)9.7/16.5
Consensus across best-of lists2.8/13.8
Awards and bestseller evidence0.0/8.3
Reader rating, sample-adjusted6.7/8.3
Amazon sales rank1.2/5.5
Recency0.0/2.8
Editorial review33.3/45
Author authority8.4/11.3
Influence10.1/13.5
Practical value7.2/9.0
Readability3.2/4.5
Currency4.4/6.8

Each criterion is scored from 1 to 10 and weighted, and the points add up to the editorial score. The rank combines observable signals (55 percent) with this score (45 percent). Read the methodology.

Verdict

A code-rich, conversational guide to modern causal designs for applied researchers, since revised and expanded as The Remix.

Strengths

  • Parallel R and Stata code for every method
  • Clear chapters on DAGs and difference-in-differences
  • Free online edition alongside the print book

Keep in mind

Readers report typos, the opening regression review is heavy going, and study design, power analysis and machine-learning methods for causal questions receive little attention.

Key takeaways

  1. Directed acyclic graphs make causal assumptions explicit, showing which variables to control for and which colliders would bias estimates if included as controls.
  2. The potential outcomes framework defines a causal effect as the difference between what happened and an unobserved counterfactual for the same unit.
  3. Regression discontinuity, instrumental variables and difference-in-differences each rest on an untestable assumption, so a study's credibility depends on defending that assumption with evidence.
  4. When treatment begins at different times for different groups, conventional two-way fixed-effects difference-in-differences estimates can be biased, a problem the book's strongest chapter addresses.

Notable facts

  • Yale University Press agreed to let Cunningham keep publishing a free HTML version of the book on his website alongside the print edition. Source (opens in a new tab)
  • Yale's book page carries endorsements from Alberto Abadie of MIT, Justin Wolfers of the University of Michigan, Guido Imbens and Judea Pearl. Source (opens in a new tab)
  • David McKenzie reviewed the book chapter by chapter on the World Bank's Development Impact blog, giving the difference-in-differences chapter his top grade of A. Source (opens in a new tab)

About the author

Scott Cunningham

Scott Cunningham is the Ben H. Williams Professor of Economics at Baylor University, where his applied microeconomics research covers drug policy, sex work markets, abortion access and mental health. He earned his PhD in economics at the University of Georgia, coedited The Oxford Handbook of the Economics of Prostitution, runs causal inference workshops for universities, firms and government agencies, and wrote the revised sequel, Causal Inference: The Remix.

Questions

Is Causal Inference: The Mixtape worth reading in 2026?
Yes, for applied analysts, though buyers should know that Cunningham has since published Causal Inference: The Remix, an expanded and fully revised edition with fuller treatment of difference-in-differences and synthetic control. The original Mixtape remains free online, which makes it a low-cost entry point, while the Remix is the better purchase for readers who want the newest methods in print.
Do I need programming experience to use The Mixtape?
Some familiarity with R or Stata helps, because each method is paired with code that reproduces published estimates. Readers also need basic statistics and comfort with regression notation. Newcomers can follow the intuition behind directed acyclic graphs and potential outcomes, but they will get far more from the book by running the examples alongside the text.
How does The Mixtape relate to machine learning and AI?
The book is not about machine learning, but it addresses a question that prediction models cannot answer: whether changing one variable will change an outcome. Data scientists who build models for pricing, policy or product decisions use these designs to test whether a pattern reflects a causal effect. The book does not cover newer machine-learning estimators for causal effects, so readers will need other sources for that material.

Sources

As an Amazon Associate, Best AI Books earns from qualifying purchases. Best AI Books is sponsored by Iternal.ai, whose founder and CEO, John Byron Hanby IV, wrote The AI Strategy Blueprint, our #1 ranked book. How we rank books.