Causal Inference: The Mixtape

by Scott Cunningham

An accessible, contemporary introduction to the methods for determining cause and effect in the social sciences
 
Causal inference encompasses the tools that allow social scientists to determine what causes what. In a messy world, causal inference is what helps establish the causes and effects of the actions being studied—for example, the impact (or lack thereof) of increases in the minimum wage on employment, the effects of early childhood education on incarceration later in life, or the influence on economic growth of introducing malaria nets in developing regions. Scott Cunningham introduces students and practitioners to the methods necessary to arrive at meaningful answers to the questions of causation, using a range of modeling techniques and coding instructions for both the R and the Stata programming languages.

Genres: Economics, Mathematics, Nonfiction, Textbooks, Science, Business, Research

328 pages, ebook

Reviews

[Martin Lábaj]

Perfect complement to other traditional (Introductory Econometrics: A Modern Approach) or modern (Mastering 'Metrics: The Path from Cause to Effect , Mostly Harmless Econometrics: An Empiricist's Companion) econometric textbooks.

[Ardyn · Rating 5 out of 5]

Intuitive explanations of causal estimation methods with practical examples in both R and Stata. Loved it!

[Hugo Salas · Rating 4 out of 5]

Just making sure I can still read an equation

[Vysloczil]

It tries to strike a balance between being a resource for advanced undergraduates in the social sciences and practitioners with some stats knowledge and a PhD-level text. For the most part, it succeeds in doing so, and it is definitely an essential text for anyone who has an interest in learning causal relationships from observational data (at least currently, in 2024). The book has a rather large introduction to basics in stats and metrics which someone who already has this knowledge might find redundant. In any case, the book is for sure not THE essential resource as it covers only selected topics and, within these selected topics, does not cover the universe of problems.

The Mixtape is more demanding than "Mastering Metrics" and "The Effect" - two nice intuitive undergrad treatments, and a bit less demanding than "Mostly Harmless Econometrics" - which is already a bit outdated as of 2024. For the basics at the grad level, thorough treatments like Bruce Hansen's new two textbooks are indispensable for anyone who wants to understand the mechanics at a deeper level.
Keep in mind that there is a free online version of the book, which is useful to do quick searches and copy some of the provided code.

For difference-in-differences, there is a large focus on triple differences with a long treatment of one of the authors' papers. When it comes to the negative weighting issues with TWFE, the treatment focuses almost exclusively on the Goodman-Bacon decomposition and only present some of the solutions for estimation in a rudimentary way. To the author's defense, that's mostly due to the fact that at the time of writing, this literature was still evolving and most papers did not fully get through peer-review yet.

[Alexander Billy · Rating 4 out of 5]

This book has the potential to become essential reading for anyone in applied economics. Scott has gone out of his way to write a tour de force on the econometrician's toolkit. He covers basic principles of regression, DAGs, RD, synthetic controls, and a host of other identification strategies and estimators.

The reading is pleasurable, too. Unlike a lot of economics papers or books, you can sit down and enjoy the writing. In some sense, it's very similar to Angrist and Pischke's Mostly Harmless Econometrics. But Scott's florid style (that I'm sure stems from his days as a poet) is striking. Words are artfully chosen but not to the extent that there are Hemingway-esque non sequiturs.

Don't sleep on this book and think it's identical to MHE. If you're a PhD student, recent PhD grad, or thinking about applying to a PhD program, read this! Scott covers topics not included in MHE, and also provides code for you to execute. The examples to motivate the concepts and highlight the code are masterfully captured.

And while the book isn't perfect, remember it's only a rough draft. You'll find typos and notational conflicts. (Hell, I'm sure this review has typos in it!) But they're not so jarring that you miss his logic. To my knowledge, Yale Press will seriously scrutinize the editorial process. So, when the book is completed, it'll be even better. Scott has even written on Twitter stating he's going to include some additional material on the new difference-in-differences techniques. So, be on the lookout for the updated version. I know I'll buy a copy to reference.

Kudos, Scott, and thanks for being a really splendid guy!

[Katie Nissen · Rating 4 out of 5]

This is a good book to cover causal inference from start to finish, and it emphasizes current trends in the field well. I knocked it a star because I think the author sometimes skips a middle step when explaining concepts. It’ll start almost insanely simple in one sentence, and then the next sentence is proceeding while skipping some crucial explanation in the middle. An example of this is the Diff-in-Diff chapter, which defines differently treated periods as k, l, and U (early, late, and never) (easy), and then skips to saying that k is treated in period 2 and takes a value of 0.4. It feels like there’s a sentence missing in between to explain where they 0.4 came from. This was pervasive throughout the book.

Besides that issue, this is a great one-stop-shop for causal inference. By personal preference, I wish it used potential outcomes lingo more than DAGs, but I might be in the minority there.

[Olivier · Rating 4 out of 5]

Best literature I read on causal inference since https://blog.echen.me/2014/08/15/prop... :) Reasonably easy to read, not too much equations.

[Leonardo]

The “Mixtape” book (also free online) by Scott Cunnigham is a good non-technical and more basic treatment of causal inference, which provides some good intuitions.

Bibliografía obligatoria del curso de Evaluación de Impacto.

[Tam · Rating 2 out of 5]

The highlight of this book is the sharing of codes and data so that learners can follow, explore, and apply themselves. Only R and Stata are supported though, yet the inclusion of R definitely widens Cunningham's audience.

The book discusses several modern tools used in Causal Inference. This falls short of my expectation though, as the exposition is not exactly clear. It struggles between keeping the book accessible to many and being rigorous, so it fails in both. But if you're a current graduate student, this book is fine as a quick crash introduction where details can fill in later. But a good definite reference text book? No, I wouldn't recommend.

[Chris]

The basic intros are useful and generally understandable, even if one doesn't scrutinize the equations or run the simulations, which I definitely did not. 🙂

"New-to-You Format" square on the 2025 Seattle Public Library Summer Book Bingo card. It's a free online textbook (thank you, Dr. Cunningham!) that's extraordinarily well designed. If you check a footnote, return to a previous chapter for a refresher, or (I imagine) stop to run a simulation, it's easy to return to your place.

[Alejandro Vicente · Rating 5 out of 5]

An amazing review of modern causal inference tools. The approach, with Stata and R code, makes learning the techniques so easy. I highly recommend the Mixtape to those grad students or professional interested in rigorous empirical work, but in my opinion it´s an intermediate level approach: there are better options for undergrads and newcomers (The Effect, Mastering Metrics), and for advanced grads and experienced proffesionals (Mostly Harmless, Imbens and Rubin).

[Francisco Yirá · Rating 5 out of 5]

A very good resource to learn causal inference techniques, but be aware that the main audience to whom the book is targeted is economics/social sciences research practitioners. If you want to apply these methodologies in industry, you may have to complement your reading with examples applied to those cases.

[Usman Khaliq · Rating 3 out of 5]

Overall, an interesting, although at times a difficult read into causal inference. Some of the chapters, such as the one on instrumental variables, synthetic control, regression discontinuity and difference-in-difference are definitely stronger than others. I would still recommend having this book in your ammunition of statistics/causal inference books.

[Cayley · Rating 4 out of 5]

I was lucky enough to take Dr Cunningham’s class in college maybe 6 years ago or so, and this book was a great refresher of that class!

The only downside as a data scientist/industry practitioner is that while I know STATA and R are widely used in academia, they are rarely used outside. I wonder if anyone has done a python version of the code yet? If not, maybe a weekend project…

[Gabor Scheiring · Rating 5 out of 5]

A friendly and powerful introduction to causal inference. Some other books out there on the topic are extremely hard to digest for most social scientists (eg. Judea Pearl's books), but this one is written from an applied economics perspective by a great educator.

[Lucia Huo · Rating 4 out of 5]

There’s a lot of casual causal inference books out there, apparently. This one is comprehensive, technical, and relatively easy to consume. I am not sure this is the best one, but I think good starting material for someone who’s really trying to learn this from scratch.

[szymborskalyte · Rating 4 out of 5]

One of the more accessible introductions to casual inference for sure.

[Rhys Williams · Rating 5 out of 5]

Very informative book full of examples (in both Stata and R) on how to apply the methods discussed

[x · Rating 5 out of 5]

Essential for any Data Scientist working with observational data to understand causal patterns

[Gwen · Rating 5 out of 5]

Dense, but very informative with great workable examples. Also appreciate that it has a free online version.

[Sam Enright · Rating 4 out of 5]

I understood about 5% of this book. I suspect I will be returning to it many more times.

[Cosimo Faeti · Rating 5 out of 5]

Thanks Scott :)

[Kasper · Rating 4 out of 5]

Nice introduction to causal inference, even for a non-math person like myself. I found it a little long in places, and occasionally I could not follow the explanations. But perhaps that is just me...

[Alejandro Shirvani · Rating 5 out of 5]

An excellent book on causal inference techniques in econometrics. Mostly fairly accessible with some more technical parts, and covers contemporary literature comprehensively.

[Lee Reynolds · Rating 4 out of 5]

a sometimes brilliant book that got a little lost, then discovered itself again then got lost. would have appreciated more of a geometric intuition at some points.

[T S · Rating 3 out of 5]

Provides a comprehensive overview of Econometrics, taking you in one book from not knowing much to being conversant in the main ideas. I feel like it is best for people that have a strong stats background but not much exposure to econometrics/economics concepts.

Even though I was familiar with most of the concepts in the book, I still had trouble following at points with some of the equations and was sometimes wondering where a certain equation came from or what certain symbols were representing and meant.

This format might not be the best for learning technical topics like this, instead might be best to stick to textbooks which can cover the concepts in greater detail and where things are much easier to follow. In other words, there is no shortcut. I have just started reading it, but The Effect may be a good alternative.

With this book, what is left is neither the advantages of a regular non-fiction book nor the textbook, being too long and cumbersome to read through like a typical non-fiction book but also despite it's length of nearly 600 pages felt like at certain parts would have benefitted from more explanation and detail.