The Causal Library
There are probably fewer than 50 books dedicated solely to the topic of causal inference as a standalone discipline. In the grand scheme of things, causal inference is a niche. That seems odd, because people encounter it from almost every other discipline. Medicine, life sciences, politics, macroeconomics & social policy, as well as business and marketing tend to be the fields that gravitate towards causal inference the most, but that’s already a very broad spectrum. With causality being a niche having relatively few books covering it, I think that causal nerds take pride in our causal library - this little corner of our bookshelves that (we insist) fights the good fight.
In celebration of the Causal Library - the greats that have mapped out this discipline and left behind their instructions - this post is a living document where I review and discuss the books in my causal library. I will come back and update this when new books are published or when I want to update something about a book I already wrote about.
The Book of Why - Judea Pearl, Dana Mackenzie
Of course this had to be first. First published in 2018, ‘the Book of Why’ synthesizes, condenses and translates the work of Judea Pearl and his colleagues since ~1988 into a very digestible popular science format suitable for light reading - something you could pick up and skim at an airport.
Functionally this book does everything it sets out to do, I just have some minor nitpicks with it’s approach. This book is attempting something fundamentally radical - it pushes a transformative way of thinking, and it tries to do that in the most accessible way possible. For that alone, Pearl and Mackenzie deserve the high praise they have received.
There is an unhurried feeling to the writing style - I don’t get the sense that an editor has taken a chainsaw to every colourful sentence in the name of brevity. I think the storytime style actually lends itself really well to this topic, because causal inference at times feels like a hostile and uncharted terrain, so having a calm and reassuring presence makes its traversal a more pleasurable experience.
One aspect of the writing style leaves me with a confusing mix of emotions. Concepts are too often delivered in parables with religious overtones. In the first chapter alone we read about the Babylonians, the book of Daniel, and about Adam, Eve and the snake. The antiquity of these stories apparently lending them weight. To my eyes, it all feels a bit tropey, but I mention that the emotions are mixed because i’m not sure if the book would be improved by removing the biblical guff. The only way I can describe it is like having a dad that puts on Deep Purple every Sunday morning - you roll your eyes and cringe, but when they’re no longer around you miss them dearly.
Overall though, a fantastic book and a triumph for the field. An absolute must read for people interested in Causal Inference.
Causal Inference In Statistics: A Primer - Judea Pearl, Madelyn Glymour, Nicholas P. Jewell
For technical practitioners both current and aspiring, this book is the absolute best one that I would recommend to anyone looking to go deep on structural causal modelling and it’s notation. It sits squarely between ‘the Book of Why’ and Pearls 1998 ‘Causality’ book which is more of an encyclopaedic tome. This is really the practitioners road manual - it is suprisingly thin clocking in it only 126 pages, but it really scales the entire ladder of causal reasoning with complete mathematical rigour.
Without a doubt this is the first book you should pick up if you want to get up and running with structural causal models. This may even be “the perfect book” on causal inference for a technical audience, because not only does it give an easy to follow set of axioms and theorems, but it also does a great job of the mandatory “convince me that I need to care about causal inference” legwork at the start of the book. So much so that if you find yourself in the position of being the resident “causal botherer” at your workplace, your job of spreading the gospel is made substantially easier by following the approach in this book - just hit ‘em with Simpson’s Paradox right out the gate and build from there.
The book is structured into sections that trace Pearl’s “ladder of causation” that he articulates well in ‘the Book of Why’. Section one is ‘Preliminaries: Statistical and Causal Inference’ which as mentioned earlier is the mandatory “why do I need to care about this” preamble and essentially covers Pearls first rung on the ladder; observation. Section two is ‘Graphical Models and Their Applications’ which lays the groundwork for rest of the book. Section three is ‘the Effects of Interventions’ (second rung on the ladder) and finally section four is ‘Counterfactuals and Their Applications’, the third and final rung on the ladder.
Despite its brevity relative to the gargantuan ‘Causality’ (Pearl, J. 1999) this book offers a complete treatment of subject from both a conceptual and a technical standpoint, though this book leans more on the technical side. For those who need more convincing of the conceptual aspects of causal inference, ‘the Book of Why’ is definitely the better place to start.
