Everyone knows correlation is not causation, and almost no one is trained in what to do about it. Causal inference is the field that fills that gap, and it is genuinely hard because the question, what would have happened otherwise, is about a world we never observe. It is also split across traditions: the graphical, do-calculus school associated with Judea Pearl, and the potential-outcomes, quasi-experimental school common in economics. Learning it well means seeing both, in the right order.
The path starts with intuition, introduces the two great frameworks, moves into applied methods, and ends with rigorous theory. Read this way, the schools complement rather than confuse each other.
Build the intuition
Start with The Book of Why by Judea Pearl, a popular-level argument for why causal thinking matters and how graphical models capture it. It is the most engaging on-ramp to the whole subject. Then step up to Causal Inference in Statistics, Pearl's slim primer that teaches the formal machinery of causal diagrams and the do-operator without overwhelming you. Together they give you the graphical worldview.
Learn the applied toolkit
The other tradition is empirical and method-driven. Mostly harmless econometrics by Angrist and Pischke is the beloved, readable guide to the quasi-experimental toolkit, instrumental variables, difference-in-differences, regression discontinuity, that powers modern applied economics. Causal Inference by Scott Cunningham, often called the Mixtape, covers the same methods with code and worked examples, making it an ideal hands-on companion. The Effect by Nick Huntington-Klein is a newer, exceptionally clear introduction that ties the intuition and the methods together, and it is a fine bridge between the two schools.
Reach the rigorous foundations
For depth, the path turns formal. Causality by Pearl is the definitive, demanding statement of the graphical framework and the theory behind the do-calculus. Causal Inference for Statistics, Social, and Biomedical Sciences by Imbens and Rubin is the authoritative treatment of the potential-outcomes approach, and reading it against Pearl shows how the traditions connect. What If by Hernán and Robins is a rigorous, freely available text especially strong for epidemiology and observational data, and Elements of Causal Inference links causality to modern machine learning.
Follow the full path and causal inference stops being a slogan and becomes a set of tools you can actually use to reason about cause and effect. You end fluent in both major frameworks and able to choose the right method for a real question. These books teach reasoning, not certainty: causal claims from observational data always rest on assumptions you must state and defend.
