Causal Inference#

Causal inference is the process of isolating and unambiguously determining the effect of a particular phenomenon acting within a larger system. An example might be the returns to education given that going to university might also act as a signalling mechanism, the effect of the minimum wage on employment, or how much a particular drug reduces the incidence of a particular disease. It’s the field of inquiry that allows you to go beyond asking ‘how?’ and lets you start to ask ‘why?’.

Causal inference has seen a huge rise in popular in economics as part of the ‘credibility revolution’. In 2021, some of the people in the vanguard of that revolution were awarded the Nobel prize for economics for their work in bringing empirical evidence to the debate—often over-turning the predictions of naive economic theory as they did so. It’s an exciting topic, and one that is becoming increasingly important in data science too.

However, like a good economist, this book recognises the value of specialisms and so, for now at least, this Chapter has been outsourced to Causal Inference for The Brave and True, also known as the Python Causality Handbook. There you will find chapters on a range of causal inference methods—plus the code to run them—including:

  • Graphical causal models and directed cyclic graphs

  • Local Average Treatment Effect (LATE)

  • Propensity Scores

  • Difference-in-Differences

  • Synthetic Control

  • Regression Discontinuity,

and more.