Lesson 2 · Causal Inference · Step 6 of 6
Instrument-based approaches
Randomisation destroys confounding by construction, because nothing upstream can influence an assignment made by a coin. When a coin is not available, the next best thing is variation in the world that behaves like one. An instrument affects the outcome only through the exposure, and nothing else. A regression discontinuity uses a threshold — an HPV vaccination programme where eligibility turned on a birth date, making those just either side of it comparable. These designs buy credibility with assumptions you must state and defend.
Also covered: instrumental variables · regression discontinuity · natural experiments · exclusion restriction
Alongside this step
- Slides and transcript — the whole video
- Read the chapter — Causal Inference
Check yourself
5 questions, marked as you go. Nothing is recorded as a grade and nothing is locked behind it — if a question catches you out, that is the useful part.
Lesson 2 · 6 steps
The Validity Toolkit · lesson 2 of 4