Causality you can believe in?
The identification challenge and taking assumptions seriously
Chad Hazlett · Summer Institute in Computational Social Science (SICSS) 2026, UCLA
A talk on causal identification: why credible causal claims rest on assumptions that cannot be checked in the data; how the main identification strategies (experiments, conditional ignorability, RDD/ITS, difference-in-differences, instrumental variables, and learning from prior outcomes) each carry their own demanding assumptions; and how sensitivity analysis and partial identification — “safe inference” — let us state honestly what we can and cannot conclude.