Causal Inference Without an Experiment

Carlos Dobkin

Economics Department, UC Santa Cruz

National Bureau of Economic Research

Unbiased estimates of treatments effects are critical for forming effective government policy. The Randomized Controlled Trial (RCT) is the gold standard approach to generating causal estimates. Unfortunately for many treatments of interest ethical or practical constraints make it impossible to implement an RCT. In some settings a Regression Discontinuity Design can be used to estimate the causal effect of a treatment. RDDs leverage rules that assign treatment eligibility based on the level of a variable. The sharp change in the probability of treatment induced by the rule make it possible to estimate causal effects as diverse as the effects of health insurance or alcohol consumption.