Randomized Experiments

Randomized Controlled Trials (RCTs) are considered the "gold standard" for causal inference. This status is well-deserved: randomization provides a remarkably elegant solution to the fundamental problem of causal inference.

Why Randomization Works

When treatment is randomly assigned, something remarkable happens: the treated and untreated groups become comparable on average with respect to all characteristics, both measured and unmeasured. This comparability is the key to causal inference.

Consider what randomization accomplishes:

Without Randomization	With Randomization
Sicker patients may get more intensive treatment	Treatment is unrelated to illness severity
Motivated patients may choose healthier options	Treatment is unrelated to motivation
Unknown factors may influence treatment choice	Treatment is unrelated to ALL factors

"Randomization ensures that the treatment groups are exchangeable, meaning that the treated and untreated groups would have had the same outcomes, on average, had they received the same treatment." — What If, Chapter 2

The Historical Context

The first recognized RCT was conducted by Austin Bradford Hill in 1948, testing streptomycin for tuberculosis. This trial revolutionized medical research by demonstrating how random allocation could provide convincing evidence of treatment effects.

Since then, RCTs have become the foundation of evidence-based medicine, required by regulatory agencies (like the FDA) for drug approval, and the benchmark against which all other study designs are compared.

The Simple Logic

If treatment is randomly assigned:

1. We cannot predict who will be treated based on any characteristic
2. Therefore, treated and untreated groups are comparable
3. Therefore, differences in outcomes must be due to treatment (or chance)

This simple logic powers an enormous enterprise of medical research and has led to countless advances in treatment.