Randomized controlled trial
The gold standard for causal evidence: randomly assign people to treatment or control.
What it means
A randomized controlled trial (RCT) is an experiment in which participants are randomly allocated to receive an intervention or to a comparison (control) condition, so that on average the groups differ only in the treatment they receive. Randomization is the engine of its power: by distributing both known and unknown confounding factors evenly across groups in expectation, it breaks the link between the treatment and any pre-existing differences, which is what licenses a causal interpretation of the outcome difference. Its strength is high internal validity, often reinforced by blinding (to control expectancy and demand effects) and by a placebo or active control, and analyzed with attention to statistical power so a real effect can be detected. Its limitations include cost, ethical and practical constraints on what can be randomized, the risk of low external validity when the sample or setting is narrow, and the fact that a single underpowered trial can mislead. RCTs are the backbone of evidence-based medicine and increasingly of applied behavioral policy. It matters because, done well, an RCT provides the most credible evidence that an intervention actually causes an effect rather than merely correlating with it.
Examples
Half of taxpayers randomly receive a redesigned reminder letter; the difference in payment rates between the two groups measures the letter's causal effect.
A hospital randomly assigns wards to the new handwashing protocol and keeps the infection assessors blind to which ward is which, so the gap in infection rates is the protocol's doing.
A charity randomly picks which supporters see the new donation page and finds a real uplift — but the trial ran only on lapsed donors, so nobody yet knows it works on anyone else.
First described in Ronald Fisher (1920s–30s).