Natural Experiment

A natural experiment is a study that exploits a naturally occurring or externally imposed variation in exposure — such as a policy change, a geographic boundary, or another event outside the investigator's control — that divides people into groups as if they had been randomly assigned. It lets researchers approximate the logic of a controlled experiment for questions where deliberately assigning exposure would be unethical, impractical, or impossible.

Definition

A natural experiment is a study in which exposure is determined by a naturally occurring or externally imposed process not controlled by the investigator, and which is analyzed as if exposure had been assigned, to estimate the effect of that exposure on an outcome.

Scope

The entry covers the idea of exploiting exogenous variation, the conditions under which such variation can act like randomization, the analytic approaches used to estimate effects, and the assumptions and threats to validity that distinguish a natural experiment from a true randomized trial. It treats the natural experiment as a methodological topic within epidemiologic study designs, not as clinical guidance.

Key concepts

• Exogenous (external) variation in exposure
• Quasi-random assignment
• Counterfactual comparison
• Plausibility of the as-if-random assumption
• Confounding from non-random exposure
• Population-level interventions and policy evaluation

Mechanisms

A natural experiment relies on a source of variation in exposure that arises outside the study — a new law, an administrative boundary, a price change, a disruption in supply — and that, crucially, is unrelated to the outcome except through the exposure itself. When this as-if-random condition holds, comparing the affected and unaffected groups approximates the counterfactual contrast a randomized trial would provide, so the effect of the exposure can be estimated despite the absence of deliberate assignment. The central methodological task is to argue and probe whether the variation really is independent of confounders; because assignment is not under the investigator's control, residual confounding is the principal threat, and the credibility of any natural experiment rests on how convincingly the as-if-random assumption can be defended. Such studies are observational in execution but experimental in logic, occupying the ground between observational designs and true experiments.

Clinical relevance

Natural experiments are a key source of evidence on the health effects of policies and population-level interventions that cannot be randomized, and interpreting them is part of evidence appraisal in public health. This entry describes how such evidence is generated and judged; it is a reference on the design and not a basis for individual diagnostic or treatment decisions.

Epidemiology

Natural experiments are most valuable for evaluating population-level and policy interventions — taxation, regulation, environmental change, programme roll-out — where randomization is impossible but a credible exogenous variation exists. Their strength is relevance to real-world conditions and their ability to study otherwise unanswerable questions; their weakness is that the as-if-random assumption can fail, so the certainty of their estimates is generally lower than that of a comparable randomized trial.

Evidence & guidelines

Methodological guidance, including Medical Research Council guidance on using natural experiments to evaluate population health interventions, sets out when the approach is appropriate and how to strengthen causal inference. In evidence hierarchies natural experiments are treated as quasi-experimental: stronger than ordinary observational comparisons when the assignment is genuinely exogenous, but generally below randomized trials because the assignment is not under the investigator's control.

History

The idea long predates its formal name: John Snow's investigation of the 1854 London cholera outbreak, which compared mortality among households supplied by water companies drawing from contaminated and cleaner sources, is often cited as an early natural experiment because the water supply divided households in a way Snow did not control. The approach was later formalized across epidemiology and the social sciences, and guidance such as the Medical Research Council framework set out how natural experiments can be used to evaluate population health interventions.

Debates

How can the as-if-random assumption be justified?

A natural experiment is only as credible as the claim that exposure was assigned independently of the outcome's other causes; because this cannot be guaranteed as in a randomized trial, defending the assumption with design and supporting evidence is the central and contested task.

Key figures

• John Snow
• Peter Craig

Seminal works

• snow-1855
• craig-2012

Frequently asked questions

How does a natural experiment differ from a randomized controlled trial?

In a randomized trial the investigator assigns exposure by chance; in a natural experiment the exposure is determined by an outside event or process the investigator does not control. The natural experiment only approximates randomization, and its validity depends on whether that external assignment was truly independent of the outcome's other causes.

Why is John Snow's cholera study described as a natural experiment?

Households in part of London received water from different companies drawing on more or less contaminated sources, dividing them by exposure in a way Snow did not arrange. Comparing cholera mortality across these as-if-assigned groups let him estimate the effect of water source, the logic of a natural experiment.