Glossary
Natural experiment
What is a natural experiment?
A natural experiment arises when an event or process outside researchers’ control produces differences in exposure between groups. Researchers can use those differences to study an intervention’s effects. Examples include policy changes, eligibility rules, or lotteries run by public authorities.
How causal inference works
The 2025 MRC/NIHR framework uses a broad definition that does not require every natural experiment to have random assignment. Researchers must examine how exposure occurred, whether groups are comparable, and which assumptions the analysis needs.
Research example
In the Oregon health insurance experiment, a 2008 lottery gave selected adults the opportunity to apply for Medicaid. Selection increased insurance coverage but did not guarantee enrollment. The lottery therefore supported a randomized comparison of the offer, with additional analysis needed for effects of coverage itself.
Natural experiment vs. naturalistic observation
A natural experiment uses externally produced differences in exposure to investigate effects. Naturalistic observation records behavior in its usual setting. The terms do not describe the same research design.
Limits
Natural experiments can offer strong evidence, but the label does not automatically remove selection bias or place every study at a fixed level below a researcher-run trial. Evidence depends on the assignment process, measurement, analysis, and validity of the causal assumptions.