Comparer des méthodes

	Expérience naturelle à simple aveugle	Expérience naturelle
Domaine	Plans d'expériences	Plans d'expériences
Famille	Process / pipeline	Process / pipeline
Année d'origine	Late 20th century (formalized practice)	1990s (formal methodological articulation); earlier in epidemiology (John Snow, 1854)
Auteur d'origine	Conceptual synthesis of natural experiment tradition (Haavelmo, 1944; Campbell & Stanley, 1963) with single-blind methodology	Varied; systematized in econometrics and political science (e.g., Meyer 1995; Angrist & Krueger 1991)
Type	Quasi-experimental design with partial blinding	Quasi-experimental research design
Source fondatrice	Dunning, T. (2012). Natural Experiments in the Social Sciences: A Design-Based Approach. Cambridge University Press. ISBN: 978-1107698000	Meyer, B. D. (1995). Natural and quasi-experiments in economics. Journal of Business and Economic Statistics, 13(2), 151–161.
Alias	single-masked natural experiment, blinded quasi-experiment, single-blind exogenous assignment study	natural quasi-experiment, naturally occurring experiment, exogenous shock design, as-if randomization
Apparentées	2	3
Résumé	A single-blind natural experiment leverages an exogenous, researcher-uncontrolled event — such as a policy change, lottery, or natural disaster — to create treatment and comparison groups, while applying single-blind procedures so that either the participants or the outcome assessors (but not both) are unaware of group assignment. This design combines the causal leverage of natural variation with reduced measurement bias from blinding.	A natural experiment exploits a real-world event, policy, or circumstance that assigns individuals to treatment and control conditions in a way that is plausibly random — or at least exogenous to the outcome of interest. Because the researcher does not control assignment, it occupies a middle ground between a true randomized controlled trial and purely observational research, offering stronger causal leverage than conventional observational designs when the as-if randomization assumption holds.