Natural “Natural Experiments” in Economics
Causal Question / Estimand
A critical survey: when do “natural natural experiments” — naturally occurring events (weather, twins, biology) used as instruments or as-if-random variation — actually identify the economic parameter of interest?
Identification Strategy
Examines a range of celebrated natural-experiment IV studies and asks what economic parameter each identifies. The central critique: interpreting a reduced-form natural- experiment estimate requires an economic model, and the Exclusion-Restriction that makes the instrument valid is often violated because the natural event affects outcomes through multiple behavioral channels (optimizing responses, general-equilibrium effects). Without the model, it is unclear which structural parameter the IV estimate corresponds to — so “atheoretical” natural experiments can be less credible than they appear. Theory is needed to map instruments to interpretable parameters; design alone does not guarantee identification of an economically meaningful quantity.
Key Assumptions
The assumptions under scrutiny: Exclusion-Restriction (the natural event affects the outcome only through the treatment), instrument independence, and an implicit economic model giving the estimand meaning. SUTVA (no general-equilibrium spillovers).
Threats to Validity
This is a threats paper: natural events typically operate through several channels, violating exclusion; behavioral and equilibrium responses confound the “experiment”; the recovered parameter may not be the policy-relevant one.
Setting / Data
n/a — critical survey of natural-experiment IV studies (twins, weather, biological events, etc.).
Key Claims
- Natural experiments do not escape the need for economic theory; an explicit model is required to know what parameter is identified.
- The exclusion restriction is frequently violated because natural events have multiple pathways to the outcome.
- Credibility from “as-if randomness” can be illusory without a structural interpretation — a structuralist counterweight to the design-based program.
Connections
- Skeptical counterpoint to AngristKrueger2001-SearchForIdentification and the design-based AngristPischke2010-CredibilityRevolution.
- Concerns the Exclusion-Restriction central to Imbens2014-IVEconometriciansPerspective and Murray2006-WeakAndInvalidInstruments. See also IV
Citation
Rosenzweig, M. R., & Wolpin, K. I. (2000). Natural “Natural Experiments” in Economics. Journal of Economic Literature, 38(4), 827–874.