9.3 What the World Can Do For Us

Natural experiment: randomization of treatment occurs without a researcher controlling the randomization

Any path from the source of exogenous variation to the outcome must be [either]

* closed
* contain our `treatment`

Natural experiments (unlike pure randomization):

• Sometimes back doors from NaturalRandomness to Outcome
• Participants might not know that they are part of the experiment
• Later: carefully define variation from Natural Randomness (ch 10)
• Harder to get approval from critics