Quasi-experiments

  • Exploits natural or accidental variation in exposure that mimics random assignment.
  • These methods do not require randomized treatment but instead leverage external structures (e.g. thresholds, timing, policies) to approximate experimental conditions.
  • They exploit natural experiments or discontinuities in treatment assignment to estimate causal effects.

Overlap: Shares conceptual ground with Experiments, Adjusted Methods (often combined with covariate control), and Independent Detection (for identifying breakpoints)

Key feature: Use of non-random but plausibly exogenous variation

Usage: For causal inference using observational data where assignment approximates randomness


Table of contents