ATLASResearch
methods
Quantitative/ Study design

Quasi-experiment

Learn from a change you did not assign.

Estimate intervention effects without randomised assignment by exploiting a credible comparison structure. Designs include difference-in-differences, regression discontinuity, and interrupted time series.

WHEN IT FITS

Randomisation is unavailable, but a policy change, cutoff, or timing pattern can support a defensible counterfactual.

Strengths

  • Can evaluate real-world policies at scale.
  • Uses existing variation or administrative data.
  • Makes causal assumptions explicit and examinable.

Limitations

  • Identification rests on design-specific assumptions.
  • Concurrent changes may undermine the comparison.
  • Some designs identify effects only near a cutoff or in specific groups.

Know the boundary

A before-and-after difference by itself is not a credible causal estimate. The counterfactual argument is the heart of the design.

USED ACROSS
Banking & financeBusiness & MBAPsychologyComputer scienceEducation