A quasi‑experimental estimation technique that identifies a treatment effect by comparing changes in outcomes over time between a treated group and a control group, attributing the differential change to the treatment under the parallel trends assumption and other identifying conditions.

Principle

The method identifies causal effects only if, absent treatment, the average outcome trajectories of treated and control groups would have followed the same trend (parallel trends). Violations of that assumption or contamination of controls (spillovers, anticipation) bias the estimator.

Demonstration

Illustrative scenario → A policy is enacted in Region A but not in Region B. Recognition → collect pre‑ and post‑policy outcome measures for both regions. Action → estimate effect as (A_post − A_pre) − (B_post − B_pre), report pre‑trend graphs, placebo tests, and robustness to alternative control groups. Consequence → if parallel trends and no interference hold, the estimate recovers the average treatment effect on the treated; otherwise the estimate may be biased.

Misapplication

Assuming parallel trends without empirical checks, using controls exposed to spillovers, or interpreting simple two‑period DiD as causal when treatment timing or group composition varies across units.

Consequence

Under the method’s assumptions, DiD provides a transparent, practical causal estimate with clear counterfactual logic; under violated assumptions it yields biased attribution of effect and misleading policy inference.

Reversal

When treatment is staggered across units or effects are heterogeneous, the classic two‑period DiD estimator can produce misleadingly weighted averages—recent methodological extensions or alternative estimators are required to recover interpretable effects.

Boundary

Appropriate for panel or repeated cross‑section data with pre‑ and post‑treatment observations and a plausible control group. Not applicable to single cross‑section comparisons without temporal variation or to randomized control trials (which provide different identification strategies).

Semantic Tension

Causal identification (credibility of counterfactual via control group) versus comparability (how similar are treated and control groups in unobserved trajectories).

Synthesis

DiD is an intuitive counterfactual device that turns temporal differences into causal estimates, but its credibility rests on testable pre‑trend evidence and careful selection of control units and robustness checks.