Difference-in-Differences: Policy Effect, Parallel-Trends Violation Bias and Standard Error
Estimate the effect of a policy on a treated group against a control group with difference-in-differences. Add a differential pre-trend to violate the parallel-trends assumption and watch the estimate drift away from the truth by delta times the window length, while noise and sample size set the standard error. Checked against the noise-free closed form, the OLS interaction coefficient and Monte Carlo panels. A Blender-built data bench.
This simulator renders a live 3D panel of individual observations for a treated and a control group across ten time periods, split by a policy introduced halfway through. Each point jitters gently around its group's true mean trend; the solid lines trace the two groups' actual outcome paths, and a dashed counterfactual line shows what the parallel-trends assumption predicts the treated group would have done without the policy. Sliders let you set the true causal effect, inject noise, resize the sample, and — most importantly — introduce a parallel-trends violation, watching the DiD estimator's bias grow away from the true effect in real time exactly as it would with a bad control group in a real policy study.
Estimate the effect of a policy on a treated group against a control group with difference-in-differences. Add a differential pre-trend to violate the parallel-trends assumption and watch the estimate drift away from the truth by delta times the window length, while noise and sample size set the standard error. Checked against the noise-free closed form, the OLS interaction coefficient and Monte Carlo panels. A Blender-built data bench.
3D · Three.js / WebGL renderer · 60 FPS target · runs fully client-side, no install