From Prediction to Impact: ML Meets Econometrics

This hands-on workshop shows how to draw credible causal insights when RCTs aren’t feasible. Using a people-analytics case on employee training - with both a staggered rollout and a voluntary global rollout - we walk through the full pipeline: design, identification choice, estimation, and stakeholder communication. We cover Differences-in-Differences, Instrumental Variables, and Regression Discontinuity; build Directed Acyclic Graphs to formalize assumptions; and implement Generalized Estimating Equations, Coarsened Exact Matching, and Double Machine Learning. We also use Causal Forests to surface heterogeneous effects, giving practitioners a robust, end-to-end toolkit for impact evaluation in real organizational settings.