Double machine learning (DML) solves a fundamental problem: machine learning models like LASSO or neural networks are great at prediction but not directly suitable for causal inference—they optimize prediction loss, which differs from valid treatment effect estimation. DML uses two samples and two stages: first, use ML on one sample to predict outcome and treatment separately, accounting for all confounders flexibly; second, regress outcome residuals on treatment residuals using the other sample. This 'orthogonalization' removes confounding, and using separate samples prevents overfitting. The final estimate is unbiased and approximately normal regardless of the ML method used. DML combines econometrics' causal…
Read the full lesson
Sign up free: one personalized lesson every day, matched to your role and goals.
Already have an account? Sign in