Causal inference when you can't experiment: difference-in-differences and synthetic controls

od Linear Digressions