Methoden vergleichen

	Maschinelles Lernen-gestützte Kausale-Auswirkungs-Analyse	Panel-Ereignisstudie
Fachgebiet	Kausale Inferenz	Kausale Inferenz
Familie	Regression model	Regression model
Entstehungsjahr	2015-2018	1990s–2020s (modern panel formulation)
Urheber	Brodersen et al. (foundational BSTS framework, 2015); Chernozhukov et al. (double ML augmentation, 2018)	Formalized by Freyaldenhoven, Hansen, Perez-Orive & Shapiro (2021); widely applied in finance (Fama et al. 1969) and policy evaluation
Typ	Quasi-experimental causal inference with ML	Quasi-experimental / causal panel design
Wegweisende Quelle	Brodersen, K. H., Gallusser, F., Koehler, J., Remy, N., & Scott, S. L. (2015). Inferring causal impact using Bayesian structural time-series models. Annals of Applied Statistics, 9(1), 247-274.	Freyaldenhoven, S., Hansen, C., Perez-Orive, J., & Shapiro, J. M. (2021). Visualization, Identification, and Estimation in the Linear Panel Event-Study Design. NBER Working Paper 29170. National Bureau of Economic Research.
Aliasnamen	ML-augmented causal impact, ML-CausalImpact, machine learning causal impact, ML-augmented BSTS	event-study regression, dynamic DiD, relative-time regression, distributed-lag panel model
Verwandt	6	4
Zusammenfassung	Machine learning-augmented causal impact analysis combines quasi-experimental counterfactual reasoning with flexible ML prediction models to estimate the causal effect of an intervention on a time series outcome. Building on Brodersen et al.'s Bayesian structural time series (BSTS) framework and extended by double/debiased ML methods, it constructs a synthetic counterfactual from donor covariates and infers the treatment effect as the gap between observed and predicted post-intervention outcomes.	A panel event study estimates the dynamic causal effect of a treatment or policy by regressing an outcome on a full set of relative-time indicators — one for each period before and after the event — while controlling for unit and time fixed effects. The resulting coefficient plot shows how the treated units diverged from untreated units at each point in calendar time relative to their treatment date, making both pre-treatment trend violations and post-treatment effect trajectories immediately visible.