A0389
Title: High-dimensional causal projection estimators under weak confounding
Authors: Alessio Sancetta - Royal Holloway, University of London (United Kingdom) [presenting]
Abstract: Abstract projection estimators are a fundamental tool in applied economics, widely used to approximate causal relationships when the true model is unknown. High-dimensional causal projection estimators that remain robust to unobserved confounding are introduced. A framework is considered in which a large number of observed covariates can offset the bias induced by pervasive, latent confounders. The approach accommodates variation in confounder strength, allowing for weak confounding, and yields estimators that are consistent and asymptotically normal, even under fat-tailed and weakly dependent data. The methodology is demonstrated in three empirical settings: Estimating the dynamic effects of monetary shocks, revisiting GDP convergence across countries, and assessing the impact of U.S. expropriation laws on house prices.
