Paper and/or Abstract  (joint work with Achim Ahrens, ETH, Zurich)

 A rich and growing literature exploits machine learning to facilitate causal inference.  A central focus is the estimation of treatment effects in the presence of high-dimensional controls and/or high-dimensional instruments.  High-dimensionality refers to situations where either many controls (instruments) are observed or where few controls (instruments) enter the model through a complex, non-linear and unknown function, or a combination of both. Double-Debiased Machine Learning (DDML), proposed by Chernozhukov et al. (2018) utilizes supervised machine learners to estimate conditional expectations with respect to controls and/or instruments. We will discuss the theory and algorithms of DDML and illustrate using examples/Monte Carlos.

 The seminar will be held both in presence and online