Economics Seminar Series with Ivan Jeliazkov

Ivan Jeliazkov,  Associate Professor, University of California-Irvine uis a Bayesian econometrician whose research focus is on modeling, model comparison, and the development of methods for simulation-based inference. He is presenting: 

The Impact of Heteroskedasticity in Observational Studies of Causal Effects

Abstract:  There is an extensive causal literature, yet little is known about the impact of heteroskedasticity in popular observational settings. While heteroskedasticity in treatment assignment or response generation can be of intrinsic interest in many contexts, its interaction with nonlinearities in treatment models can bias estimators in ways that standard error adjustments alone cannot correct. To address these challenges, we extend current methodology in regression discontinuity designs, potential outcome regressions, propensity score matching, and inverse probability weighting using a Bayesian framework. This approach incorporates flexible modeling and customized parameter estimation and model comparison algorithms to account for model uncertainty and improve estimation efficiency. Simulation studies assess the consequences of omitted heteroskedasticity and gauge the performance of the proposed modeling and estimation methods, while real-world applications demonstrate their practical implications. Specifically, we examine the effect of academic probation on subsequent academic performance, the influence of Medigap on healthcare expenditures, and the impact of COVID-19 vaccination on mental well-being in the UK. The results provide compelling evidence that the risks of ignored heteroskedasticity in causal analyses should not be overlooked.