Laura Liu from the University of Pittsburgh will present "Binary Outcome Models with Extreme Covariates: Estimation and Prediction" Joint work with Yulong Wang (Syracuse)

Abstract: This paper presents a novel semiparametric method to study the effects of extreme events on binary outcomes and forecast future outcomes, which is particularly relevant given the recent occurrences of extreme events. Our approach, based on Bayes' theorem and regularly varying (RV) functions, facilitates a Pareto approximation in the tail while imposing no parametric assumptions on the relationship between covariates and outcomes beyond the tail. We analyze cross-sectional as well as static and dynamic panel data models, incorporate additional covariates in both setups, and accommodate the unobserved unit-specific tail thickness and RV functions in the panel setup. We establish consistency and asymptotic normality of the proposed tail estimator. We demonstrate that, under regularity conditions, our objective function converges to that of a panel Logit regression on tail observations with the log of the extreme covariate as a regressor, which simplifies implementation and facilitates empirical research. We evaluate the finite-sample properties of the proposed tail estimator in Monte Carlo simulations. In the empirical application, we examine a panel of small banks to assess whether they become riskier when local housing prices experience a significant decline, a crucial channel in the 2007--2008 financial crisis.