Prof. Benjamin Liengaard (Aarhus University)

Date: Wednesday, November 22nd
Time: 4:00pm - 5:30pm
Location: Ludwigstr. 28 RG/III; Room 329
Title: “Dealing with regression models’ endogeneity by means of an adjusted estimator for the Gaussian copula approach”
Abstract: “Gaussian copulas are an instrumental variable (IV)-free approach to deal with endogeneity issues in linear regression models. Although this approach is beneficial in several ways, previous research has revealed performance issues in regression models with an intercept, which, in practice, is almost always necessary. This research will explain the performance differences between models with and without an intercept, showing that the endogenous regressor’s density estimation is the component that most affects the Gaussian copula approach’s performance. We use this insight to develop an adjusted estimator that performs consistently and substantially better than established ones. Finally, we suggest a simple correlation measure that supports researchers to meet the Gaussian copula’s sample size and endogenous regressor’s non-normality requirements to obtain significant bias reduction. Consequently, this research contributes to the Gaussian copula approach’s applicability regarding handling the important issue of endogeneity problems in regression models.”