Join us for a Joint Quantitative Psychology Brownbag with Dr. Christian M. Ringle (Institute of Management and Decision Science, Hamburg University of Technology (Germany)
Join the talk using this Zoom link.
Title: A Multi-Copula Approach to Deal with Endogeneity in Regression Models
Abstract: Endogeneity is a persistent problem in regression models estimated with observational data and leads to biased and inconsistent parameter estimates when regressors correlate with the structural error term. Park and Gupta propose an instrumental-variable–free approach that addresses endogeneity by modeling the dependence between an endogenous regressor and the error term using a Gaussian copula. Identification relies on three assumptions: a Gaussian dependence structure, normally distributed structural errors, and a non-normally distributed endogenous regressor. Subsequent research shows that the original framework implicitly restricts correlations between endogenous and exogenous regressors and develops extensions that relax this assumption, thereby broadening the range of applicable models.
Despite these advances, limited attention has been paid to the sensitivity of results to the Gaussian copula assumption itself. Recent evidence indicates that violations of this assumption can generate substantial bias. This concern is particularly relevant because the dependence between endogenous regressors and unobserved errors is inherently unobservable, making the choice of copula difficult to justify empirically.
This research extends the copula-based endogeneity correction framework by allowing for alternative dependence structures beyond the Gaussian copula. We consider a set of commonly used copulas and introduce a model selection procedure that facilitates the evaluation and selection of an appropriate specification. The proposed framework improves the robustness of empirical analysis by reducing the risk of copula misspecification and enabling valid statistical inference in the presence of endogeneity. In marketing applications, this robustness can yield more reliable estimates of key quantities such as demand elasticities.
About Dr. Ringle: Christian Ringle is Chaired Professor and Director of the Institute of Management and Decision Sciences at Hamburg University of Technology (Germany) and Adjunct Professor at James Cook University (Australia). He earned his PhD from the University of Hamburg and holds an MBA from the University of Kansas. His research spans management, marketing, technology and innovation management, business analytics, machine learning, artificial intelligence, and methodological advances in multivariate data analysis and decision-making research. Dr. Ringle is the co-founder and co-developer of SmartPLS, one of the world’s most widely used software platforms for partial least squares structural equation modeling (PLS-SEM).
Additional event details are forthcoming.
Joint Quantitative Psychology Brownbags are organized by psychology programs at universities around the nation, including: The Ohio State University, University of Notre Dame, University of Maryland, College Park, University of North Carolina at Chapel Hill, University of Virginia, Vanderbilt University and University of South Carolina. Participating programs also include those from York University, McGill University, University of Missouri and University of British Columbia.
