Workshop: Addressing Endogeneity in Marketing Models
Many empirical research projects in marketing that use non-experimental data are struggling with the proper identification of causal effects of independent variables (e.g., price, management decisions) on dependent variables (e.g., demand). The reason is that the identification of a causal effect hinges on the untestable assumption that the error term of a model is uncorrelated with the independent variables. If this assumption is not met, a model is plagued by endogeneity.
The topic of endogeneity has received considerable attention, and it is probably the most frequently encountered troublemaker in a review process at an academic journal.
This course therefore has the goal of making students familiar with the problem of endogeneity and potential remedies. This implies that it will cover the opportunities and problems associated with traditional approaches (e.g., Instrumental Variable estimation) as well as more recent developments (e.g., Gaussian Copulas; Bayesian implementation). The course will also cover how the data structure (e.g., panel data) can be utilized to address the problem. Because the literature on endogeneity is often quite technical, this course aims at providing an easily accessible approach to this topic. Special emphasis will also be given to understanding when endogeneity indeed poses a real problem as compared to settings in which endogeneity is less likely to be a real threat to the validity of the findings.