Endogeneity of marketing variables in multicategory choice models

Abstract A regressor is endogenous if it correlated with the unobserved residual of a model. Ignoring endogeneity may lead to biased coefficients. We deal omitted variable bias that arises firms set marketing variables considering factors (demand shocks) researchers do not observe. Whereas publications on sales response or brand choice models frequently take potential into account, multicategory provide different picture. To consider in models, we follow two-step Gaussian copula approach. The first step corresponds an individual-level random coefficient version multivariate logit analyze yearly shopping data for one specific grocery store, referring 29 product categories. If assumption correlation structure met, approach indicates category-specific about 31% majority rated as are positively variable, implying ignoring leads overestimation coefficients respective variable. Finally, investigate whether taking account by managerial implications. In this regard, demonstrate our often suggests level activity too high.