Simultaneous Estimation of Heterogeneous, Flexible Distance Decay Functions to Better Understand and Predict How Far People Will Go to Be Entertained
We incorporate purchase data for 2.3 million households along with organization, trade area, and household characteristics to simultaneously estimate heterogeneous, flexible decay functions for 84 visual and performing arts organizations. To smooth noise in the purchase data attributable to spatial disaggregation, we introduce an iteratively reweighted algorithm that produces robust estimates and better out-of-sample predictions than four comparison models. We demonstrate the predictive validity and usefulness of the heterogeneous, flexible decay function estimates by: (1) generating counterfactual simulations that provide insights to (a) policy-makers as to how changes in socioeconomics, population, population density, and commute times impact attendance at entertainment venues, and (b) managers as to how to effectively grow attendance and program revenues; and (2) using the results to build a spatial model that more effectively controls for market-level variations in the study of purchase behavior and retail performance; results demonstrate that spatialized trade area variables explain, on average, 20% of the variation in organization-level measures that are commonly used in retail performance studies.
SMU Cox: Marketing (Topic)