Choosing Response Models for Budget Allocation in Heterogeneous Markets with Carry-Over Effects: Why Simple is Sometimes Better

Abstract

Calibrating a sales response model for budget allocation is a complex task because sales responses may vary across allocation units and usually involve carry-over effects. Many applications are characterized by small samples or other data restrictions, thus raising the question of whether the available data are consistently rich enough for complex econometric estimation techniques. As a consequence, researchers may sometimes be better off using simpler estimation techniques or allocation heuristics, even if these methods are theoretically inferior. We test this proposition by conducting a comprehensive Monte Carlo simulation study using a large number of data conditions ranging from rich to very poor. To evaluate the success of each model calibration strategy, we use the expected profits resulting from the proposed allocation instead of the usual goodness of parameter recovery. Using this novel measure, we find that when the data quality is questionable, using simple response models such as fixed effects or even allocation heuristics can be very effective from an allocation decision maker’s point of view.