Stern Center for Research Computing

New York University • Leonard Stern School of Business

Measuring Asymmetric Persistence and Interaction Effects of Media Exposures Across Platforms

Bryan Bollinger, Michael Cohen, and Lai Jiang

In this paper, we explicitly model and estimate the effect of paid, owned and earned media exposures, including television, online banner ad, and Facebook exposures, on purchase behavior at the household level. We use an advertising goodwill model, allowing for asymmetric decay rates for channel-specific goodwill stocks, and incorporate two levels of interactions. First, we include interaction effects between these goodwill stocks in the consumer utility function. Second, we allow for interactions in exposures across channels in the goodwill production functions. We use hierarchical Bayesian methods to estimate the model, incorporating channel-specific models of exposures to control for endogeneity due to firms’ ability to set aggregate levels of advertising as a function of expected demand, as well as their ability to target specific types of consumers. Due to the computational demands of the model, we use NYU Stern’s High Performance Computing servers.

Our single source data allow us to assess both the short-term and long-term marginal contributions of paid, owned and earned media on sales at the consumer level; we find no meaningful interactions in the consumer utility function, but we do find a positive interaction between TV and online exposures in the creation of goodwill. On average, Facebook exposures have an insignificant effect on purchases although there is considerable heterogeneity in its effect.