Validation of Causal Transportability for Advertisement Exposure Using Observational Data.

Abstract

Using the Z-Transportability framework for causal transportability for observational data, we examine the practical feasibility of causal transport methods from one domain to another, and validate the accuracy of causal effects for each domain and the causal validation of their differentials. Using tri-phasic data of purchase data, ad viewing data, and individual attribute data from the Intage Single-Source Panel Data, we estimate the causal effects of the presence or absence of an app intervention on the amount of relevant product purchases. We apply the Z-Transportability methodology of Bareinboim and Pearl (2012) to transfer the causal effects estimated for men in their 20s to different age groups based on demographic information. By comparing the results with the causal effects directly estimated for each age group, we refer to the feasibility of causal transportability in practice. Through this validation, we aim to conduct an empirical analysis in the domestic marketing field and establish a highly practical approach and process for causal transportability.