Leveraging Differential Privacy for Targeted Advertisements

Abstract

Differential Privacy offers the online advertising industry new means to increase consumer privacy by obfuscating consumer data. While achieving the same privacy, these means decrease targeting accuracy differently, subsequently reducing advertisers’ willingness to pay (WTP) for targeted advertisements. Up to now, it has proven challenging to trade-off between privacy and advertisers’ WTP because of the unknown impacts of (i) privacy on targeting accuracy and (ii) targeting accuracy on advertisers’ WTP. This article presents these impacts based on (i) an analytical model explaining how privacy affects targeting accuracy and (ii) an empirical study estimating the impact of targeting accuracy on advertisers’ WTP. The analytical findings show heterogeneity of the privacy’s impact on targeting accuracy depending on how the industry obfuscates data. The empirical results indicate that a 1% decline in targeting accuracy reduces advertisers’ WTP by 1.16%. Combining the analytical and empirical results reveals that different means to obfuscate data that yield the same 5% increase in privacy could reduce advertisers’ WTP by 10.44% (i.e., -$0.31 CPM) or 16.24% (i.e., -$0.48 CPM). Therefore, these findings could provide an opportunity for the industry to improve consumer privacy without needlessly decreasing advertisers’ WTP.

Keywords: Online Advertising, Targeting Accuracy, Differential Privacy

JEL Classification: M3, M31, M37, M15, C1, L5