This is an online seminar using Cisco Webex. You will receive the login data with the invitation to the talk.
Abstract: The tracking of online user behavior is considered essential for the construction of consumer profiles, which help platforms monetize their services. Prominent examples are advertising in online search or social media, but also online retailing in which matching consumers to products and services is key. Measurement of the relevance of data is important for firms collecting data but also for policy, for example to understand the competitive effects of increasing data collection. Yet, measuring how much information platforms can extract from online behavior is difficult. We show how prediction quality of a range of consumer profiles varies across platforms of different size and scope of user data. Using data on users' clickstreams and website-specific trackers, we link tracker ownership to varying prediction quality based on the extent of tracked website visits. We find decreasing returns to the number of users observed and the number of websites tracked. For some groups, the largest trackers such as Google, Facebook, and AppNexus have a visible advantage in prediction quality. As a result, we aim to inspect the role of this prediction advantage over time, relating changes in prediction quality specifically to important acquisitions such as Google/Doubleclick and historical changes in data combination policies.
Joint work with Christian Peukert, HEC Lausanne