Showing to be Seen: Using Data Science to Discover TV Programs for Public Health Announcements
Professor, Information Systems & Decision Sciences
Muma College of Business, University of South Florida
Friday, April 26, 2019
10:30 AM – noon
Speakman Hall Suite 200
Television is a prominent channel for educating the public about chronic health conditions. This study presents a methodology for selecting TV programs for public health campaigns targeted to the at-risk individuals. Through high-dimensional analysis of a large dataset on TV viewership of the entire U.S. panel in 2016 the methodology first inductively discovers programs whose popularities are correlated with eight chronic conditions and risk factors. A series of nonparametric tests then examine the robustness of findings and verify that a significant portion of the correlations is genuine—that is, not all the discovered correlations are accidental due to the “curse of dimensionality.” We then use Facebook’s split testing platform and run a series of online experiments to compare the effectiveness of targeting the shows discovered by the methodology with those that were targeted by the major 2016 public health campaigns. The experimental results corroborate the potential value of the methodology, which opens up a potentially new set of programs for public health officials to consider in their efforts to combat a range of conditions that are significantly expensive both in human lives and cost to the economy.