Generating Correlated Synthetic Binary Indicators of Radio Listening Behavior for Long-Term Projections

• Activity Number: 492
• Type: Contributed
• Date/Time: Wednesday, August 3, 2016 : 8:30 AM to 10:20 AM
• Sponsor: Survey Research Methods Section
• Author(s): Yanna Yan* and Michael Elliott and Brady West and William Waldron
• Companies: University of Michigan and University of Michigan and University of Michigan and Nielsen
• Keywords: Synthetic Data ; Correlated Bernoulli Variables ; Radio Listening Behavior

Abstract

Radio panels, such as the Nielsen Audio panel, are a reliable source of information on the population's radio listening behavior. Nielsen panels consist of a representative probability sample of a target population, and panel members constantly carry a wearable device that extracts radio signals 24 hours a day. This device generates a binary variable (0/1) to indicate if a respondent listened to a particular radio station for a given 15-minute time window. This raises the question of can limited capture of exposure data (e.g., one week) be used to project radio listening behavior for longer periods of time? This study evaluates a proposed beta-binomial modeling approach for generating correlated synthetic indicators of radio listening behavior for an entire month based on one week of panel data. The presentation compares synthetic estimates of "cumulative reach" (i.e., number of people were exposed to a radio station) for one month to direct estimates of reach based on a full month of panel data, and finds that the proposed methodology works quite well.