The Reliability and Temporal Stability of Self-reported Media Exposure: A Meta-analysis

The measurement of media exposure is essential to not only traditional audience research, but also media effects research which relies on accurate estimates of media exposure. Even in the age of digital trace data and passive audience measurement, the workhorse of basically all communication research is self-report data. This article presents a meta-analysis of the reliability and temporal stability of media exposure self-reports in 33 panel studies. Results from a Bayesian multilevel analysis show that self-reported media exposure was moderately reliable (0.69, 90% HDI: 0.65, 0.72) and highly stable (0.90, 90% HDI: 0.88, 0.92). In line with previous studies, the reliability of media exposure measures was higher in adult samples (0.72, 90% HDI: 0.69, 0.76) compared to adolescents (0.60, 90% HDI: 0.53, 0.66). Rank-order stability of media exposure was comparable in adult (0.91, 90% HDI: 0.89, 0.93) and adolescent samples (0.85, 90% HDI: 0.81, 0.90). Moderation analyses showed that self-reported exposure to specific outlets was more reliable than general media use in adult samples. Media-specific differences in reliability were only found in adolescent samples. Overall, moderate reliability in combination with high temporal stability poses important challenges for scholars investigating causes and consequences of media exposure.