Reevaluating the Stability of Media Use

Abstract

Measuring media exposure is notoriously difficult due to the unreliability of self-reports. Current methods often suggest that media use is highly stable after correcting for measurement error, but this paper argues that such stability estimates may be inflated due to problematic assumptions in common reliability models. Specifically, the quasi-simplex model, widely used to measure stability, rests on assumptions that are frequently violated in communication research. These include the lag-1 assumption, which, if violated, can lead to underestimated reliability and overestimated stability. Despite reports of extreme stability, changes in media use are often associated with theoretically expected outcomes such as political polarization and participation. This suggests that meaningful variation may be overlooked. I propose alternative strategies for assessing stability, including modeling lag-2 effects and using polychoric correlations, to provide a more accurate picture of media behavior over time.