We study media value in theory and practice. We focus on situations in which a media outlet observes the realization of a state of the world and must decide whether to report this realization to a consumer who pays an opportunity cost to consume the report. The probability of optimal reporting for the consumer is monotonic in an appropriate scoring rule, a statistical measure of the amount of “new” in the achievement compared to the consumer’s previous. We show that a particular scoring rule from the statistical literature parsimoniously captures major trends in reporting probabilities across several areas of American television news. We argue that the scoring rule can serve as a useful control variable in contexts where a researcher wishes to test for bias in news reporting. Score monitoring significantly reduces the appearance of bias in our applications.
It is a new paper by Luis Armona, Matthew Gentzkow, Emir Kamenica and Jesse M. Shapiro. I guess this means the actual bias is more towards surprising news than negative news per se? Via Paul Novosad.
