Agenda Chasing and Contests Among News Providers

We model competition among news providers as a contest where each firm chooses to publish on a topic from a large pool of topics with different prior success probabilities. If a topic is successful, firms that chose to publish on it share a fixed reward. We explore how increased competition (as measured by the number of firms and/or the share-structure of the reward) and the prior distribution of topics affect the diversity of published news. We relate our findings to current trends in news media, characterized by lower barriers to entry and the increased use of sophisticated technologies to identify successful topics from the large amount of, both professional and user-generated content available on the Internet. We show that the contest nature of competition tends to lead to a broader set of published media themes with a higher representation for marginal topics. The breadth of topics increases the more topics follow a “fat-tail" prior distribution and the more a priori popular topics' success are correlated. It also increases with the number of competing firms but only if the share of the reward in the contest dissipates rapidly. We also explore the effect of asymmetry on competition. First, we assume that some firms have a `brand', i.e. a capability to attract a loyal audience. We show that branded publishers are more likely to choose topics with high prior success probabilities, while unbranded publishers tend to choose a priori `unlikely' topics. Second, we assume that some firms have better forecasting capability for the topics' success. Surprisingly, in this case, less informed firms choose topics in a conservative way (i.e. publish topics with the highest prior probabilities). When many firms reporting on the same topic increases the topic's rate of success, marginal topics may emerge but only if the contest is not too competitive and competing firms are neither too few nor too numerous. These findings are related to current trends in the news media industry.