Levi Boxell, 13 October 2021

With the array of television news networks, the expansion of internet services and the rise of social media, voters see more images daily than at any previous moment in history. And yet, the literature on media bias tends to focus far more on text than photos. This column constructs a dataset of nearly one million image files from the 2016 US election cycle and finds deeply partisan coverage of different candidates. Media bias also varied in response to electoral incentives – in particular when switching from the primaries to the general election.

Harry Mamaysky, 08 August 2020

The COVID-19 crisis has been characterised by extremely volatile markets and extremely negative news coverage. Using all relevant Reuters news articles from January to June 2020, this column shows that a 12-topic model effectively tracks the evolution of crisis news flow. In the early stages of the crisis, markets frequently reacted to uninformative news. This dynamic underwent a structural break in mid-March, likely due to Fed interventions, after which markets became more ‘normal’. Investors, lacking early hard evidence on the effects of the crisis, interpreted many news stories as being informative about future pandemic outcomes, even though they were not.

Events

CEPR Policy Research