Local News Online and COVID in the U.S.: Relationships among Coverage, Cases, Deaths, and Audience

We present analyses from a real-time information monitoring system of online local news in the U.S. We study relationships among online local news coverage of COVID, cases and deaths in an area, and properties of local news outlets and their audiences. Our analysis relies on a unique dataset of the online content of over 300 local news outlets, encompassing over 750,000 articles over a period of 10 months spanning April 2020 to February 2021. We find that the rate of COVID coverage over time by local news outlets was primarily associated with death rates at the national level, but that this effect dissipated over the course of the pandemic as news about COVID was steadily displaced by sociopolitical events, like the 2020 U.S. elections. We also find that both the volume and content of COVID coverage differed depending on local politics, and outlet audience size, as well as evidence that more vulnerable populations received less pandemic-related news. Introduction As health and medical professionals continue to fight COVID on the front lines, a different battle is playing out on Americans’ screens. Public health officials and politicians, aiming to spread critical health-related information and to shape the narrative around the pandemic, are fighting for space and attention in both new and traditional media. In order to understand fulfilled and unfulfilled information needs during the crisis, tools and data are needed to rapidly characterize what information the public is receiving, and how that information varies across populations. To this end, scholars have begun to analyze the information environment surrounding COVID, showing partisan variation in coverage in national news outlets in the U.S. (Muddiman et al. 2020; Krupenkin et al. 2020), and how urban/rural divides in news have been associated with behavior (Kim, Shepherd, and Clinton 2020) during COVID. However, despite the critical role they have played in information consumption during COVID (Ritter 2020; Kim, Shepherd, and Clinton 2020) and their consistent track record of shaping opinion and behavior (Schulhofer-Wohl and Garrido 2013; Hayes and Lawless 2015; Gentzkow, Shapiro, and Sinkinson 2011), no work has yet looked at what information has been produced from local news outlets durCopyright © 2021, Association for the Advancement of Artificial Intelligence (www.aaai.org). All rights reserved. ing the pandemic, or how that coverage has varied across outlets. Local news, moreover, is important because at least early in the pandemic, the majority of Americans turned to local news outlets to understand the impact of COVID in their community (Frank 2020). Further, almost no published work has looked at content of any kind during later stages of the pandemic, especially during the deadly ”third” wave that began in November 2020. How the media responded to the prolonged impact of the pandemic is therefore still largely unclear. We address these questions using a unique dataset of the complete set of 757,053 news articles written by 310 local newspaper outlets in the U.S. who provide RSS feeds of their online content. Our dataset spans a breadth of local communities around the country (see Figure 2), and captures articles written between April, 1st, 2020 and February, 17th, 2021, allowing for analyses regarding local information environments over a long time span. To the best of our knowledge, our data is the most comprehensive corpus of local news outlets available for research. While there are similarly complete news datasets at a national level (Nørregaard, Horne, and Adalı 2019; Gruppi, Horne, and Adalı 2021), no such collections exist at the local level. Using these data, we address two primary research questions: • RQ1: What factors are associated with the degree of local news coverage, over time, of the COVID-19 pandemic? • RQ2: What were the primary COVID-related topics covered during the pandemic, and how did this coverage vary across outlets? With respect to RQ1, it would be reasonable to expect that coverage of COVID varied with the severity of the pandemic in the local area. However, many other factors (McCollough, Crowell, and Napoli 2017), e.g. economic issues (Peterson 2021), are at play in determining what local news outlets choose to cover. For example, Figure 1 shows that coverage of COVID was not equally distributed over time—on average, local newspaper agencies shifted their attention away from COVID as time progressed. We use a set of COVIDrelated keywords informed by prior work (Shugars et al. 2021) to identify articles that were or were not related to the pandemic. Using the county an outlet is located in as a proxy for the outlet’s audience, we then estimate the rear X iv :2 11 1. 08 51 5v 1 [ cs .S I] 1 6 N ov 2 02 1 30.0% 40.0% 50.0% 60.0% 70.0% Ap ril M ay Ju ne Ju ly Au gu st