Assessing the Quality and Distribution of News in Local Television and Newspaper Markets Using Data Science Methods

📰 Increasingly over the last two decades, U.S. newspapers have struggled with declining readership, decreasing revenues and competition from social media companies, among other challenges. More Americans have found themselves living in areas known as news deserts, or communities with limited access to credible and comprehensive news sources.

📺 Television stations have fared somewhat better financially than newspapers and have generally not faced the same monetary headwinds as newspapers. There has been little research directly comparing the relative quality and quantity of coverage of newspapers and television stations within the same regions, as the bulk of studies on news deserts have traditionally focused on newspaper and digital outlets. Additionally, most studies of quality and quantity of coverage have been limited in their ability to scale in that they depend, in large part, on manual techniques to code news content and classify the content into categories.

🎯 In this project, I investigated whether local television stations in Missouri tended to cover the news with the same quantity and quality as local newspapers in the same coverage area. Through geospatial analysis, natural language processing (NLP) techniques and machine learning models, I proposed a broad and scalable approach to uncover gaps in news coverage. I also proposed a resiliency model — one that could be transferred to other regions across the country — that provides local television stations with concrete metrics to match the quantity and quality of coverage of the newspapers in their areas should local newspapers continue to downsize or even close at the same rapid pace that has been observed in recent years.

💡 The project showed that television stations, on average, tended not to cover their regions as thoroughly as their newspaper counterparts, typically missing populations in far-out border counties. Television newsrooms also tended to produce half or even a third as much civic/political content and sports content as their newspaper counterparts in the same region, while producing a much higher share of content related to crime and roads/car crashes.