Local journalism has long been understood as a central pillar of democratic life in the United States, providing citizens with coverage of local elections, civic engagement opportunities, and community events. Local news has historically facilitated accountability and participation by lowering the barriers to accessing reputable information. However, over the past few decades, economic pressures, media consolidation, and the rise of new digital media platforms has made it increasingly difficult for local news outlets to survive, producing what scholars call “news deserts” — communities with little or no access to the vital institution of local news. Scholars and journalists alike have warned that these news deserts may have consequences, weakening civic engagement by limiting citizens’ ability to access the information necessary for political participation. While prior research demonstrates that local news promotes civic engagement, this study examines county-level voter turnout in the context of the 2020 election, accounting for demographic controls and highlighting potential complexities introduced by digital news proliferation.

A large body of research studies the critical role that local journalism plays in democratic life. More specifically, these institutions increase political knowledge, increase the likelihood that public officials are held accountable, and facilitate methods of civic engagement by connecting people to issues that directly impact their own communities (Hayes & Lawless, 2015). When local outlets become few and far between, citizens are less informed, less trusting of institutions, political or otherwise, and ultimately are less likely to turnout to vote in elections. National news trends are compounding this issue as the increasingly fragmented nature of widespread news programs are allowing national and digital news platforms to push local news out of its once significant civic functions (Bennett & Livingston, 2018). 

Despite growing concern surrounding expanding news deserts, the empirical relationship between local news availability and election turnout is still pretty contested. Some studies, like those referenced above, discuss how loss of local newspapers reduces turnout and civic engagement, however, some studies point out that the proliferation of digital news sources creates an entirely different ball game, complicating an already complex relationship. This project aims to explore whether counties with fewer local news outlets exhibit systematically lower voter turnout than those with more robust local news ecosystems.

My hypothesis is that counties with fewer local news outlets will experience lower voter turnout than those with more access to local outlets. This is grounded in multiple political and social theories. First, research on political accountability demonstrates that local news can increase voter participation by lowering the cost, monetarily and mentally, of accessing information and monitoring elected officials (Snyder & Strömberg, 2010). Second, studies of civic engagement found that institutions—including local news—were vital in maintaining the norms and willingness of citizens to participate in many aspects of society (Putnam, 2000). Third, studies on media fragmentation suggest that when local news networks weaken, citizens may disengage from local politics in favor of more national or entertainment oriented broadcasts (Prior 2007). 

To test this hypothesis, I conducted a cross-sectional analysis of U.S. counties combining data from many sources into a new dataset. Data on local news outlets was found from the The University of North Carolina at Chapel Hill’s News Deserts database, which provides comprehensive, county level information on all active news organizations in the United States. I found voter turnout data from the National Neighborhood Data Archive (NaNDA). Additional demographic data to use as controls, including median income, age and gender levels were found from the U.S. Census Bureau’s American Community Survey. The primary independent variable was access to local news, operationalized as a count of active news sources. The dependent variable was voter turnout, operationalized as the percent of eligible voters that cast a ballot in the 2020 general election.

To manage the datasets I had to make sure they all labeled counties identically. To do this I used standardized Federal Information Processing Standards (FIPS) codes, which are five digit codes ensuring that each county has an individual ID number and won’t be confused with a similarly-named county in another state. Additionally I filtered all the datasets to the year 2020 to get a better cross-sectional look at the country. Because the analysis only focuses on one election year (2020) the dataset captures variations across counties not changes over time. This design allows comparison between counties with different local news environments while holding potential effects of time constant. Lastly, I found it impossible to distinguish if a county had missing news data or no news stations in the county, which may have caused a flaw in my regression, but I got rid of as many irregularities as I could by excluding counties with missing voter turnout information to help insure consistency and that a number of counties wouldn’t become outliers incorrectly having 0% voter turnout. 

The model I used was a multivariate ordinary least squares (OLS) regression with voter turnout as the dependent variable, regressed on the number of news outlets and other control variables (income, median age, gender breakdown, etc.). This analysis would not be able to determine causality as there is no way to randomly apply treatments to different counties. There is also no way to rule out reverse causality, such as politically active counties supporting more local newspapers, or potentially omitting an important control variable, consequentially, the results will only establish a correlation between the variables, not a causal relationship.

When I ran the regression, the model indicated a statistically significant relationship between the number of news outlets and voter turnout, however, contrary to my original hypothesis the coefficient for news outlets was negative, suggesting that counties with more news outlets have slightly lower voter turnout rates, holding the demographic control variables constant. I also noticed that there was one outlier county in Chicago which had over twice as many news outlets as any other county and had decently low voter turnout. I ran a regression only excluding this county and discovered that the relationship was no longer statistically significant, maybe pointing towards an outsized impact from a few counties on the rest of the data.