Did a Lack of Local News Cause Trump’s Election? It’s Hard to Say.

A new analysis from Politico has shown a correlation that many suspected. But the direction of causation remains murky.

Crumpled local newspaper.
Photo illustration by Slate. Photo by Thinkstock.

A new study from Politico brings cold hard data to an idea that many observers of the 2016 election have already put forward—it suggests the erosion of local news across America may have contributed to Trump’s victory. According to the analysis, which compared results from the past two presidential elections with subscription information from the Alliance for Audited Media, low subscription rates correlated with a greater vote share for Trump at the county level relative to both Hillary Clinton in 2016 and Mitt Romney in 2012. In so-called news deserts, the counties with the lowest news circulation, Trump tended to outperform both; counties in the top 10 percent of subscription rates, by contrast, were twice as likely to go for Clinton than those in the bottom 10 percent.

The findings have been much-discussed, in part because, as one ProPublica reporter put it, “[It] totally affirms what I saw on the ground in 2016.” The reasoning offered up by Politico is that Trump got away with his lies partially due to the dearth of locally trusted sources that could critically evaluate his claims. But it’s important not to overstate the study’s results just yet: So far, what we have is a correlation, not evidence of causation—meaning that low subscription rates could easily be a proxy for other factors. It’s unclear if not reading local news contributed to a vote for Trump or if the people who lacked access to such outlets were more likely to vote for him due to other characteristics common to news-desert communities. As Politico acknowledges, news subscription rates largely tracked with the urban-rural divide that was pored over extensively in the wake of the election; the reporters note that they controlled for college education and employment, but there is no mention of other potentially vital factors such as race, or broadband penetration, or low-information voting.

Another important consideration that goes unaddressed in the piece is the kinds of stories that are typically given column inches or airtime in local newspapers and on local TV stations. There are certainly regional outlets that do vital investigative work, exposing corruption and holding powerful figures in their communities to account (and they shouldn’t have to be the last bastion of informed political discourse to be worth maintaining), but a great deal of local news coverage simply isn’t of this nature. As one former employee of a small Southern newsroom noted on Twitter, some remaining papers “are little more than newsletters for the local chamber of commerce. Deliver the obits, police report[s] and local sports to keep the few subscribers happy.” In other words, not all local papers regularly cover national politics or fact-check stump speeches to begin with—so subscribing to one doesn’t guarantee a greater awareness of inconsistencies or inaccuracies in Trump’s campaign statements. Nikki Usher, an associate professor at George Washington University, corroborated his personal experience with her own extensive research:

The piece also cites local newspapers’ refusal to endorse Roy Moore and independent verification of the Washington Post’s reporting on his sexual misconduct as evidence supporting the news-influences-political-outcomes theory. But it doesn’t tell us which newspapers conducted investigations or withheld endorsements in which counties, so the relationship between local coverage and outcome isn’t straightforward. And unlike with Trump, it’s not immediately clear that low subscription rates corresponded to a Republican win at the county level. Wilcox, Alabama, a news desert per Politico’s map, voted for Jones; about 40 nondesert counties preferred Moore.

This isn’t to suggest the Politico study is without merit or utility, just that it is an early step in assessing direction of causation. Ultimately, more work needs to be done in order to get the whole picture. (Some counties were excluded due to lack of data, and not all local papers report their circulation to the Alliance for Audited Media.) The move from speculative “Trump safaris” toward data that can formalize what was previously anecdotal is important. Transparency is too. The best course of action would be for Politico to follow the examples set by the likes of ProPublica (which publishes its data sets online and invites readers to share their own stories and insights) and the Upshot (whose code and raw data are often hosted on GitHub) so that readers and external experts can draw meaningful conclusions more quickly and with greater confidence. That will leave us better equipped to get to the heart of the problem of disinformation—and hopefully to start solving it.