In the wake of the 2016 election, there was a lot of finger-pointing about what the media did and didn’t do. There was warranted criticism of giving one candidate free ongoing airtime for his campaign rallies; there was justified questioning of the amount of attention paid to emails stolen from another candidate’s campaign chairman. And there was no shortage of arguing that the media’s lack of familiarity with blue-collar America led it to miss that Donald Trump would emerge the victor.
That latter criticism is further from the mark. Hillary Clinton won more votes than Trump, in line with polls right before Election Day. Some state polls overstated her support, throwing off aggregated assessments of who was likely to win the electoral college vote. As the FiveThirtyEight website put it shortly before Election Day, Trump was within a normal polling error of winning the presidency. And so he did. He was deeply unpopular, a position driven in part by the accurate coverage of his campaign and track record, but he won anyway.
On Sunday, Politico offered a related rationale for Trump’s victory.
“An extensive review of subscription data and election results,” Shawn Musgrave and Matthew Nussbaum wrote, “shows that Trump outperformed the previous Republican nominee, Mitt Romney, in counties with the lowest numbers of news subscribers, but didn’t do nearly as well in areas with heavier circulation.”
In a nutshell, Musgrave and Nussbaum argue that places in which fewer people had a print or digital news subscription were also places that were not only more likely to have voted for Trump in 2016 but are places that voted for Trump more heavily than they had voted for Mitt Romney in 2012. This, the authors hypothesize, left an information vacuum into which Trump’s social media presence and other social media stepped.
The only industry that loves stories about itself more than Hollywood is the media, and so the Politico story caught fire in the media’s public chat room, Twitter. Here was a doubly exonerating thesis: The media’s truth-telling was simply unseen in places that opposed Trump — and those places that the media was criticized for ignoring had ignored the media right back.
It’s just that this story downplays at least one correlation that probably explains many of the effects it claims.
“The results,” the authors wrote, “show a clear correlation between low subscription rates and Trump’s success in the 2016 election, both against Hillary Clinton and when compared to Romney in 2012. Those links were statistically significant even when accounting for other factors that likely influenced voter choices, such as college education and employment, suggesting that the decline of local media sources by itself may have played a role in the election results.”
Noticeably missing is another factor that should immediately leap to mind: Urban vs. rural.
Particularly east of the Rockies, Politico’s map of “news desert” counties — “places with minimal newspaper subscriptions, print or online” — overlaps heavily with places with larger rural populations. Local print news outlets require significant population density to be successful; delivering newspapers to hundreds of people scattered over thousands of square miles is necessarily more expensive than dropping off the New York Times to subscribers in Manhattan.
Consider Politico’s graph of vote results relative to the density of news subscriptions.
Compare that to a graph of the 2016 results by county — relative to the percentage of the population that lives in a rural area.
Our graph excludes counties where nearly all of the population is rural, since it skews the graph. (Ours shows counties in which 99 percent of the population is rural to counties in which zero percent is.) About a fifth of counties are almost entirely rural, and 93 percent of them voted for Trump. Many of those counties are ones for which Politico had no data, probably in part because the populations are so small in those places.
The Politico article does address the question, mind you.
“Many of the counties in POLITICO’s analysis of news subscription rates tracked with the urban-rural divide that was widely noted in the 2016 election,” the authors wrote, “with higher-density population centers having higher subscription rates.” In other words, there was, in fact, a correlation between rural populations and “news deserts.”
This is waved away.
“But Trump appeared to struggle even in rural areas with a higher-than-average presence of mainstream media,” they wrote. Such as Yuma County, Ariz., which is “a heavily rural county on the Mexican border” with a high news subscription rate where Trump did worse than Romney. Compare that with Robeson County, N.C., which has similar economic characteristics as Yuma — but which has lower subscription rates and voted more heavily for Trump.
The catch, here, is that while Yuma County does indeed have a lot of empty space, much less of its population lives in rural areas. Yuma County is home to Yuma, Ariz., which has about half the county’s population. Census Bureau data indicates that about 10 percent of the county’s inhabitants live in rural areas, compared to Robeson County’s 63 percent.
A better example for the authors would be Sweetwater County, Wyo. It’s identified as a “news desert” on Politico’s map (despite being home to the Rock Springs Daily Rocket-Miner). Only about 10 percent of its population live in rural parts of the county, and it voted 14 points more Republican in 2016 than 2012.
But, then, we’re just cherry-picking, as the authors were when identifying Yuma and Robeson Counties. We could as easily pick the juiciest cherry: That little red dot at the far right of Politico’s dot-chart. The county with the highest rates of news consumption in the country … also voted for Trump? That doesn’t really help the argument.
There’s a good reason in Politico’s piece to downplay the idea that the Trump vote was a function of news subscriptions: Very few people actually subscribe in even the most subscription-saturated places. Politico’s data show that the most-dense counties have subscription rates of about 35 percent — meaning that two-thirds of the homes don’t have any subscriptions at all. That explains why Trump did so poorly in those places?
It’s also worth noting that the argument that these news-light areas voted more heavily for Trump than Romney — an argument that implies that the president was better able to leverage that vacuum than the preceding Republican candidate — also overlaps with the urban-rural divide. More rural counties voted more heavily for Trump than they did for Romney as a rule.
The challenge with a lot of data analysis is that it’s hard to pick out causality. We could flip our argument on its head: How do we know that the reason more-rural places voted for Trump wasn’t because they had fewer news subscribers? Or, as a critic of the Politico argument told its authors, perhaps it was lower rates of education in those counties that led to more Trump support.
There are a lot of factors that make rural counties more likely to vote Republican than urban centers — too many to articulate casually here. It’s a pattern that was established before 2016 and which, along with other political divides, has been exacerbated over time.
For example: We could as easily argue that the defining factor for Trump was the ability to buy a firearm. Counties that vote Trump have a lot more gun dealers per 100,000 residents than places that voted for Clinton, as I wrote in February.
This correlates to being more rural, of course, but we can still make the argument.
For some reason, though, the media isn’t highlighting that relationship.