It’s a tension as old as media: Is a news outlet more in the business of giving the audience what it wants, or giving it what it needs?

For just a moment, set aside the reality that different people will always define wants and needs differently. Every editorial decision is made with two variables floating, however inchoate, in the deep recesses of a journalist’s brain:

Is this story going to do well — generate clicks, dominate social, make Page 1, lead the 11 o’clock news? And: Is this story important to tell — inform voters, expose wrongdoing, catch hypocrisy, inspire civic good?

No story is all of one and none of the other, but they’re all on that spectrum somewhere.

Broadly speaking, the Internet has made more stories available at every point on this wants/needs spectrum. But the most common editorial critique of digital news is that it’s gotten too focused on the “wants” end — mindless clickbait, partisan poison, misinformation that makes someone feel they’re on the right side of history.

Often, the stakes are low enough that how an outlet approaches a story doesn’t make much of an impact. But what happens when the subject at hand is critically important — even a matter of life and death?

That’s a question raised by a new paper just published in The International Journal of Press/Politics. It’s titled “Do Local Newspapers Mitigate the Effects of the Polarized National Rhetoric on COVID-19?” and it’s by Catie Snow Bailard of George Washington University. Here’s the abstract:

This analysis tests two distinct predictions regarding local newspapers’ coverage of the COVID-19 pandemic in the United States.

A public service view of local newspapers predicts that a robust local newspaper sector would mitigate the politicized national partisan rhetoric surrounding COVID-19; reducing the disparity in social-distancing behaviors between predominantly Republican and predominantly Democratic counties by increasing compliance in Republican counties. The alternative hypothesis, informed by a demand-side view of the market pressures local newspapers face, predicts that increased competition between local newspapers will increase the degree to which local newspapers amplify the rhetoric of national officials in line with the partisan composition of their community, further polarizing adherence to social-distancing behaviors across predominantly Republican versus predominantly Democratic counties.

The results of this analysis offer strong support for the second hypothesis; but, an additional analysis of vaccination rates offers a more nuanced perspective than a simple public service versus demand-side dichotomy would imply.

You may have heard that how people think about COVID-19 is strongly connected to their partisan identity. No, really, it’s true! Your preferred political party is more important in determining how you feel about virus safety measures (like vaccines and masking) than whether or not you’ve actually had COVID yourself.

Many of the on-the-ground conflicts those differences have spawned have been fundamentally local — think school board meetings gone wild. But from a media perspective, the messaging is primarily national. It’s national outlets like Fox News and OANN, the national political figures who appear on them, and national digital outlets on the right that have fueled all the skepticism and mistrust — which then gets recycled into local Facebook groups and local communities.

So could an institution like your local newspaper — broadly considered less partisan and more trustworthy than national media, even in its now-shrunken state — play a role in counter-programming all the national noise? Or would it instead lean into what its readers are already hearing on cable news at night?

Here’s how Bailard frames her two hypotheses:

Hypothesis 1: Relative to their counterparts in counties with fewer to no local newspapers, predominantly Republican and Democratic counties with local newspaper abundance will be less polarized in their adherence to social-distancing, by increasing compliance in predominantly Republican to be more commensurate with that of predominantly Democratic counties.

Got that? There are more Republican counties and more Democratic counties. But some of those counties on both sides have stronger local newspapers, and some of them don’t. This hypothesis is saying the R and D counties with stronger local papers will be closer to one another on COVID issues than the R and D counties with weaker local papers. If true, Hypothesis 1 would mean, roughly, that local newspapers were giving people the information they need.

Hypothesis 2: Relative to their counterparts in counties with fewer to no local newspapers, predominantly Republican and Democratic counties with robust local newspaper sectors will be more polarized in their adherence to social-distancing behaviors.

And this is the opposite. Stronger local newspapers will do more to feed into their audience’s existing beliefs. In red counties, the local daily will play up Ron Johnson and Rand Paul; in blue counties, they’ll play up Joe Biden and Anthony Fauci. Hypothesis 2 says local papers are giving people the information they want.

Bailard’s method for capturing these ideas with are interesting, if limited. To measure a county’s whole vibe on COVID, she uses data from Unacast, a location-data company that you probably didn’t realize has been keeping track of your location all throughout the pandemic. (Ever tap “OK” when an app on your phone wants access to your location? There’s a pretty good chance that app then sells your data to someone else, like Unacast.) Orwell aside, Unacast’s location data can show how places reacted in the early days of COVID in terms of social distancing. Did people really stay home when government asked them to, or did they keep having crowded ragers in Dad’s basement? Unacast knows. She also pulls in data on local COVID deaths and the status of a county stay-at-home order during a given week.

The data on local newspaper “robustness” is a bit weaker, to my mind. Bailard uses a metric derived from UNC’s “news deserts” database, which aims to gather information on every single newspaper (and some other forms of local media) in the United States. This sort of data-gathering is a tremendous effort, but I am almost annoyingly on the record about its flaws.

A lot of places get missed. And measuring media by county makes about as much sense as the Electoral College. Los Angeles County has 10.1 million people; Hawaii’s Kalawao County has 86. But they’re each one county for “news desert” purposes. And focusing on print days — specifically, “the square-root of the number of days of local newspaper content that is published in a given county per week” — misses a ton of important information. Did The Local Gazette cutting to three days of print per week mean they instantly became 57% less “robust”? Or did they do it to cut printing costs and keep a few extra newsroom jobs? And does anyone read The Local Gazette, anyway? Shouldn’t it matter at least as much how many people are going to localgazette.com?

But if I adjust my crankiness filter down a few notches — you go to war with the data you have. It’s far from perfect, but might well still be interesting.1

So, what did Bailard find? “The results of the analysis offer strong support for the demand-side hypothesis” — that is to say, the dreaded Hypothesis 2.

Rather than function as a mitigating influence, a more robust local newspaper sector is correlated with greater polarization of social-distancing behaviors across predominantly Republican versus Democratic counties, reflective of the rhetoric espoused by their respective national party leaders.

As counties move along the spectrum from being a local news desert to having a robust local newspaper sector, the more polarized predominantly Republican counties are relative to predominantly Democratic counties in terms of adhering to social-distancing.

In substantive terms, in a county that skews Republican by a 40%-vote share, the average marginal effect of moving across the inter-quartile range for the local newspaper sector measure — from being a local newspaper desert (i.e., 0) to a relatively saturated local newspaper market (i.e., 7.5) — predicts a decrease of approximately 1.9 points in that county’s social-distancing grade, out of a 5-point scale. This is equivalent to a dropping from a grade of B- to a D-.

Data qualms aside, that’s a pretty serious gradient on that chart. To orient you, the left side of the chart represents counties with no local newspapers at all. The right side represents counties with lots of newspapers printing lots of days.

Want to be depressed? In none of these scenarios does having more robust local newspapers make a positive difference worth a damn. That top-most line represents counties that voted for Clinton over Trump in 2016 by 60 percentage points — so, mostly large urban counties. That line has a sliiiiiiight upward slant, indicating more social distancing coming along with more local newspapering. But in all the others — counties that went +20 Clinton, +20 Trump, and +60 Trump — more robust local newspapers correlate to less social distancing, not more.

You can see the partisan difference here:

Very Democratic counties are on the left here, very Republican ones on the right. You can see there is a positive relationship in very Democratic counties — but it only really kicks in once you get to really really Democratic counties, ones Clinton won 80-20 or more. Anything less than that level of Democratic support and more newspapers correlate with less social distancing.

Let me put my data-skeptic hat back on for a minute. Bailard does a lot of good work here controlling for variables that could be interfering with the clarity of the signal here: demographics, education levels, median income, urban vs. rural. But I’m not convinced it was enough. I’ve already aired my grievances about the newspaper data. But the spatial differences between a county that went 80-20 Clinton (like Philadelphia County, Pa., population 1,603,797) and one that went 80-20 Trump (like Rockcastle County, Kentucky, population 16,037) are massive. Social distancing means wildly different things in downtown Philly than in the Daniel Boone National Forest, and while I’m sure the Unacast folks did their best, I don’t think their rating system is strong enough to hold the weight of this sort of cross-county comparison.

It’s possible that having a more “robust” set of local newspapers leads to less social distancing. But it’s also possible that being the sort of place with less social distancing correlates with a more “robust” set of local newspapers by this sort of measure. (For example, a county of 50,000 people would be very likely to have one daily paper, in the county seat. But a county of 500,000 isn’t likely to have 10 daily papers, and a county of 5,000,000 isn’t likely to have 50. That’s not least because counties have public-notice obligations that require them to put legal ads somewhere, so every county’s incentivized to have at least one newspaper.) Basically, both newspaper “robustness” and social distancing “score” both scale up irregularly, a result of weirdly drawn county lines, population density, and a host of other factors — so I think it’s worth taking these results with whatever grains of salt you have handy. I think it’s more likely that there are underlying factors that drive both the newspaper robustness and the social distancing score.2

(To her credit, Bailard acknowledges this with a similarly saline analogy: “Since the estimated average marginal effects of interactions are sensitive to model specification and reliant on extrapolation where the moderating variable’s data is sparse, the size of these estimated marginal effects should be taken with a grain of salt.”)

I also just don’t believe local newspapers are powerful enough to make a county move from a B- to a D- in social distancing. Remember, as of 2020, only 10% of Americans said they get news from print media “often.” (And “print media” includes magazines and national newspapers, too.) Another 22% say they get news from print “sometimes”; more than two-thirds of Americans say they get news from print “rarely” or “never.” Only 5% of people said print was their preferred way to get news, versus 35% for TV, 26% for news sites, and 11% for social media. That small a group of people can’t have that big of an impact. (And, of course, the median print newspaper reader is much older than the median, say, Twitter user.)

Finally, remember that this study isn’t attempting to measure the actual editorial content of these newspapers. That’s not how it’s calculating “robustness.” But I’ve read a lot of local news the past couple of years — in blue states like Massachusetts and in red states like Louisiana — and I frankly haven’t seen any sort of wild ideological swing against social distancing, masking, or vaccines on their pages. I’m sure there’s a metro columnist or two who’s been irresponsible, and we can debate edge cases all day, but if the tone of local newspaper coverage had shifted this radically, I think we would have heard about it.

All that said, what I like about this paper is that it is a useful reminder that — unlike what it’s natural for journalists to think — more news isn’t always better news. Usually, yes! But more competition in a market can lead individual players to seek their own niche — perhaps an ideological one — that might make business sense but not serve the public interest.

1. I should also note that Bailard also runs crosschecks with a number of other variables, including some, like Fox News viewership, that make sense as a covariate. ↩︎
2. Free idea for any journalism-funding foundation: Fund research to create a more, er, “robust” measure of local newspaper health than just counting papers and print days. Specifically, I’d love to see a time-bound analysis counting how many original news stories these newspapers produce in a week, or a month. A chain-stripped daily that slips one local byline into a bundle of wire copy isn’t more “robust” than the Tampa Bay Times, just because it now only runs the presses twice a week. ↩︎