What would you do to know what news consumers actually want?

The costs of knowing the public.

“Solar Eclipse from Caroline Island,” H. A. Lawrence and C. Ray Woods, 1883. (The Metropolitan Museum of Art)

Human beings rarely behave or think the way you might expect. This uncertainty principle is one of the reasons journalism is a social necessity, though it makes it harder for journalists to gauge what the public actually wants from us.

When I covered the 2020 Democratic presidential primaries for the Los Angeles Times (before COVID-19), I interviewed scores of primary voters about the field of more than 20 candidates. It was pretty rare to run into voters with totally consistent worldviews. I met a twentysomething utility worker and self-described socialist from Nebraska who (of course) really liked Bernie Sanders but also (?) Pete Buttigieg.

A less polite way of saying this is that the average voter’s politics were not especially coherent — to me, anyway — nor are they required to be. My own views aren’t all that coherent either. Subjectivity is the essence of politics.

I don’t raise all this to get into the social science of decision-making in elections, but to make a more basic point about the fundamental chaos and unpredictability of human preference and our most reliable methods of navigating it. We are not all utility-maximizers or Kantian rule-followers but walking, talking mystery machines filled with galaxies of possibility.

Well, if you’re a journalist, how do you find out what your audience wants? Surveys and focus groups will show you people’s stated preferences, and data analytics will show you people’s revealed preferences. They frequently do not show the same thing, and in the end you’ll have to listen to one or the other.

Before the 2020 election, I did a Google Forms survey asking people what they wanted from my coverage, and more than 3,000 of my followers from social media responded. I had initially felt proud of surveying my audience to gauge their policy interests before the primary race, in which many Democratic followers identified climate change as a top policy issue.

Then I watched Democratic primary voters almost totally ignore the field’s highly qualified and leading environmentalist candidate, Jay Inslee. Analytics data said that not all that many readers seemed to be engaging with my meatiest campaign trail policy stories, let alone buying subscriptions for them.

I saw two empirical problems in my own experience of trying to do better.

My own surveyed followers were certainly not representative of the broader electorate, or even the L.A. Times readership at large (they were probably much more progressive). But even if my followers were representative of the universal public, should I pay more attention to what my followers said they wanted, or what the data said they actually consumed? What if my audience really liked the idea of being surveyed because it showed respect for their intelligence and relevance, BUT ALSO actually preferred consuming my calorie-free horserace stuff about who was going to win?

One thing I liked about my followers is that they were quite self-aware about this common contradiction between stated preferences and revealed preferences. NiemanReports:

In 11 questions developed in one day, Pearce surveyed respondents’ deal-breaker issues, voting strategies, thoughts on triaging coverage and determining frontrunners, “one big question” they would ask every candidate, and the difference between the stories they would assign reporters and the ones they would want to read. (They were honest: Policy explainers were “assigned” two-thirds of the time and only chosen to “read” a third.)

In the end, the survey results were more useful to me as a vox populi sourcing document — I had a list of potential voters who had told me their hot-button issues and about their willingness to be interviewed, which I did when I was looking for regular voter voices on certain subject issues. A couple years later, I tapped this database to interview some of my conservative respondents to ask them how they felt about being duped by Fox News after it settled its 2020 election results defamation lawsuit with Dominion Voting Systems. The people whose opinions I asked about how to do the news just… became the news, for me.

Would I do more proper and fully interactive focus-grouping in the future? I found it to be an enriching experience, but that’s a lot of work, and scaling a quality survey to an appropriate approximation of the general public could be prohibitively labor-intensive. One of the basic lessons of the internet is that it’s just a hell of a lot easier and more efficient to spy on people and monitor their commercial activity to get a reliable bead on their revealed preferences.

On platform services like Substack — dear reader, you should probably know this if you don’t already — you can see who subscribes, which one of your individual readers engage with these newsletters the most, and who clicks on the most links. And the smart creator journalist will know from this data surveillance that those people are your super-customers who deserve the highest-quality service, because they are the likeliest to pay the most for the actual content and to keep the lights on in our super-spender economy.

After all, one of the fundamental insights of data-mining isn’t that the same commodity has different value to different consumers, but that it’s now possible to cheaply identify and segregate those consumers and treat them differently at enormous scale, whether it’s through discriminatory pricing, premium subscription tiers or even just nonprofit fundraising.

This is the central premise of Google and Meta’s multitrillion-dollar ad tech business models, but it’s not actually something that makes them special relative to most entities trying to do anything online. Being anything less than invasive is leaving money on the table under the principle of what economists call consumer surplus.

You may have guessed by this point that I feel a little conflicted about the obvious market implications of these tensions and don’t spend a lot of time looking at my analytics as a result, which are pretty clear in saying that I should write less about my evergreen intellectual curiosities and take more sharp opinions on trending news events.

It requires more intellectual fidelity to say that most of the time I’m not sure where I landed on some problem that’s bothering me, at the cost of being able to simplify things and reward the people who managed to make it this far down the essay. My apology is that I’m a human being, after all. We’re complicated. And sometimes being a little frustrating, when I could be clearer, is my own way of showing respect.