Data? We Hardly Knew Ya: Has data journalism lost its oomph?
I hope you’ve had a nice week and have exciting weekend plans. I am continuing my occasional attempts to gain more premium subscribers to Future Proof, who can join those ranks for the price of one (1) coffee per month. If enough people do it, I’ll stop asking.
“It’s tough being in a business while having a constant anvil over your head,” wrote Nate Silver this week, responding to the news that ABC, the owner of 538, the site he founded back in 2008, would be dissolved, resulting in the elimination of all the publication’s remaining jobs. The site had, since 2013, existed under the auspices of Disney, first as part of ESPN (reflecting a sports focus) and then with ABC News (representing more strategic interest in its political coverage). But in recent years, it had begun to feel like Disney’s belief in the project was waning. Layoffs were followed by layoffs, and Silver himself left the site in 2023.
Silver went here – to Substack! – and launched The Silver Bulletin, one of the biggest, most profitable newsletters on the site (if not quite on the same scale as the artist formerly known as FiveThirtyEight). There’s a joke in The Office where Michael Scott reveals himself to be the perfect enemy of the big-box paper company he had previously worked for. “If tomorrow my company goes under,” he reveals in negotiations over the ‘Michael Scott Paper Company’, “I will just start another paper company, and then another and another and another. I have no shortage of company names.” Having left the New York Times to launch FiveThirtyEight, Silver has timed his departure well from 538 to launch The Silver Bulletin. A gambler, by trade and instinct, not to mention an outsider, his preferred position is running his own show, free from the influence of the Sulzbergers or Igers of this world.
But the decline and fall of 538 is a troubling thing. Silver assigns the blame to a lack of focus – by both Disney and his own team – to developing the product, as well as the content. But the truth is that data journalism (and he doesn’t like the term, calling it “a dumb name for what we were doing”) is a difficult product. It’s unsexy, boring and peddles in an objectivity that’s hard to sell in the current climate.
Personal anecdote alert: back in 2018 when I set up my little podcast company, Podot, I wanted to launch a flagship show. I had been working for a political magazine and that was obviously the direction the show would head in, but how to make it distinctive? There were already a number of successful shows on the right and the left, but the idea of being too dogmatic made me queasy (after all, I also needed to make some income in the corporate world, where being seen as institutionally left-wing or right-wing was unlikely to do me favours). The best idea I could alight upon was a show called Polling Politics, pitched as a British version of 538, in podcast form. I collaborated with a superb pollster, Joe Twyman, who had just launched his own polling company, Deltapoll, which has gone on to be one of the UK’s most credible market research providers. And we brought in Marie Le Conte, an excellent Westminster hack (though she is currently reforming herself in that department), to balance out the nebbishness.
Clearly this anecdote is going to end with me not being wildly pleased with the outcomes. The show managed to attract partnerships with a few different bookmakers, which is a synergistic position for political polling (and, indeed, the granular data journalism that is currently à la mode in sports). But there was a very thin line to straddle. Too data focused, too numerical, too nerdy, and we pushed away the core audience for political content. But too chatty, too bolshy, too mainstream, and we disavowed our niche. Your credibility as a product comes from the presentation of disinterested data, but, without a bit of dynamic analysis, that would sit blandly on the plate, like a shrivelled new potato. In the end, the show became, I think, too generic.
But data journalism is a tricky thing to do. You are feeding readers their vegetables and they are often unwilling to swallow the stems. The covid-19 pandemic was a significant moment in the modern history of data journalism, because suddenly Joe Everyman was sat on his couch each night with a forensic interest in the graphs that were showing “reported cases” and “expected deaths” and things like that. Journalists, like John Burn-Murdoch of the Financial Times, became superstars because of their ability to take data, digest it, and spit it out in well-presented, parseable graphs and diagrams.
Because of this, I think that a lot of media organisations bet quite big on data being an important part of their offering going forward. But as with so many of the social changes initiated by the pandemic, it has proved only temporary. Silver, in his valediction to 538, speaks of the problems the site encountered during fallow years, when there were no elections for them to pore over. The truth is that data has to be really pressing for it to be compelling. This is something we encountered with Polling Politics: frankly the UK’s political polling ecosystem doesn’t produce enough good data to sustain a weekly chat-show, even in an election year. In the General Election here last year, we saw endless models of Tory destruction and Labour landslides, each using slightly different sets of variables, each claiming a revolution in methodology. All produced near identical headlines: “CONSERVATIVES FACING ELECTORAL OBLIVION”, “POLITICAL WIPEOUT LOOMS FOR RISHI SUNAK”, “LABOUR ON COURSE FOR RECORD MAJORITY” etc.
The narratives were so simple and so similar that you almost didn’t need to keep doing the polling. We weren’t learning anything new, and readers of the Sun or the Mirror seemed satisfied with these top lines. Once every pollster agrees on the expected outcome of an election, the other analytic variables become less interesting. Policy polling, while interesting and informative, is kind of anathema to the role of the media, which likes to think that it sets that agenda rather than simply reflecting it. Why report that the public are worried about immigration, the NHS and the economy, when you could run stories about migrants washing up in small boats, people dying in hospital corridors after deferred surgeries, and bags of lettuce now costing the same as a Nissan Qashqai? Somehow none of this ‘data journalism’ felt as urgent as that question – the most pressing journalistic question of the past several decades – of whether we were getting the pandemic under control.
Today, I saw a post on Bluesky celebrating the work the Financial Times has put into a recent story about the profits being made from the Trump memecoin. “Our stories often downplay the computational lift that goes into getting them,” Martin Stabe, the FT’s Data Editor wrote, appending a note on the methodology which allowed the team to access blockchain records for $TRUMP. It was a salient reminder of the extent to with forensic data journalism underpins simple headlines. Of course, the FT’s own headline – “Donald Trump’s crypto project netted $350mn from presidential memecoin” – elides much of the legwork done by data journalists. The interesting thing there is that figure – $350mn!! – and, having arrived at it, the picture it paints of presidential integrity. It’s all in the answer, not the working.
Which perhaps goes back to Silver’s idea of “product”. The danger of data journalism is that you treat the data as the journalistic act in itself, leaving others to sweep in and establish the narrative. This issue is compounded by the tendency of people who really care about data journalism to actively wish to avoid editorialising. Let the data do the talking, comes the mantra. But that relies upon a degree of active engagement that feels unlikely in the modern era. I’m something of a news junkie, but even I struggle to engage with data-heavy stories. I’d much rather the data provided the foundations, built the narrative, and that, then, the act of journalism took over.
I still think that data-based journalism is an important antidote (or counterpoint) to the opinion-based journalism that predominates these days. Opinion is easy: you can get affirmatory clicks from within your bubble and hate clicks from without, and because an opinionated column takes a couple of hours to write, it costs a couple of hours worth of a single writer’s labour. Data is hard: not only is the process of accumulating and verifying the data slow, it requires a dedicated team to extract the nutritional florets before it can be fed, stem and all, to the general public. Expensive and of unreliable interest is not a great combo in the current climate.
But at the same time as data journalism seems to be contracting, we are ever more subject to a data-driven world. More data is being accumulated all the time. British mathematician Clive Humby famously remarked that “data is the oil of the 21st century". And political polling is going to happen, whether it’s commissioned by newspapers and news channels or not (the ‘not’ meaning that it will be overseen by political parties, special interest groups and shadowy lobbying firms). The data is going to be created, inexorably. The task of journalism, therefore, is one of communication. Our world is ever more taxonomical, ever more encoded. Do you want this information to be reserved for plutocrats and faceless corporate operators? Or do we want a middle-person who can parse these trends so that Joe Everyman, recovering from covid-induced graph addiction, can understand the world he lives in?
Data journalism is an act of communication. Along the way, too many people have lost sight of the words amidst the numbers, the picture within the diagram. Data journalism needs to be resold, both to the public and to the industry, now that it’s no longer life and death, because otherwise it becomes a niche pursuit to whet the blade of professional gamblers, rather than a method of shaping how we see our world.
Follow me on Bluesky, if you’re using that platform. Otherwise, log off and have a nice weekend.