How news gets decided

Something like a hundred thousand things happened in the world today. You'll hear about nine of them.

The interesting question isn't whether that filtering is biased in the direction you suspect. It's that a filter has to exist at all, that it's built out of deadlines, staffing, habit, money and a set of professional rules of thumb that are barely a hundred years old, and that most of it is visible if you know where to look.

This is a description of how the machinery works. It's not a complaint about anyone.

The hole is a fixed size

Start with the constraint that shapes everything else. A television bulletin is thirty minutes minus adverts, which is about eighteen minutes of speech, which is maybe three thousand words. A newspaper front page holds one lead, two or three secondary stories and a photo. A radio headline sequence is ninety seconds.

The hole doesn't shrink on a quiet day and doesn't grow on a busy one. That's not a metaphor — it's a production reality, and it means "was this important enough" is never the actual question being asked in a newsroom. The actual question is "was this more important than the other things competing for the same slot, given what we already have shot, written and legally cleared".

Websites broke the physical limit and immediately rebuilt it, because attention is scarce even when pixels aren't. There's still a top of the page, still a running order, still an editor deciding what leads. Infinite space just moved the bottleneck from paper to eyeballs.

Somebody is standing at the gate

The psychologist Kurt Lewin came up with the idea of a "gate keeper" in the 1940s while studying, of all things, how food gets from the shop to a family's table. Things move along channels, he said, and at certain points somebody decides whether they go further.

In 1950 a researcher called David Manning White borrowed the term and did something obvious that nobody had done: he asked a wire editor at a small American paper to keep every piece of copy he rejected for a week, and to write down why. The editor, anonymised as "Mr Gates", threw out about nine-tenths of what came in. His reasons included "not interesting", "too vague", "propaganda", "he's too red", "don't care for suicide stories" and, repeatedly, "no space".

It was a study of one man, and later work complicated it a lot — showing that editors are far more constrained by routine, by what the competition is running and by what fits the shape of the paper than by personal taste. But the basic finding survived. Selection is the job. Somebody sits at each narrowing point in the pipe, and by the time a story reaches you it has been through several of them.

The dozen tests a story has to pass

In 1965 two Norwegian researchers, Johan Galtung and Mari Holmboe Ruge, took a set of foreign crises and asked what made some of them into news abroad. They ended up with a list of factors, and it's held up remarkably well as a description of what newsrooms actually respond to.

The main ones, in plain language:

• Frequency. An event that fits the news cycle gets covered; something unfolding over thirty years doesn't have a day when it happens.
• Threshold. It has to be big enough — and the bar rises with distance.
• Unambiguity. Stories that can be told cleanly beat stories that need six caveats.
• Meaningfulness. Cultural closeness. Somewhere you can pronounce, with people whose lives look like yours.
• Consonance. Things that match what the newsroom already expected. Anticipated stories get reported more easily than surprising ones, which is the opposite of what most people assume.
• Unexpectedness. Within the consonant range, the rare and the sudden win.
• Continuity. Once a story is running it stays running, at a lower bar than it took to start it.
• Composition. A bulletin needs a mix. A perfectly good foreign story gets dropped because there are already two foreign stories.
• Elites. Powerful countries and prominent people clear the threshold at a fraction of the size.
• Personification. Events attached to an individual travel; abstractions don't.
• Negativity. Bad news is more sudden, more unambiguous and more consensual than good news, which tends to be a gradual process nobody can date.

Tony Harcup and Deirdre O'Neill went back over the list against actual British newspaper content in 2001 and again in 2017, and found the shape was right but the modern set needed additions — celebrity, entertainment, good news in the human-interest sense, audience shareability, and the outlet's own agenda, meaning stories that fit the paper's declared campaigns and interests.

Read that list again and notice something. Not one of those factors is "true", "important" or "useful to know". They're all about shape.

News comes from where the reporters already are

Newsrooms don't scan the world. They can't. They station people at institutions that reliably produce events, and that decision, made once, determines most of what gets found for years afterwards.

Courts, police, parliament, the town hall, the stock exchange, the big clubs, the emergency services. These places have press offices, publish diaries, hold briefings at convenient hours and generate documents on a schedule. Sociologists who sat in newsrooms through the 1970s — Gaye Tuchman, Herbert Gans, Philip Schlesinger among them — all came back with the same observation: the great majority of what runs is pre-scheduled. Not accidents and revelations. Announcements, hearings, results, launches, verdicts, votes.

Which has a hard consequence. If a problem has no address — no institution obliged to publish about it, no press officer, no annual figures — it can be enormous and still generate almost no coverage. Not because anyone suppressed it. Because nothing about it ever landed on the diary.

The other half of that is public relations. As newsroom budgets fell, the number of people paid to supply pre-packaged material rose, and the ratio between them shifted sharply. Research done for Nick Davies's 2008 book Flat Earth News, carried out by researchers at Cardiff University, went through a large sample of British national newspaper stories and found the majority consisted wholly or mainly of agency copy or PR material, with only a minority showing evidence of independent reporting. You can argue about the exact numbers. The direction isn't seriously disputed.

The wire services wrote the rules everyone still uses

Two of the deepest habits in news exist because of the telegraph.

The Associated Press was set up in 1846 when New York papers pooled the cost of getting news from the Mexican war. Paul Julius Reuter opened in London in 1851, having previously run a service using carrier pigeons to bridge a gap in the telegraph network between Aachen and Brussels. Havas in Paris, founded 1835, became what's now Agence France-Presse. These agencies gathered once and sold to everyone.

That business model produced the inverted pyramid — most important fact first, then the next, tapering to detail — partly because telegraph lines failed mid-transmission and partly because an editor at the far end needed to cut from the bottom without rereading. It's an unnatural way to tell a story. Nobody talks like that. It survives because it was a wire protocol before it was a style.

The same model also made neutrality commercially necessary. If you're selling one dispatch to papers of every political stripe in every country, the dispatch cannot take a side, or half your customers won't buy it. The historian Michael Schudson traced how detachment hardened into a professional ideal in American journalism in the 1920s, complete with training, ethics codes and the idea of the reporter as a person without views. Whether it's achievable is argued about endlessly. Where it came from isn't especially mysterious.

Worth knowing when you read: a great deal of what appears under many different mastheads is the same wire copy, lightly rewritten. Two outlets agreeing doesn't always mean two outlets checked.

News is event-shaped, and most things aren't events

The format wants a moment. Something that happened, ideally yesterday, ideally with pictures, ideally with someone to interview.

Enormous amounts of what determines how life goes have no moment. Soil losing fertility, a currency slowly drifting, a demographic shift, antibiotic resistance building, a road that gradually becomes unsafe, a bureaucracy quietly getting worse at answering the phone. There's no day on which any of that occurred, so there's no peg, so there's no story, until it produces a disaster — at which point it gets covered as the disaster and not as the thirty years.

The old jibe about foreign coverage being "coups and earthquakes" is exactly this. It isn't that editors think a coup matters more than a decade of institutional decay. It's that a coup happens on Tuesday.

One knock-on effect is worth knowing about yourself. People estimate how common something is partly by how easily examples come to mind, so a rare event covered heavily feels frequent and a common one covered rarely feels rare. Your sense of what's dangerous is calibrated to newsworthiness, and newsworthiness is calibrated to unusualness. The two run in opposite directions by design.

Attention has an economy, and it changed

For most of the twentieth century, general news was cross-subsidised. Papers made money on classified advertising, display advertising and being the only bundle in town; some of that money paid for a court reporter and a foreign bureau, neither of which paid for itself.

The internet unbundled all three, and it hit local reporting hardest, because local papers had the least distinctive product and the most substitutable adverts. The Pew Research Center's count of American newspaper newsroom employees found it fell by more than half between 2008 and 2020. Similar contractions happened in a lot of countries. Whole categories of story — council meetings, magistrates' courts, planning decisions — stopped being covered by anyone at all, not through any editorial judgement but because the desk was closed.

What replaced the old economics is measurable. Every headline can be tested. Every story has a read-through rate. That's genuinely useful for finding out what audiences actually want as opposed to what editors assumed, and it also applies constant pressure towards whatever gets clicked, which is not the same thing as whatever's worth knowing.

Subscription models push differently again. When revenue comes from committed readers rather than passing traffic, the incentive shifts towards giving those particular readers something they value enough to keep paying for — which rewards depth, and also rewards knowing your audience very well. Every funding model shapes output. There isn't a neutral one.

Corrections never get the seat the error had

An error runs at the top of the page. The correction runs in six-point type on page two, two days later, and you've never read one.

There's a mechanical reason, and it's the same news values list as before. A correction fails almost every test. It's not new, not unexpected, not personified, not visual, and it has an institutional cost to the outlet running it. Nothing in the machinery is set up to give a retraction the same prominence as the claim, and pretending otherwise would require a rule that no commercial pressure supports.

Then psychology makes it worse. Experiments on what's called the continued influence effect consistently find that people who read a claim and then read a clear retraction still use the original claim in their later reasoning. The correction is understood, remembered as a correction, and the belief persists anyway. Repeating the false statement in order to deny it can even reinforce it.

The practical upshot is unglamorous. First reports of anything chaotic — casualty figures, causes, who did what — are wrong at a high rate, and the wrong version is the one that gets shared hardest, because it's arriving during peak interest. Waiting forty-eight hours costs you nothing.

Balance and accuracy pull against each other

One genuine dilemma, stated without a verdict, because reasonable people land in different places on it.

The neutrality norm says: present the competing positions and let the audience judge. Applied to a question where evidence is genuinely contested, that's exactly right. Applied to a question where the evidence is lopsided, giving two speakers equal time can leave an audience with the impression of a fifty-fifty split that doesn't exist in the underlying research. Critics call that false balance.

The reply is that deciding which questions are settled is itself an editorial judgement, one that outlets have got wrong before, and handing journalists that power has costs too. Broadcast regulators in some countries impose impartiality duties; in others broadcasting is openly partisan and nobody pretends otherwise. Print and online in most democracies has no such duty at all and never has — the tradition of newspapers having declared political positions is older than the tradition of them not having any.

Both failure modes are real. Any newsroom is choosing between them several times a week.

Reading it better

None of this needs you to distrust everything. It needs a few reflexes.

Ask where the story came from. A named reporter at the scene, a document, a press release, a wire, or another outlet's report — the piece usually tells you, in the byline, the dateline, and phrases like "it is understood" or "sources said", which are doing real work.

Ask what the peg was. Why today? If the answer is a report published today, the report is the story and someone chose to publish it now.

Ask what would have to happen for the opposite to be reported. If a study finding no effect would never have made the page, then seeing effects reported tells you less than it seems.

And separate the event from the framing. Two outlets can run identical facts with different emphasis, different quotes and different headline verbs, and both be accurate. That's usually what you're arguing about with someone when you think you're arguing about facts.

Nine stories out of a hundred thousand. Somebody picked them, on a deadline, out of what they had.