Last week we wrote about bias by omission — the facts an outlet leaves out within a story it chooses to cover. The natural follow-up question is: what about the stories an outlet never covers at all? That is selection bias. It is the most upstream form of media bias, and it is also the easiest to miss, because there is nothing on the page to argue with. The story simply isn't there.
This post defines selection bias, distinguishes it from framing bias and bias by omission, and walks through five concrete examples from this week's news cycle. The examples come straight from Signal/noise's multi-source coverage data — no abstractions, no hypotheticals. We picked them on both sides of the aisle on purpose, because selection bias is not a left problem or a right problem. It is a feature of every newsroom that has finite reporters and infinite stories.
What is selection bias in news?
Selection bias is what happens when a news outlet decides which events qualify as news. Editors triage incoming stories every hour. Some get the front page, some get a wire brief, and some get nothing. The pattern of those decisions — over weeks, over months — produces a worldview. If you only read outlets that consistently skip the same kinds of stories, you end up with a coherent but partial map of reality.
Three types of bias show up in news, and they are easy to confuse:
- Selection bias — Did this story get covered at all? Decided at the assignment desk. - [Bias by omission](https://s2n.news/blog/what-is-bias-by-omission-types-of-media-bias) — Within a story that did get covered, which facts were left out? Decided at the reporting and editing stage. - Framing bias — How is the same set of facts characterized? Decided at the headline and lede.
You can think of them as a funnel. Selection bias narrows the universe of stories. Bias by omission narrows the universe of facts inside each story. Framing bias shapes how those facts land. To detect bias well, you need to watch all three layers — and selection bias is the one most readers never see, because spotting it requires comparing what one outlet ran against what another outlet ran on the same day.
This is exactly the kind of pattern Signal/noise's news blind spot tracking is built to surface. When we mark a story with a "blindspot: right" or "blindspot: left" tag, we are flagging selection bias in real time.
5 real examples from this week
1. Trump's portrait on limited-edition US passports
The U.S. government announced limited-edition passports featuring a portrait of President Trump tied to the 250th anniversary of the Declaration of Independence. The Washington Post, the Guardian, Al Jazeera, Deutsche Welle, and the Straight Arrow News all ran versions of the story. Right-leaning U.S. outlets ran zero coverage in our source set. Center-right outlets also went dark on it.
The framing on the left side was relatively neutral — the Washington Post used "featured inside new passports," and the Guardian noted the visual context of "declaration text and flag motifs." But the framing-versus-no-framing comparison misses the bigger point. The reason this story has a left lean is not because left-leaning outlets covered it harshly. It is because right-leaning outlets did not pick it up. That's selection bias.
2. Musk vs. Altman trial begins
A federal trial began in San Francisco in Elon Musk's lawsuit against Sam Altman and OpenAI, with Musk alleging that Altman diverted what was founded as a public-benefit nonprofit into a for-profit enterprise. The New York Times and the BBC covered the opening day. Right-leaning outlets, including Fox News and the New York Post — both of which have given Musk extensive coverage on other beats — published nothing in our source set.
Selection bias often shows up in stories that don't fit an outlet's existing frame for a public figure. Musk-as-litigant against a tech CEO is harder to slot into an existing narrative than Musk-as-Twitter-owner or Musk-as-government-efficiency-commissioner. That awkwardness is exactly the kind of friction that produces silence.
3. The PEPFAR / HIV funding cuts data
Vox and the Guardian reported on downstream data from the Trump administration's cuts to PEPFAR and the dismantling of federal HIV surveillance infrastructure. No right-leaning outlet in our source set published anything. No center-right outlet did either.
This is one of the cleanest selection-bias signals you can ask for: a policy story with quantifiable health consequences, covered with primary data on one side of the spectrum and ignored entirely on the other. There is no framing to compare. There is only a story and a non-story.
4. Trump's alleged assassination plot and left-wing social media reaction
Fox News and the New York Post both covered an alleged assassination attempt at the White House Correspondents' Dinner and the social media reaction that followed — Fox cited an unnamed analysis claiming roughly one in five left-wing posts characterized the plot as a hoax. Left and center-left outlets in our source set did not cover this angle of the story.
This is the mirror image of the first three examples. Selection bias on the left looks like declining to engage with stories whose premise itself is contested — in this case, "left-wing reaction to a plot against Trump." The Fox piece's evidentiary base is genuinely thin, as our analysis flagged. But thin evidence is a reason to cover a story carefully, not to leave it uncovered. When one side of the spectrum produces all the coverage of an event, the resulting record skews — regardless of which side it is.
5. The Dunmurry bombing and Iran-Hezbollah connection
The New York Post and Fox News led on a bombing at a Police Service of Northern Ireland station in Dunmurry, claimed by a "pop-up" group called HAYI with suspected links to Iran and Hezbollah. The Week ran a center-left summary. No left-leaning U.S. outlet covered it in our source set. No center-right outlet covered it either.
Stories that fit a "global militant network tied to Iran" frame travel further on the right than on the left, especially during an active U.S.-Iran war. That is not a conspiracy. It is editorial judgment about which threats are most newsworthy. But the cumulative effect across a year is a readership that has either been told a great deal about Iran-linked terror networks, or almost nothing — depending on which outlets they read.
How to spot selection bias yourself
Three habits help. First, scan headlines from at least one outlet outside your usual rotation every day. You will not always learn that the other side has the better take — sometimes you'll learn that the other side has no take, which is the data point. Second, when you read a story that strikes you as obviously important, search for it in two or three other outlets before you share it. If only outlets that share your political lean covered it, that's a flag — not necessarily that the story is wrong, but that the rest of the press did not consider it news. Third, pay attention to source counts, not just source identities. A story with five right-leaning sources and zero left-leaning sources is usually a different kind of story than one with five sources from each side.
Tools help too. Our practical guide to detecting media bias walks through the workflow we use, and Signal/noise's story view shows the source spread on every story so you can see selection bias the moment it happens. Selection bias is the hardest type of bias to see precisely because it shows up as silence. Once you start looking for the silence, you start hearing it everywhere.