USA Times’ Data Desk runs an automated, deterministic sentiment score — no human rating, no editorial judgment in the scoring itself — across every headline from five major outlets that names the sitting U.S. president by surname. This piece consolidates that ongoing project: one comparison table across outlets, and a second view of how the same outlet’s tone shifts from one administration to the next.
The headline finding: all five outlets cluster tightly. Scored on a 0–100 scale where 50 is neutral, every outlet in this dataset falls between 46.7 and 48.5 — a spread of under two points, and all five sit slightly below neutral. Outlets with very different public reputations for political lean produced nearly identical aggregate headline-tone scores.
Across outlets: all-time headline sentiment
| Outlet | Score (0–100) | Headlines (all-time) | Negative | Neutral | Positive |
|---|---|---|---|---|---|
| Fox News | 48.5 | 8,619 | 35.5% | 34.8% | 29.7% |
| Forbes | 48.4 | 5,168 | 32.0% | 42.1% | 25.8% |
| Washington Post | 48.1 | 2,233 | 34.8% | 37.4% | 27.8% |
| CNN | 47.7 | 10,876 | 33.5% | 41.6% | 24.8% |
| NPR | 46.7 | 7,693 | 37.1% | 38.5% | 24.4% |
Every outlet’s headlines skew negative more often than positive when discussing a sitting president — unsurprising given that political headlines are disproportionately about controversy, conflict, and problems rather than routine good news. What’s notable is how little separates the five: NPR, the lowest scorer, is only 1.8 points below Fox News, the highest.
Batmandir · The Circle Full access to theatres, museums & exhibitions. S2 · ongoing membership, without a Founder’s seat. Explore membership →The same outlet, four administrations
A second, complementary Data Desk project tracks whether an individual outlet’s tone shifts depending on which president is in office — comparing the same publication’s headlines across Obama’s second term, Trump’s first term, Biden’s term, and Trump’s second term to date. Two of the three outlets in this project have produced usable results so far:
| Outlet | Administration | Headlines matched | Mean tone score |
|---|---|---|---|
| Fox News (in progress) | Obama II | 5,833 | −8.0 |
| Trump I | 29,358 | −6.9 | |
| Biden | 23,275 | −13.0 | |
| Trump II | 12,228 | −5.4 | |
| NPR (final census) | Obama II | 2,532 | see note |
| Trump I | 9,749 | see note | |
| Biden | 3,889 | see note | |
| Trump II | 5,734 | see note |
On Fox News’s numbers, the pattern is striking: coverage of Biden (−13.0) ran measurably more negative than coverage of either Trump term (−6.9 and −5.4) or Obama’s second term (−8.0). That is the clearest within-outlet partisan skew in the data collected so far. NPR’s completed census of 21,904 headlines found wording “measurably more negative for both Trump periods” than for Obama or Biden — the opposite direction from Fox’s pattern — though NPR’s per-administration mean scores are still being finalized for direct numeric comparison. CNN’s equivalent per-administration breakdown remains in progress and has not yet reached a large enough completed sample to report reliably; we’ll fold in CNN’s numbers here once that crawl finishes rather than publish a partial figure.
One correction worth preserving: what actually happened to NPR’s funding
While auditing NPR’s headline archive, USA Times’ Data Desk ran into a factual claim worth stating precisely, since it’s frequently garbled in political discussion: Congress did not abolish NPR or eliminate NPR’s operating budget. Public Law 119-28 rescinded approximately $1.1 billion in fiscal 2026–2027 advance appropriations to the Corporation for Public Broadcasting — a separate entity that funds local stations and system infrastructure, some of which those stations use to buy NPR programming. NPR has historically reported that direct CPB and federal grants make up roughly 1% of its own annual operating budget. The cut to CPB funding was real; “Congress directly defunded all of NPR” is not an accurate description of it.
Methodology & limits
Each outlet’s dataset is built from that publication’s own first-party article sitemaps, inspecting every URL for a headline that explicitly names, by surname, the person who was the sitting U.S. president when the article was published. This is a strict name-match corpus, not a semantic “aboutness” study — a headline can discuss a president’s policies without naming him, and those headlines are excluded here. Sentiment is scored by an automated, deterministic lexical model (VADER-style compound scoring) with no human rating or editorial judgment applied to individual headlines. A negative or positive lexical score describes word-level tone only; it does not by itself establish favorable or unfavorable treatment, accuracy, fairness, or partisan intent, and it combines many different stories and contexts into one number. Some of these crawls (marked “in progress” above) are still collecting; figures will be updated as each dataset’s completed-span sample grows. Every number here is reproducible from the linked source data.
