No difference

"Full moons bring grim headlines" — a working hunch

"Nothing good happens on a full moon" — you hear it in newsrooms, precinct break rooms, and from anyone coming off a night shift. The phrase *lunar lunacy* exists for a reason: the link between full moons and a world gone sideways has been felt since antiquity. People who face the news every night have earned their hunches, and we won't laugh them off from an armchair. Instead, this page tests the hunch on its own turf — the actual record of news coverage.

You can build a plausible story, too. Maybe crime and accidents rise on full-moon nights and darken the next day's front page. Maybe shorter sleep makes reporters and readers reach for grimmer stories. But the question is not whether a story can be told — it's whether decades and hundreds of millions of news records actually carry its fingerprints.

How do you measure "dark news"?

GDELT (the Global Database of Events, Language, and Tone) machine-reads news coverage worldwide and scores each event with a "tone" — how negative or positive the language around it is. It runs from 1979 to the present, accumulating hundreds of thousands of events per day.

For this page we take events located in Japan and in the United States, and compute, day by day, the share of events that fall in the darkest 20% of that year's tone distribution. That share is our "darkness of the day's news." Both the volume of coverage and the tone scale itself have changed over the decades (GDELT's historical archive uses a different tone scale than the current feed), so we compare year-relative shares, not counts or fixed thresholds.

How the verdict is computed

• From GDELT 1.0 events, we aggregate daily counts of events located in Japan and in the United States (since 1979)
• The metric is the share of negative events (events in the darkest 20% of the same year's tone distribution ÷ all events), which is insensitive to changes in coverage volume and tone scaling
• News has a strong weekly rhythm (weekend editions read differently) and long-term drift in volume and tone. So the expected value is the average for the same weekday × same year, and we compare observed ÷ expected
• Using the moon's age at each country's noon (Japan = JST, US = Eastern), days are classified as full-moon days (within ±24h of the instant), new-moon days, or neither, and each group's mean index is compared against normal (1.00)

See the methodology for the verdict thresholds.

A caveat: what this data does not measure

GDELT's tone measures the language of coverage, not the total amount of misfortune in the world. Also, full-moon days see more coverage of the moon itself — festivals, supermoons — as confirmed in our Wikipedia topic, and those stories tend to be cheerful. If anything, that should nudge full-moon days slightly *brighter*, not darker.

Where the hunch comes from

Even if the verdict says "no effect," the hunch isn't a lie. A full moon glimpsed after a night of grim headlines sticks in memory; the full moon over an uneventful night is forgotten. Confirmation bias works hardest on the most conscientious news-watchers — the memorable nights cut the deepest. All the data can do is correct that bias with a few decades of counting.

Sources

• The GDELT Project (GDELT 1.0 Event Database, since 1979)
• Tone definition: GDELT average tone; "negative" is counted as the darkest 20% of the same year's tone distribution (absorbing the scale difference between the historical archive and the current feed)
• Lunar phases computed in-house from Jean Meeus, "Astronomical Algorithms" (UTC)