Pre-registration

Does the Full Moon Make People Act Up? A Pre-Registration

October 26, 2026

Results are in: Does the Full Moon Make People Act Up? The Results. This box was added after the analysis ran. Nothing else on this page has changed since it was published.

Ask someone who works nights in an ER, a police dispatch center or a classroom whether the full moon matters, and many will say yes without hesitating. This is a test of whether the data agrees with them.

This page is written before any of the data has been compared to the moon. I checked which datasets exist, how far back they go and how they’re structured, and nothing else. The method below is fixed now so I can’t steer the answer later, the same way I ran my Mercury retrograde test.^[22]

Numbers in brackets point to the sources listed at the bottom of the page.

The claim being tested

Belief is strongest among the people who see the most chaos up close:

• ER staff: 80% of emergency nurses and 64% of emergency doctors in a 1987 survey.^[1]
• Surgical nurses and mental health professionals: 69% and 81%.^[2]
• Teachers and childcare staff: 65% in a UK poll.^[3]

The claimed effects are wide: busier ERs, psychiatric crises, more violence, more police calls, more births, wilder kids, worse sleep. The reasons offered range from the moon’s pull on the water in our bodies and extra light at night^[5] to astrology’s view of the full moon as an emotional “culmination and release.”^[4]

I’m testing four versions that public data can reach:

1. People are moodier.
2. There’s more violence.
3. Police get busier.
4. The ER gets busier.

Births are tested as a secondary claim.

What has been done before

• Most large studies find nothing. A 1985 review of about 37 studies found that the moon explains 1% or less of the variation in behavior.^[6] A follow-up found that about half of the positive studies had statistical errors.^[7]
• Single-claim checks agree:
  • 150,999 ER visits showed no effect.^[8]
  • 908,000 Australian police calls showed no effect.^[9]
  • Decades of births showed no effect.^[10]
• Sleep is the open question. Two small studies found people sleep less near the full moon; a larger one found no difference.^[11] I can’t test sleep here, because there’s no large public daily sleep record.
• Mood hasn’t been tested this way. Researchers have used daily Twitter happiness to study weekly and holiday patterns,^[13] but I found no published test of it against the moon.

The data

Notes on each:

• Hedonometer stopped in May 2023, when Twitter cut off research access. It reflects people who tweet, not everyone, and its days are most likely UTC, which the three-day window below absorbs.
• New York keeps records by the date a crime was reported. I count them by the date it happened, and stop in March 2026 so late reports have time to come in.
• Seattle logs one row per responding unit, so calls are de-duplicated to one per event.
• New Orleans only marks officer-started calls from 2014, so it starts in 2014.
• NEISS only covers injuries involving a consumer product (falls, bikes, sports gear, tools and so on), not every ER visit. Its download site blocks automated access, so I still have to retrieve the files by hand. If I can’t get them, the ER test is dropped and reported as a change to this plan.
• Births after 2014 are not published by exact date. Many births are also scheduled (inductions and C-sections), which pushes against any moon effect.

How the full moon is defined

• The full moon is the moment the moon is opposite the sun, computed from NASA JPL’s planetary data.^[20]
• The full-moon window is the local calendar date of that moment plus the day before and the day after: three days out of each 29.5-day cycle, about 10% of all days. Believers usually talk about “around” the full moon, not one exact night.
• Local means each city’s own time zone. Hedonometer uses UTC; NEISS and births use US Central time.

How it gets compared

• The core number: how much higher or lower each measure is inside the full-moon window than outside it. For mood, the gap is reported in happiness points and as a share of a typical day’s swing. For counts, it is reported as a percentage. Every number comes with a range of plausible values.
• A fair comparison: each day is compared with others in the same city and month, and on the same day of the week. Major holidays such as New Year’s, the Fourth of July, Halloween, Thanksgiving and Christmas are accounted for separately. Weekends and holidays move all of these measures far more than any plausible moon effect, so they can’t be allowed to masquerade as one.
• Could chance do this? In each lunar cycle, the three-day window is moved to a random spot, 10,000 times. I report how often a random placement shows a gap as large as the real one.
• Four main tests: mood, violence, police calls and the ER. With four chances to find something, the bar for “unlikely to be chance” is raised (Holm correction).^[21]
• No minimum effect size. I report the size, the range and how often chance matches it.
• Year by year: the same gap is estimated separately for every calendar year, and I report how many years lean in the direction believers expect. If the full moon really did something, it should lean the same way most years, not flip back and forth.
• Is it predictive? For each measure, the model is built on the earlier two-thirds of the years and asked to forecast the last third, with and without the full-moon calendar. The full moon counts as predictive only if the real calendar forecasts better than at least 95% of the randomly placed calendars.
• Causation is out of scope. This can show whether these things line up with the moon, not why.

Secondary tests

These are adjusted for multiple testing among themselves, and reported whichever way they come out:

• The full-moon day alone, without the days on either side, for each of the four measures.
• Moon brightness: the fraction of the moon lit each night, as a sliding scale instead of a window.
• Full moon against new moon: the two extremes compared directly.
• Night only: violence and police calls between 6 p.m. and 6 a.m., when moonlight could plausibly matter.
• Births: the same three-day window.

Anything else I look at, such as individual cities or the zodiac sign the full moon falls in, will be labelled as exploration, not as a test.

Not tested here: sleep, suicides and psychiatric admissions. There’s no public daily data for any of them in recent years. “The full moon affects me” is also out of scope: that would need people tracking themselves day by day.

My prediction

Some years will lean toward the full moon and some will lean away, depending on the year. Across the whole record the differences will be small, inconsistent and no bigger than chance produces, and the full moon won’t help forecast any of these measures. In short: the full moon is not predictive of mood, violence, police calls or ER visits.

What happens next

I’ll pull the data, run exactly what’s above, and publish the results whichever way they fall, along with the code, the data and any place I had to change the plan and why.

A note on blinding

While checking the datasets, I downloaded the birth files to confirm their columns and dates. An automated check also printed Hedonometer’s daily tweet-volume column while counting missing days. Neither was compared to the moon, and tweet volume isn’t one of the measures.

A note on how this was written

Based on the way this is written, you can probably tell it was written heavily by AI. The structure of everything is built by me in conjunction with AI, and nothing is published without my final approval.