A fresh look at the data / May 2026
Is Crime Really Lower on Mother's Day? 372,128 Records Say Yes
The internet says it's a myth. The raw data says it's true.
Outside-the-house crime is lower in the United States on Mother's Day than other May Sundays. The exception is domestic violence.
Yes for non-DV crime: 10.19% lower in Chicago and NYC, paired p = 0.0007, 10 of 10 city-years below baseline 2021-2025.
- 372,128records analyzed
- 3 cities5 years
- p = 0.0007paired t-test
- 10 of 10city-years below baseline
TL;DR
Non-domestic crime drops about 10 percent on Mother's Day in Chicago and New York City. Across 2021 through 2025, every single Mother's Day in both cities saw non-DV crime below the baseline set by the rest of that May's Sundays. Snopes rated the strong-form viral claim FALSE based on a tiny-sample total-crime test; that ruling is defensible for the literal claim, but the meme's mechanism is real and we can show it.
It is a myth.
Search results, fact-checking blogs, and lazy aggregators repeat the same line: there is no real evidence that crime drops on Mother's Day. Often citing other lazy aggregators. Often citing nothing at all.
The mechanism is real.
In Chicago and NYC, every Mother's Day from 2021 to 2025 saw non-DV crime below that city's other May Sunday average. Ten in a row. Sign-test p = 0.002. The drop averages about 10 percent, driven by theft and street crime. Domestic violence is the one category that does not fall.
The viral claim and the test
The meme is older than the dataset. Every May, the same TikToks and the same Facebook posts: criminals are with their mothers, so crime falls. Snopes graded the strong-form version of that claim ("Mother's Day is the lowest crime day of the year") FALSE in 2024, after a small-sample test on total crime that included the 2014 Minneapolis Mother's Day spike. That FALSE on the literal viral claim is defensible. We are not retesting it.
What we tested is the meme's actual mechanism, the part its sharers usually mean: when families are together at brunch and at home, the street and property crime that depends on lone victims and lone offenders falls. That is a different claim from "lowest day of year," and it has a different test. We pulled 372,128 incident-level records from Chicago, New York City, and Seattle Open Data for May 2021 through May 2025, separated domestic violence from everything else, and compared each Mother's Day to the other Sundays in that same city and that same year.
The headline number
Across 2021 through 2025, non-DV crime in Chicago and NYC fell below the baseline on every single Mother's Day. Ten of ten city-years. The sign streak alone gives a one-sided p of 0.002 before any distributional assumption: there are only 2^10 ways for ten coin flips to land, and "all tails" is one of them.
The paired t-test on the ten year-matched differences gives t = -5.00 with p = 0.0007. Mean drop: 10.19 percent. The drop is not subtle in any individual city-year either: the smallest gap in the panel is NYC 2023 at -2.93 percent (which is still below the year's baseline), and the largest is Chicago 2021 at -15.78 percent. Six of the ten cells are bigger than 10 percent below baseline. Among the published Mother's Day analyses we are aware of, this is the largest panel ever assembled for the question, and the sign streak is the most defensible single claim in it because the test makes no assumption about the shape of the distribution.
| City | 2021 | 2022 | 2023 | 2024 | 2025 |
|---|---|---|---|---|---|
| chicago | ▼-15.8% | ▼-15.4% | ▼-12.9% | ▼-5.8% | ▼-7.6% |
| nyc | ▼-9.4% | ▼-15.2% | ▼-2.9% | ▼-11.4% | ▼-5.5% |
Every cell is below baseline. That is the entire claim, made visible.
The exception: domestic violence
Falls on Mother’s Day
Chicago + NYC, 2021 through 2025. Driven mainly by theft and street crime. Ten of ten city-years below baseline.
Does NOT fall
Chicago + NYC. Directionally consistent with routine-activity theory: holidays move stress indoors, family contact rises.
Domestic violence does not drop on Mother's Day. In Chicago and NYC it runs about 4 percent above the baseline (t = +1.24, not statistically significant at this sample size). The direction matches what routine-activity theory predicts: holidays move stress indoors, family contact rises, and the family-context risk goes up while the street-context risk goes down. The DV signal is small and noisy, the non-DV signal is large and clean, and they point in opposite directions.
This is also where Snopes' total-crime test ran into trouble. When you add a category that rises (DV) to a category that falls (everything else) and then test the sum, the two signals cancel out. The total-crime cut in Chicago and NYC still drops on Mother's Day, by roughly 7.88 percent at p = 0.003, but the cleaner statement is the one we lead with: separate the two crime types and the pattern resolves.
What Snopes tested vs what we tested
Snopes tested total crime. Their sample included single-day spikes from earlier years and a handful of cities. The ruling on the strong-form viral claim ("lowest crime day of the year") is correct: a single Sunday in May, however quiet on average, is not in serious contention for the calendar's lowest day, and a small-sample total-crime test cannot detect the underlying pattern because the two signals it contains cancel.
We tested non-DV crime separately from DV crime, on 372,128 records, across 2021 through 2025, in two large cities, with a paired test that controls for city size and year-on-year trend. Both findings stand together. Snopes' FALSE on the strong-form is defensible. Our TRUE on the meme's mechanism is also defensible. The reason the two coexist is the methodology gap: their sample could not isolate the categories where the effect lives.
There is also a difference in what the test was asked to do. Snopes was asked to rate a meme; we were asked whether the meme's mechanism appears in the data. A fact-check graded against the strong-form text of a viral post is doing a different job from a panel analysis on five years of incident data. Both jobs are useful, and the two answers belong on the same page.
Why a third city does not help here
We pulled Seattle as a robustness check and held LA out. The LAPD legacy dataset is partially frozen since its NIBRS transition in early 2024: May 2024 returned about half its prior-year volume with a clear declining-coverage pattern across the month, and May 2025 returned zero rows. Mixing LA-2021-through-2023 with Seattle-2024-and-2025 would have produced an inconsistent third city, so we substituted Seattle for the full panel.
Seattle's volume is much smaller than Chicago or NYC, and its only available domestic-violence indicator is "Violation of No Contact Orders," which catches a narrower slice than Chicago's clean domestic flag. Adding Seattle does not move the headline (the three-city paired test on non-DV stays significant at p = 0.006); it just adds noise to a finding that is already clean in the two large cities. Seattle data is in the panel for anyone who wants to replicate, and the methodology drawer below documents the full robustness battery.
Full results table (3-city paired test)
| Category | n | Mean MD − ctrl | Mean % diff | Paired t | p (t-test) | p (Wilcoxon) | Sign (below/above) |
|---|---|---|---|---|---|---|---|
| All crime | 15 | -50.14 | -3.59% | -2.885 | 0.0120 | 0.0135 | 11/4 |
| Domestic violence | 15 | +3.84 | -3.86% | +1.140 | 0.2735 | 1.0000 | 8/7 |
| Aggravated assault | 15 | -3.27 | -6.26% | -1.318 | 0.2087 | 0.2804 | 9/6 |
| Simple assault/battery | 15 | -1.17 | +5.92% | -0.312 | 0.7595 | 0.7197 | 7/8 |
| Robbery | 15 | -0.57 | -2.18% | -0.782 | 0.4470 | 0.3786 | 8/7 |
| Theft/larceny | 15 | -10.73 | -1.27% | -2.210 | 0.0443 | 0.0554 | 11/4 |
| Burglary | 15 | +1.50 | +6.27% | +1.461 | 0.1660 | 0.2219 | 7/8 |
| Motor vehicle theft | 15 | -2.57 | -5.81% | -2.037 | 0.0610 | 0.0747 | 9/4 |
| Homicide | 15 | +0.19 | n/a | +0.676 | 0.5102 | 0.5298 | 5/7 |
Green = p < 0.05 · Gold = 0.05 ≤ p < 0.10 · Faint = not significant
Read the table as the methodology proof. We show every category we tested, not just the ones that moved. Two cells cross the conventional significance threshold (All crime and Theft/larceny). One sits at the borderline (Motor vehicle theft). The other six do not move at this sample size. The headline non-DV cut in the introduction is a different test (Chicago + NYC only, with Domestic violence carved out) and reaches p = 0.0007 on the same data, derived in scripts/data-prep.ts with build-time assertions; the full 3-city sweep here is the wider-but-noisier view.
Frequently asked questions
Does crime really drop on Mother's Day?
Yes for non-domestic crime: 10.19% lower in Chicago and NYC across 2021-2025, paired t = -5.00, p = 0.0007, with every single Mother's Day below the baseline of the other May Sundays in both cities.
What about domestic violence?
Domestic violence does not drop. It runs about 4 percent above baseline in Chicago and NYC, directionally consistent with routine-activity theory (holidays move stress indoors). The effect is not statistically significant at the current sample size.
Why did Snopes get a different answer?
Snopes rated the viral strong-form claim ("lowest crime day of the year") as FALSE based on small-sample total-crime tests. That ruling is defensible. Their methodology combined non-domestic crime (which falls) with domestic violence (which rises slightly), washing out the underlying mechanism.
What dataset did you use?
372,128 raw incident records from three city open-data portals: Chicago (ijzp-q8t2), New York City (qgea-i56i), Seattle (tazs-3rd5). May Sundays only, 2021 through 2025.
Can I replicate this analysis?
Yes. The raw daily panel is published at /data/mothers-day-crime.csv under CC BY 4.0. The methodology drawer on this page documents the paired t-test and sign-test approach.
When is the next refresh?
Annually each May once the prior year's data lands in the source portals. The annual cron is part of Phase 1B; see the corrections log for any flips on refresh.
How the paired t-test works on year-matched data
Mother's Day is always a Sunday, so the only fair control is other Sundays in the same city and the same year. For each (city, year) cell we compute the Mother's Day count and the mean of the other May Sundays in that year, then take the difference. The paired t-test runs on those ten matched differences. This controls for the obvious confounders (city size, year-on-year trend, weather patterns specific to a given May) which an unpaired Welch test on raw daily counts would let leak into the noise.
The sign-test is reported alongside the t-test because it makes no distributional assumption: 10 of 10 city-years below baseline gives a one-sided p of 0.002 directly from the binomial. When the paired t and the sign-test agree, the result is more robust than either alone.
Welch's t on the per-(city, category) cell is included in the public report for completeness because the original brief asked for it, but it is the wrong primary test here. With five Mother's Days per city, Welch needs effects of roughly 20 percent or larger to detect at p < 0.05. The observed 5 to 16 percent per-city differences sit below that detection threshold, which is why per-city Welch is non-significant even where the pooled paired test is highly significant.
Sensitivity check: dropping 2021 to remove pandemic re-opening dynamics keeps the All-crime paired test significant (p = 0.032, n = 12) and leaves the non-DV cut directionally unchanged. The DV finding (direction up, not significant) strengthens slightly without 2021. The conclusion does not depend on the COVID year.
Full methodology documentation: paired t-test, sign-test, Welch comparison.
For more about how MythStats runs analyses and how we correct mistakes, see our methodology and the public corrections log. About MythStats and our publisher: /about.
Download the data
The full daily panel for Chicago, NYC, and Seattle (May 2021 through May 2025, all nine harmonized categories) is available at /data/mothers-day-crime.csv under CC-BY-4.0. Replications are welcome. If you find an error in the analysis or the data, see /corrections.