Every birthday.

The default is France’s complete 2025 birth year, from INSEE’s September 2026 release. The selector also offers 31 years of England and Wales daily births through 2025 (ONS Table 12, released September 8, 2026), provisional January–July 2026, Denmark through 2025, metropolitan French history, and 42 observed U.S. years. The source table lists coverage and limitations. Birth events, living-population stocks and surveys are kept separate. France’s all-years comparison includes complete 2025 only. Provisional January–July 2026 remains separately selectable; rarity cards are withheld for that incomplete calendar.

Two daily U.S. birth datasets published by FiveThirtyEight: CDC/NCHS for 1994–2003 and the Social Security Administration for 2000–2014. The original U.S. view uses only CDC/NCHS 1994–1999, then SSA 2000–2014: 85,712,738 recorded births on 7,670 dates. The overlapping 2000–2003 years are counted once.

The sources have different administrative coverage and totals. Use the source selector to check whether a pattern survives that choice. The early U.S. extension uses summed record weights from the upstream natality query. ONS birth-registration tables can include late-registration effects. These are historical U.S. birth counts, not global counts, a census of people alive today, or current live estimates.

Request data access ↗

FiveThirtyEight data: CC BY 4.0. Raw numeric records are preserved; this app adds aggregation and modeling. Retrieved September 9, 2026.

Average per occurrence divides a date’s total births by the number of times that date appears in the selected data. In the combined view, September 9 appears 21 times; February 29 appears five times. This makes seasonal comparisons fair, but it is not the probability that a randomly chosen person has that birthday.

Total births in sample keeps the actual count. February 29 is far rarer by this measure. Ranks follow your chosen measure, use unrounded values, and give equal ranks to exact ties. Dates absent from a selected year are marked as missing, not zero.

Card colors use fixed labels: Ultra rare (top 1%), Very rare (top 5%), Rare (top 10%), Uncommon (the remainder of the rarer half), Very common (top 25% most common), and Common (the remaining common half). A top-25% rarity badge is displayed only when its exact threshold is met. Exact ties are kept together; thresholds compare observed calendar dates, not people.

Share-card rarity always uses total births: #1 is the rarest date. The percentile is the proportion of other observed calendar dates with strictly more births. Ties receive the same rank and do not count as more common; a top-X% badge includes the full tied group. This compares calendar dates, not percentiles of people. Unobserved dates are excluded.

Month bars always show births per observed calendar day, so a long month does not win just for being long. “1 in N” divides total births in the sample by births on the selected date, rounded to the nearest whole number. It is not a current world-population estimate. The birthday picker asks for month and day only. Use “Recorded year” to filter the birth records.

The default is a conventional approximation: 40 weeks minus an assumed two weeks before conception = 266 days. The NIH’s pregnancy overview dates the usual 40 weeks from the last menstrual period. That is not the same clock as conception.

A 2013 study by Jukic and colleagues followed 125 naturally conceived singleton pregnancies and reported a median of 268 days from ovulation to birth, with a 37-day range even after excluding preterm births. It was a small, specific cohort, not a nationally representative duration distribution. Try the 268-day study median ↗

Our ±14-day default is an illustrative assumption, not a fitted gestation distribution or a confidence interval. Each lag from 252 to 280 days gets equal weight. Change the center and spread, including a literal single-lag rewind, to see how fragile precise dates are. The window does not cover every possible pregnancy length.

For each real calendar date c, average the recorded births at dates c + lag − spread through c + lag + spread. Retain only dates whose entire window has birth data. Group these modeled daily values by month and day, and average across their observed occurrences.

Divide each date’s modeled average by the overall modeled daily mean and multiply by 100. 100 = an average modeled day. An index of 110 means 10% above that mean, not 110 conceptions. Dates use UTC arithmetic, including real leap days; incomplete windows at the dataset edges are excluded.

This is a shifted, smoothed birth-season pattern. It does not recover actual conception dates. Birth scheduling, premature delivery, pregnancy losses, fertility treatment, and changing population size are not corrected. The calendar model does not use individual gestational ages or dates of sex; clinical records are analyzed separately. Smoothing cannot remove those confounders. Read the calculation code ↗

No. The calendar’s birth files contain dates of birth, not sexual activity or conception. The CDC’s 2013 birth-timing analysis shows that delivery timing differs by interventions and setting. A birth calendar carries hospital scheduling as well as biology.

Even time-of-day birth data cannot reveal time-of-day sex. The app explores seasonality among births that were recorded; it cannot estimate all conceptions, all sexual activity, or the cause of a holiday pattern. It is not a way to time, prevent, or diagnose a pregnancy.

Additional data: the research includes birth date × gestation × delivery for a 2019 Brazilian regional mirror, and recent clinical detail from the full 2024 U.S. file at month/weekday/hour resolution. A newer nationwide exact-date clinical file remains the main acquisition gap. The current NSFG public pregnancy file groups gestation and releases pregnancy dates as years. Source notes and additional detail are available by request.